Tailoring and reversing m⁶A editing with sequential RNA bioorthogonal chemistry

Abstract

Many existing methods for post-transcriptional RNA modification rely on a single-step approach, limiting the ability to reversibly control m⁶A methylation at specific sites. Here, we address this challenge by developing a multi-step system that builds on the concept of sequential RNA bioorthogonal chemistry. Our strategy uses an azide-based reagent (NAI-N₃) capable of both cleavage and ligation reactions, thereby allowing iterative and reversible modifications of RNA in living cells. By applying this approach in CRISPR (clustered regularly interspaced short palindromic repeats)-based frameworks, we demonstrate tailored editing of m⁶A marks at targeted RNA sites, overcoming the one-way restriction of conventional bioorthogonal methods. This sequential protocol not only broadens the scope for fine-tuned RNA regulation but also provides a versatile platform for exploring dynamic m⁶A function in genetic and epigenetic research.

Graphical Abstract

Open in new tabDownload slide

Introduction

The intricate regulation of RNA functionality underpins vital processes within cellular biology [1–3], where advanced chemical methodologies have increasingly become crucial [4]. Traditional approaches for controlling RNA functionality have predominantly relied on solid-phase synthesis modifications, a method limited by the synthetic RNA’s length and high cost. In contrast, post-synthetic modification techniques emerge as a formidable alternative, offering a versatile platform for the introduction of regulatory handles onto RNA molecules post-synthesis [5, 6]. This approach leverages the principles of biorthogonal chemistry [4], which enables the incorporation of functional groups into biomolecules without interfering with the native biochemical processes. We believe that sequential bioorthogonal reactions will represent an evolution of this concept, potentially offering a more flexible method of control over RNA functionality. Unlike single-step bioorthogonal reactions [7, 8], which provide a one-time modification opportunity [9, 10], sequential reactions allow for multiple, successive modifications [11]. Such sequential modifications are particularly advantageous in the context of dynamic cellular environments, where the temporal and spatial regulation of RNA function can have profound implications. However, the development and application of sequential bioorthogonal reactions remain nascent, primarily because executing a bioorthogonal tag typically necessitates two distinct sequential bioorthogonal reactions: first, a cleavage reaction between the tag and an introduced chemical probe, followed by a ligation reaction with another chemical probe.

Dynamic RNA post-transcriptional modifications, especially N⁶-methyladenosine (m⁶A) [12, 13], play pivotal roles in various cellular processes [14–16]. The discovery in 2011 of FTO-mediated oxidative demethylation of m⁶A in nuclear RNA [17], along with the identification and mapping of m⁶A through m⁶A-seq in 2012 [15, 16], significantly advanced the understanding of its roles in numerous biological processes [18, 19]. The orchestration of m⁶A deposition, recognition, and removal by “writer,” “reader,” and “eraser” enzymes is critical for maintaining the delicate balance of m⁶A modifications [20–22], which are key to post-transcriptional gene control [23]. However, pinpointing the specific contributions of individual m⁶A sites is challenging due to their widespread presence across the transcriptome [24]. While recent innovations of clustered regularly interspaced short palindromic repeats (CRISPR) have facilitated site-specific m⁶A modification by combining dCas (dead Cas) proteins with m⁶A regulatory proteins [25–28], these methods could benefit from enhanced fine-tuning and reversible control to more comprehensively explore the dynamic nature of RNA modifications. Drawing on recent progress on chemical regulation of CRISPR–Cas9 via structural manipulation of guide RNA (gRNA) [29, 30], we propose leveraging these RNA control techniques with m⁶A editing tools for precise, programmable modification of specific m⁶A sites. The fusion of sequential bioorthogonal reactions with CRISPR-guided systems potentially represents a promising avenue for achieving detailed, reversible control over RNA m⁶A methylation.

In this study, we introduce a strategy termed sequential RNA bioorthogonal chemistry (SRBC), which builds upon azide-based chemistry to enable versatile and reversible editing of m⁶A sites within CRISPR frameworks. The crux of this strategy lies in the molecule NAI-N₃, which carries an azide group to facilitate both Staudinger reduction (for the removal of modifications) and click ligation (for the addition of bulky groups). Our SRBC approach proceeds in three phases (Fig. 1A and B). First, RNA is transiently modified with NAI-N₃, effectively suspending its function by introducing steric or structural hindrances. Second, Staudinger reduction selectively removes most of these modifications, restoring RNA activity. Third, an optional in situ strain-promoted azide–alkyne cycloaddition (SPAAC) targets any residual NAI-N₃, providing the opportunity to irreversibly inactivate the RNA if desired. By coupling this methodology with CRISPR-based m⁶A editing—examined here via SELECT [31] and MeRIP–RT-qPCR (methylated RNA immunoprecipitation-reverse transcription–quantitative polymerase chain reaction)—we demonstrate robust and programmable site-specific methylation control in both M3M14-dCas9 [25] and dCas13a-M3M14 systems [32]. Our data confirm that SRBC retains efficacy across varied RNA transcripts and cellular contexts. By adopting sequential, reversible reactions that better mirror the dynamic nature of RNA biology, SRBC stands to advance programmable RNA modifications.

Materials and methods

Human cell lines

HeLa cells and HEK293T cells were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. The cells were incubated at 37°C with 5% CO₂ in a humidified environment. Cells were passaged with 0.25% trypsin–EDTA and phenol red. To ensure mycobacteria-free cultures, the cells were screened for mycoplasma contamination using the MycoBlue Mycoplasma Detector Kit following the manufacturer’s instructions.

Dot blot assay

The nylon membrane was prepared to the required dimensions. A total of 100 ng of either native or NAI-N₃-modified R-32nt RNA was treated with varying concentrations of DPPEA (Supplementary Fig. S3E) in a 5 μl reaction at 37°C for 30 min. Following this, 0.5 μl of 2 mM DBCO-biotin was added, and the mixture was incubated at 37°C for an additional 30 min. RNA purification was performed using the RNA Clean & Concentrator™ as per the manufacturer’s instructions. Approximately 15 ng of each sample was then carefully applied to the membrane, allowed to air dry, and subjected to UV cross-linking twice using a HybriLinker HL-2000 Digital Hybridization Unit (UVP). The membrane underwent blocking with 5% bovine serum albumin (BSA) at 37°C for 1 h, followed by three washes in 1× TBST (Tris-Buffered saline with Tween 20) and incubation with high-sensitivity streptavidin-HRP at 37°C for 1 h. After another series of three washes in 1× TBST, the membrane was treated with ECL chemiluminescent substrate and visualized using a ChemiDocMP Imaging System (Bio-Rad Laboratories, USA). For a quality check, methylene blue staining was also performed. As a control, biotin-modified DNA (DNA-biotin) was utilized, with its sequence provided in Supplementary Table S2.

RT-qPCR analysis

Approximately 1.5 × 10⁵ HeLa cells or 293T cells (per well) were seeded into 12-well plates and cultured overnight before transfection with 1.5 μg of M3M14-dCas9 or dCas13a-M3M14 using Lipofectamine™ 3000. Transfected cells were cultured for 6 h, washed once with warm DMEM, and then subjected to a second transfection with 1.0 μg of gRNA using Lipofectamine 3000. The medium was changed to complete DMEM (with indicated concentrations of DPPEA) after 4 h post-transfection. Cells were then cultured for an additional hour before the medium was changed to complete DMEM [containing 8 μM t-DBCO (Supplementary Fig. S2B)], and continued for 36 h (unless otherwise noted, the following experiments treated with DPPEA and t-DBCO were under these conditions). After 36 h, cells were washed with phosphate buffered saline (PBS), and total RNA was extracted using the SteadyPure Universal RNA Extraction Kit II following the provided protocol. The concentration and quality of the total RNAs were determined using NanoDrop (Thermo Fisher). The extracted total RNAs were employed as templates for complementary DNA (cDNA) synthesis using the Hifair^® V one-step RT-gDNA digestion SuperMix for qPCR. The generated cDNA templates were subjected to qPCR analysis using the Hieff^® qPCR SYBR Green Master Mix in 96-well plates. The qPCR reactions were carried out using a C1000™ Thermal Cycler (Bio-Rad). All samples were run in technical duplicate. Gapdh was used as the internal reference control for Actb, Foxm1, and Sox2 analysis, while Actin was used as the internal reference control for Gapdh analysis. All used single-guide RNA (sgRNA), PAMmer RNA, and CRISPR RNA (crRNA) are listed in Supplementary Table S3.

SELECT for detection of m⁶A

To prepare the SELECT assay, 2 μg of total RNA was combined with 40 nM each of up and down primers, along with 5 μM dTTP in a 17 μl mixture containing 1× CutSmart buffer (50 mM KAc, 20 mM Tris–HAc, 10 mM MgAc₂, 100 μg/ml BSA, pH 7.9, at 25°C). This mix underwent a series of annealing steps: starting at 90°C for 1 min, then sequentially dropping by 10°C every minute until reaching 40°C, where it was held for 6 min. After annealing, 3 μl of an enzyme blend comprising 1× CutSmart buffer, 0.01 U Bst 2.0 DNA polymerase, 0.5 U SplintR ligase, and 10 nmol ATP was introduced to the reaction. The combined 20 μl reaction was then incubated at 40°C for 20 min, denatured at 80°C for 20 min, and chilled to 4°C pending further analysis. qPCR (initial denaturation at 95°C for 5 min, followed by 40 cycles of 95°C for 10 s and 60°C for 45 s) was employed to evaluate the SELECT assay yields, with the results normalized against the RNA level of the targeted transcript at each specific site. The primers utilized for SELECT analysis are detailed in Supplementary Tables S1 and S2.

mRNA stability

HeLa or HEK293T cells underwent transfection with M3M14-dCas9 or dCas13a-M3M14 systems, paired with the respective gRNAs. Subsequent to DPPEA and t-DBCO treatment, cells were exposed to the transcription inhibitor actinomycin D and collected at designated intervals (0, 1, 3, and 6 h). Cells were then rinsed with PBS, harvested, and total RNA was isolated and analyzed with subsequent RT-qPCR analysis.

m⁶A immunoprecipitation and RT-qPCR

HeLa and HEK293T cells were plated in 10-cm dishes until reaching ∼75% confluence, after which they were transfected with 6 μg of either the M3M14-dCas9 or dCas13a-M3M14 plasmid using Lipofectamine™ 3000 (Thermo Fisher) for 6 h, followed by transfection of NAI-N₃-modified gRNA, and treated with DPPEA and t-DCBO. The cells were then incubated for an additional 36 h before collection. Cells were washed with PBS, harvested, and total RNA was extracted.

m⁶A immunoprecipitation was conducted according to a well-established protocol [33]. Initially, 20 μl of protein A/G magnetic beads (Thermo Fisher Scientific) underwent two washes in 1× m⁶A IP buffer (150 mM NaCl, 10 mM Tris–HCl, pH 7.5, 0.1% NP-40, RNase-free H₂O) and were then suspended in 300 μl of the same buffer. To this, 3 μg of m⁶A antibody (New England Biolabs) was added, and the mixture was incubated overnight at 4°C with gentle rotation. Following two buffer washes, the antibody-coated beads were combined with 300 μl of a solution containing 1× m⁶A IP buffer, 90 μg of fragmented total RNA, 4 μl of RNase inhibitor (ABclonal), and 2 mM RVC (ribonucleoside vanadyl complexes) [26], and incubated for an additional 6 h at 4°C with continuous gentle rotation. The beads were then washed once with 1× m⁶A IP buffer, twice with low-salt IP buffer (50 mM NaCl, 10 mM Tris–HCl, pH 7.5, 0.1% NP-40, RNase-free H₂O), and twice with high-salt IP buffer (500 mM NaCl, 10 mM Tris–HCl, pH 7.5, 0.1% NP-40, RNase-free H₂O). m⁶A-enriched RNA fragments were eluted using RLT buffer (Qiagen) and subsequently purified with RNA Clean & Concentrator-5 kits (Zymo). The purified m⁶A-enriched RNA was analyzed with subsequent RT-qPCR analysis [34].

Western blot analysis

Around 1.5 × 10⁵ HeLa or 293T cells were plated per well in 12-well plates and cultured overnight. They were transfected with 1.5 μg of M3M14-dCas9 or dCas13a-M3M14 plasmids. Transfected cells were cultured for 6 h, washed once with warm DMEM, and subjected to a second transfection with 1.0 μg of gRNA using Lipofectamine 3000, treating with DPPEA and t-DBCO subsequently. After 36 h, cells were washed with PBS and lysed with 100 μl of RIPA buffer containing protease inhibitors, incubated on ice for 10 min, and then scraped and collected into microcentrifuge tubes. Centrifugation at 12 000 × g for 15 min at 4°C separated the supernatants, which were then moved to new tubes, mixed with 5× SDS loading buffer, heated at 95°C for 5 min for denaturation, and subjected to 12% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) using a 10 μl sample volume. Electrophoresis settings were initiated at 80 V and increased to 120 V once the marker bands clearly separated.

The proteins were then transferred to a pre-activated polyvinylidene fluoride (PVDF) membrane at a steady 200 mA current, timing adjusted by protein size. After transferring, membranes were rinsed with TBST, blocked with 5% BSA in TBST for 1 h at 37°C, washed four times with TBST, and then incubated with primary antibody overnight at 4°C. Following another four TBST washes, membranes were incubated with HRP Goat Anti-Rabbit IgG (H + L) for 1 h at 37°C, washed three times with TBST, and visualized using the ChemiDoc^MP Imaging System. All antibodies were diluted in 5% BSA–TBST.

Results

Tailoring and reversing RNA functions with SRBC in RNA-only systems

Our study pioneers the SRBC strategy, an advanced method that employs azide chemistry for dynamic and iterative RNA manipulation. Initially, we densely modify RNA with NAI-N₃ groups, a key step that temporarily inhibits RNA activity by obstructing its function. Subsequently, Staudinger reduction is employed to meticulously remove most of the NAI-N₃ groups (refer to Fig. 1A and Supplementary Fig. S1), effectively restoring RNA’s activity. Intriguingly, we further integrate an optional in situ SPAAC reaction that capitalizes on the residual NAI-N₃ groups not entirely removed from RNA (illustrated in Fig. 1B). This step leads to a permanent inhibition of RNA activity, offering a versatile approach to selectively and reversibly control RNA function with precision.

We first explored this SRBC strategy using a short 32-nt single-stranded RNA. By employing the RNA 2′-OH acetylation technique, NAI-N₃ activated with CDI (1,1′-carbonyldiimidazole) was incubated with the model RNA at 37°C to prepare NAI-N₃ modified R-32nt. Analysis of the modified RNA via denaturing PAGE assay revealed that with the extension of reaction time, the migration rate of the modified RNA gradually decreased, indicating an increase in the NAI-N₃ groups tagged on the RNA (see Supplementary Fig. S2A). To test the ligation activity of the modified RNA, RNAs with varying degrees of modification were also reacted with a multimeric dibenzocyclooctyne (t-DBCO in Supplementary Fig. S2B) at room temperature for 15 min before further analysis by denaturing PAGE assay. It was observed that after treatment with t-DBCO, the migration rate of the modified RNA bands significantly reduced, and a set of cross-linked RNA bands, which barely moved, appeared. This phenomenon indicates that the modified RNA exhibits robust click reaction activity. Conversely, native RNA was unaffected by the t-DBCO treatment (Supplementary Fig. S2C).

To achieve optimal control effects with the SRBC strategy, we selected seven different triphosphine compounds and tested their efficiency in releasing NAI-N₃ groups from modified RNA (Fig. 2A–H and Supplementary Fig. S3). Among these, DPA, TPPMS, and DPP showed no release effect even at 1024 μM (Supplementary Fig. S3A, C, and D), while THPP and TCEP exhibited only weak release effects (Supplementary Fig. S3F and G). DPBA and DPPEA significantly released NAI-N₃ groups (Supplementary Fig. S3B and E), particularly DPPEA, which demonstrated exceptional efficiency with an EC₅₀ of only 22.9 μM. Based on these results, we chose DPPEA for subsequent SRBC strategy experiments involving bioorthogonal cleavage.

Next, we began testing the feasibility of the SRBC strategy by treating modified RNA with varying concentrations of DPPEA to achieve different degrees of release. After incubating at 37°C for 30 min, we immediately added DBCO-biotin for the second step of the bioorthogonal ligation reaction, and finally analyzed the differently treated samples using a dot blot assay (Fig. 2I and J). It was evident that in the DMSO group, neither unmodified nor modified RNA could detect a biotin signal under any DPPEA concentration treatment. However, in the DBCO-biotin (structure in Supplementary Fig. S4A) treatment group, modified RNA treated with SA-HRP exhibited bright signal spots. More importantly, as the concentration of DPPEA increased, the signal intensity gradually decreased, and at 256 μM DPPEA treatment, the signal was almost completely lost. These results indicate that we can control the number of NAI-N₃ groups in RNA by adjusting the concentration of DPPEA, and these remaining NAI-N₃ groups still possess good click reaction performance.

We also substituted DBCO-biotin used in the second step of the bioorthogonal ligation reaction with t-DBCO for testing. Denaturing PAGE analysis showed that as the concentration of DPPEA treatment increased, the reactivity of modified RNA with t-DBCO gradually decreased. After 128 μM DPPEA treatment, the modified RNA released almost all NAI-N₃ groups, and bands reacting with t-DBCO were nearly unobservable (Supplementary Fig. S4B). Moreover, when reacting different concentrations of t-DBCO with RNA treated with 8 μM DPPEA, it was observed that RNA, which had released a portion of NAI-N₃ groups after low concentration DPPEA treatment, exhibited similar reactivity to RNA modified without DPPEA treatment. As the concentration of t-DBCO increased, the degree of reaction gradually rose, and cross-linked bands increased (Supplementary Fig. S4C). These results imply that through the SRBC strategy, we can achieve sequential manipulation of RNA structure and function.

To validate our hypothesis, we tested the SRBC technique in the Broccoli-DFHBI and TERRA-ThT systems, as illustrated in Supplementary Fig. S5A and B. When Broccoli RNA is correctly folded, it binds to the DFHBI ligand, resulting in bright fluorescence [35]. TERRA, a classic rG4 sequence, forms stable complexes with the small molecule ThT, also emitting bright fluorescence [36]. We observed that NAI-N₃ modifications significantly disrupted the function of these ligand molecules, greatly diminishing the fluorescence signals. After treatment with DPPEA, the fluorescence of both DFHBI and ThT was substantially enhanced. However, subsequent treatment with t-DBCO reversed the effects of DPPEA, causing the fluorescence signals of DFHBI and ThT to be lost once again. Furthermore, using qRT-PCR in conjunction with a dye-binding assay, we examined changes in the melting temperature (T_m) between R-20nt (a 20-nt sequence from the sg-Gapdh spacer region) and its complementary strand R-20nt-c under various conditions, as shown in Supplementary Fig. S5C–G. We found that as the degree of NAI-N₃ modification increased, the T_m values for ds R-20nt/R-20nt-c gradually decreased. At 180 min of modification, the T_m was effectively undetectable, indicating nearly complete disruption of RNA double-strand pairing. Subsequently, we selected the 180 min mod. R-20nt for further study. After treatment with DPPEA, its T_m value recovered to 55°C, but subsequent use of t-DBCO again disrupted the ds R-20nt/R-20nt-c structure, reducing the T_m to unmeasurable levels. Notably, neither individual nor combined treatments of DPPEA or t-DBCO affected the T_m value of native R-20nt. These results also demonstrate the excellent versatility of the SRBC technique, showing its applicability across various RNA systems.

SRBC-CRISPR-dCas9 integration for reversible RNA methylation

Subsequent to confirming SRBC strategy’s potential in RNA-only systems, we proceeded to explore its integration with CRISPR frameworks, aiming to exert tailored control over m⁶A methylation at predetermined RNA sites. The effectiveness of CRISPR technologies hinges on gRNA’s role in targeting and efficiency, implying that steering gRNA’s structural and functional properties essentially steers CRISPR’s outcomes. In 2019, the Qian group introduced a programmable RNA m⁶A editing system by fusing CRISPR-dCas9 with either a single-chain m⁶A methyltransferase or demethylase [25]. We therefore utilized the M3M14-dCas9 platform to assess our SRBC strategy. We targeted adenine A1216 in the 3′-untranslated region (UTR) of β-actin (Actb), a recognized m⁶A site with a modest methylation level (21% in HeLa cells) [25], allowing potential for enhanced m⁶A incorporation. A previously reported sgRNA (sgActb, with a protospacer adjacent 2 bp 5′ to the Actb A1216 site) along with its corresponding PAMmer was employed to direct M3M14-dCas9 [37]. We modified sgActb with 20 mM NAI-N₃ at 37°C for 10–30 min to ascertain optimal modification levels that could entirely inhibit M3M14-dCas9 function without excessive modification hindering subsequent release.

Initially, HeLa cells were transfected with a plasmid expressing M3M14-dCas9 for 4 h, followed by the transfection of either native or modified sgActb, and further cultured for 24 h. An antibody-independent assay, SELECT method [31], is utilized to quantify site-specific m⁶A editing, leveraging m⁶A’s ability to interfere with the single-base extension by Bst DNA polymerase and the ligation activity of SplintR ligase at single-base resolution (Fig. 3A). Our findings indicated a >70% reduction in ligation products for the targeted Actb A1216, while the modification of sgActb similarly impeded m⁶A incorporation in a time-dependent manner (Fig. 3B and C), with 30 min of modification showing no effect on ligation products, akin to nontargeting sgRNA (sgNT). To validate these observations, an antibody-dependent assay was conducted to measure the m⁶A abundance under each condition. The total RNAs extracted were fragmented and enriched with m⁶A antibodies. RT-qPCR quantified the enrichment of immunoprecipitated RNA fragments (MeRIP–RT-qPCR). A >50% increase in m⁶A methylation levels at the target Actb A1216 region was observed for native sgActb, whereas modified sgActb progressively reduced methylation with extended modification time, and 30 min of NAI-N₃modification completely inhibited RNA methylation (Fig. 3D).

To further validate these findings, we extended the application of NAI-N₃ modification to another endogenous transcript, targeting adenine A690 within the coding region of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) messenger RNA (mRNA), identified as another site of low methylation extent [32]. Aiming for optimal methylation efficiency, we designed a series of three sgRNAs with their protospacer 5′ ends positioned at varying distances (2, 6, and 10 bp; see Supplementary Fig. S6A) from the Gapdh A690 site. Both SELECT and MeRIP–RT-qPCR analyses demonstrated significant A690 methylation of Gapdh by all three sgRNAs, with sgGapdh-6 (6 bp distance, referred to as sgGapdh) exhibiting optimal efficiency (Supplementary Fig. S6B–D). This sgRNA was subsequently selected for further release and click evaluations. In line with our previous observations, introducing NAI-N₃ modifications to sgGapdh markedly inhibited the site-specific methylation mediated by M3M14-dCas9, achieving complete inhibition (fully turning off m⁶A writing) after 40 min of modification (Fig. 3E and G). These results collectively indicate that NAI-N₃ modification can effectively hijack the function of sgRNA, enabling the blockade of the M3M14-dCas9 m⁶A writer system.

Subsequently, we utilized the 40-min modified sgGapdh as a model sgRNA to assess the release efficiency of NAI-N₃ groups by different triphosphine reducing agents. Cells were treated with the indicated triphosphine 4 h after sgRNA transfection and washed following a 2-h incubation. The m⁶A levels within the A690 site of Gapdh were assessed using the SELECT assay (Supplementary Fig. S7). As anticipated, DPPEA proved to be the most effective compound, completely restoring RNA methylation with a 200 μM treatment (Supplementary Fig. S7D). Other triphosphine compounds also demonstrated a significant restoration of m⁶A writing.

After identifying the triphosphine compound with the most effective release capabilities, we proceeded to evaluate the dosage response of DPPEA in facilitating the recovery of m⁶A installation by the M3M14-dCas9 system to determine the optimal concentration required (Fig. 4A). Cells were incubated with specified concentrations of DPPEA for 2 h, then washed twice with warm medium, treated with or without 8 μM t-DBCO, and cultured for an additional 36 h. Results from the SELECT assay clearly demonstrated that, for both sgActb and sgGapdh, the restoration of RNA methylation exhibited a DPPEA dose-dependent pattern and was significantly inhibited in all samples treated with DBCO, indicating that the remaining NAI-N₃ groups had indeed conjugated with DBCO, significantly altering the original sgRNA structure (Fig. 4B and D, and Supplementary Fig. S8). Remarkably, m⁶A writing in samples treated with t-DBCO also progressively resumed when DPPEA concentrations exceeded 120 μM, suggesting that the modified NAI-N₃ groups were completely released from sgActb, which would not interact with t-DBCO treatment. These findings suggest that 120 μM DPPEA represents an optimal concentration for release and further click regulation.

MeRIP–RT-qPCR analysis, consistent with SELECT method results, showed that methylation at the A1216 site of Actb was substantially restored by 120 μM DPPEA treatment and nearly completely suppressed following t-DBCO treatment in the modified sgActb groups. Meanwhile, DPPEA and t-DBCO treatments had no effect on the RNA methylation mediated by native sgActb (Fig. 4C). At the A690 site in Gapdh, a remarkable 8.2-fold increase in m⁶A levels was observed in the modified sgGapdh group after treatment with 80 μM DPPEA, which was significantly reduced to 1.7-fold by t-DBCO treatment. Conversely, native sgGapdh exhibited no response to DPPEA and t-DBCO treatments (Fig. 4E). Together, these results unequivocally demonstrate the SRBC strategy’s capability to temporally control site-specific m⁶A modification. More importantly, releasing active sgRNA with DPPEA treatment followed by t-DBCO addition at varying time intervals led to different m⁶A levels at the target site (Fig. 4H and I). This implies that the SRBC method enables precise control over the operational duration of the m⁶A writer, allowing for the regulation of m⁶A levels at specific sites to desired extents to investigate and explore the various biological functions associated with different m⁶A levels.

To demonstrate the versatility of our SRBC-integrated dCas9-M3M14 platform in broad-spectrum m⁶A editing, we selected the transcripts of forkhead box protein M1 (FOXM1) and SRY box 2 (SOX2) as our targets due to their biological significance. Methylation at these sites is known to correlate with glioblastoma [38] and stem cell differentiation [39], respectively. Furthermore, both transcripts exhibit endogenous methylation in HEK293T cells, presenting an opportunity for enhanced m⁶A incorporation via overexpressed methyltransferases. Mirroring the approach used in previous studies, we designed a series of sgRNAs and their corresponding PAMmers to identify the optimal proximity between their 5′ protospacer ends and the target sites. The effectiveness of each sgRNA was determined through MeRIP–RT-qPCR to select the most suitable candidates for subsequent experiments (Supplementary Fig. S9). In the SRBC protocol, sgFoxm1 (sgFoxm1–9)/sgSox2 (sgSox2–7) underwent modification with NAI-N₃ for various durations to determine the point of saturation. Consistent with the outcomes for Actb and Gapdh target transcripts, DPPEA treatment successfully restored the inhibited m⁶A writing in modified sgFoxm1/sgSox2 transcripts (achieving an 8.9-fold increase for sgFoxm1 and a 2.7-fold increase for sgSox2), which was subsequently reduced upon t-DBCO treatment (down to 1.9-fold for sgFoxm1 and 1.2-fold for sgSox2; Fig. 4F and G). Cumulatively, the data from MeRIP–RT-qPCR and SELECT assays across four distinct target transcripts and two different cell types underscore the SRBC technology’s capability to precisely manipulate RNA-guided site-specific m⁶A editing efficiently.

Tailored mRNA stability and translation control with SRBC-dCas9-M3M14

Having successfully demonstrated controlled site-specific m⁶A modification of endogenous mRNA, we proceeded to investigate whether the artificially incorporated m⁶A exerted biological functionality, capable of triggering downstream effects (Fig. 5A). It has been documented that methylation of A1216 in Actb transcript leads to a reduction in its stability [37]. To verify this, we assessed the stability of Actb mRNA by measuring its half-life under various treatment conditions. HeLa cells, corresponding to each experimental group, were collected at designated intervals following actinomycin D application (a DNA transcription inhibitor) to quantitatively analyze endogenous Actb mRNA levels using RT-qPCR. In line with expectations, native sgActb-mediated m⁶A installation significantly shortened Actb’s half-life, whereas Actb’s stability was preserved when modified sgActb was utilized (Fig. 5B and Supplementary Fig. S10). Crucially, we observed that the destabilization effect on Actb mRNA, attributed to sgRNA activity deficiency, was reversible upon DPPEA application and could be subsequently neutralized through t-DBCO treatment within the modified sgActb cohort (Fig. 5D and Supplementary Fig. S10). Conversely, native sgActb displayed no susceptibility to either DPPEA or t-DBCO interventions (Fig. 5C and Supplementary Fig. S9). These findings suggest the feasibility of artificially modulating mRNA stability in vivo via SRBC-driven site-specific m⁶A editing.

We further explored whether the SRBC-based approach for RNA methylation regulation could be harnessed to modulate mRNA functionality, given that m⁶A modifications within eukaryotic mRNA are known to impact RNA transcript stability and translation efficiency. We confirmed that SRBC-driven precise control of site-specific m⁶A editing at A1398 and A1405 (3′-UTR) of Sox2, as well as A3488 and A3504 (3′-UTR) of Foxm1, effectively influenced mRNA stability and subsequent translation in HEK293T cells. Initially, we transfected cells with native sgRNA and a spectrum of modified sgRNAs, incrementally extending modification durations, then assessed relative mRNA levels as indicators of transcript stability. Native sgRNAs indeed facilitated mRNA destabilization, which was counteracted by NAI-N₃ modification (Supplementary Fig. S11A and B). Treatment with DPPEA released the sgRNA, markedly diminishing relative mRNA levels, while subsequent t-DBCO treatment recaptured the reactivated sgRNA, thereby preserving mRNA levels (Fig. 5E and F, and Supplementary Fig. S11C and D). Western blot analysis indicated that m⁶A installation reduced FOXM1 and SOX2 protein levels; however, modified sgRNA lacked observable impact on protein levels due to structural and functional deficits. DPPEA treatment activated the m⁶A writing system, leading to targeted protein downregulation. Furthermore, the SRBC system’s induced reduction in protein expression was reversible upon t-DBCO treatment (Fig. 5G for FOXM1 and Fig. 5H for SOX2). Collectively, these findings underscore that SRBC-based site-specific m⁶A editing regulation can effectively govern endogenous protein expression within a native cellular context.

Broad applications of SRBC within a compact CRISPR/dCas13b framework

One of the SRBC system’s most notable advantages lies in its RNA-centric approach to modulate structure, activity, and consequent function. This versatility allows for its universal application across multiple RNA-guided functional platforms [32], including CRISPR/dCas13b systems [26], characterized by smaller protein sizes and guided by shorter, PAMmer-free gRNAs (Fig. 6A). We synthesized corresponding gRNAs targeting the previously mentioned four transcripts in vitro, modifying them as per our established protocol. Their methylation activities were assessed using the SELECT (for Actb and Gapdh) and RT-qPCR (for Foxm1 and Sox2) methods. NAI-N₃ modifications severely impaired the dCas13-M3M14 function in a dose-responsive manner (Supplementary Fig. S12), highlighting the effectiveness of our SRBC system. The conditions showing complete blockage were further analyzed to determine the optimal parameters for the SRBC strategy. DPPEA treatment demonstrated a dose-dependent reactivation of all four modified gRNAs. Furthermore, subsequent t-DBCO treatments effectively nullified the restored gRNA functions, switching off m⁶A editing once more (Supplementary Fig. S13). Orthogonal confirmation of the efficient toggling of site-specific m⁶A writing was obtained via MeRIP–RT-qPCR assays. These assays revealed that modified gRNA led to a significant reduction in m⁶A incorporation at each target site, which was substantially recovered following DPPEA treatment and almost entirely inhibited again with t-DBCO incubation (Fig. 6B–E). The mRNA stability of Actb (Supplementary Fig. S14), Foxm1 (Fig. 6F), and Sox2 (Fig. 6G), along with protein expression levels of Foxm1 (Fig. 6H) and Sox2 (Fig. 6I), was scrutinized, showcasing the SRBC system’s capacity for high-efficiency and reversible control of endogenous mRNA functionality. Conversely, native gRNAs remained unaffected by DPPEA and t-DBCO treatments, underscoring the high selectivity of the SRBC regulatory framework.

Discussion

Our primary goal was to develop a platform that achieves precise and reversible m⁶A editing in living cells—an essential step toward dissecting the complex roles of RNA modifications in gene expression and cellular function. The SRBC method addresses this need by building on azide-based chemistry for multi-phase control over RNA structure and function. Unlike many existing approaches, which rely on irreversible chemical modifications or protein engineering, SRBC harnesses two core steps—Staudinger reduction for partial removal of azide tags and SPAAC for click ligation—to enable a stepwise cycle of functional “ON” and “OFF” states. Previous strategies for RNA modification often manage only one-time changes, offering limited control. In contrast, SRBC can repeatedly apply cleavage and ligation reactions at chosen time points, more accurately reflecting the nuanced, spatiotemporal control that RNA modifications undergo in vivo. This feature is particularly valuable for investigating rapid changes in m⁶A status that are linked to complex cellular processes.

While the SRBC approach shows promise, it also faces noteworthy challenges. First, the efficiency of Staudinger reduction and SPAAC may vary according to RNA secondary structure, local chemical environment, and modification site accessibility. This variability can lead to incomplete reversibility or inconsistent levels of modification across different transcripts. Second, the possibility of off-target effects, such as partial modifications at unintended RNA sites, remains a concern—particularly in highly structured regions. Addressing these issues will require further refinement of reaction conditions, possibly through engineering more selective reducing agents or click ligation reagents. Third, although we demonstrate compatibility with M3M14-dCas9 and dCas13a-M3M14, the performance of SRBC in other CRISPR systems remains to be thoroughly evaluated.

Moving forward, improvements in reagent design could enhance both the specificity and the kinetics of each chemical step, potentially reducing the concentrations and times needed to achieve robust editing. We also envision that SRBC could be expanded to include orthogonal reaction pairs, offering even greater flexibility. For instance, coupling azide chemistry with another fully independent bioorthogonal reaction might enable multiplexed, concurrent modifications on multiple RNA sites. Additionally, adopting photocleavable groups could mitigate potential toxicity concerns, making the approach more amenable to sensitive cell types or in vivo applications. By adjusting the “ON-time” and “OFF-time” for specific m⁶A marks, researchers can dissect the immediate and downstream effects on RNA stability, translation, and protein expression.

Conclusions

In summary, SRBC provides a stepwise and reversible means of controlling m⁶A modifications at specific RNA sites. Our results indicate that SRBC is broadly compatible with various CRISPR platforms, enabling researchers to selectively toggle m⁶A writing in a predictable manner. The improvements in precision and reversibility open new possibilities for investigating how m⁶A modifications shape complex RNA function in cells.

Acknowledgements

We would like to thank Dr Zhixian Qiao and Xiaocui Chai at the Analysis and Testing Center of Institute of Hydrobiology, Chinese Academy of Sciences for their assistance with RNA-seq and data analysis.

Author contributions: T.T. and X.Z. conceptualized the original idea, designed the studies, and led the project. X.Y.L., Q.Q.Q., and W.S. conducted the biological experiments. W.X. was responsible for the synthesis of the compounds. Y.Y.Z., X.Y.X, Y.T.Z., E.Y.Z., and M.L. performed gel electrophoresis analysis. T.T. wrote the initial draft and revised the manuscript. All authors reviewed and approved the final manuscript.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

The authors acknowledge financial support from the National Natural Science Foundation of China (Grant Nos. 22037004, 22377094, 22177089, 91853119, 91753201, 22177088) and the Fundamental Research Funds for the Central Universities (Grant No. 2042023kf0204). Funding to pay the Open Access publication charges for this article was provided by National Natural Science Foundation of China (Grant No. 22037004).

Data availability

The data underlying this article are available in the article and in its online supplementary material.

References

Author notes