Site-Specific m⁶A Erasing via Conditionally Stabilized CRISPR-Cas13b Editor

Abstract

N6-methyladenosine (m⁶A) on RNAs plays an important role in regulating various biological processes and CRIPSR technology has been employed for programmable m⁶A editing. However, the bulky size of CRISPR protein and constitutively expressed CRISPR/RNA editing enzymes can interfere with the native function of target RNAs and cells. Herein, we reported a conditional m⁶A editing platform (FKBP*-dCas13b-ALK) based on a ligand stabilized dCas13 editor. The inducible expression of this m⁶A editing system was achieved by adding or removing the Shield-1 molecule. We further demonstrated that the targeted recruitment of dCas13b-m⁶A eraser fusion protein and site-specific m⁶A erasing were achieved under the control of Shield-1. Moreover, the release and degradation of dCas13b fusion protein occurred faster than the restoration of m⁶A on the target RNAs after Shield-1 removal, which provides an ideal opportunity to study the m⁶A function with minimal steric interference from bulky dCas13b fusion protein.

Graphical Abstract

A conditional m⁶A erasing platform was developed. The FKBP*-dCas13b-ALKBH5 fusion protein can only be stabilized by Shield-1 to achieve site-specific m⁶A erasing. The removal of Shield-1 causes rapid degradation and release of the dCas13b-m⁶A eraser from target RNAs, while the reversal of m⁶A editing is much slower, providing an opportunity to study the m⁶A with minimal steric interference from the bulky dCas13b fusion proteins.

Introduction

Recent advances in the studies of chemical modifications on RNAs have shed light on a new layer of post-transcriptional regulation on genetic information in eukaryotes^[1]. So far, more than 170 distinct chemical modifications have been identified on RNAs from all living species^[2], among which, N6-methyladenosine (m⁶A) is the most prevalent in eukaryotic RNAs^[3]. The m⁶A modification is dynamic and reversible in the cells, regulated by the methyltransferase (m⁶A writers), demethylases (m⁶A erasers) and recognition proteins (m⁶A readers)^{[1, 4]}. In mammalian cells, methyltransferase-like 3 (METTL3), methyltransferase-like 14 (METTL14) and Wilms tumor 1-associated protein (WTAP) are responsible for m⁶A deposition^[5]. The S-adenosyl-L-methionine (SAM) binding pocket of METTL3 transfers the methyl group from the SAM to the N6 position of adenosine in the DRACH motif (D represents A, G, U; R represents A, G; and H represents A, C, U)^{[3b, 5a, 6]}, while the METTL14 and WTAP facilitate the methylation process. Fat mass and obesity-associated protein (FTO) and ALKB homologue 5 (ALKBH5) from ALKB subfamily of α-oxoglutarate-dependent dioxygenase remove m⁶A modification through a ferrous-dependent oxidation pathway^{[5c, 7]}. The m⁶A readers include proteins of YTH family and heterogenous nuclear ribonucleoproteins (HNRNPs) that mediate m⁶A-associated downstream effects^{[5c, 7]}. It is known that m⁶A directly impacts RNA properties and activities, for example, altering the structure of RNA^[8], modulating RNA metabolism and stability^{[3a, 4b, 9]}, enhancing the phase separation of mRNA^[10], and affecting mRNA maturation^[11], translocation^[12] and translation^[13], which further regulate various cellular activities^{[3b, 13a]}. It has also been shown that the roles of m⁶A are context-dependent and can vary when m⁶A modifications are located within different regions on an RNA transcript, on different RNAs, or under distinct cellular or tissue types and environments^[14]. Genetic approaches manipulating m⁶A-associated enzymes/proteins have been applied to study m⁶A functions, however they alter m⁶A distribution patterns globally and irreversibly, which are not effective in dissecting the dynamic and context-dependent roles of m⁶A in cells.

To address these issues, advanced m⁶A editing tools based on the clustered regularly interspaced short palindromic repeat (CRISPR) methods have been recently reported. These approaches use programmable single guide RNAs (sgRNAs) for site-specific recruitment of m⁶A writers, erasers, or readers, either constitutively or temporally, to the target RNA transcripts/loci to achieve m⁶A editing^{[6, 15]}. In these CRISPR-based m⁶A editing methods, the bulky nuclease-dead Cas9/Cas13b (dCas9 or dCas13b) fusion protein remains constitutively bound with the RNA targets even after the m⁶A editing is completed, raising the concern about its steric interference with the target RNAs. Although strategies have been reported to address this issue by developing a miniature or split RNA targeting CRISPR system^[16], these exogenous RNA regulating fusion protein components are constitutively expressed. The constitutive activity of dCas13 fusion proteins has a higher potential to induce off-target effects as well as undesirable immune response in the long-term^[17]. Strategies enabling dosage and temporal controls over the RNA-editing dCas13 fusion proteins will have great potential to tackle these problems.

It has been reported that protein destabilization domains (DDs) can be utilized to confer reversible destabilization to the DD-fused proteins of interest (POI)^[18], providing a precise control of the POI level in both inducible and reversible manners^{[18a, 18c, 19]}. For example, the DD of FK506 binding protein (FKBP) (FKBP*, with F36V and L106P mutations) can be stabilized by the small molecule, Shield-1, which enabled conditional expression of the FKBP*-fused proteins^[19–20]. The DDs have been coupled with the CRISPR technology to control Cas9 protein expression and activities^[21]. We envision that the integration of FKBP DD with the dCas13b proteins can provide the conditional and reversible stabilization of the dCas13b fusion protein under the control of Shield-1, which can be utilized for inducible and reversible m⁶A editing. To this end, we designed FKBP*-dCas13-ALKBH5 system, which will be referred to as FKBP*-dCas13b-ALK hereafter, by fusing the FKBP* DD with the dCas13b-ALKBH5 (core domain, 66–286 aa) for site-specific m⁶A erasing (Figure 1a). Our results demonstrated that the inducible site-specific m⁶A erasing can be achieved and reversed by the addition and removal of Shield-1. Moreover, the FKBP*-dCas13b-ALK fusion protein was released from the target RNA and degraded rapidly, while the altered m⁶A status was restored at a much slower pace, which can facilitate the investigation of m⁶A-deficient RNAs without the interference form dCas13 fusion proteins. This strategy has the advantages to provide temporal control of targeted RNA manipulation with minimal steric interference of dCas13b with the target RNAs, while avoiding constitutive expression of exogenous RNA editing enzymes.

Figure 1.

The establishment of inducible FKBP*-dCas13b-ALK. a) The fusion protein of FKBP*-dCas13b-ALK will be degraded without Shield-1 treatment, whereas it can be stabilized to induce the site-specific m⁶A erasing in the presence of Shield-1. The removal of Shield-1 will cause rapid degradation of FKBP*-dCas13b-ALK to release it from the RNA. b) The construct of FKBP*-dCas13b-ALK-HA. The Shield-1 dosage- (c) and incubation time-dependent (d)-protein stabilization of FKBP*-dCas13b-ALK-HA. The corresponding uncropped images of blots can be found in Figure S2.

Results and Discussion

To investigate whether the FKBP* DD and Shield-1 could be utilized to regulate the expression level of dCas13b-ALK for conditional m⁶A erasing, we cloned the FKBP*-dCas13b-ALK-HA plasmid with the FKBP* fused to the N-terminus of dCas13b-ALK and an HA tag to the C-terminus for detection purposes (Figure 1b). The Shield-1 dependent conditional expression of FKBP*-dCas13b-ALK was examined in HeLa cells by western blotting. We found that the FKBP*-dCas13b-ALK expression was barely detectable in the absence of Shield-1 (Figure S1), while its expression was rescued in the presence of Shield-1, indicating the robust stabilization efficiency of the fusion protein mediated by Shield-1. To further characterize the Shield-1 dependent conditional stabilization of FKBP*-dCas13b-ALK, the dosage-dependent and the kinetics of FKBP*-dCas13b-ALK expression controlled by Shield-1 were studied by monitoring the level of FKBP*-dCas13b-ALK expression in HeLa cells. We observed that the FKBP*-dCas13b-ALK protein level increased in a Shield-1 dosage-dependent manner, reaching maximum expression at 2 μM Shield-1 induction (Figure 1c). As a result, this concentration was used in the following time course experiments. We also found that the expression of FKBP*-dCas13b-ALK was observed rapidly after Shield-1 addition (within 3 h) and increased over time (Figure 1d). Collectively, these findings indicated that the level of FKBP*-dCas13b-ALK can be precisely controlled by Shield-1 in both dosage- and time-dependent manners.

Encouraged by these results, we further investigated the Shield-1 dependent targeted recruitment of the m⁶A eraser on the target RNAs. HeLa cells were transfected with FKBP*-dCas13b-ALK-HA and sgRNA targeting the A2577 site of Malat1 long non-coding RNA (lncRNA)^{[15d, 22]}, followed by Shield-1 or DMSO treatment. The condition of dCas13b-PYL (pyrabactin resistance (PYR)/PYR1-like component) with sgRNA was used as the negative control for normalization, while the dCas13b-ALK was used as the positive control^[15g]. The enrichment of FKBP*-dCas13b-ALK-HA at the target RNA site was then determined by RNA crosslinked immunoprecipitation (CLIP) using the anti-HA antibody and the quantitative polymerase chain reaction (qPCR) analysis. We observed an increased enrichment of FKBP*-dCas13b-ALK on the target RNA only in the presence of Shield-1, but not in the case of DMSO treatment (Figure 2a). These results demonstrated that the m⁶A eraser can be recruited to the targeted site on RNA of interest under the control of Shield-1.

Figure 2.

Site-specific m⁶A erasing on Malat1 mediated by the FKBP*-dCas13b-ALK-HA. a) The enrichment of FKBP*-dCas13b-ALK-HA at the Malat1 A2577 area. b) The relative m⁶A level at the A2577 site on the Malat1. c) The biological impact of m⁶A erasing on the binding with HNRNPC on Malat1. d) The kinetics of targeted m⁶A erasing on Malat1 induced by FKBP*-dCas13b-ALK system. All relative results were calculated by normalizing to the dCas13b-PYL+sgRNA (Ctr) group. Values and error bars represent the mean and s.e.m. of 3 independent biological experiments. P values were determined by one-way ANOVA.

To further investigate the conditional site-specific m⁶A erasing, HeLa cells were transfected and treated as in the experiments described above. The m⁶A levels were quantified by m⁶A-RNA immunoprecipitation (MeRIP) and qPCR analyses. We found that the FKBP*-dCas13b-ALK induced a significant decrease in the m⁶A level at the Malat1 A2577 site only in the presence of Shield-1, with an erasing efficiency comparable to the non-inducible version of dCas13b-ALK (Figure 2b), while the Shield-1 itself and the FKBP*-dCas13b-ALK with non-targeting sgRNA (NT-sgRNA) in the presence of Shield-1 did not cause obvious changes of m⁶A (Figure S3). To confirm that the m⁶A erasing indeed occurred at the A2577 site, we validated the site-specific reduction of the m⁶A level upon Shield-1 induction by the single base elongation- and ligation-based PCR amplification method (SELECT)^[22] (Figure S4). To examine the editing specificity on the same transcript, we measured the m⁶A level changes at a distal A2415 site within the m⁶A consensus sequence under the Shield-1 induction condition and did not observe any changes of m⁶A at this site (Figure S5). To further validate that the observed m⁶A erasing was also transcript-specific, we quantified the m⁶A level changes at other loci (e.g., A48 on the coding sequence of CYB5A mRNA, A180 on CTNNB1 mRNA)^[15b] that are not targeted by the Malat1 sgRNA, when editing the Malat1 A2577 site. Results showed that the m⁶A levels at these loci had no observable changes upon the addition of Shield-1 (Figure S6). Collectively, these results demonstrated that the designed FKBP*-dCas13b-ALK can achieve site and transcript-specific m⁶A erasing under the control of Shield-1.

Next, we investigated whether the edited m⁶A status can lead to expected biological outcomes. It has been reported that the m⁶A modification at A2577 of Malat1 lncRNA can expose the U-tract that can be recognized and bound by the heterogenous nuclear ribonucleoprotein C (HNRNPC)^[8], while the m⁶A removal disrupts this binding. To validate whether the m⁶A erasing on Malat1 mediated by this conditional m⁶A erasing platform can cause this biological effect, we quantified the abundance of HNRNPC at the Malat1 A2577 area under each treatment condition by CLIP and qPCR analyses in cells transfected and treated as described above. We observed a significant decrease of HNRNPC binding with Malat1 lncRNA only when Shield-1 was added (Figure 2c). In addition, we tested if the m⁶A removal at A2577 will impact the stability of Malat1 lncRNA and found that the RNA stability has no observable changes (Figure S7). These results verified that the artificially altered m⁶A status on A2577 site of Malat1 by the conditional editing platform was biologically relevant and this method can regulate the m⁶A-dependent cellular process on the targeted RNA.

To determine the kinetics of Shield-1 inducible m⁶A erasing by the FKBP*-dCas13b-ALK platform, time course experiments were performed to measure the m⁶A level at the A2577 site of Malat1 lncRNA at different time points after Shield-1 addition. HeLa cells were transfected with the FKBP*-dCas13b-ALK-HA and Malat1 sgRNA for 24 h, followed by Shield-1 treatment for different time points before being harvested for the quantification of m⁶A levels using MeRIP and qPCR analyses. We observed that the m⁶A level started to decrease after 3 h following Shield-1 treatment and continued to decrease over time (Figure 2d), which was consistent with the timeline of Shield-1-induced stabilization of FKBP*-dCas13b-ALK (Figure 1d). These results confirmed that the m⁶A erasing mediated by FKBP*-dCas13b-ALK can be rapidly induced in a time-dependent manner.

It is known that the induced stabilization of fusion proteins can be readily reversed after removing Shield-1^[19]. To investigate the reversibility of our m⁶A erasing system, we first examined the protein level of FKBP*-dCas13b-ALK at 24 h after Shield-1 removal in HeLa cells. We observed that Shield-1 stabilized FKBP*-dCas13b-ALK was completely degraded after Shield-1 removal (Figure S8). Moreover, we tested the reversibility of m⁶A erasing and FKBP*-dCas13b-ALK recruitment at the Malat1 lncRNA after Shield-1 removal. We transfected cells with the FKBP*-dCas13b-ALK-HA and Malat1 sgRNA for 24 h. Cells were then treated with Shield-1 for 24 h, followed by incubation with Shield-1-free media for another 24 h (i.e., the “wash” step) before being harvested for the quantification of the abundance of FKBP*-dCas13b-ALK and m⁶A level. We observed that the enrichment of FKBP*-dCas13b-ALK on Malat1 lncRNA returned to the background level at 24 h after Shield-1 removal (Figure 3a) and the reduced m⁶A level at A2577 site was fully restored at this point (Figure 3b). Moreover, we found that the observed reduction of HNRNPC binding accompanying m⁶A erasing was recovered as well after Shield-1 removal (Figure 3c). These results demonstrated that the m⁶A erasing and its biological impact mediated by Shield-1 and FKBP*-dCas13b-ALK were completely reversible.

Figure 3.

Reversibility of m⁶A erasing mediated by Shield-1 and FKBP*-dCas13b-ALK. a) The enrichment of FKBP*-dCas13b-ALK-HA on Malat1 lncRNA at 24 h post Shield-1 removal. b) The enrichment of m⁶A at the A2577 site of Malat1 lncRNA at 24 h post Shield-1 removal. c) The level of HNRNPC protein bound to Malat1 lncRNA at 24 h post Shield-1 removal. All relative results were calculated by normalizing to the dCas13b-PYL+sgRNA (Ctr) group. Values and error bars represent the mean and s.e.m. of 3 independent biological experiments. P values were determined by one-way ANOVA.

To further investigate the reversal kinetics of Shield-1-induced m⁶A erasing, we performed time course experiments to measure the m⁶A eraser enrichment and m⁶A level at the Malat1 A2577 site at different time points after Shield-1 removal. We transfected HeLa cells with the FKBP*-dCas13b-ALK-HA and Malat1 sgRNA for 24 h. Cells were then treated with Shield-1 for 24 h, followed by incubating with Shield-1-free media for different time before being harvested for quantifying the levels of FKBP*-dCas13b-ALK and m⁶A. We found that the release of m⁶A eraser from the Malat1 RNA occurred rapidly within 1 h and reached the background level within 3 h (Figure 4a). To confirm that the FKBP*-dCas13b-ALK-HA fusion protein was degraded after removing Shield-1, its protein level at different time points after Shield-1 removal was determined by western blotting. We observed a rapid decrease in the FKBP*-dCas13b-ALK-HA protein level within 1 h after Shield-1 removal and the fusion protein was completely degraded after 3 h (Figure 4b). Moreover, we noticed that the Shield-1-induced decreases of m⁶A at the Malat1 A2577 site persisted for an extended time period, which only slightly increased at 6 h after Shield-1 removal and not fully restored until 24 h after Shield-1 removal (Figure 4c). These results revealed a wide temporal gap between the release of dCas13b-m⁶A eraser and the restoration of m⁶A level on the target RNAs. It provided an optimal opportunity to study the impact of m⁶A alteration on the RNA in its native environment without potential interferences from exogenously introduced CRISPR/m⁶A editor components. Overall, our results demonstrated that Shield-1 and FKBP*-dCas13b-ALK can achieve inducible and reversible m⁶A erasing, while minimizing the constitutive steric interference from the dCas13b fusion proteins.

Figure 4.

The reversal kinetics of the m⁶A erasing mediated by FKBP*-dCas13b-ALK after 24-h Shield-1 induction followed by the Shield-1 removal. a) The enrichment of FKBP*-dCas13b-ALK-HA on the Malat1 at different time points after Shield-1 removal. b) Representative images of FKBP*-dCas13b-ALK-HA expression after Shield-1 removal. The uncropped images of blots can be found in Figure S9. c) The m⁶A level on Malat1 after Shield-1 removal. d) Inducible and reversible m⁶A erasing on CTNNB1 mRNA mediated by FKBP*-dCas13b-ALK upon Shield-1 addition and removal. All relative results were calculated by normalizing to the dCas13b-PYL+sgRNA (Ctr) group. Values and error bars represent the mean and s.e.m. of 3 independent biological experiments. P values were determined by one-way ANOVA.

To demonstrate that the conditional m⁶A erasing platform can be applied to editing other RNAs, we investigated the site-specific m⁶A erasing on CTNNB1 mRNA. We cloned the sgRNA targeting the A180 site of CTNNB1 mRNA^[15b]. Cells were transfected with FKBP*-dCas13b-ALK-HA and CTNNB1 sgRNA, followed by Shield-1 or DMSO treatment before being harvested to quantify the m⁶A levels. We observed that FKBP*-dCas13b-ALK-HA induced a significant decrease in the m⁶A level at the target CTNNB1 A180 site only in the presence of Shield-1 and the decreased m⁶A level was restored in 24 h after Shield-1 removal (Figure 4d). This inducible m⁶A erasing at the A180 site of CTNNB1 was further confirmed by SELECT assay (Figure S10). In addition, to confirm the m⁶A erasing was transcript specific, the m⁶A level changes on Malat1 and CYB5A when editing CTNNB1, we quantified the corresponding m⁶A levels and observed minimal changes of m⁶A at those non-targeted loci (Figure S11). Moreover, we studied the biological impact of m⁶A removal on the CTNNB1 A180 site by evaluating the changes in the stability and translation of the CTNNB1 mRNA. We observed that the mRNA stability increased when m⁶A was erased at the A180 site (Figure S12a), indicating that m⁶A at this site destabilized the CTNNB1 mRNA. We also observed that the protein expression level of CTNNB1 was elevated upon m⁶A removal (Figure S12b and c), which was likely due to the stabilization of the mRNA. We also confirmed that these Shield-1 induced biological effects were reversible after Shiled-1 was removed (Figure S12).

To demonstrate the applicability of this Shield-1 controlled m⁶A erasing platform, we applied it to edit m⁶A on additional RNA transcripts in different cell lines. We used the Shield-1/FKBP*-dCas13b-ALK system to edit m⁶A at the A48 site of CYB5A mRNA in HeLa cells and the A1211 site of H1F0 mRNA in HEK293T cells. We observed that the m⁶A levels at both target sites were reduced by FKBP*-dCas13b-ALK only when cells were treated with Shield-1, which is comparable to the editing efficiency of the non-inducible dCas13b-ALK (Figure S13). Similarly, the reduced m⁶A levels were completely recovered after Shield-1 removal. Taken together, these findings demonstrated that this conditionally stabilized Shield-1/FKBP*-dCas13b-ALK platform can be easily tailored to achieve inducible and reversible site-specific m⁶A editing on different RNAs of interest in different cell types.

Conclusion

In summary, we reported a small molecule-dependent conditionally stabilized m⁶A editing system by integrating the FKBP protein destabilization domain with the dCas13b system for targeted m⁶A erasing on cellular RNAs under the control of Shield-1 molecule. We demonstrated that the expression of dCas13-m⁶A eraser fusion proteins can be precisely controlled by Shield-1 in both dosage- and time-dependent manners. The targeted recruitment of dCas13-m⁶A eraser and the site-specific m⁶A erasing on target RNAs can be achieved under the control of Shield-1. Moreover, we validated that the dCas13-m⁶A eraser fusion protein can be rapidly released from the target RNA and degraded after Shield-1 removal and the reduced m⁶A at the targeted RNA site can be recovered. Importantly, we observed that the release of dCas13-m⁶A eraser fusion protein occurred much faster than the restoration of m⁶A on the target RNA, which provides an ideal opportunity to study the m⁶A function in its native environment without interference from exogenously introduced dCas13b or m⁶A editor components. This strategy opens a new avenue for the targeted manipulation of RNA modifications, which will contribute to the advance of the RNA epigenetics field.