Precise in vivo RNA base editing with a wobble-enhanced circular CLUSTER guide RNA

Abstract

Recruiting the endogenous editing enzyme adenosine deaminase acting on RNA (ADAR) with tailored guide RNAs for adenosine-to-inosine (A-to-I) RNA base editing is promising for safely manipulating genetic information at the RNA level. However, the precision and efficiency of editing are often compromised by bystander off-target editing. Here, we find that in 5'-UAN triplets, which dominate bystander editing, G•U wobble base pairs effectively mitigate off-target events while maintaining high on-target efficiency. This strategy is universally applicable to existing A-to-I RNA base-editing systems and complements other suppression methods such as G•A mismatches and uridine (U) depletion. Combining wobble base pairing with a circularized format of the CLUSTER approach achieves highly precise and efficient editing (up to 87%) of a disease-relevant mutation in the Mecp2 transcript in cell culture. Virus-mediated delivery of the guide RNA alone realizes functional MeCP2 protein restoration in the central nervous system of a murine Rett syndrome model with editing yields of up to 19% and excellent bystander control in vivo.

Fig. 1. Wobble base pairs modulate RNA editing in an orientation-dependent manner.

a, Design of the cis-acting editing reporter and illustration of the applied triplet base-pairing motifs. The orientation of wobble base pairs at nearest neighbor positions or the presence of G or C counter bases modulate Δ-editing at the central A of a triplet in comparison to its fully Watson–Crick base-paired counterpart. All triplet base-pairing motifs with ocher background color were installed at the empty dotted outline within the cis-acting editing construct to generate the results in b,c. b, Suppression of RNA editing in different target triplets using G•U wobble base pairs or G•A mismatches. c, Enhancing RNA editing in different target triplets using U•G wobble base pairs. The Sanger sequence analysis in b and c was performed after transfection of editing reporter plasmids into ADAR1 p110 Flp-In T-REx cells. Data in b and c are shown as the mean ± s.d. of n = 3 biological replicates. For statistical analysis, a Student’s t-test (two-tailed, parametric) was applied.

Fig. 2. G•U wobbles improve efficiency and precision of trans-acting LEAPER guide RNAs for exogenous targets.

a, Schematic of 111-nt-long unstructured linear LEAPER guide RNA. b–d, Editing heat maps of the LEAPER guide RNA-binding sites within the indicated transcripts: AHI1 (b), COL3A1 (c) and BMPR2 (d). The basic design column, LEAPER, lacks G•A mismatches and wobble base pairs. The other guide RNAs contain either G•A mismatches or G•U wobbles at G•U-amenable sites or a combination of both solutions at all bystander sites. In the latter case, G•A mismatches are placed at sites not amenable to G•U wobbles. The triplet context for each listed editing event is given with the target A highlighted in green and all off-target A bases in blue. The position of each site is given relative to the transcript and the target A (±0 position). Editing was performed with plasmid-borne guide RNA and target in HeLa cells (endogenous ADAR). Data are shown as the mean editing percentage ± s.d. of n = 3 (AHI1 and COL3A1) or n = 5 (BMPR2) biological replicates.

Fig. 3. G•U wobble base pairs are broadly applicable to numerous side-directed RNA base-editing systems.

a, Schematic of the λN-ADAR editing system. The 2× boxB motif-containing guide RNA (84 nt) binds its target mRNA (through 49 bp) to recruit the engineered hyperactive λN-ADAR2Q editase. b, Editing heat map of exogenous AHI1 W725>amber mRNA targeted by the λN-ADAR system. c, Schematic of the Cas13b-ADAR system. The DR motif-containing guide RNA (87 nt) binds its target mRNA (through 51 bp) to recruit the engineered PspCas13b-ADAR editase carrying a hyperactive ADAR2 E488Q deaminase domain. d, Editing heat map of exogenous AHI1 W725>amber mRNA targeted by the Cas13b-ADAR system. e, Schematic of a symmetric ASO bound to its target mRNA (through 58 nt). The ASO is end-blocked by 2′-OMe, contains phosphorothioate linkages and recruits endogenous ADARs. f, Editing heat map of the ASO binding site within exogenous PEX1^G843D transcript. In b,d,f, the basic designs (columns 2× boxB, DR guide RNA and ASO) do not contain G•A mismatches, wobble base pairs or 2′-OMe modifications beyond the end-blocks. The other guide RNAs contain additional 2′-OMe modifications, G•A mismatches, G•U wobbles at G•U-amenable sites or a combination of these solutions at all bystander sites. In the case of a combined solution, G•A mismatches or 2′-OMe modifications are placed at sites not amenable to G•U wobbles. The triplet context for each listed editing event is given with the target A highlighted in green and all off-target A bases in blue. The position of each site is given relative to the transcript and the target A (±0 position). Editing was performed in HeLa cells using plasmid-borne guide RNAs (2× boxB and DR guide RNA) and editase (λN-ADAR2Q and Cas13b-ADAR2Q) or ASOs recruiting endogenous ADAR. Data are shown as the mean editing percentage ± s.d. of n = 3 biological replicates.

Fig. 4. Wobble base pairing improves the design of CLUSTER guide RNAs.

a, Results from an in silico search for RS-binding sites within the murine Mecp2 W104>amber ORF when applying the previous filter settings to the GuideRNA-Forge tool. The binding site for the SD is displayed in turquoise and outlined in black. It contains the target A in green and underlined. All potential RS-binding regions ≥ 20 nt length, in which binding sites for RSs can be selected, are displayed in light blue. The previous filter excludes all A bases except for those within a 5′-GAB (B = U, C or G) triplet context. b, Results for the same in silico search with the latest filter that included A bases in a 5′-GAB, 5′-UAB, 5′-BAU, 5′-KAAU and 5′-CAC triplet context, in addition to allowing A bases independent of their sequence context at the edges of a detected binding region. c, Illustration of binding regions in which specific binding sites of RSs of three circular CLUSTER guide RNAs are located within the Mecp2 ORF, generated using the previous filter (V1 and V2) or the latest filter (V3). d, Editing was performed with plasmid-borne guide RNA and murine Mecp2 transcript in HeLa cells (endogenous ADAR). e, Secondary structure prediction of guide RNA V1–V3 generated using the ViennaRNA Package 2.0 (ref. ). The mean free energy (MFE) of the antisense part (SD and CLUSTER of RS) is given in kcal per mol. The dot–bracket ratio (DBR) indicates the number of dots (unpaired bases) per bracket (paired bases) within the dot–bracket annotation of the antisense part. Data are shown as the mean editing percentage ± s.d. of n = 3 biological replicates. For statistical analysis, a Student’s t-test (two-tailed, parametric) was applied.

Fig. 5. Circular CLUSTER versus circular LEAPER Mecp2 guide RNAs.

Editing heat map of guide RNA-binding sites within the murine Mecp2 W104>amber transcript. The triplet context for each listed editing event is given with the target A highlighted in green and all off-target A bases in blue. The position of each site is given relative to the transcript and the target A (±0 position). Editing was performed with plasmid-borne guide RNA and murine Mecp2 transcript in HeLa cells (endogenous ADAR). a, Circular CLUSTER guide RNA design. b, Circular LEAPER guide RNA design. c, Yields achieved using circular unstructured LEAPER guide RNAs containing a 111-nt-long antisense part. The basic design (column circular LEAPER) does not contain bystander solutions. The other guide RNAs either contain G•A mismatches or apply U depletion at all bystander sites, G•U wobbles at G•U-amenable sites or a combination of G•A mismatching or U depletion with G•U wobbles at amenable sites. d, Yields achieved using circular CLUSTER guide RNAs containing a 100-nt-long antisense part split into a targeting sequence (20 nt) and four RSs (each 20 nt). The basic design (column circular CLUSTER) does not contain bystander solutions. The other guide RNAs contain G•U wobbles at positions −49, −27 and +48, as well as no bystander solution, G•A mismatch or depleted U at position −5. Data in a,b are shown as the mean editing percentage ± s.d. of n = 3 biological replicates.

Fig. 6. Transcript repair in a mouse model of Rett syndrome using circular wobble-optimized CLUSTER guide RNAs and endogenous murine Adars.

a, Editing yields in different brain regions, after delivery of the AAV-PHP.eB-encoded guide RNA through retro-orbital injection into Rett mice carrying the Mecp2 W104>amber mutation and quantification by Sanger sequencing 4 weeks later. b, Correlation between the median editing in a and the geometric mean of the RT–qPCR targets in c,g,i,k. Olfactory bulb, Ob; cerebellum, Cb; hippocampus, Hi; cortex, Cx; midbrain, Mb; thalamus, Th; brainstem, Bs. c, Guide RNA expression quantified by RT–qPCR (normalized to the geometric mean of Actb, Rps29 and Rnu6. d, As in b but correlating a,c. e, Absolute quantification of AAV episomes per cell by standard curve qPCR (normalized to Actb). f, As in b but correlating a,e. g, As in c but for Adar1. h, As in b but correlating a,g. i, As in c but for Adar1 p150. j, As in b but correlating a,i. k, As in c but for Adar2. l, As in b but correlating a,k. m, Bystander off-target events at the guide RNA-binding sites and the TS. n, All amplicon reads with on-target editing binned according to their number of bystander events (based on Supplementary Fig. 14). o, Global RNA editing at 2,533 endogenous sites (coverage ≥ 50 reads, REDIportal). p, Thalamus sections stained for MeCP2 and DAPI (nuclei). For a, data are shown as the median editing percentage ± 95% CI determined in n = 3 mice (nontargeting virus) and n = 5 mice (targeting virus). For statistical analysis, a Mann–Whitney U-test (two-tailed, nonparametric) was applied. For c,e,g,i,k,m,n, data are shown as the median fold change ± 95% CI or median number of copies per cell ± 95% CI determined in n = 2 mice (three technical replicates each). The NGS analysis in o is based on results from n = 1 (nontargeting virus) or n = 2 (targeting virus) mice. The results in p are derived from n = 1 mouse per group. For b,d,f,h,j,l, the R² values were determined by simple linear regression.

Extended Data Fig. 1. The enhancing and suppressing effects of G·U and U·G wobble base pairs on RNA editing apply to all catalytically active ADAR isoforms.

(a) Schematic of the utilized cis-acting editing reporter construct. (b) Editing yields when applying G·A mismatches or G·U wobbles for editing suppression at the 5’-UAG triplet in either ADAR1 p110-, ADAR p150- or ADAR2-Flp-In T-REx cells. The exact position of each mismatch or wobble is indicated below each bar. The target adenosine is highlighted in green, bold and underlined. Bases that induce wobbles or mismatches are highlighted in dark blue and bold. The 5’ and 3’ orientation of the triplet and the used ADAR isoform are shown above the grouped bars. (c) As b) but regarding editing yields when applying enhancing U·G wobbles. Data for b) and c) are shown as the mean ± s.d. of N = 3 biological replicates per bar. For statistical analysis, a student’s t-test (two-tailed, parametric) was applied.

Extended Data Fig. 2. G·U wobbles improve efficiency and precision of trans-acting LEAPER guide RNAs for endogenous targets.

(a-b) Editing heat-map of the LEAPER guide RNA binding site within the indicated transcripts. The basic design column, LEAPER, lacks G·A mismatches and wobble base pairs. The other guide RNAs contain either G·A mismatches or G·U wobbles at G·U amenable sites or a combination of both solutions at all bystander sites. In the latter case, G·A mismatches are placed at sites not amenable to G·U wobbles. The triplet context for each listed editing event is given with the target adenosine highlighted in green and all off-target adenosines in blue. The position of each site is given relative to the transcript and the target adenosine (±0 position). The editing events detected at position 2546 ( + 47) and 2548 ( + 49) of the NUP43 transcript were natural events and were thus skipped in regard to bystander suppression. Editing was performed with plasmid-borne guide RNA and endogenous target transcripts in HEK293FT cells (endogenous ADAR). Data are shown as the mean editing percentage ± s.d. of N = 3 biological replicates.

Extended Data Fig. 3. G·U wobbles significantly outperform previous solutions in bystander suppression at nearest neighbor (NN) sites.

(a) Design of the cis-acting RNA editing reporter constructs used in e) and f). (b) Design of the linear trans-acting CLUSTER guide RNAs used in g). (c) Illustration of the base-pairing motifs applied to the constructs in a) (ocher) and b) (grey) that were used to suppress bystander editing at the 5´-U[AAG] sequence motif in e) and g). (d) As c) but for the 5´-[UAA]U sequence motif and the results in f). (e) Suppression of bystander editing at the 5´-NN A and concurrent effect on the target A editing yield in a 5’-U[AAG] sequence motif. (f) Similar to e) but targeting the A in a 5’-[UAA]U sequence motif and suppressing bystander editing at a 3´-NN A. Editing in e) and f) was performed in ADAR1 p110 Flp-In T-REx cells using plasmid-borne reporters as displayed in a). (g) Suppression of bystander editing at a 5´-NN A when editing a 5’-U[AAG] site (K984) of the human BMPR2 transcript using trans-acting CLUSTER guide RNAs. (h) Meta-analysis of bystander suppression at 5´- and 3´-NN sites over three different target transcripts (AHI1, BMPR2 and COL3A1), three different A-rich target triplets flanked either 5’ and/or 3’ by uridines (5’-[UAA]U, 5’-U[AAG], 5’-U[AAA]U), and applying two trans-acting site-directed RNA editing systems (CLUSTER/endogenous ADAR, boxB/λN-ADAR2Q). The full data set of the meta-analysis is shown in Supplementary Fig. S5. Editing in g) and h) was performed in HeLa cells using plasmid-borne editase (λN-ADAR2Q) and/or guide RNA (2x boxB, CLUSTER). Data in e) – g) are shown as the mean editing percentage ± s.d. of N = 3 biological replicates. Data in h) are shown as the mean editing percentage ± s.d. of three 5’-NN off-target sites, seven on-target sites, and five 3’-NN off-target sites containing N = 3 biological replicates per site. The center line represents the median, the plus-sign the mean. The whiskers extend to the minimum and maximum values. The box limits the 25^th and 75^th percentile. For statistical analysis, a student’s t-test (two-tailed, parametric) was applied.

Extended Data Fig. 4. Circularization of a split-R/G CLUSTER guide RNA using the Tornado expression system.

The guide RNA is expressed from either a polymerase III (for example U6) or a polymerase II (for example CAG) promoter with a corresponding terminator. 5’ and 3’ ribozymes auto-cleave the primary transcript resulting in a 5’-OH and a 2’,3’-cyclic phosphate overhang, respectively, which are subsequently ligated by the endogenous RTCB ligase. In the final guide RNA, the ligation stem (orange) forms a nearly continuous duplex with the split-R/G motif (light green), only interrupted by a five-nucleotide bulge. The circular guide RNA binds to its target transcript. Strategic placement of wobble base pairs within the recruitment sequences (RS) and/or the specificity domain (dark green) suppresses off-target editing, while a C-A mismatch within the specificity domain (SD) promotes on-target A-to-I RNA editing.

Extended Data Fig. 5. Extended data set for editing Mecp2 W104amber with LEAPER and CLUSTER guide RNAs.

Panels (a) to (d) give schematic representations of the used linear and circular LEAPER and CLUSTER guide RNAs and their binding to the target mRNA. LEAPER guide RNAs were 111 nt long, centered with a C-A mismatch on the target adenosine. CLUSTER guide RNAs combined an ADAR-recruiting domain (R/G motif), a 20 nt long specificity domain containing the C-A mismatch with the target adenosine at position 8, and a cluster of three recruitment sequences of 20 nt length each. The circular CLUSTER guide RNA combines the ligation stem for circularization with the ADAR-recruiting motif into a split-R/G motif. (e) Editing heat-map using linear and circular LEAPER guide RNAs as seen in a) and b). (f-h) Editing heat map using linear and circular CLUSTER guide RNAs as seen in c) and d). Notably, the linear R/G version 21 guide RNA achieved already highly precise editing with an impressive 38 ± 7% on-target editing yield, clearly better than the best circular LEAPER guide RNA design (25 ± 2% on-target yield) with similar precision. A circular CLUSTER guide RNA with six recruitment sequences, but lacking the ADAR recruitment motif, achieved 54 ± 7% on-target editing with very high precision after applying the GU wobble strategy. However, adding a split-R/G recruiting motif achieved even better on-target editing yields of 75 ± 1% to 82 ± 1%, even though only 4 recruitment sequences were applied. GU wobbling was required to eliminate bystander editing at the binding sites of recruitment sequences. The triplet context for each listed editing event is given with the edited adenosine highlighted in bold green. The position of each bystander site (bold blue) is given relative to the target adenosine (±0 position). Editing was performed with plasmid-borne guide RNA and murine Mecp2 in HeLa cells (endogenous ADAR). Data are shown as the mean editing percentage ± s.d. of N = 3 biological replicates.

Extended Data Fig. 6. Verifying guide RNA circularization via quantitative RT-qPCR.

The circular state of the split-R/G CLUSTER guide RNA (V27.2.4, pTS2108) targeting mMecp2 W104Amber was verified by using quantitative RT-PCR on in vitro samples from transfected HeLa cells. As reference a linear version of the same guide RNA was used. The dataset was normalized to human RNU6 snRNA. The amplicon of the outward primer pair in a) is 187 bp long and requires reverse transcription and PCR through a strong hairpin structure. It can thus not be directly compared to the inward primer pair in d) which is smaller (75 bp) and does not contain inhibitory secondary structures. (a) Binding sites of the outward primer pair, which can only amplify circular guide RNAs. (b) Fold change of linear and circular split-R/G CLUSTER guide RNAs (V27.2.4, pTS2108, pTS2228) relative to endogenous human RNU6 as detected by using the outward primer pair. The 2984-fold difference between linear and circular shows that the endogenous circularization of the guide RNAs though the RTCB ligase (Extended Data Fig. 4) was successful and that the outward primer pair can identify circular guide RNAs with high confidence (low type-I error rate). (c) Type-I error rate calculated from b) amounts to only 0.03%, thus 99.97% of circular guide RNAs are correctly identified. (d) Binding sites of the inward primer pair, which can amplify both linear and circular guide RNAs. (e) Fold change of linear and circular split-R/G CLUSTER guide RNAs (V27.2.4, pTS2108) relative to endogenous human RNU6 snRNA as detected by using the inward primer pair. The 235-fold difference between linear and circular shows the high level of total guide RNA enrichment achieved after circularization. (f) Percentage of successfully circularized guide RNAs calculated from e). Under the assumption that the 235-fold increase detected for circular guide RNAs compared to linear ones solely depends on circularization, 99.6% of all guide RNAs were successfully processed by the ribozymes of The tornado expression system and then ligated by RTCB. Data in b) and e) are shown as the mean editing percentage ± s.d. of N = 3 biological replicates.

Extended Data Fig. 7. Amplicon sequencing results before background correction.

(a) Editing heat map of the indicated brain tissue of Rett syndrome mice treated with either the scrambled circular CLUSTER guide RNA virus (nontargeting virus) or (b) the circular CLUSTER guide RNA virus (targeting virus). The triplet context for the target-site row is boxed, and the target adenosine underlined and green. All other triplet contexts contain an off-target adenosine in blue. The position of each site is given relative to the transcript and the target adenosine (±0 position). The dataset covers the complete guide RNA binding region including all binding sites (BS) and the target sequence (TS) containing the target adenosine. Data are shown as the mean editing percentage of N = 2 mice.

Extended Data Fig. 8. Analysis of global off-target events by RNA-seq.

(a) Rett mice injected with wobble-optimized circular CLUSTER guide RNA (targeting virus) versus scrambled guide RNA (nontargeting virus). (a) (top) A-to-G RNA editing index in coding sequences. An index of 1 means 1% of RNA nucleotides mapped to genomic adenosines are guanosines. The index was stable (no change in global editing activity due to the targeting guide RNA). (a) (bottom) Global RNA editing at 2,528 endogenous sites (coverage ≥ 50, REDIportal). With only 3 sites showing > 25% Δ-editing (beyond blue dotted lines), no clear off-target editing was detectable in comparison to panel b. (b) Reanalyzed RNA-seq data (previous study). With respect to the tissue selected in panel c), data from the brainstems of two untreated Rett mice are compared. (b) (top) Both untreated mice show a stable editing index. (b) (bottom) Global RNA editing at 2,301 endogenous sites (coverage ≥ 50). These sites are also displayed in panel a) (bottom) and c) (bottom). The 7 sites showing > 25% Δ-editing are considered normal mouse to mouse and technical variability. (c) Reanalyzed RNA-seq data (previous study). Rett mice injected with boxB/λN-ADAR2 virus were compared to untreated ones. The ADAR2 catalytic domain in the virus is from the native ADAR2 protein. (c) (top) The A-to-G index increase by 0.007 in the coding space reflects a substantial and somewhat concerning increase in global editing, probably due to the overexpressed ADAR2 catalytic subunit. (c) (bottom) Global RNA editing at 2,533 endogenous sites that include the 2,528 sites shown in panel a) (coverage ≥ 50 reads). 25 sites showing > 25% Δ-editing after injection of the λN-ADAR2 targeting virus. This is in clear contrast to the very high precision seen for CLUSTER guide RNAs recruiting endogenous ADARs in a) (bottom) or the untreated mice in b) (bottom). Importantly, the potential off-target sites included two ADAR2-specific, highly evolutionary conserved sites in GRM4 (glutamate transmission) and NOVA1 (splicing). Data in a), b) and c) are based on results from N = 1-2 mice per group. Thalamus and brainstem were selected for showing the highest editing yields in their respective study.

Extended Data Fig. 9. Immunohistochemical evidence supporting repair of MeCP2 protein and function after treatment of Rett mice with the circular CLUSTER guide RNA.

(a) Confocal images of sections of thalamus stained for MeCP2 protein and DAPI to indicate nuclei. MeCP2 puncta staining, co-localizing with DAPI, only seen in middle and lower panels indicating restoration of endogenous protein expression. Nuclei (DAPI) were pseudo-colored to highlight colocalization puncta. (b) Each bar represents the total percentage of MeCP2 protein foci positive cells relative to DAPI positive cells in each mouse group. Each datapoint represents one microscopic field of view. Non-injected Rett mice show background levels of immunofluorescence. The distribution of positive nuclei in non-injected wild-type mice reflects the cellular heterogeneity within the brain. Analysis was performed by confocal imaging in three sections from each brain region indicated in the figure. We selected the four brain regions with the highest editing yields. Data is shown as the median percentage of cells showing MeCP2 protein foci ± c.i. (confidence interval 95%) from N = 2 mice per bar, as counted in 3 (non-injected Rett mice) or 6 (injected Rett mice) fields of view per animal. (c) High correlation between the median editing yields (amplicon sequencing) and the percentage of cells with MeCP2 protein foci in the four brain regions in a). The abbreviations stand for Cortex (Cx), midbrain (Mb), thalamus (Th), and brainstem (Bs). Data is shown as one datapoint per brain region. Each datapoint represents the median result of the respective tissue from N = 2 mice injected with targeting virus per readout. For technical reasons the correlated editing yields and protein foci percentages are from different animals.

Extended Data Fig. 10. Proposed mechanisms for the suppressive effect of G·U wobble base pairs on RNA editing when placed at nearest neighbor sites.

(a) Superposition of a 5´ G U wobble base pair (blue) on top of a Watson-Crick AU base pair (black) arranged according to panel b) and c). The shifted Guanine (blue) results in the projection of an exocyclic amino group into the minor groove, which is a distinctive feature of the GU wobble base pair. (b) hADAR2d/Bdf2 crystal structure indicating the close proximity of the 5’ nearest neighbor AU base pair and Glycine 489. The position of the observed base pair (black, bold) and the target adenosine (green, bold, underlined) is highlighted in the Bdf2 duplex sequence. (c) A 5´ G·U wobble base pair fitted into the hADAR2d/Bdf2 crystal structure at the 5’ nearest neighbor position predicts a steric clash between the exocyclic amino group of the Guanine and the Glycine 489. The clash would require extensive accommodation of the complex and thus cause structural perturbations, that might negatively affect the deamination rate. The position of the observed wobble base pair (blue, bold) and the target adenosine (green, bold, underlined) is highlighted in the Bdf2 duplex sequence. (d) As a) but arranged according to panel e) and f). (e) As b) but showing the 3’ nearest neighbor AU base pairs proximity to Serine 486. The AU base pair was fitted into the position of the original CG base pair of the hADAR2d/Bdf2 crystal structure. (f) As c) but showing the Guanosine of the 3’ G·U wobble base pairs clashing with Serine 486. The 3´ G·U wobble base pair was fitted into the position of the original CG base pair of the hADAR2d/Bdf2 crystal structure.