The 18S rRNA m⁶A methyltransferase METTL5 promotes mouse embryonic stem cell differentiation

Abstract

RNA modifications represent a novel layer of regulation of gene expression. Functional experiments revealed that N⁶‐methyladenosine (m⁶A) on messenger RNA (mRNA) plays critical roles in cell fate determination and development. m⁶A mark also resides in the decoding center of 18S ribosomal RNA (rRNA); however, the biological function of m⁶A on 18S rRNA is still poorly understood. Here, we report that methyltransferase‐like 5 (METTL5) methylates 18S rRNA both in vivo and in vitro, which is consistent with previous reports. Deletion of Mettl5 causes a dramatic differentiation defect in mouse embryonic stem cells (mESCs). Mechanistically, the m⁶A deposited by METTL5 is involved in regulating the efficient translation of F‐box and WD repeat domain‐containing 7 (FBXW7), a key regulator of cell differentiation. Deficiency of METTL5 reduces FBXW7 levels and leads to the accumulation of its substrate c‐MYC, thereby delaying the onset of mESC differentiation. Our study uncovers an important role of METTL5‐mediated 18S m⁶A in mESC differentiation through translation regulation and provides new insight into the functional significance of rRNA m⁶A.

METTL5, the m6A methyltransferase for 18S rRNA, is critical for Fbxw7 translation and mESC differentiation.

Introduction

To date, over 150 different types of chemical modifications have been identified in cellular RNAs, which brings out a new field termed as “epitranscriptome” (Helm & Motorin, 2017). Methylation of the N⁶ position of adenosine (m⁶A), as one of the most abundant mRNA modifications, also occurs on noncoding RNAs (ncRNAs) such as ribosomal RNAs (rRNAs), transfer RNAs (tRNAs), and small nuclear RNAs (snRNAs) (Roundtree et al, 2017). The m⁶A modification on mRNAs has been extensively studied and demonstrated to regulate stability, splicing, nuclear export, and translation of mRNAs (Liu et al, 2015). Functional studies of m⁶A modification revealed that it is crucial for various physiological processes such as gametogenesis, neurogenesis, embryogenesis, and cell differentiation (Batista et al, 2014; Geula et al, 2015; Lin et al, 2017; Shi et al, 2018; Lan et al, 2019).

rRNAs are the most abundant RNA species in cells and were reported to undergo a series of post‐transcriptional modifications (Polikanov et al, 2015; Sloan et al, 2017). In general, eukaryotic rRNAs are highly decorated by 2′‐O‐methylation (Nm) and pseudouridine (Ψ) modifications, which are mostly introduced by box C/D and box H/ACA small nucleolar (sno)RNPs, respectively (Polikanov et al, 2015; Sloan et al, 2017). Previous study observed that loss of snoRNA‐guided rRNA modification causes developmental defects and embryonic lethality in zebrafish (Higa‐Nakamine et al, 2012). In addition, eukaryotic ribosomes contain several different types of base modifications, which are often installed by stand‐alone rRNA‐modifying enzymes (Sloan et al, 2017). Two site‐specific m⁶A modifications were found at position 1,832 (m⁶A_1,832) on 18S rRNA and position 4,220 (m⁶A_4,220) on 28S rRNA in human cells and other vertebrates (Maden, 1986, 1988). Recently, ZCCHC4 was identified to be the human 28S rRNA m⁶A_4,220 methyltransferase and is frequently overexpressed in tumors (Ma et al, 2019; Pinto et al, 2020). The depletion of ZCCHC4 significantly disrupts proper functions of translation in ribosomes, which in turn affects cell proliferation and cancer progression. Thus, rRNA modifications are critical in regulating ribosome function and intricately linked to development and tumorigenesis.

Mouse embryonic stem cells (mESCs) are derived from the inner cell mass (ICM) of late pre‐implantation blastocysts and have the capability to differentiate into a variety of cell types comprising the derivatives of three germ layers (Evans & Kaufman, 1981). The in vitro differentiation system has been used extensively to study stem cell pluripotency and cell fate decisions (Pedersen, 1994; Keller, 1995). Multiple regulatory mechanisms involving transcriptome, epitranscriptome, translatome, and proteome precisely orchestrate cell fate determination (Szutorisz et al, 2006; Lu et al, 2009; Meissner, 2010; Kojima et al, 2014; Shi et al, 2017). Among them, ubiquitin–proteasome system (UPS) emerges as an essential regulator of stem cell function (Buckley et al, 2012; Strikoudis et al, 2014). For example, F‐box and WD‐40 domain 7 (FBXW7), a recognition component of SCF (complex of SKP1, CUL1 and F‐box protein)‐type ubiquitin ligase, was identified as an important factor for ESC differentiation and cellular reprogramming in UPS‐targeted RNAi screens (Buckley et al, 2012). Following‐up studies showed that silencing of FBXW7 expression inhibits ESC differentiation by stabilization of its key substrate c‐MYC, which positively regulates pluripotency‐related networks to maintain ESC self‐renewal and hinders differentiation (Smith & Dalton, 2010; Fagnocchi et al, 2016). However, how Fbxw7 itself is regulated by upstream factors during these biological processes is still open to question.

Recently, several methyltransferase‐like (METTL) family proteins have been reported to play vital roles in regulating ESC pluripotency and early embryonic development. For instance, METTL3‐METTL14 methyltransferase complex ensures pluripotent cell differentiation, which is dependent on its m⁶A catalytic activity (Batista et al, 2014; Geula et al, 2015). mRNA m³C writer METTL8 inhibits c‐Jun N‐terminal kinase (JNK) pathway and affects mESC differentiation (Xu et al, 2017; Gu et al, 2018). U6 snRNA m⁶A methyltransferase METTL16 is required for mouse embryonic development via controlling the splicing of SAM synthetase MAT2A (Pendleton et al, 2017; Mendel et al, 2018).

Though a previous study firstly analyzed the atomic‐resolution structure of METTL5 and identified the 18S rRNA m⁶A role of METTL5 in human cancer cells during the preparation of our work (van Tran et al, 2019), the exact roles of 18S rRNA m⁶A on translation regulation are still largely unknown. Here, we showed that METTL5 is responsible for 18S rRNA methylation both in vivo and in vitro. Functionally, m⁶A‐dependent METTL5 endows mESCs with efficient translation for triggering multi‐lineage differentiation timely, which depicts the importance of rRNA m⁶A in cell fate transition.

Results

METTL5 is the N⁶‐adenosine methyltransferase targeting 18S rRNA

METTL5, as a METTL family member, has been reported to bind RNA rather than DNA (Franke et al, 2015). Through bioinformatic analysis, we identified that there is a typical [DNSH] PP [YFW] motif of m⁶A‐specific methyltransferase in METTL5 protein (Fig EV1A), which inspired us to search for its potential targets and biological function. Previous functional studies uncovered the crucial roles of METTL family members in regulating pluripotency of mESCs (Batista et al, 2014; Geula et al, 2015; Gu et al, 2018). We were thus prompted to use the mESC model to determine the function of METTL5.

Figure EV1. METTL5 is a putative N6‐adenosine methyltransferase.

1. Schematic representation of the functional domain of mouse METTL5. The NPPF motif was highlighted with a red box. APPA motif was the catalytically inactive mutant after changing key residues (126–129) “NPPF” to “APPA”.
2. PAGE gel electrophoresis showing the PCR amplification of the targeted region of Mettl5 locus from WT and Mettl5 KO mESCs.
3. Sequencing results of the targeted region Mettl5 locus detected in WT and KO clones in mESCs. Red colored letters indicate Mettl5 sgRNA sequence. Green colored and underlined letters indicate Mettl5 PAM sequence. Blue colored dots and letters indicate deletions and insertions, respectively. 15 TA clones of the PCR products were analyzed by Sanger sequence.
4. LC/MS analysis of the m⁶A level of total RNA from the WT and Mettl5 KO (KO‐6B) mESCs. Data are shown as mean ± SD from three biological replicates. Student's t test, two‐tailed. ***P < 0.001.
5. Real‐time fluorescence amplification curves of qPCR showing the SELECT results for detecting m⁶A_1,832 and A_1,825 sites (for input control).
6. Histogram showing the threshold cycles of qPCR (qPCRC_T) for detecting m⁶A_1,832 and A_1,825 sites in 18S rRNA of WT and Mettl5 KO (KO‐6B) mESCs, respectively. Data are shown as mean ± SD from three biological replicates. Student's t test, two‐tailed. ns, not significant; ***P < 0.001.
7. Coomassie blue staining showing the purified WT and catalytically mutant METTL5 proteins.

Source data are available online for this figure.

First, CRISPR/Cas9 system was used to target the second exon of Mettl5 to generate two Mettl5 knockout (KO) colonies (number KO‐2C and KO‐6B; Fig 1A). The successful depletion of Mettl5 was validated both by Sanger sequencing and Western blot (WB) using METTL5‐specific antibody (Figs 1B and EV1B and C). Consistent with our hypothesis that METTL5 may function as a m⁶A methyltransferase, we observed a dramatic decrease of m⁶A level in total RNA from Mettl5 KO cells compared with that in wild‐type (WT) cells (Fig EV1D). Considering the rRNA is the most abundant RNA type in the cell, we wondered whether the loss of m⁶A is actually from rRNA. To test this hypothesis, we quantified the m⁶A levels of purified 18S and 28S rRNA from WT and KO cells with LC‐MS/MS (QQQ). As expected, a dramatic decrease of m⁶A was detected in 18S rRNA from Mettl5 KO cells. In line with recent reports that ZCCHC4 is a specific 28S rRNA m⁶A methyltransferase (Ma et al, 2019; Pinto et al, 2020), we did not detect any noticeable changes of m⁶A level in 28S rRNA (Fig 1C). Furthermore, we employed SELECT method with the accuracy of single‐base resolution to confirm 18S rRNA m⁶A at position 1,832 indeed installed by METTL5 (Xiao et al, 2018). Consistently, the SELECT assay data indicated that m⁶A_1,832 level on 18S rRNA was significantly decreased upon Mettl5 KO, enabling the ligation products more sufficient for amplification (Figs 1D and EV1E and F). Then, we set out to investigate whether METTL5 can install a methyl group into the 18S rRNA in vitro. We purified the WT and catalytically inactive mutant (“NPPF” to “APPA”, Mut) recombinant METTL5 proteins from E. coli and then incubated them with synthetic 18S rRNA oligos, respectively (Fig EV1G). The in vitro assay results showed that WT METTL5 proteins have considerable methyltransferase activity toward 18S rRNA oligos. In sharp contrast, there was barely any activity detected in Mut METTL5 group (Fig 1E), indicating that METTL5 is a bona fide m⁶A methyltransferase for 18S rRNA in vivo and in vitro, which is similar with the recent report (van Tran et al, 2019).

Figure 1. METTL5 is the N⁶‐adenosine methyltransferase targeting 18S rRNA in vivo and in vitro .

1. Schematic diagram of sgRNA targeting mouse Mettl5 locus. sgRNA targeting site is highlighted in red. PAM sequence is underlined and highlighted in green.
2. Western blot analysis confirming two Mettl5 KO clones by specific METTL5 antibody. α‐TUBULIN was used as a loading control.
3. LC‐MS/MS analysis of m⁶A/A levels in purified 18S rRNA and 28S rRNA from the WT and Mettl5 KO (KO‐6B) mESCs.
4. PAGE gel electrophoresis showing the PCR amplification of the elongated and ligated products of SELECT method for detecting m⁶A_1,832 site and A_1,825 site (for input control) in 18S rRNA of WT and Mettl5 KO (KO‐6B) mESCs, respectively.
5. LC‐MS/MS analysis of d3‐m⁶A/A levels in 18S rRNA oligos after being incubated with recombinant WT or catalytically mutant mouse METTL5 proteins in vitro. No enzyme group was used as a negative control.

Source data are available online for this figure.

Mettl5 KO mESCs exhibit unaffected pluripotency and self‐renewal

Upon deleting Mettl5, we firstly noticed that mESCs still maintained nest, round colony morphologies (Fig 2A). Alkaline phosphatase (AP) staining showed that Mettl5 KO mESCs had clearly positive reactivities similar to those of WT mESCs when cultured in the naive 2i/Lif condition (Fig 2B). Through detecting the mRNA and protein levels of pluripotency markers such as POU5F1, SOX2, NANOG, and ESRRB by real‐time quantitative PCR (RT–qPCR) and WB, respectively, we found that expressions of pluripotency‐related factors were almost unaltered in Mettl5 KO mESCs (Fig 2C and D). Moreover, immunostaining analysis of POU5F1, NANOG, and SOX2 in mESCs indicated that depletion of Mettl5 does not disrupt the undifferentiated status of mESCs (Fig 2E). Additionally, Mettl5 KO mESCs also displayed similar pluripotency as WT mESCs when cultured in the conventional serum/Lif condition (Fig EV2A–D). Hence, these results indicated that METTL5 is dispensable for maintaining pluripotency and self‐renewal of mESCs.

Figure 2. Characterizations of the Mettl5 KO mESCs.

• A, B
  Representative bright‐filed (A) and AP staining (B) images of the WT and Mettl5 KO mESCs cultured in 2i/Lif medium. Scale bars represent 100 and 50 μm, respectively.
• C
  RT–qPCR analysis of the expression of pluripotency genes in WT and Mettl5 KO mESCs. Data are represented as mean ± SD from three biological replicates. Student's t test, two‐tailed. ns, not significant; ***P < 0.001.
• D
  Western blot analysis of the expression of pluripotency markers in WT and Mettl5 KO mESCs. α‐TUBULIN was used as a loading control.
• E
  Immunostaining analysis of the expression of pluripotency markers (POU5F1, SOX2 and NANOG) in WT and Mettl5 KO mESCs, respectively. DAPI (blue) was used as a nuclear counterstain. Scale bars represent 10 μm.

Source data are available online for this figure.

Figure EV2. Characterizations of the Mettl5 KO mESCs in serum/Lif condition.

• A, B
  Representative bright‐filed (A) and AP staining (B) images of WT and Mettl5 KO mESCs. Scale bars represent 100 and 50 μm, respectively.
• C
  RT–qPCR analysis of the expression of pluripotency genes. Data are represented as mean ± SD from three biological replicates. Student's t test, two‐tailed. ns, not significant; ***P < 0.001.
• D
  Western blot analysis showing the expression of pluripotency markers in WT and Mettl5 KO mESCs. α‐TUBULIN was used as a loading control.

Source data are available online for this figure.

Depletion of Mettl5 impairs mESC early differentiation

We next sought to explore the role of METTL5 in differentiation potential using a three‐dimensional (3D) model of inducing embryoid body (EB) formation. Due to the greater differentiation capacity than serum/Lif‐grown mESCs, 2i/Lif‐grown mESCs were chosen to induce differentiation (Marks et al, 2012). The mESCs can be spontaneously differentiated into multi‐lineage cells under the defined serum‐free suspension culture (SFEB) condition after withdrawal of all cytokines (Fig EV3A) (Pedersen, 1994; Keller, 1995). As expected, POU5F1 expression gradually decreased during the course of differentiation, indicating that the cell aggregates were switching off pluripotency gene regulatory networks. Intriguingly, the endogenous METTL5 protein expression increased to its peak at the early differentiation stage of Day3 and then recovered to a previous level (Fig 3A). Consistently, the m⁶A_1,832 modification level was transiently increased in Day3 EBs, compared with that in other time points detected by SELECT (Figs 3B and EV3B and C), which implied that METTL5 might play a certain role in initiating early differentiation.

Figure EV3. Induction of embryoid body formation in vitro .

1. Representative bright‐filed images of EBs at different time points. Scale bars represent 20 μm.
2. Real‐time fluorescence amplification curves of qPCR showing the SELECT results for detecting m⁶A_1,832 and A_1,825 sites (for input control) during the indicated time points of EB induction.
3. Histogram showing the threshold cycles of qPCR (qPCRC_T) for detecting A_1,825 site in 18S rRNA during the indicated time points of EB induction. Data are shown as mean ± SD from three biological replicates. Student's t test, two‐tailed. ns, not significant.
4. Immunostaining analysis of the mesoderm marker T in control and Mettl5 KO EBs at different time points. DAPI (blue) was used as a nuclear counterstain. Scale bars represent 100 μm.
5. Quantification of percentages of positive cells shown in (D). Quantitative analysis was based on at least three independent experiments. Data are presented as mean ± SD. Student's t test, two‐tailed. **P < 0.01; ***P < 0.001.
6. Representative AP staining images of the control and Mettl5 KO Day7 EBs. Scale bars represent 20 μm.
7. Western blot analysis confirming overexpression of HA‐tagged WT or catalytically mutant (aa126–129: “NPPF” to “APPA”) METTL5 in KO mESCs, respectively. α‐TUBULIN was used as a loading control.
8. Western blot analysis showing the expression of pluripotency (POU5F1) and three germ layer markers (FOXA2, T and NESTIN) in Day7 KO, KO + HA‐MT5_WT, and KO + HA‐MT5_Mut EBs. GAPDH was used as a loading control.
9. H&E staining and immunostaining analysis revealing the differentiated structures of three germ layers in teratomas derived from WT and Mettl5 KO mESCs. The mESCs were injected subcutaneously into the flanks of immunodeficient mice (n = 6 per group). Scale bars represent 50 μm (H&E staining) and 20 μm (immunostaining), respectively.

Source data are available online for this figure.

Figure 3. Mettl5 KO impairs early differentiation of mESCs.

1. Western blot analysis showing dynamic expression patterns of POU5F1 and METTL5 during the indicated time points of EB induction. α‐TUBULIN was used as a loading control.
2. Histogram showing the threshold cycles of qPCR (qPCRC_T) for detecting m⁶A_1,832 site in 18S rRNA during the indicated time points of EB induction. Data are shown as mean ± SD from three biological replicates. Student's t test, two‐tailed. ns, not significant; *P < 0.05.
3. Immunostaining analysis of the expression of pluripotency marker POU5F1(left), endoderm marker FOXA2 (middle), and neuroectoderm marker NESTIN (right) in control and Mettl5 KO EBs at different time points, respectively. DAPI (blue) was used as a nuclear counterstain. Scale bars represent 100 μm.
4. Quantification of the percentages of positive cells shown in (C). Quantitative analysis was based on at least three independent experiments. Data are represented as mean ± SD. Student's t test, two‐tailed. **P < 0.01; ***P < 0.001.
5. RT–qPCR analysis of the expressions of pluripotency and lineage‐specific markers in Day7 EBs. Data are represented as mean ± SD from three biological replicates. Student's t test, two‐tailed. ns, not significant; *P < 0.05; **P < 0.01; ***P < 0.001.
6. Western blot analysis showing pluripotency (POU5F1) and three germ layer markers (FOXA2, T and NESTIN) expression in Day7 EBs. α‐TUBULIN was used as a loading control.

Source data are available online for this figure.

To investigate the potential role of METTL5 during the differentiation, we quantified the expressions of pluripotency and lineage‐specific markers in EBs by immunostaining analysis at the indicated time points of differentiation. Encouragingly, we found that expression levels of pluripotency marker POU5F1 were strikingly higher than the controls after Day3, and lineage‐specific markers such as FOXA2 (for endoderm), T (for mesoderm), and NESTIN (for neuroectoderm) failed to be timely expressed in Mettl5 KO EBs (Figs 3C and D, and EV3D and E), suggesting that Mettl5 KO mESCs were resistant to the stimuli of triggering differentiation.

Then, Day7 EBs were harvested for more detailed analysis. AP staining showed the control EBs were undergoing differentiation and maturation, whereas Mettl5 KO EBs still remained high pluripotency with dark purple staining (Fig EV3F). Both of the RT–qPCR and WB analysis confirmed that the naive pluripotency markers (POU5F1, NANOG, and ESRRB) were still highly expressed, while the endoderm markers (FOXA2, SOX17, GATA4, and GATA6) and mesoderm markers (MIXL1, WNT3, and T) were severely suppressed in Mettl5 KO EBs, compared with the controls (Fig 3E and F). To determine whether the deficient differentiation was indeed caused by the loss of 18S rRNA m⁶A, we performed a rescue experiment by overexpressing either HA‐tagged WT or catalytically mutant (“NPPF” to “APPA”, Mut) METTL5 in KO cells (KO + HA‐MT5_WT or KO + HA‐MT5_Mut; Fig EV3G). As shown in Fig EV3H, differentiation block was rescued by re‐expressing HA‐tagged WT METLL5 rather than Mut METTL5, demonstrating that the regulation is dependent on the m⁶A catalytic activity of METTL5.

Nevertheless, we analyzed the 6‐week teratomas derived from WT and Mettl5 KO mESCs. Histological analysis of sections revealed that both groups were composed of characteristic lineage‐specific morphologies and markers such as epithelia (for endoderm), muscle (for mesoderm), and epidermis (for ectoderm) (Fig EV3I). Therefore, Mettl5 KO does not fully inhibit differentiation potential in vivo and mESCs can ultimately differentiate into three germ layers.

Collectively, these data suggested that disruption of m⁶A‐dependent METTL5 delays pluripotency exit and impairs germ layer specification of mESCs.

Depletion of Mettl5 affects translation efficiency of mRNAs

Considering that 18S A_1,832 is localized in the decoding center (DC) (Piekna‐Przybylska et al, 2008), we thus focused on the role of METTL5 in the translation regulation. Ribosome profiling sequencing (Ribo‐seq) was performed to chart the translational profiles in WT and Mettl5 KO mESCs. The initial quality control showed that a majority of footprints mapped to coding sequences (CDS) with the strong 3‐nt periodicity distribution (Fig EV4A–C). Thus, the data quality was good enough for downstream analysis (Calviello & Ohler, 2017). The translation efficiency correlation analysis showed there was a good consistency between two replicates in both WT and Mettl5 KO cells (Fig EV4D and E). Principal component analysis (PCA) revealed that two replicates from WT and KO groups were clustered, respectively, whereas the two groups were separated clearly from each other, reflecting a large divergence between the WT and Mettl5 KO cells (Fig EV4F). We identified 1,339 differentially translated genes (1,169 down‐regulated and 170 up‐regulated genes) between WT and KO cells in a total of 9,093 actively transcribed genes with high confidence (≥ 1.5‐fold change and P < 0.05) (Fig 4A and Dataset EV1). Notably, core stemness genes such as Pou5f1, Nanog, Sox2, and Esrrb exhibited non‐significantly translational changes in Mettl5 KO mESCs compared with controls, in agreement with the findings as described in Fig 2.

Figure EV4. Quality control of Ribo‐seq data in mESCs.

• A
  The portions of mapped reads in different RNA species in WT and Mettl5 KO mESCs.
• B
  The statistics of mapped reads in different locations including CDS, 5′UTR, 3′UTR, intron, and intergenic regions in Ribo‐seq library.
• C
  The distribution of ribosome‐protected fragments (RPF) around translation starting site (TSS) and translation ending site (TES) in WT and Mettl5 KO cells.
• D, E
  Scatter plots of correlation analysis showing the translation efficiency correlation between two replicates in WT and Mettl5 KO cells, respectively. P values were inferred using Pearson correlation test implemented in cor.test function of R (https://www.r-project.org).
• F
  PCA analysis of the variability between WT and Mettl5 KO cells.

Figure 4. Depletion of Mettl5 affects the translation efficiency of Fbxw7 .

1. Volcano plot depicting down‐regulated (left) and up‐regulated (right) genes in Mettl5 KO mESCs compared with WT mESCs in Ribo‐seq. Significantly altered genes are defined as those ≥ 1.5‐fold change and P value < 0.05 (shown in red dots). Blue dots represent genes with < 1.5‐fold change and P value < 0.05. Green dots represent genes with ≥ 1.5‐fold change and P value ≥ 0.05. Black dots represent genes with < 1.5‐fold change and P value ≥ 0.05. P values were inferred from two‐sided Wald tests (DESeq2) and P value adjustments were made for multiple comparisons. Examples of up‐ or down‐regulated or non‐significantly changed genes are annotated in the graph.
2. Track plot showing translational changes of Fbxw7 in Mettl5 KO mESCs compared with controls.
3. RT–qPCR analysis of Fbxw7 mRNA levels in WT and Mettl5 KO cells during the indicated time points course of EB induction. Data are represented as mean ± SD from three biological replicates. Student's t test, two‐tailed. ns, not significant.
4. Western blot analysis showing FBXW7 expression in control and Mettl5 KO cells during the indicated time points of EB induction. GAPDH was used as a loading control.
5. RT–qPCR analysis showing Fbxw7 mRNA levels in Day7 KO, KO + HA‐MT5_WT, and KO + HA‐MT5_Mut EBs. Data are represented as mean ± SD from three biological replicates. Student's t test, two‐tailed. ns, not significant.
6. Western blot analysis showing FBXW7 expression in Day7 KO, KO + HA‐MT5_WT, and KO + HA‐MT5_Mut EBs. GAPDH was used as a loading control.

Source data are available online for this figure.

After exploiting the Ribo‐seq data, we noticed the translation efficiency of the Fbxw7 mRNA was severely suppressed in mESCs (Fig 4B). Given that FBXW7 was demonstrated as an important regulator in ensuring cell differentiation through degrading its substrate c‐MYC by UPS (Buckley et al, 2012), we speculated that Fbxw7 may be one of candidate genes that account for the compromised differentiation. In addition, there was a severe decrease of FBXW7 protein without a distinct alteration of mRNA levels, suggesting the translation inefficiency of Fbxw7 also existed in Mettl5 KO EBs (Fig 4C and D). Moreover, FBXW7 protein expression was rescued in KO + HA‐MT5_WT cells, but not in KO + HA‐MT5_Mut cells, which provided evidence that translation of Fbxw7 is directly dependent on m⁶A catalytic activity of METTL5 (Fig 4E and F).

FBXW7 induction partially rescues the defects and c‐MYC induction mimics differentiation block

To further address whether the differentiation defects were caused by the inefficient translation of Fbxw7, we examined the endogenous expression patterns of FBXW7 and c‐MYC in our differentiation system. The FBXW7 protein was weakly detected in Day1 EBs, but increased from Day3 onward. In contrast, c‐MYC protein was most highly expressed in Day1 EBs and subsequently declined to a nearly undetectable level in Day7 EBs (Fig 5A), suggesting there is an inverse regulation of FBXW7‐c‐MYC axis for the normal differentiation program. Importantly, the re‐expression of FBXW7 was concomitant with a decreased expression of c‐MYC in KO + HA‐MT5_WT cells, compared with that in Mettl5 KO cells (Fig 5B).

Figure 5. FBXW7 induction partially rescues the defects in Mettl5 KO mESCs while c‐MYC induction mimics the differentiation block.

1. Western blot analysis showing dynamic expression patterns of FBXW7 and c‐MYC during the indicated time points of EB induction. α‐TUBULIN was used as a loading control.
2. Western blot analysis showing FBXW7 and c‐MYC expression in control, Mettl5 KO (KO‐6B) and KO + HA‐MT5_WT Day7 EBs, respectively. α‐TUBULIN was used as a loading control.
3. Schematic diagram depicting induction of FBXW7 expression using rtTA‐dependent system in Mettl5 KO mESCs. Reverse tetracycline‐controlled trans‐activators (rtTA) can recognize the TET response element (TRE) and express FBXW7 at the presence of a derivative of tetracycline‐doxycycline (Dox).
4. RT–qPCR analysis of the expression of pluripotency and lineage‐specific markers in Day7 FBXW7‐induced EBs at the presence or absence of 10 ng/ml Dox. Data are represented as mean ± SD from three biological replicates. Student's t test, two‐tailed. *P < 0.05; **P < 0.01; ***P < 0.001.
5. Immunostaining analysis of the expression of pluripotency marker POU5F1, lineage‐specific markers FOXA2, T and NESTIN in Day7 FBXW7‐induced EBs at the presence or absence of 10 ng/ml Dox. Scale bars represent 100 μm.
6. Schematic diagram depicting induction of c‐MYC expression using rtTA‐dependent system in WT mESCs.
7. RT–qPCR analysis the expression of pluripotency and lineage‐specific markers in Day7 c‐MYC-induced EBs at the presence or absence of 10 ng/ml Dox. Data are represented as mean ± SD from three biological replicates. Student's t test, two‐tailed. *P < 0.05; **P < 0.01; ***P < 0.001.
8. Immunostaining analysis of the expression of pluripotency marker POU5F1, lineage‐specific markers FOXA2, T and NESTIN in Day7 c‐MYC-induced EBs at the presence or absence of 10 ng/ml Dox. Scale bars represent 100 μm.

Source data are available online for this figure.

Furthermore, we attempted to induce HA‐tagged FBXW7 expression in Mettl5 KO mESCs using rtTA‐dependent system (Fig 5C). As expected, the induction of FBXW7 by addition of Dox compensated the loss of METTL5 and EBs differentiated into three germ layers, preferentially the neuroectoderm (Fig 5D and E). On the other hand, we expressed Dox‐inducible c‐MYC in WT cells to test whether c‐MYC alone can mimic the Mettl5 KO phenotype using rtTA‐dependent system (Fig 5F). As shown in Fig 5G and H, there was a similar differentiation block in c‐MYC‐induced Day7 EBs confirmed both by RT–qPCR and immunostaining analysis. Hence, these data support the hypothesis that the precise expression of FBXW7‐c‐MYC axis is important for pluripotency exit, and METTL5 is required for the initial differentiation in vitro by regulating the translation of Fbxw7.

Discussion

Ribosomes are conserved protein‐synthesis machines across all kingdoms of life. The rRNA modifications such as 2′‐O‐methylation, pseudouridine, and base methylation are flexibly installed at different stages during ribosome biogenesis to ensure the stabilization of rRNA structure, thereby facilitating the efficiency and accuracy of translation (Sharma & Lafontaine, 2015). Importantly, rRNA modifications are not randomly distributed on the ribosomal subunits. They are preferentially concentrated in conserved functional sites including the DC, the peptidyl transfer center (PTC) and the inter‐subunit interface, suggesting that rRNA modifications may play important roles in promoting subunit maturation and function of ribosome (Polikanov et al, 2015; Sloan et al, 2017). In our work, we found that METTL5 catalyzes A_1,832 methylation on 18S rRNA (summarized in Fig 6A), which is located near the DC, prompting us to hunt for the functional significance of METTL5 in translation.

Figure 6. Summary model for the essential roles of METTL5.

1. Representation of METTL5's role in 18S rRNA. Large ribosomal subunit (LSU) and small ribosomal subunit (SSU) of the ribosome are shown in light pink and antique pink, respectively (left). The secondary structure of 18S rRNA in the SSU (middle) was adapted from (http://apollo.chemistry.gatech.edu/RibosomeGallery/). METTL5 (shown in brown) catalyzes A_1,832 methylation on 18S rRNA in the SSU of the ribosome.
2. Schematic illustration depicting the regulation of differentiation potential for endoderm (endo), mesoderm (meso), and ectoderm (ecto) by METTL5 via the translational control of FBXW7‐ c-MYC axis. Dotted arrows indicate inefficient differentiation induction.

Identification and characterization of novel factors to initiate differentiation cascades is a central issue in stem cell biology. We found that METTL5 is dispensable for the maintenance of pluripotency but required for proper differentiation. Given the specificity of 18S m⁶A_1,832 site catalyzed by METTL5, it is tempting to ask whether the differentiation resistance is caused at the translation layer in Mettl5 KO mESCs. Our data revealed the translation of Fbxw7 regulated by m⁶A‐dependent METTL5 is important for the proper mESC differentiation (summarized in Fig 6B). Previous studies have evidenced that some rRNA modifications are crucial for the translation of particular subsets of mRNAs by altering the affinity of ribosomes (Yoon et al, 2006; Baxter‐Roshek et al, 2007; Basu et al, 2011). For example, knocking down Dkc1, a rRNA pseudouridine synthase, has little changes in total protein synthesis rates, but impairs translation of mRNAs containing internal ribosome entry sites (IRES) such as p27, Xiap, and Bcl‐xL (Yoon et al, 2006). To investigate whether the affected mRNAs share common structural features that are sensitive to 18S m⁶A_1,832, we analyzed the differentially translated genes affected by Mettl5 KO and tried to identify their common features. Codon frequency analysis showed that A/T‐rich codons such as “GAA” and “AAA” are significantly enriched in down‐regulated genes, while G/C‐rich codons such as “CGC” and “CTG” are frequently enriched in up‐regulated genes (Fig EV5A). Analyses of sequence features like GC content, length and minimum free energy (MFE) of secondary structure of 5′UTR, CDS, and 3′UTR showed that the sequence features are significantly distinct between the down‐regulated genes and unchanged genes, especially on CDS. The MFE in CDS of down‐regulated genes are noticeably higher than non‐differentially translated genes, implying the potential high structure composition on their CDS (Fig EV5B). Besides, the consensus sequence motif analysis identified a sequence motif significantly enriched in CDS of down‐regulated genes, which was also detected in the CDS of Fbxw7 (Fig EV5C and D). However, no significant enriched motif was detected in 5′UTR and 3′UTR regions of both up‐regulated genes, down‐regulated genes, and CDS of up‐regulated genes. Therefore, there may be some special features in CDS of down‐regulated genes that are highly sensitive to the 18S m⁶A for translation, which needs further investigation in future.

Figure EV5. Feature analysis of the differentially translated genes.

1. Codon frequency analysis identifying codons significantly enriched in down‐regulated and up‐regulated genes, respectively.
2. Sequence feature analysis of GC content, sequence length, and MFE in 5′UTR, CDS, and 3′UTR, respectively. The genes are divided into three groups (up‐regulated, down‐regulated and non‐regulated) from two replicates. Box plots display the full range of variations based on five number summaries (minimum, first quartile, median, third quartile, and maximum). P values were inferred from a two‐sided Wilcoxon rank sum test. ns, not significant; **P < 0.01; ***P < 0.001; ****P < 0.0001.
3. Sequence motif analysis identifying a significantly enriched motif (P = 10⁻²⁸) in CDS of down‐regulated genes. The motif analysis was performed using findMotifs.pl in homer software based on hypergeometric test.
4. A sequence motif in the CDS of Fbxw7 detected by FIMO. The P value is based on a zero‐order Markov model of the input sequences.

Interestingly, a recent report revealed that individuals with truncating variants in METTL5 displayed behavioral and physical abnormalities, such as intellectual disabilities, microcephaly, facial dysmorphisms, and short stature (Richard et al, 2019). During the revision of our manuscript, another two METTL5 papers were published and showed maldevelopment and behavioral abnormalities in Mettl5 KO mice and Drosophila, respectively (Ignatova et al, 2020; Leismann et al, 2020). Consistent with our observations, Ignatova et al demonstrated that Mettl5 KO disrupts the process of mESC differentiation. Though it seems contradictory between the successful gestation of Mettl5 KO newborns and impaired differentiation observed in the culture system, it should be noticed that the differentiation potential is compromised rather than totally lost in Mettl5 KO mESCs. Lineage‐specific markers could be gradually detected in Mettl5 KO groups during differentiation, and differentiated structures were observed in 6‐week Mettl5 KO teratomas (Figs 3D and EV3I). Additionally, it is worth noting that in vivo embryonic morphogenesis and development of a viable organism involve multiple regulatory mechanisms. Intrinsic and extrinsic factors potentially serve as compensatory mechanisms for alleviating developmental abnormalities, thus contributing to formation of a living organism. Indeed, it is not rare that the mice survive under the depletion of factors important for differentiation. For example, knocking out Whsc1 (encoding a histone methyltransferase) impairs the formation of mesendoderm during mESC differentiation (Tian et al, 2019), but does not lead to fully embryonic lethality in mice, despite various developmental defects (Nimura et al, 2009).

Taken together, our work here demonstrated that METTL5 is a m⁶A methyltransferase targeting 18S rRNA A_1,832. m⁶A‐dependent METTL5 regulates the robust differentiation by efficient translation of Fbxw7, which highlights the importance of m⁶A‐mediated translation regulation in development.

Materials and Methods

Plasmids and protein purification

For the expression of METTL5 protein, Mettl5 cDNA was cloned into pPB or pET28a expression vector (Invitrogen). Mettl5 mutants were generated using the QuikChange Site‐Directed Mutagenesis Kit (Stratagene, 200518) according to the manufacturer's protocol. Mettl5‐specific 20nt guide RNA sequences were cloned into lentiCRISPR v2 plasmid.

To express recombinant METTL5 protein, cells were incubated at 37°C until OD600 reached 0.6–1 and then cooled down to 16°C. IPTG was added to 0.2 mM final concentration, and cells were further incubated at 16°C for 16 h. Cell pellets were lysed in a buffer containing 300 mM NaCl, 25 mM Tris pH 7.5, 10% Glycerol, and 0.5% NP‐40. Total lysate was incubated with HisPur Ni‐NTA Resin at 4°C for 5 h. His‐METTL5 protein was eluted with elution buffer (20 mM Tris pH 7.5, 150 mM NaCl, 200 mM Imidazole) in 0.5 ml aliquots until color change was no longer observed according to a Bradford assay.

Assays for m⁶A methyltransferase activity in vitro

In vitro methyltransferase activity assay was performed in a 30 μl of reaction mixture containing the following components: 1 μg RNA probes, 0.5 μg fresh recombinant protein, 0.8 mM d3‐SAM, 50 mM Tris–HCl, 5 mM MgCl₂, 1 mM DTT, and pH 8.0. The reaction was incubated at 37°C for 1 h. After incubation, samples were treated with proteinase K at 50°C for 20 min, and resultant RNA was desalted and then digested with nuclease P1 and AP. Nucleosides were quantified by using nucleoside‐to‐base ion mass transitions of 285–153 (d3‐m⁶A) and 268–136 (Adenosine). Adenosine (A) served as an internal control to calculate the amount of RNA probe in each reaction mixture for QQQ LC‐MS/MS analysis.

LC‐MS/MS

Two hundred nanogram of RNA were digested by nuclease P1 (1 U, Sigma) in 25 μl of buffer containing 25 mM NaCl and 2.5 mM of ZnCl₂ at 42°C for 2 h, which was followed by addition of NH₄HCO₃ (1 M, 3 μl, freshly made) and AP (1 U, Sigma) and additional incubation at 37°C for 2 h. Samples were then diluted to 60 μl and filtered (0.22 μm pore size, 4 mm diameter, Millipore); 5 μl of solution was loaded into liquid chromatography‐tandem mass spectrometry (LC‐MS/MS; Agilent 6410 QQQ triple‐quadrupole mass spectrometer). Nucleosides were quantified by using retention time and nucleoside‐to‐base ion mass transitions of 282.1–150.1 (m⁶A), and 268–136 (A).

SELECT

For quantitatively detecting the m⁶A status in 18S rRNA A_1,832 and A_1,825 (as input control) locus, the SELECT (single‐base elongation‐ and ligation‐based PCR amplification method) was performed as described (Xiao et al, 2018). In brief, we used 1 ng total RNA extracted from different cells, 40 nM up/down primer and 5 μM dNTP, 1 × CutSmart buffer in the 17 μl reaction mixture, which were annealed at a temperature gradient. Then, 3 μl of 0.01 U Bst 2.0 DNA polymerase (NEB, M0537S), 0.5 U SplintR ligase (NEB, M0375S), and 10 nmol ATP (NEB, P0756S) were added in the former reaction mixture and incubated at 40°C for 20 min, denatured at 80°C for 20 min, and hold at 4°C. Afterward, we used 2 μl elongated and ligated products as a template to perform quantitative real‐time PCR (qPCR) amplification or PCR amplification. qPCR was performed on ABI Q5 or ABI Q7 (Thermo Fisher Scientific, USA). Data were analyzed by QuantStudio TM Real‐Time PCR Software v1.4 and PCR products were analyzed by electrophoresis. The primers used in this study are listed in Appendix Table S1.

Ribo‐seq

For ribosome profiling (Ribo‐seq), mESCs were pre‐treated with 50 mg/ml of cycloheximide (CHX; Sigma, C7698) for 10 min and then washed twice with ice‐cold PBS supplemented with CHX. Cells were collected and resuspended in 400 μl lysis buffers (20 mM Tris–HCl pH 7.4, 150 mM NaCl, 5 mM MgCl₂, 1 mM DTT, and 100 mg/ml CHX). After incubation on ice for 10 min, lysate was triturated five times through a 25‐gauge needle and then lysate was centrifuged at 16,000 g for 10 min. To generate ribosome‐protected fragments, the lysates (30 mg) were first mixed with 200 μl DEPC‐H₂O then incubated with 15 U RNase I for 45 min at room temperature. The reaction was stopped with 10 μl SUPERase*In RNase inhibitor. 0.9 ml of sucrose‐supplemented lysis buffer was added to the digestion mixture and ultracentrifuged at 240,000 g, 4°C for 3 h. Pellets were resuspended in 300 μl of water and after phenol–chloroform extraction, precipitated with ethanol. The RNA was then run on a 15% 8 M urea TBE gel, stained with SYBR Gold, and a gel fragment between 17 and 34 nucleotides was excised. RNA was eluted for 2 h at 37°C in 300 μl RNA extraction buffer (300 mM NaOAc pH5.5, 1 mM EDTA, 0.25%v/v SDS) after crushing the gel fragment. RNA was ethanol precipitated and resuspended in 26 μl water and treated with Ribo Zero Gold kit. Purified RNA was treated with PNK and subjected to library construction with NEB Next Small RNA Library Prep Kit.

mESC culture

All the cell lines were cultured at 37°C with 5% CO₂ in the cell incubator. The mESC line in this study was a kind gift from Sha Jiahao's laboratory (Nanjing Medical University, China). The mESCs were cultured in 2i/Lif conditions, which contained N2B27 medium with 2i (PD0325901: 1 μM, Selleck; CHIR99021: 3 μM, Selleck) and leukemia inhibitory factor Lif (1,000 U/ml, Millipore). mESCs were cultured in serum/Lif conditions, which contained N2B27 medium with 15% serum (GIBCO, 10100139C) and leukemia inhibitory factor Lif (1,000 U/ml, Millipore). N2B27 medium consisted of 50% DMEM/F12(GIBCO, 10565‐018), 50% Neurobasal medium (GIBCO, 21103‐049) supplemented with 1 × N2 (GIBCO, 17502048), 1 × B27 (GIBCO, 17504044), 1 × Glutamax (GIBCO, 35050061), 1 × NEAA (GIBCO, 11140050), 1 mM Sodium Pyruvate (GIBCO, 11360070), and 0.1 mM 2‐mercaptoethanol (Sigma‐Aldrich, M3148). The mESCs were cultured in dishes coated with poly‐L‐ornithine solution (Sigma‐Aldrich, P4957) and Laminin (Invitrogen, 23017015, 10 ng/ml) one after another. Cells were routinely passaged every 2 days.

Mouse embryoid body generation from mESCs

The differentiation medium contained GMEM (GIBCO, 11710035), 10% KOSR (GIBCO, 10828028), 1 × Glutamax, 1 × NEAA, 1 mM Sodium Pyruvate, 1% Penicillin‐Streptomycin (Hyclone, SV30010), and 0.1 mM 2‐mercaptoethanol. The mESC differentiation was performed as described previously (Kamiya et al, 2011). Briefly, mESCs were disaggregated to single cells using TrypLE Select Enzyme (GIBCO, 12563‐011) and seeded into the wells of an ultra‐low attachment multiple 96‐well plate (4,000 cells per 150 μl medium per well; Corning, 7007). Medium was changed every other day.

Generation of Mettl5 knockout mESC colonies by CRISPR/Cas9

The small guide RNA (sgRNA) was designed to target the second exon of Mettl5. Cas9 plasmid and sgRNA were constructed as described (Niu et al, 2014). To generate Mettl5 knockout mESCs, a mixture of 2 μg of sgRNA and 2 μg of Cas9 plasmids were transfected to mESCs using Lipofectamine Stem Transfection Reagent (Invitrogen, STEM00003). After 2 weeks’ selection with 2 μg/ml puromycin (Invivogen, ant‐pr‐1), colonies were screened out by PCR amplification and Sanger sequence and confirmed by WB.

To re‐express METTL5 protein into KO cells: Mettl5 ORF cDNA was cloned from mouse cDNA pools and then inserted into PB‐CAG expression vector (Invitrogen). The successfully constructed vector together with a PB transposase (PBase) vector (Invitrogen) was transfected to KO mESCs using Lipofectamine Stem Transfection Reagent.

The length of PCR amplification in the targeted region of Mettl5 locus was 434 bp. The primers for genotyping include:

• Forward primer: 5′‐ AATCTGGCGCTACTGCAGATTAC‐3′;

• Reserve primer: 5′‐ TTCTAGCATGGCAGGTATAAGGG‐3′.

Teratoma formation

3 × 10⁶ mESCs cultured in the 2i/Lif condition were resuspended in 200 μl DMEM and injected subcutaneously into the flanks of female immunodeficient NCG mice (purchased from Nanjing Biomedical Research Institute of Nanjing University). After 6 weeks, all injected mice were sacrificed. Teratomas were excised and subjected to histological analysis.

All animal experiments were conducted in accordance with the guidelines of the Institutional Animal Care and were approved in advance by the Use Committee of Nanjing Medical University (China).

AP staining

Alkaline phosphatase staining was performed with AP Staining kit (Stemgent, 00‐0055). Briefly, cells were fixed with Fix Solution for 2 min and washed with PBST (1 × PBS with 0.05% Tween‐20). The freshly prepared Staining Solution were then added to the wells, followed by an incubation of 10–15 min in the dark at room temperature (RT). The Staining Solution was removed immediately when the color turned bright. Images were captured with Nikon Dsl132 Microscope.

Immunohistofluorescence and H&E staining

The samples were fixed in freshly prepared 4% paraformaldehyde/1 × PBS (Sigma, 158127) at 4°C for 2 h, washed three times in PBS with 10 min each, dehydrated, and embedded in paraffin using a paraffin‐embedding station (Leica EG1150H). 5 μm sections were cut on a microtome (Leica RM2235), mounted on glass slides, dried in 65°C oven overnight, and prepared for immunofluorescence staining or H&E staining. For immunofluorescence, sections were deparaffinized, rehydrated, and boiled for 15 min in sodium citrate buffer for antigen retrieval. When cooled down to room temperature, slides were washed in PBS and permeabilized in PBST (1 × PBS with 0.1% Triton X‐100) for 20 min. The slides were blocked in blocking buffer (1% (w/v) bovine serum albumin in PBST) for 1 h at RT and then incubated in the indicated primary antibodies diluted in the blocking buffer at 4°C overnight. The next day, after three times washes with PBST (10 min each), the slides were incubated with appropriate diluted fluorescence‐conjugated secondary antibodies and DAPI in the dark for 1 h at RT. Subsequently, slides were washed again and mounted with anti‐fade mounting medium (Vectorlabs, H‐1000). All images were obtained on ZEISS LSM800 confocal microscope (Carl Zeiss AG, Jena, Germany) and processed with Zeiss Zen Desk (Blue Edition).

For hematoxylin and eosin (H&E) staining, the slides were deparaffinized, rehydrated, and immersed in hematoxylin solution for 8–10 min, and then washed in running tap water for 10 min. After destained using 1% HCl for 30 s, the slides were washed, counterstained with eosin‐phloxine solution for 1 min, dehydrated, and then mounted for histological analysis. The images were captured on Zeiss Axio Imager A1.

Real‐time quantitative PCR analysis

Total RNA was extracted from the cells using TRIzol reagent (Invitrogen, 15596018) according to the manufacturer's instruction. 500 ng of total RNA was used to synthesize cDNA with the RT reagent Kit (Takara, RR047B). Resultant cDNA was diluted in water and 10 ng was used in each RT–qPCR. Reactions were run on ABI Q5 using ChamQ SYBR qPCR Master Mix (Vazyme, Q341‐03). Actin gene was used as an internal control for the RT–qPCR analysis of gene expression. The relative expression level was calculated using the $2^{-\mathrm{\Delta }\mathrm{\Delta }{\mathrm{C}}_{\mathrm{t}}}$ method. Each experiment was performed in triplicate and repeated three times. The primer sequences used in this study are listed in Appendix Table S2. Statistical analyses were performed with GraphPad Prism version 8.

Western blot analysis

Cells were washed, collected, and lysed in ice‐cold RIPA buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA,1% NP‐40, 0.5% sodium deoxycholate and 1% SDS) with 1 × complete protease inhibitor cocktail (Sigma, P5726). After incubating on ice for 30 min, the cells were centrifuged at 16,200 g for 15 min at 4°C. The supernatants were collected and the protein concentration was measured by the Pierce BCA Protein Assay kit (Thermo Scientific, 23227) following the manufacturers protocol. The equal amounts of proteins were loaded on SDS–PAGE gels (Epizyme, PG112) and then transferred to nitrocellulose filter membranes (GE, 10600001). Membranes were blocked with 5% non‐fat milk in TBST (Tris‐buffered saline with 0.1% Tween‐20) for 1 h followed by an incubation with primary antibodies overnight at 4°C. The next day, membranes were washed 3–4 times with TBST and incubated with the appropriate anti‐mouse/rabbit secondary antibodies conjugated to HRP (Jackson) at RT for 1 h. After washed again, the membranes were visualized by ECL reagent (Millipore, WBKLS0100) on Tanon 5200.

Antibodies

The antibodies were mentioned as follows: Mouse anti‐POU5F1 (Santa Cruz, sc‐5279), Mouse anti‐α‐TUBULIN (Santa Cruz, sc‐8035), Mouse anti‐FOXA2 (Santa Cruz, sc‐374376), Goat anti‐T (R&D, AF2085), Mouse anti‐SMA (R&D, MAB1420), Mouse anti‐NESTIN (Santa Cruz, sc‐58813), Goat anti‐NANOG (R&D, AF1997), Mouse anti‐ESRRB (R&D, PP‐H6705‐00), Mouse anti‐SOX2 (Santa Cruz, sc‐365823), Mouse anti‐ACTIN (Santa Cruz, 2Q1055), Rabbit anti‐c‐MYC (CST, D84C12), Rabbit anti‐METTL5 (ProteinTech, 16791‐1‐AP), Mouse anti‐HA (MBL, M180‐3), Rabbit anti‐FBXW7 (ProteinTech, 28424‐1‐AP), Donkey anti‐Rabbit secondary antibody Alexa Fluor 488 (Invitrogen, A‐21206), Donkey anti‐Mouse secondary antibody Alexa Fluor 555 (Invitrogen, A‐31570), Goat anti‐Rabbit IgG (H + L) secondary antibody HRP (Invitrogen, 31466), and Goat anti‐Mouse IgG (H + L) secondary antibody HRP (Invitrogen, 32230).

Ribo‐seq data analysis

Trim galore (https://www.bioinformatics.babraham.ac.uk/projects/trim_galore) was first used to trim adapters and low‐quality sequence from raw FASTQ sequences of Input and Ribo‐seq data. There were two replicates for both WT and Mettl5 KO cells. The clean sequences for Input samples were directly mapped to mouse reference genome sequences (mm10) using STAR (Dobin et al, 2013). The raw number of reads mapped to CDS region of each gene was then counted by using Subread package based on GENCODE gene annotation (version M21) (Liao et al, 2014). For Ribo‐seq, the clean data containing ribosome‐protected fragments (RPFs) in FASTQ format were first converted into collapsed FASTA format using fq2collapsedFa.pl (Liu et al, 2017). The collapsed sequences and gene expression raw read count from Input sample were then uploaded to RiboToolkit (Liu et al, 2020) to perform various analyses, including tRNA and snRNA removing, RPF quality control, translation efficiency analysis, and gene functional enrichment for differentially translated genes. In brief, after removing the RPF tags aligned to mouse rRNA sequences, tRNA sequences and snRNA sequences by Bowtie v1.2.2 with a maximum of two mismatches (Langmead et al, 2009), the filtered RPFs were mapped to mouse reference genome sequences using STAR (Dobin et al, 2013). The unique mapped RPFs were used to count the RPF abundance on CDS region of each gene mode of GENCODE annotation (M21) using Subread (Liao et al, 2014). The counts were then normalized as RPF per kilobase per million mapped RPFs (RPKM). The translation efficiency (TE) was calculated by dividing normalized RPF count by its normalized gene read count. The TE list was further filtered using the threshold of “> 10 RPFs per transcript for each sample and > 100 total RPFs for all four samples”. A threshold of 1.5‐fold change and P value < 0.05 was used to define the differentially translated genes (Zinshteyn & Gilbert, 2013).

The motif analysis of differentially translated genes was performed using findMotifs.pl in homer software (Heinz et al, 2010). The motif scan in specific sequence was performed using FIMO (Grant et al, 2011). The MFE of RNA secondary structure was conducted using RNAfold (Lorenz et al, 2011).

Author contributions

MX and HC performed most of the experiments, analyzed the data, and prepared figures. QL carried out the bioinformatics analysis. CM constructed most of plasmids in the study. XZ and SQZ conducted the experiments of in vitro transcription of sgRNA and microinjection of mice. HYZ was responsible for ESC culturing. LC was responsible for providing essential experimental materials. MXL, BS, and XJG assisted MX with EB experiments. MX and HC wrote the manuscript with the assistance of all other authors. JZ, HC, and HHM supervised the project. All authors contributed to the final version of the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Dataset EV1

Source Data for Expanded View

Review Process File

Source Data for Figue 1

Source Data for Figue 2

Source Data for Figue 3

Source Data for Figue 4

Source Data for Figue 5

EMBO Reports (2020) 21: e49863

Data availability

Raw and processed data from Ribo‐seq analysis have been deposited at the GEO database (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE138341) with the accession number GSE138341.