Alternative splicing of METTL3 explains apparently METTL3-independent m6A modifications in mRNA

Abstract

N6-methyladenosine (m6A) is a highly prevalent mRNA modification that promotes degradation of transcripts encoding proteins that have roles in cell development, differentiation, and other pathways. METTL3 is the major methyltransferase that catalyzes the formation of m6A in mRNA. As 30% to 80% of m6A can remain in mRNA after METTL3 depletion by CRISPR/Cas9-based methods, other enzymes are thought to catalyze a sizable fraction of m6A. Here, we reexamined the source of m6A in the mRNA transcriptome. We characterized mouse embryonic stem cell lines that continue to have m6A in their mRNA after Mettl3 knockout. We show that these cells express alternatively spliced Mettl3 transcript isoforms that bypass the CRISPR/Cas9 mutations and produce functionally active methyltransferases. We similarly show that other reported METTL3 knockout cell lines express altered METTL3 proteins. We find that gene dependency datasets show that most cell lines fail to proliferate after METTL3 deletion, suggesting that reported METTL3 knockout cell lines express altered METTL3 proteins rather than have full knockout. Finally, we reassessed METTL3's role in synthesizing m6A using an exon 4 deletion of Mettl3 and found that METTL3 is responsible for >95% of m6A in mRNA. Overall, these studies suggest that METTL3 is responsible for the vast majority of m6A in the transcriptome, and that remaining m6A in putative METTL3 knockout cell lines is due to the expression of altered but functional METTL3 isoforms.

Fig 1. Previously described Mettl3 KO mESC lines express shorter isoforms of Mettl3.

(A) Mettl3 KO mESCs from two groups have different m⁶A levels. To reconfirm the m⁶A levels using quantitative methods, we performed mass spectrometry to estimate the m⁶A in mRNA. Exon2 Mettl3 KO mESCs show persistence of 40.2% (exon2 Mettl3 KO mESC-a) and 55.6% (exon2 Mettl3 KO mESC-b) m⁶A, respectively, while exon4 Mettl3 KO mESCs show 1.45% m⁶A compared to WT. This confirms that exon4 Mettl3 KO mESCs have near-complete loss of m⁶A, but not the exon2 Mettl3 KO mESCs. Error bars indicate standard error (n = 3 for all, except n = 2 for exon2 Mettl3 KO mESC-b). * = p-value < 0.5, ** = p-value < 0.01, *** = p-value < 0.005, n.s. = not significant. Underlying data can be found in S1 Data. (B) Exon2 Mettl3 KO mESCs exhibit new anti-METTL3-immunoreactive bands. To investigate the effectiveness of the Mettl3 knockout, we measured the loss of METTL3 via WB. Full-length METTL3 (75 kDa, arrowhead) was lost in both KO cell lines, but new bands, which were reactive to the anti-METTL3-antibody, appeared at approximately 50 kDa in exon2 Mettl3 KO mESC-a and approximately 55 kDa in exon2 Mettl3 KO mESC-b (arrowheads). This indicates the possibility that a novel smaller METTL3 protein was expressed in the exon2 Mettl3 KO mESCs. In contrast, exon4 Mettl3 KO mESCs have no proteins reactive to anti-METTL3-antibodies. 30 μg per lane. (C) 5′ RACE reveals the expression of shorter Mettl3 mRNAs in the Mettl3 KO mESCs. We used 5′ RACE to identify novel Mettl3 mRNAs in the Mettl3 KO mESCs. The full-length RACE product (approximately 1,500 bp) was lost in the Mettl3 KO cells, but novel products at approximately 1,000 bp and approximately 700 bp were found in exon2 Mettl3 KO mESC-a and at approximately 1,500 bp and approximately 1,300 bp in exon2 Mettl3 KO mESC-b. These shorter mRNAs may encode the smaller METTL3 proteins seen in the KO cells. (D) Sequencing of 5′ RACE products show Mettl3 mRNAs with exon skipping or alternative transcription-start sites. We sequenced the 5′ RACE products to characterize the Mettl3 mRNA transcripts that are expressed by the exon2 Mettl3 KO mESCs. All Mettl3 mRNAs expressed in the KO cells skipped the guide RNA deletion region by exon skipping, or by using alternative transcription-start sites downstream of the deletion. The longest ORFs that are in-frame with the WT Mettl3 mRNAs are shown as solid lines below each mRNA. The encoded protein is also represented, with the domains required for METTL3 activity shown. m⁶A, N⁶-methyladenosine; mESC, mouse embryonic stem cell; NLS, nuclear localization signal; ORF, open reading frame; pAb, polyclonal antibody; RACE, rapid amplification of cDNA ends; WB, western blot; WT, wild-type; ZFD, zinc finger domain.

Fig 2. Shortened isoforms of METTL3 catalyze m⁶A formation in Mettl3 KO mESCs.

(A) The predicted domain structure of proteins encoded by the altered METTL3 ORFs expressed in exon2 Mettl3 KO mESCs suggests they may be functional. The domains that are known to be necessary for m⁶A formation by METTL3 include the WTAP-binding domain [39], ZFD [19,36], and methyltransferase domain [19,20]. To determine if the METTL3 ORFs from exon2 Mettl3 KO mESCs encode functional METTL3 proteins, we predicted the domain structure of the METTL3 protein isoforms from their ORFs. While all the predicted proteins have the methyltransferase domain, only METTL3-a.ii and METTL3-b.ii have all the known critical domains for m⁶A-writing. (B) WB of transfected FLAG-tagged METTL3 isoform ORFs. To determine if the METTL3 ORFs we found in the knockout cells can synthesize m⁶A, we expressed the METTL3 ORFs in exon4 Mettl3 KO mESCs that exhibit no METTL3 protein and essentially no baseline m⁶A signal. After 48 h, the alternatively spliced METTL3 proteins can be detected by immunoblotting with an anti-METTL3 antibody. FLAG-METTL3-a.ii (50 kDa, red arrowhead) and FLAG-METTL3-b.ii (55 kDa, blue arrowhead) have similar sizes to the anti-METTL3-antibody-reactive protein seen in exon2 Mettl3 KO mESC-a (red arrowhead) and exon2 Mettl3 KO mESC-b (blue arrowhead), respectively. 30 μg per lane. (C) Isoforms of METTL3 proteins can write m⁶A. After 48 h of transfection, RNA from each sample was processed, and m⁶A was measured using mass spectrometry. Expression of full-length WT METTL3 was able to rescue 19.7% of the m⁶A. METTL3-a.i was unable to rescue m⁶A, but METTL3-a.ii was able to rescue 18.3% of m⁶A. METTL3-b.i could only rescue 8.5% of m⁶A, whereas METTL3-b.ii rescued 24.4% of m⁶A, respectively. Thus, METTL3-a.ii and METTL3-b.ii proteins that are expressed in the exon2 Mettl3 KO mESCs are able to catalyze the formation of m⁶A. Error bars indicate standard error (n = 3). * = p-value < 0.5, ** = p-value < 0.01, *** = p-value < 0.005, n.s. = not significant. Underlying data can be found in S1 Data. (D) A METTL3-specific inhibitor leads to loss of m⁶A even in exon2 Mettl3 KO mESCs. Exon2 WT and Mettl3 KO mESCs were treated with 30 μM STM2457, a METTL3-specific inhibitor, and m⁶A levels were measured by mass spectrometry after 48 h. STM2457 treatment reduced m⁶A by 82.8% in the WT mESCs. In exon2 Mettl3 KO mESC, m⁶A was reduced by 85.4% in Mettl3 KO mESC-a after STM2457 treatment and by 94.8% in Mettl3 KO mESC-b. Thus, METTL3 is responsible for the remaining m⁶A in the exon2 Mettl3 KO mESCs. Residual m⁶A after STM2457 treatment may reflect incomplete inhibition of METTL3 at 30 μM. Error bars indicate standard error (n = 3). * = p-value < 0.5, ** = p-value < 0.01, *** = p-value < 0.005, n.s. = not significant. Underlying data can be found in S1 Data. m⁶A, N⁶-methyladenosine; mESC, mouse embryonic stem cell; NLS, nuclear localization signal; ORF, open reading frame; pAb, polyclonal antibody; WB, western blot; WT, wild-type; ZFD, zinc finger domain.

Fig 3. METTL3 knockout in U2OS cells also appears to be incomplete.

(A) METTL3 KO U2OS cells have persistent m⁶A. METTL3 KO U2OS cells have been reported to have 60% the levels of m⁶A found in control U2OS cells [24]. We reconfirmed this with mass spectrometry measurements of m⁶A, which showed that METTL3 KO U2OS cells have 75.2% remaining m⁶A compared to WT. Thus, m⁶A levels remain high in METTL3 KO U2OS cells. Error bars indicate standard error (n = 3). * = p-value < 0.5, ** = p-value < 0.01, *** = p-value < 0.005, n.s. = not significant. Underlying data can be found in S1 Data. (B) METTL3 KO U2OS cells express a novel anti-METTL3-antibody-reactive protein. As m⁶A levels were not completely ablated in the METTL3 KO U2OS cells, we assessed METTL3 protein expression in these cells to confirm if the knockout was effective. We found that WT METTL3 protein was lost in the METTL3 KO U2OS cells, but a larger protein that was reactive to the anti-METTL3 antibody was found in the METTL3 KO U2OS cells. This suggests the METTL3 KO U2OS cells express a novel METTL3 protein that is slightly larger than the WT METTL3. (C) Confirmation of a novel METTL3-like protein in METTL3 KO U2OS using a second METTL3 antibody. To confirm that the METTL3-immunoreactive band we saw in METTL3 KO U2OS cells in Fig 3B was METTL3, we used a second anti-METTL3 mAb to confirm the result. The same protein band is immunoreactive to the second anti-METTL3 antibody, thus suggesting that the METTL3 KO U2OS cells express a novel, larger METTL3 protein. (D) A METTL3-specific inhibitor leads to loss of m⁶A even in METTL3 KO U2OS cells. WT and METTL3 KO U2OS cells were treated with 30 μM STM2457, and m⁶A levels were measured by mass spectrometry after 48 h. m⁶A was reduced by 89.8% in the WT U2OS cells and 92.1% in the METTL3 KO U2OS cells after STM2457 treatment. It should be noted that 30 μM may not fully inhibit METTL3, so some of the residual m⁶A after STM2457 treatment may still derive from METTL3 isoforms. Thus, a METTL3 isoform is responsible for most of the remaining m⁶A in the METTL3 KO U2OS cells. Error bars indicate standard error (n = 3). * = p-value < 0.5, ** = p-value < 0.01, *** = p-value < 0.005, n.s. = not significant. Underlying data can be found in S1 Data. mAb, monoclonal antibody; m⁶A, N⁶-methyladenosine; pAb, polyclonal antibody; WB, western blot; WT, wild-type.

Fig 4. Conditional METTL3 knockouts can be used to study m⁶A when stable METTL3 knockouts are not viable.

(A) Most cell lines are dependent on METTL3 for growth. Mouse studies previously indicated that Mettl3 is an essential gene for early embryonic survival [5], so we wanted to know which cell lines are dependent on METTL3. Using the CRISPR gene dependency probability score from the DepMap 21Q4 dataset [–47], we found that 801 of 1,054 cell lines are dependent on METTL3 (dependency probability score >0.5). Therefore, most cells lines will not survive after METTL3 knockout. The density plot shows the overall distribution of dependency probability scores, while each individual cell line is represented as a dot. U2OS and A549 cell lines (red), where METTL3 has been previously knocked out, are shown here to be dependent on METTL3 and thus should not be viable after METTL3 knockout. Underlying data for this figure was extracted from the DepMap 21Q4 dataset [47]. (B) A small subset of cell lines may be m⁶A independent. Although most cell lines are dependent on METTL3, a small subset of cell lines can survive despite METTL3 knockout. To identify cell lines that we can confidently consider m⁶A-independent, we obtained a list of cell lines whose survival is also independent of other members of the m⁶A writer complex, METTL3, METTL14, and WTAP. This approach suggests that 65 cell lines may be able to proliferate in an m⁶A-independent manner (S4 Table). Underlying data for this figure was extracted from the DepMap 21Q4 dataset [47]. (C) Mettl3 conditional knockout MEFs do not express METTL3 protein. To determine the amount of m⁶A in mRNA that can be attributed to Mettl3, we generated a tamoxifen-inducible Mettl3 conditional knockout MEF cell line. We used WB to validate the loss of METTL3. After 5 days of 4-hydroxytamoxifen treatment (500 nM), we observed loss of the WT METTL3 protein. 30 μg per lane. (D) Mettl3 conditional knockout MEFs show near-complete loss of m⁶A. We measured m⁶A levels in mRNA derived from tamoxifen-inducible Mettl3 conditional knockout MEFs. Eight days after 4OHT treatment (500 nM), the Mettl3 KO MEFs showed 3.6% remaining m⁶A. Hence, m⁶A is almost completely lost after Mettl3 knockout in MEFs. Error bars indicate standard error (n = 3). * = p-value < 0.5, ** = p-value < 0.01, *** = p-value < 0.005, n.s. = not significant. Underlying data can be found in S1 Data. m⁶A, N⁶-methyladenosine; MEF, mouse embryonic fibroblast; pAb, polyclonal antibody; WB, western blot; 4OHT, 4-hydroxytamoxifen.