Inducible and reversible RNA N6-methyladenosine editing

Design of the ABA-inducible RNA m⁶A writing system

To achieve the temporal control of m⁶A editing, we integrated the CIP method with the RNA-targeting dCas13b CRISPR system to develop an inducible editing platform. In the CIP system, the inducer (e.g., ABA) triggers the association between two unique inducer-binding adaptor proteins (e.g., PYL and ABI) that are fused individually to two proteins of interest (POIs). Depending on the choice of POIs, a variety of downstream biological events can be specifically controlled by these chemical inducers^30,31,32,33. For example, the ABA-based CIP systems have been successfully combined with the CRISPR/dCas9 or CRISPR/Cas-inspired RNA-targeting system (CIRTS) system to achieve inducible epigenome editing or RNA A-to-I base editing^34,35. To develop an inducible m⁶A writing system, we cloned DNA constructs expressing fusion proteins of dCas13b-PYL and ABI-METTL3 (ABI-M3) (Fig. 1b). HA and FLAG tags were fused to dCas13-PYL and ABI-M3 individually for detection purposes and a nuclear localization signal (NLS) sequence was inserted in these fusion proteins to facilitate their nuclear localization, which has been shown to increase editing efficiency²⁵. We anticipated that dCas13b-PYL can be targeted, through the custom guide RNA (gRNA), to the desired locus on specific RNA transcripts while ABI-M3 cannot localize on its own. Upon ABA addition, ABI-M3 will be recruited to the gRNA/dCas13b-PYL targeting site, through ABA-induced ABI-PYL heterodimerization, resulting in the increase of the m⁶A level at the targeted site on the specific RNA transcripts (Fig. 1a). We have previously demonstrated that ABA-induced dimerization and effects can be rapidly reversed by removing ABA^34,36. We expect that ABA withdrawal will lead to the release of ABI-M3 from the RNA sites and result in a decrease in the m⁶A level. To eliminate any unwanted effects except the methyltransferase activity, we chose to include only the core catalytic domain, which was reported to be sufficient to catalyze m⁶A writing^37,38,39. To confirm that the observed m⁶A level changes were caused by the recruited m⁶A-writing enzymatic activities, we also cloned a catalytic inactive version of METTL3, the ABI-M3* (D395A mutant), which lacks the m⁶A writing activity^24,25. The constitutively active dCas13b-METTL3 fusion (dCas13b-M3) was also cloned for comparison.

ABA-inducible m⁶A writing

We confirmed the expression and nuclear localization of dCas13b-PYL and ABI-M3 fusion proteins in HEK293T cells by western blotting and immunostaining (Supplementary Figs. 1 and 2). To examine if the designed ABA-inducible editing system can induce m⁶A writing on a specific RNA transcript, we tested the installation of m⁶A at the A1216 site within the 3′- untranslated region (UTR) of Actb mRNA. This locus has been reported to have low m⁶A abundance⁴⁰ and has shown to be a validated m⁶A locus in previous m⁶A editing studies²⁵. A gRNA (gRNA-2, with its 3′-end two nucleotides (nts) away from the A1216 site on Actb mRNA) was used to target dCas13b-PYL. We transfected HEK293T cells with plasmids expressing dCas13-M3, or dCas13-PYL plus ABI-M3 (or ABI-M3*), and gRNA-2 for 24 h and then treated cells with or without ABA (100 μM) for another 24 h. Total RNAs were then isolated from cell lysates and fragmented, followed by the enrichment of m⁶A-containing RNA sequences through immunoprecipitation (IP) using the anti-m⁶A antibody. The levels of m⁶A at the Actb A1216 site under different experimental conditions were quantified by reverse transcription and quantitative PCR (RT-qPCR). We observed that the m⁶A level at the Actb A1216 region probed by the qPCR primers increased only when ABA was added and when the catalytically active METTL3 was used (Fig. 2a), and the increased m⁶A level was comparable to the constitutively active dCas13-M3 fusion. No m⁶A enrichment was observed when the mutant METTL3 domain was used with or without ABA. To confirm the m⁶A enrichment indeed occurred at the A1216 locus, we performed experiments using the SELECT method to probe the m⁶A level changes specifically at the A1216 site under each editing condition with single-base resolution⁴¹. We designed 2 synthetic DNA oligonucleotides with PCR adapters that were complementarily annealed to the Actb RNA surrounding the A1216 region but left a single nucleotide gap opposite to the A1216 locus. The presence of m⁶A at A1216 prohibited the single-base elongation of DNA polymerases and the nick ligation activity of ligases, resulting in reduced full-length ligation products (quantified by qPCR) when compared to the unmethylated RNA⁴¹. From the SELECT experiments specifically design to probe the A1216 locus, we indeed observed increased m⁶A levels at A1216 only when ABA was added to the inducible m⁶A writing system or when dCas13-M3 was used (Supplementary Fig. 3). In addition, we investigated the dosage response of ABA in inducing m⁶A writing. We observed that the m⁶A levels increased with increasing ABA levels (Supplementary Fig. 4). Minimal changes were observed when using ABA concentrations higher than 100 μM. These results indicated that 100 μM ABA was an optimal concentration for inducing m⁶A editing, and therefore, was used in the following studies.

a The m⁶A level changes under different treatment conditions using the ABA-inducible or non-inducible m⁶A writing system. b The m⁶A level changes using the ABA-inducible m⁶A writing system with different gRNAs. c The ABI-M3 level changes with or without ABA at the target site. d The m⁶A level (left) and the ABI-M3 level (right) change before or after ABA removal (for 24 h). e The m⁶A enrichment at different time points after adding ABA. f The m⁶A enrichment at different time points after washing out ABA. ABA was added to induce m⁶A writing for 24 h before the time-course experiments of ABA removal. All results were calculated by normalization of data from each sample to that from the condition of dCas13b-PYL plus gRNA-2 and ABA. Values and error bars reflect the mean, s.e.m. of three independent biological replicates. P values are shown in the charts are determined by one-way ANOVA. Source data are provided as a Source Data file.

To examine if the gRNA targeting locus affects the editing efficiency and to confirm the editing was achieved by transcript/site-specific targeting, we cloned additional gRNAs designed to target different sites with varying distances from the A1216 site, including gRNA-8 and gRNA-22 (with their 3′-end 8 or 22 nucleotides (nts) away from A1216), as well as a non-targeting gRNA (NT-gRNA) as a negative control²⁵. After transfecting cells with plasmids expressing dCas13-PYL, ABI-M3, and individual gRNAs for 24 h, ABA was added for another 24 h. RNAs were purified and analyzed for m⁶A level changes as described above. We observed that all tested gRNAs, except NT-gRNA, resulted in m⁶A enrichment at the Actb A1216 site (Fig. 2b), although with different levels of increase. Based on the increased m⁶A fold changes, gRNA-2, which targeted a region closest to the A1216 site, was chosen and used in the following studies.

To confirm that ABI-M3 was recruited upon ABA addition to the targeted region on Actb mRNA to enable m⁶A writing, we transfected cells with dCas13-PYL, ABI-M3, and gRNA-2 expressing plasmids and treated cells with or without ABA, followed by performing cross-linking immunoprecipitation (CLIP) experiments with cell lysates (using anti-FLAG antibody) and RT-qPCR assays to quantify the ABA-dependent enrichment of ABI-M3. We observed that the level of ABI-M3 was increased at the targeted site on Actb mRNAs only when ABA was added (Fig. 2c), which verified the ABA-dependent recruitment of the METTL3 domain.

Reversibility and the temporal control of m⁶A writing by the ABA-inducible system

To investigate the reversibility of the induced m⁶A writing and ABI-M3 recruitment, HEK293T cells were transfected with dCas13b-PYL, ABI-M3, and Actb gRNA-2 for 24 h and treated with ABA for another 24 h. Cells were then washed with fresh cell culture media (without ABA) to remove ABA and incubated for 24 h before cells were lysed and processed as described above to quantify the level changes of m⁶A and ABI-M3 at the Actb A1216 site. We observed that both m⁶A and ABI-M3 levels were significantly reduced after ABA removal (Fig. 2d), confirming that the induced m⁶A writing and ABI-M3 recruiting were reversible. To obtain kinetic information regarding the inducing and reversing processes of this m⁶A writing system, we performed time-course experiments monitoring changes of the m⁶A levels at the Actb A1216 site at different time points after adding ABA (Fig. 2e), or removing ABA for different time periods after a 24-h ABA induction (Fig. 2f). We observed a rapid induction of m⁶A writing at the Actb A1216 site within 3 h, and the m⁶A levels continued to increase within the 24-h observation period. After ABA was removed, the m⁶A levels were significantly reduced in 3 h and eventually reached the background level within 24 h. These results demonstrated that the ABA-inducible m⁶A writing system enabled precise temporal control for both inducing and reversing m⁶A modification by the addition or removal of ABA.

Feasibility of the ABA-inducible m⁶A editing system on other RNA transcripts

To test if this ABA-inducible m⁶A writing platform can be applied to edit m⁶A on other RNA transcripts, we cloned DNA constructs of gRNAs targeting other RNAs at sites with low m⁶A levels but susceptible to m⁶A writing, including Gapdh A690, Foxm1 A3488, and A3504, Sox2 A1398 and A1405 sites²⁵. We transfected cells with dCas13b-PYL, ABI-M3 combined with corresponding gRNAs of Actb, Gapdh, Foxm1, or Sox2, and treated cells with ABA. The samples were processed and analyzed as described above to quantify the m⁶A level at each site under different gRNA targeting conditions. When using the specific gRNA targeting one of the four tested RNA sites, we expected that only the targeted RNA site showed an increased m⁶A level but not at any other tested sites. As expected, when the Actb gRNA was used, we observed that only Actb A1216 site showed enrichment of m⁶A but not at the sites on Gapdh, Foxm1, or Sox2 (Fig. 3a). The same observations were found in the cases of using gRNAs for Gapdh, Foxm1, and Sox2, respectively. These results suggested that this ABA-inducible m⁶A writing system can be robustly applied to editing m⁶A on different RNAs by choosing different gRNAs and the editing was selective among the tested editing studies. We expect that the off-target editing rate of this method will be similar to other dCas13b or dCas9-based methods, which have been reported to have a ~2.8% off-site editing rate out of all RNA transcripts through comprehensive RNA sequencing studies^24,25.

a The relative m⁶A level changes at different mRNA sites when using different mRNA-specific gRNAs for editing. b The relative abundance of Actb mRNAs in HEK293 cells at different time points after actinomycin D treatment under different m⁶A writing conditions. Gapdh was used as the internal control gene. c The relative abundance of Sox2, Foxm1, and Gapdh mRNAs in HEK293 cells under different m⁶A writing conditions. The mRNA levels were quantified after 6-h actinomycin D treatment. P values are shown in charts determined by one-sample t and Wilcoxon test. d Western blotting of Sox2, Foxm1, and Gapdh protein levels (top panel) under different m⁶A writing conditions. Endogenous Vinculin or Gapdh were used as the internal control (lower panel). All results of groups in a–c with different treatments were normalized to the group of dCas13b-PYL with gRNA-2 and ABA treatment. Values and error bars reflect the mean, s.e.m. of three independent biological replicates. P values shown in the charts are determined by one-way ANOVA. Three independent trials were performed for d and representative images were shown. Source data are provided as a Source Data file.

Induced Actb A1216 m⁶A writing reduced Actb mRNA stability

The m⁶A at the Actb A1216 site has been shown to reduce its stability²⁴. To examine if the artificially installed m⁶A through the inducible editing method was biologically relevant and resulted in the expected downstream effects, we investigated the stability of Actb mRNA under different m⁶A editing conditions in HEK293T cells. We transfected HEK293T cells with plasmids expressing dCas13-M3, or dCas13-PYL plus ABI-M3, with gRNA-2 or NT-gRNA for 24 h. After treating cells with or without ABA for another 24 h, actinomycin D, an RNA polymerase inhibitor, was added to block the synthesis of new RNAs¹⁴. At different time points after actinomycin D treatment, cells were harvested and subjected to RT-qPCR assays to quantify the Actb mRNA abundance under different experimental conditions. We observed that the stability of Actb mRNA was significantly reduced when the inducible m⁶A writing system was supplied with ABA and when gRNA-2 was used (but not without ABA or with NT-gRNA) (Fig. 3b). The observed reduction in Actb mRNA stability caused by the ABA-induced m⁶A writing showed a reduction level similar to the results when using the non-inducible dCas13-M3 system. To confirm that reversing the m⁶A writing can also lead to the reversal of m⁶A-dependent destabilization of Actb mRNA, we repeated the experiments described above, except that, after the 24-h ABA induction period, we removed ABA and incubated for another 24 h (or without removal as the control) before adding actinomycin D. We observed that when ABA was removed, the Actb mRNA indeed become stabilized and returned to the endogenous level (Supplementary Fig. 5). These results confirmed that the ABA-induced m⁶A writing system-generated functional RNA modification changes that resulted in the expected effect, supporting that the ABA-controlled m⁶A editing system can be a valid tool for studying m⁶A cellular functions.

Investigation of the impact of m⁶A on other mRNAs

We next applied the inducible m⁶A writing system to examine the effects of m⁶A located in different regions on other mRNAs, including the Sox2 A1398 and A1405 (3′-UTR), Foxm1 A3488 and A3504 (3′-UTR), and Gapdh A690 (gene body) sites, where the function of m⁶A have not been examined. We specifically investigated the effects of m⁶A on RNA transcript stability and translation, which are known to be affected by the m⁶A modification¹. We applied the inducible system to install m⁶A at the indicated sites on these mRNAs and determined the relative mRNA levels as an indication for transcript stability (as described above) or the protein levels of each mRNA product for the effects on translation (by western blotting) under different m⁶A writing conditions. We observed that the writing of m⁶A at these 3′-UTR sites on Sox2 and Foxm1 caused destabilization of transcripts and decreased RNA levels (Fig. 3c), similar to the effects of m⁶A within 3′-UTR observed on other mRNAs²⁴. We also observed reduced Sox2 and Foxm1 protein levels with m⁶A installed (Fig. 3d), which we believed were due to reduced mRNA levels of these transcripts. On the contrary, the methylation on Gapdh A690 (located in the gene body) did not result in any changes in RNA levels (Fig. 3c), which was consistent with previous observations²⁵. It has been shown that m⁶A in the gene body can have different effects on different transcripts, including promoting or impeding translation^15,42,43. However, we did not observe any effects on Gapdh translation with m⁶A installed at Gapdh A690 (Fig. 3d). More future experiments will be required to determine the biological function of m⁶A at the Gapdh A690 site. Nevertheless, the context-dependent variations of m⁶A function highlight the need for precision RNA modification editing tools as in this study.

The ABA-inducible m⁶A erasing system

After validating the inducible m⁶A writing system, we investigated the feasibility to apply the same strategy for inducible m⁶A erasing. We expected that by recruiting an m⁶A eraser instead of a writer, ABA-controlled m⁶A erasing can be achieved. The A2577 site on the nuclear non-coding RNA MALAT1, known to have a high m⁶A abundance in HeLa cells⁴⁰, was used to test the ABA-inducible m⁶A erasing system. We cloned fusion proteins of ABI-ALKBH5 containing either the catalytic core domain of ALKBH5 (ABI-ALK)⁴⁴ or the inactive H204A/H267A mutant⁴⁵ (ABI-ALK*) (Fig. 4a). The constitutively active dCas13b-ALKBH5 fusion (dCas13b-ALK) and a gRNA (gRNA-1) designed to target the MALAT1 A2577 site-based on previous studies²⁷, were also cloned. We confirmed the expression and subcellular localization of dCas13b-PYL and ABI-ALKBH5 in Hela cells (Supplementary Figs. 6 and 7). To test the ABA-induced m⁶A erasing at the MALAT1 A2577 site, we transfected Hela cells with plasmids encoding the dCas13-ALK, or dCas13-PYL plus ABI-ALK (or ABI-ALK*) and MALAT1 gRNA-1 for 24 h and then treated cells with or without ABA (100 μM) for another 24 h. The resulting cell samples were processed and analyzed by m⁶A IP and RT-qPCR assays as described above to quantify the m⁶A levels under different conditions. We observed that m⁶A demethylation at the MALAT1 A2577 site occurred only when ABA was added to the inducible system and only when the catalytically active ALKBH5 domain was used (Fig. 4b). No m⁶A erasing was observed without ABA or when mutant ALKBH5 was used. The level of induced m⁶A reduction was comparable to the erasing caused by the non-inducible dCas13b-ALK, indicating that the ABA-induced recruiting of ALKBH5 activity and the m⁶A erasing were effective. To confirm the editing occurred at the A2577 locus, we performed SELECT experiments to probe the m⁶A specifically at the MALAT1 A2577 site and observed decreased m⁶A levels at this locus only when ABA was added to the inducible m⁶A erasing system or when dCas13-ALK was used (Supplementary Fig. 8). These results confirmed that the ABA-inducible m⁶A editing strategy can be extended to control m⁶A erasing by recruiting proper enzymatic activity.

a DNA constructs for the ABA-inducible m⁶A erasing system, including dCas13b-ALKBH5, ABI-ALK (with the wild-type ALKBH5), and ABI-ALK* (ALKBH5 with H204A and H267A mutations). b The m⁶A level changes under different editing conditions using the ABA-inducible or non-inducible m⁶A erasing system. c The m⁶A level changes using the ABA-inducible m⁶A erasing system with different gRNAs. d The m⁶A level changes at different RNA sites (MALAT1 A2577, MYC A5553, CYB5A A135, CTNNB1 A126) using an inducible m⁶A erasing system targeting the MALAT1 A2577 site. All results of groups in a–c with different treatments were normalized to the group of dCas13b-PYL with gRNA-1 and ABA treatment. Values and error bars reflect the mean, s.e.m. of three independent biological replicates. P values are shown in the charts are determined by one-way ANOVA. Source data are provided as a Source Data file.

To examine the effects of locations that gRNAs targeted in the inducible m⁶A erasing system, we cloned additional gRNAs targeting 50 or 100 nts away from the A2577 site (gRNA-50 and gRNA-100) and compared the editing efficiency using different gRNAs in experiments similar to those described above. We observed that although all these gRNAs led to m⁶A level decreases, using MALAT1 gRNA-1 resulted in more effective m⁶A erasing than other gRNAs (Fig. 4c). On the contrary, the m⁶A level did not change when using NT-gRNA. These results showed that gRNA targeting locations or the distance to the m⁶A site could be important for editing efficiency.

To investigate the selectivity of the ABA-induced m⁶A erasing system, we transfected each component of ABA-inducible m⁶A erasing system with MALAT1 gRNA-1 in Hela cells. After adding ABA for 24 h, cells were harvested and total RNA was extracted, followed by m⁶A IP and RT-qPCR assays. We observed that only m⁶A A2577 site showed a significant decrease in m⁶A level, but not at other reported sites with high m⁶A levels including MYC A5553, CYB5A A135, and CTNNB1 A126 (Fig. 4d)^27,28. This result suggested that the ABA-inducible m⁶A erasing system can robustly and selectively remove m⁶A.

Light-inducible m⁶A writing by incorporating the photo-caged ABA

A light-activatable m⁶A editing system has been reported using light-inducible heterodimerization protein pairs²⁶, which demonstrated that light can be utilized to control m⁶A editing. To introduce the light control capacity in the ABA-CIP-based m⁶A editing platform, we incorporated the photo-caged ABA that allowed the activation of ABA, and therefore, inducing the RNA m⁶A editing, by light. We have shown that ABA can be caged with sensor units responsive to different stimuli and ABA can only be activated to induce CIP-controlled effects when exposed to the corresponding signals including light^46,47,48. To create a light-inducible m⁶A editing system, we used 4,5-dimethoxy-2-nitrobenzyl (DMNB)-caged ABA, which can be uncaged by UV light⁴⁸. It has been reported that caged ABA-DMNB cannot induce dimerization between PYL and ABI⁴⁸, therefore, no m⁶A editing was expected with ABA-DMNB treatment. However, with UV light irradiation, the DMNB group can be removed to generate functional ABA and subsequently achieve m⁶A editing (Fig. 5a).

a The light-inducible m⁶A writing system incorporated photo-caged ABA-DMNB that can only be activated by UV light to regenerate functional ABA, which recruits ABI-M3 to enable m⁶A writing. b The relative m⁶A levels at the Actb mRNA A1216 site were edited by the inducible m⁶A writing system using ABA or ABA-DMNB with or without light irradiation. All results of groups with different treatments were normalized to the group of dCas13b-PYL with gRNA-2 and ABA treatment. Values and error bars reflect the mean, s.e.m. of three independent biological replicates. P values shown in the charts are determined by one-way ANOVA. Source data are provided as a Source Data file.

We synthesized ABA-DMNB as previously reported⁴⁸ and its UV light-dependent uncaging to generate ABA was confirmed using HPLC analysis (Supplementary Fig. 9). To test light-inducible m⁶A editing, we tested using ABA-DMNB and light to control m⁶A writing at the Actb A1216 site. We transfected cells with constructs expressing dCas13b-PYL, ABI-M3, and Actb gRNA-2 for 24 h, and treated cells with ABA or ABA-DMNB and with or without UV light irradiation (2 min) for another 24 h before cells were harvested to quantify the m⁶A level. We observed that the m⁶A level increased only when cells were treated with ABA or with ABA-DMNB plus UV light, but not with ABA-DMNB without light (Fig. 5b). These results confirmed that by incorporating a photo-caged CIP inducer, ABA-DMNB, a light-controlled m⁶A editing platform can be created, which provides further spatiotemporal controls than using the chemical ligand alone, and can potentially provide dosage-dependent control of effects⁴⁸, which is difficult to achieve using light-inducible dimerization proteins.