Striated muscle-specific base editing enables correction of mutations causing dilated cardiomyopathy

Abstract

Dilated cardiomyopathy is the second most common cause for heart failure with no cure except a high-risk heart transplantation. Approximately 30% of patients harbor heritable mutations which are amenable to CRISPR-based gene therapy. However, challenges related to delivery of the editing complex and off-target concerns hamper the broad applicability of CRISPR agents in the heart. We employ a combination of the viral vector AAVMYO with superior targeting specificity of heart muscle tissue and CRISPR base editors to repair patient mutations in the cardiac splice factor Rbm20, which cause aggressive dilated cardiomyopathy. Using optimized conditions, we repair >70% of cardiomyocytes in two Rbm20 knock-in mouse models that we have generated to serve as an in vivo platform of our editing strategy. Treatment of juvenile mice restores the localization defect of RBM20 in 75% of cells and splicing of RBM20 targets including TTN. Three months after injection, cardiac dilation and ejection fraction reach wild-type levels. Single-nuclei RNA sequencing uncovers restoration of the transcriptional profile across all major cardiac cell types and whole-genome sequencing reveals no evidence for aberrant off-target editing. Our study highlights the potential of base editors combined with AAVMYO to achieve gene repair for treatment of hereditary cardiac diseases.

Fig. 1. Molecular and physiological characterization of P635L and R636Q mouse lines.

a Confocal images of isolated adult murine cardiomyocytes. Scale bar: 20 μm. ACTN1 was used as cardiomyocyte marker. b, c RBM20 granule size (b) and amount (c) in adult mouse cardiomyocytes. N = 21 (WT), 16 (P635L HET), 28 (P635L HOM), 16 (R636Q HET) and 39 (R636Q HOM) images with 1–4 cells each obtained from three mice per genotype. Boxplots depict the median with the box including the 25–75th percentile and the whiskers ranging from the smallest to the largest value. d Number of DEGs (P_adjust < 0.05) in bulk RNA-seq of Rbm20 mutant mice compared to WT. N = 5 mice per genotype. e GO analysis (biological function) of DEGs overlapping for both P635L and R636Q HOM mice with a stringent cut-off of P_adjust < 1e⁻¹⁰ to reduce the number of DEGs for display in Supplementary Fig. 2. f Number of differentially spliced events compared to WT detected and categorized by rMATS: alternative 5‘ or 3‘ splice site (A5SS or A3SS), mutually exclusive exons (MXE), retained intron (RI), skipped exon (SE). g Averaged ΔPSI ( = percent spliced-in) values relative to WT of significant differentially spliced events (P_adjust < 0.01, ΔPSI > 0.1) overlapping in both HOM Rbm20 mutant mice. Multiple splice events per gene are depicted if they match the selection cut-off. Genes in red were validated by RT-PCR or qPCR. Grey squares indicate that the splice event was not detected by rMATS. h RT-PCR of RBM20 target genes Ttn, Ryr2, Ldb3 and the housekeeping gene Gapdh. i Kaplan–Meier survival curve of mutant mice monitored for 120 days. P value obtained by Log-rank test between each mutant and WT indicated next to the curves. Percentage of survival indicated for HOM mice. j Percentage of ejection fraction determined by narcosis echocardiography of mutant mice. N = 13 (WT), 5 (P635L HET), 6 (P635L HOM), 11 (R636Q HET) and 11 (R636Q HOM) mice. P values in (b, c, j) obtained from one-way ANOVA with Tukey’s multiple comparison test: ****P < 0.0001, ***P < 0.001, **P < 0.01, n.s. = not significant. All data were obtained in 16-week-old mice except in (j) where data of 24-week-old mice is shown. Error bars depict the standard error of the mean (SEM) in all panels.

Fig. 2. Base editing of RBM20 in human iPSC-CMs and in mice.

a–c Transient expression of base editor and gRNA in human iPSCs and iPSC-CMs. Experimental outline in (a), editing efficiency of P633L in (b) and R634Q in (c). “CP” labels the circular permuted base editor CP-1041. Purple line indicates average repair efficiency in iPSC-CMs. d Generation of stable base editor expression in R634Q iPSCs with a repair efficiency of 34.26 ± 2.36% as determined by amplicon-seq. N = 3 independent differentiations. e Expression of spliced and unspliced isoforms of TTN and IMMT in parental, R634Q and edited R634Q iPSC-CMs differentiated for 15 and 32 days. Significant changes compared to R634Q indicated when present and analyzed by unpaired, two-tailed t tests. *P < 0.05, **P < 0.01, ****P < 0.0001. f Experimental outline of AAVMYO-mediated base editing in mice. g Percentage of editing of P635L HOM mice injected with AAVMYO carrying different gRNA-base editor combinations or PBS as empty control. For NRCH-gRNA1, AAV9 was also used as vector. Significance was assessed using unpaired, two-tailed t tests ***P < 0.001, **P < 0.01, *P < 0.05. Sequence shows location of the on-target edit in blue and the two bystander edits in red. Numbers depict the position of the nucleotides within the targeting gRNA (gRNA2 was used as reference) with the PAM sequence in position 21–23. h Allele frequency of repaired DNA in the muscle tissues heart, diaphragm and quadriceps femoris (quadriceps f.), as well as the liver, plotted for ten mice with the highest editing events in (g). i Percentage of editing of Rbm20 mRNA in mice treated with AAVMYO-SpRY for 6 or 12 weeks. Editing was assessed by amplicon-seq of cDNA isolated from the whole heart. Most base editors contain the deaminase variant Abemax except when indicated by “8e”, which are base editors with the Abe8e version. Percentage “repaired” in (b, c, g–i) is defined by NGS reads from amplicon-seq with only the wild-type sequence. The number of biological replicates i.e., independent differentiations in (b, c, e) or mice in (g, i) is indicated in brackets above the bars. Error bars depict the SEM in all panels.

Fig. 3. Phenotypic characterization of mice after AAVMYO-ABE treatment.

a Allele frequency of repaired Rbm20 mRNA in mice treated with AAVMYO-ABE determined by RNA-seq. N = 3 (R636Q), 4 (P635L) and 8 (WT) mice per condition. b, c RBM20 staining in whole heart tissue sections in WT and Rbm20 mutant mice treated with PBS or AAVMYO-ABE. Representative images in (b) and quantification of nuclear and cytoplasmic RBM20 localization in (c). Scale bar: 20 μm. Arrows highlight nuclear restored RBM20 (magenta) and cytoplasmic RBM20 (white) in base-edited mice. Manual quantification of >200 nuclei in 2 mice per condition. d Isoform expression of RBM20 target genes Ttn, Ryr2, Ldb3 and the housekeeping gene Gapdh determined by RT-PCR. N = 2–4 mice per condition. e, f Vertical agarose gel (e) and quantification (f) of titin protein isoforms in WT and Rbm20 mutant mice treated with PBS or AAVMYO-ABE. Based on gel images in Supplementary Fig. 4e. N = 3 mice per condition except for WT and P635L/SpRY where four mice were analyzed. g RNA-seq data showing changes of the ΔPSI values relative to WT in P635L, or R636Q HOM mice injected with saline or base editor. See 'Methods' section (bulk RNA sequencing and analysis) for definition of the three categories. Rescue splice events are labeled in red, all Ttn splice events in blue. N = number of splice events per category. R636Q was sequenced deeper compared to P635L explaining the difference in number of DSGs detected. N = 3 mice per condition except fornWT (4 mice) and P635L SpRY (5 mice). h–j Percentage of ejection fraction (h), LVID (i) and cardiac volume (j) determined by narcosis echocardiography of mutant mice treated with PBS or AAVMYO-ABE. N = 5 mice per condition. Same WT cohort used as in Supplementary Fig. 1m–o (16-week time point). P values obtained from one-way ANOVA with Tukey’s multiple comparison test: ***P < 0.001, **P < 0.01, *P < 0.05, n.s. = not significant. All data were obtained 12 weeks after AAVMYO-ABE injection. Only P635L or R636Q HOM mice were treated. Error bars depict the SEM in all panels.

Fig. 4. Cell type-specific profiling of cells after base editing by snRNA-seq.

a UMAP projection integrating all datasets and annotated based on their gene expression profile. b Expression of known marker genes defining the main cell types. c UMAP projection of ventricular cardiomyocytes from WT, P635L HOM, and base-edited mice. d Histograms depicting the distribution of pairwise Euclidean distances of ventricular cardiomyocytes from P635L HOM and base-edited mice relative to WT using the two largest principal components (PC). e UMAP projection showing the activity score (see 'Methods' section snRNA-seq analysis) of cardiomyocytes using a subset of genes that were up- or downregulated in P635L HOM relative to WT. Maximum 15 significantly up- or downregulated were used. DEGs are listed in Supplementary Data 3. f Threshold of activity score values based on (e) relative to percentage of cells above the threshold for genes upregulated (upper panel) or downregulated (lower panel) in P635L HOM cardiomyocytes relative to WT. g Percentage of cells above the critical threshold for genes upregulated or downregulated in P635L HOM cells relative to WT. VCMs ventricular cardiomyocytes, ACMs atrial cardiomyocytes, SMCs smooth muscle cells. Data were generated by snRNA-seq of isolated nuclei from two mice per condition.

Fig. 5. WGS of mouse tissue before and after AAVMYO-ABE treatment.

a Mean number of tissue-specific and variants detected in all tissues (common). b Mean relative distribution of distinct nucleotide conversions for tissue-specific and common SNVs. Tissue overlap represent variants that were common to all three tissues. c Allele frequency of tissue-specific T > C/A > G variants. N = 33 (heart), 32 (liver), 146 (tail). d Mean number of mismatches to the gRNA and PAM sequence in the area of ± 30 bases around the variant start site. Error bars depict the SEM in all panels. N = 3 mice in (a, b, d).