Allele-Specific Suppression of Variant MHC With High-Precision RNA Nuclease CRISPR-Cas13d Prevents Hypertrophic Cardiomyopathy

Abstract

BACKGROUND:

Familial hypertrophic cardiomyopathy has severe clinical complications of heart failure, arrhythmia, and sudden cardiac death. Heterozygous single nucleotide variants (SNVs) of sarcomere genes such as MYH7 are the leading cause of this type of disease. CRISPR-Cas13 (clustered regularly interspaced short palindromic repeats and their associated protein 13) is an emerging gene therapy approach for treating genetic disorders, but its therapeutic potential in genetic cardiomyopathy remains unexplored.

METHODS:

We developed a sensitive allelic point mutation reporter system to screen the mutagenic variants of Cas13d. On the basis of Cas13d homology structure, we rationally designed a series of Cas13d variants and obtained a high-precision Cas13d variant (hpCas13d) that specifically cleaves the MYH7 variant RNAs containing 1 allelic SNV. We validated the high precision and low collateral cleavage activity of hpCas13d through various in vitro assays. We generated 2 HCM mouse models bearing distinct MYH7 SNVs and used adenovirus-associated virus serotype 9 to deliver hpCas13d specifically to the cardiomyocytes. We performed a large-scale library screening to assess the potency of hpCas13d in resolving 45 human MYH7 allelic pathogenic SNVs.

RESULTS:

Wild-type Cas13d cannot distinguish and specifically cleave the heterozygous MYH7 allele with SNV. hpCas13d, with 3 amino acid substitutions, had minimized collateral RNase activity and was able to resolve various human MYH7 pathological sequence variations that cause hypertrophic cardiomyopathy. In vivo application of hpCas13d to 2 hypertrophic cardiomyopathy models caused by distinct human MYH7 analogous sequence variations specifically suppressed the altered allele and prevented cardiac hypertrophy.

CONCLUSIONS:

Our study unveils the great potential of CRISPR-Cas nucleases with high precision in treating inheritable cardiomyopathy and opens a new avenue for therapeutic management of inherited cardiac diseases.

Clinical Perspective.

What Is New?

• We developed a high-precision, high-fidelity Cas13d variant, hpCas13d, with single nucleotide variant resolution and low collateral cleavage activity.

• RNA editing with hpCas13d is an effective and safe approach for preventing and treating hypertrophic cardiomyopathy (HCM) in 2 new HCM mouse models.

• hpCas13d could resolve various human MYH7 pathogenic variants, including single nucleotide variants not accessible to base editors.

What Are the Clinical Implications?

• The Myh6 p.R872H sequence variation leads to mild HCM in mice; compound heterozygous variants with Myh6 p.R872H and p.R404Q result in early-onset and severe HCM.

• In vivo application of hpCas13d precisely cleaves the pathogenic Myh6 RNAs and prevents cardiac hypertrophy without apparent adverse effects, demonstrating it is a safe and effective approach for treating genetic disorders.

• Suppression of 1 variant rather than 2 is sufficient to prevent HCM with 2 compound MYH7 sequence variations.

Editorial, see p 299

With a prevalence of ≈1 in 500 to 1 in 200, hypertrophic cardiomyopathy (HCM) is the leading cause of sudden death affecting young people.^1,2 The clinical phenotype of HCM is manifested by left ventricular hypertrophy, myocardial hypercontractility, myofibrillar disarray, and fibrosis,² along with severe clinical complications including heart failure, arrhythmia, and sudden cardiac death. Although >1000 HCM-causing variants have been found in various sarcomere genes, more than one-third of cases are caused by missense variants in β-cardiac myosin heavy chain (MHC; MYH7),³ resulting in amino acid substitutions in myosin that perturb the contractile function of sarcomere. Among them, 78% of MYH7 variants are heterozygous single nucleotide variants (SNVs).⁴

The majority of MYH7 pathogenic variants are located on the myosin head, which increases actomyosin interactions leading to hypercontractility and disease progression.⁵ Sequence variations in the head region, such as MYH7 p.R403Q, a well-studied point mutation, correlate with a high incidence of cardiac events and a more malignant prognosis.⁶ Pathogenic variants located in the rod domain, such as c.2609G>A (p.R870H), have also been reported but lack investigation. p.R870H is identified in 0.26% of patients with HCM,^4,7 causing a mild form of HCM with ≈50% disease penetrance in adults.⁸ MYH7 p.R870H has also been reported in childhood-onset HCM, which presents with much more severe hypertrophic phenotypes than adult-onset HCM.⁹ Compound heterozygous sequence variations with other MYH7 variants might account for the phenotypic heterogeneity of MYH7 p.R870H, but remain unexplored in both human and animal genetic studies.

Gene therapy recently has emerged as a compelling approach to treat MYH7-causing HCM in animal models. Selective suppression of an MYH7 missense sequence variation allele using RNA interference had a substantial effect on HCM prevention.¹⁰ However, achieving allele-specific knockdown requires a single nucleotide polymorphism (SNP) nearby the pathogenic SNV, limiting the usage of this strategy. More recently, correcting the genomic SNV using a base editor also effectively prevented MYH7-causing HCM in transgenic mice.¹¹ However, because of the constraint of the protospacer adjacent motif sequence and limited editing window, many SNVs cannot be effectively targeted. Moreover, the well-tested base editors are adenine base editors and cytosine base editors that only allow A:T to G:C and C:G to T:A transitions.¹² Therefore, nearly 30% of SNVs requiring other types of transitions are unable to be edited. In addition, the inherent bystander toxicity and genomic off-target risk further limit their applications for in vivo gene therapy.

The type VI RNA endonuclease CRISPR-Cas13 (clustered regularly interspaced short palindromic repeats and their associated protein 13) has been identified recently as a valuable programmable sequence-specific RNA endonuclease.^13,14 Thus far, 4 families of Cas13 enzymes (Cas13a through Cas13d) and their corresponding CRISPR RNA (crRNA) arrays have been described.¹⁵ Because of its high RNA cleavage activity and apparently greater safety compared with DNA editors, CRISPR-Cas13 has been used recently to develop potential disease therapies.^16–18 Unlike DNA CRISPR nucleases, many Cas13 nucleases do not have the protospacer adjacent motif or protospacer flanking site and therefore should be able to target any RNA sequence. Cas13d, the most compact Cas13 (≈930 amino acids), has high RNA nuclease activity, especially for Ruminococcus flavefaciens XPD3002 Cas13d (RfxCas13d), a recently described variant.¹⁹ However, it remains unclear whether Cas13d could allelically cleave MYH7 pathogenic transcripts, which differ from wild-type (WT) transcripts by a single nucleotide. This capability is necessary for Cas13d to treat MYH7-causing HCM.

In this study, to explore the possibility of using RNA nuclease Cas13d to treat cardiomyopathy with heterozygous MYH7 SNVs, we first established an SNV allelic discrimination reporter system. By screening guide RNAs (gRNAs) targeting human MYH7 p.R870H and p.R403Q analogous variants (mouse Myh6 p.R872H and p.R404Q), we demonstrated that WT RfxCas13d lacks the capability to discriminate variant from WT Myh6 RNAs. Rational evolution of Cas13d yielded a high-precision Cas13d variant, hpCas13d, that precisely and efficiently cleaves variant Myh6 while preserving the WT Myh6. In vivo application of hpCas13d to 2 distinct genetic HCM mouse models (Myh6^R872H/+ and Myh6^R872H/R404Q) showed that it repressed Myh6 RNA missense variants accurately and selectively and prevented the hypertrophic phenotype. Delivery of hpCas13d to the developed hypertrophic hearts exhibited rescue effects as well. A large-scale screening demonstrated that hpCas13d discriminated various MYH7 pathogenic variants, including SNVs unavailable to base editors, from WT MYH7. Our study demonstrates the great potential of RNA nucleases in treating genetic cardiac diseases and providing medications with translational prospects for patients with HCM.

METHODS

Animal Experiments

All animal protocols and procedures were approved by the institutional animal care and use committee of Shanghai Jiao Tong University. ARRIVE guidelines (Animal Research: Reporting of In Vivo Experiments) were used in our animal study.²⁰ Animals were randomly assigned to receive hpCas13d. During hpCas13d administration, investigators were blinded to drug information. Operators were blinded to animal genotypes and treatments during echocardiography and electrocardiography analyses.

Statistics

Statistical details can be found in the figure legends. The effect size was estimated from previous publications, and G Power 3.1 was used to calculate the sample size a priori.²¹ All experiments required a minimum of 3 biological replicates (independent experiments or mice), with additional technical replicates as appropriate. All data are presented as mean±SD. The Shapiro-Wilk test and Kolmogorov-Smirnov test were used to assess data normality, and P≥0.05 was considered as a normal distribution. Comparisons between 2 groups were analyzed by a 2-tailed parametric t test if the data were normally distributed; otherwise, the Mann-Whitney U test was used. Differences between multiple groups were performed using 1-way ANOVA followed by Tukey post hoc tests for normally distributed data; the nonparametric Kruskal-Wallis test with Dunn multiple-comparisons test was used for non-normally distributed data. Two-way ANOVA with post hoc Sidak multiple comparison tests was performed for data with 2 independent variables. Statistical analysis was performed using GraphPad Prism 9. A difference was considered significant if P<0.05.

Data Availability

All raw and processed RNA sequencing data sets have been deposited to the Gene Expression Omnibus database with a specified accession number (GSE260500). The materials that support the findings of this study are available from the corresponding author on request.

RESULTS

WT Cas13d Was Insufficient to Resolve Pathogenic HCM SNVs

MYH7 pathogenic variants have a dominant negative effect on the myosin contractile function. Allele-specific silencing of MYH7 variant transcript by CRISPR-Cas13 nuclease holds the possibility to alleviate HCM. To corroborate this, we adopted a Cas13d variant, RfxCas13d, which has an inherently small size and is conducive for in vivo gene therapy. We further optimized the activity of RfxCas13d by replacing A at the 5′ terminus of the crRNA with U to shorten its direct repeat (DR) stem, as reported²² (Figure S1A). We evaluated this DR variant in an MYH7-dEGFP (EGFP-PEST, fast-degradable EGFP) reporter system in HEK293T cells. Compared with WT DR, the variant DR exhibited substantially greater repression (1.23- to 2.06-fold relative to WT DR) with 6 different crRNAs (Figure S1B and S1C). We also examined the variant DR on 3 endogenously expressed genes and found similar improvements, up to 3.2-fold greater than WT DR for PLK1 (Figure S1C). Long crRNA spacers could increase crRNA–target RNA mismatch tolerance. To minimize this possibility, we tested spacers from 20 to 27 nt in length and found that spacers ≥23 nt were able to maintain nuclease activity (Figure S1D and S1E). Therefore, 23-nt spacers were used for subsequent experiments.

Next, to probe the allelic specificity of RfxCas13d, we developed a dual-fluorescence single nucleotide discrimination assay for Myh6 variants, analogous to human MYH7 p.R870H and p.R403Q. WT mouse Myh6 was fused to dEGFP, and variant Myh6 with a substitution of c.2615 G>A (p.R872H) or c.1211 G>A (p.R404Q) was fused to mCherry-PEST (dmCherry) and cotransfected with RfxCas13d and gRNAs designed to target these variant sequences (Figure 1A). WT Myh6-dEGFP and variant Myh6-dmCherry had comparable transfection efficiency and expression levels (Figure S2A). With this assay, we tested the capacity of RfxCas13d to discriminate between the WT and variant sequences by tiling each variant with 12 possible gRNAs that cover the SNV (Figure S2B). For R872H, 7 of 12 gRNAs comparably depleted WT and R872H Myh6 variant transcripts (P>0.05), and 5 of 12 gRNAs depleted variant Myh6 more efficiently, but the discrimination was inadequate (3.6%–15.5%; Figure S2C and S2D). For R404Q, 8 of 12 gRNAs comparably cleaved the WT and variant Myh6, and 4/12 gRNA cleaved WT Myh6 even more efficiently than the variant (Figure S2E through S2H). These results demonstrate that WT RfxCas13d lacks the ability to discriminate between MYH7 alleles that differ by a single nucleotide.

Figure 1.

Structure-directed evolution of high-precision Cas13d. A, Schematic diagram of the dual-fluorescence selection system. Wild-type mouse Myh6 was linked to fast turnover EGFP-PEST, and variant Myh6 (Myh6^R872H or Myh6^R404Q) was linked to fast turnover mCherry-PEST. RfxCas13d variants, wild-type Myh6, and variant Myh6 reporter plasmids were cotransfected into HEK293T cells. Guide RNAs (gRNAs) were completely complemental to variant Myh6. B, Predicted RfxCas13d ternary structure with potential RNA interaction amino acid. Brown sticks indicate amino acid residues that interacted with CRISPR RNA or target RNA in the model. C, Interaction sites of gRNA:target RNA duplex with RfxCas13d. Asterisks indicate conserved interactions between EsCas13d and RfxCas13d. D through F, Knockdown efficiency and discrimination ratio of Cas13d. gRNA15 (D) and gRNA17 (E) were used for screening. The variants (N348A, K350A, T486A, R529A, N641A, and R648A) and a combinational variant (R648A/N641A/T486A) are labeled in red. Wild-type Cas13d is labeled in blue. F, Discrimination ratio (%) = knockdown efficiency on variant Myh6 (Kd_mu)−knockdown efficiency on wild-type Myh6 (Kd_wt). Data represent the mean of 4 independent experiments.

Evolving Cas13d for Resolving Allelic HCM SNVs

To enhance the discrimination ability of RfxCas13d on HCM SNVs, we used structure-directed mutagenesis to evolve WT Cas13d (wtCas13d). Because there is no RfxCas13d structure available, we inferred a structure of the RfxCas13d/crRNA/template ternary complex using a homology structure model on the basis of the cryogenic electron microscopy structure of the homologous Eubacterium siraeum Cas13d (EsCas13d; 34% analog to RfxCas13d)/RNA complex (PDB: 6e9f²³; Figure S3A). Manipulating the interplay between the Cas9 protein and its crRNA/DNA substrates has been shown to improve editing fidelity.^24,25 We first identified 16 Cas13d/RNA interaction sites in the homologous structural model. Twelve sites (K323, N324, G326, R333, K350, K411, Y415, T486, R601, Y604, R648, and Y649) were conserved in the EsCas13d cryogenic electron microscopy structure, and 4 sites (N348, R529, K585, and N641) were only predicted by our model. These sites spread across the entire Cas13d-RNA cleft (Figure 1B and 1C).

We altered each of these 16 sites to alanine and analyzed the effect of these variants in the dual-fluorescence single nucleotide discrimination assay. We used 2 gRNAs (gRNA15 and gRNA17) complementary to R872H variant Myh6, as these gRNAs had high knockdown efficiency in our initial studies (Figure S2B and S2C). Six alanine substitutions increased the allelic discrimination ratio (knockdown efficiency on variant Myh6 minus knockdown efficiency on WT Myh6). Among them, N641A and R648A increased allelic discrimination most considerably (for gRNA15 and gRNA17, respectively: 43% and 35% for N641A, 37% and 30% for R648A, and 14% and 17% for WT) and retained comparable knockdown efficiency on Myh6-R872H to WT Cas13d (Figure 1D through 1F; Figure S3B and S3C). T486A, located in the catalytic HEPN1-II (higher eukaryotes and prokaryotes nucleotide) domain, also increased the allelic discrimination but mildly decreased knockdown efficiency. K350A in the Helical-1 linker domain mildly improved allelic discrimination and had high cleavage activity. R529A and N348A improved discrimination efficiency but substantially reduced knockdown efficiency. Other sequence variations had either no substantial effect on discrimination ratio or low knockdown efficiency (Figure 1D through 1F; Figure S3B and S3C).

To further improve allelic discrimination, we generated 9 additional variants with variable combinations of N641A, R648A, T486A, or K350A (Figure 1D through 1F; Figure S3B and S3C). Among them, only one variant carrying T486A/N641A/R648A (TNR) exhibited higher allelic discrimination than any single sequence variation with both guides (gRNA15, 55%; and gRNA17, 49%). TNR showed comparable knockdown efficiency to WT Cas13d on variant Myh6 (TNR versus WT: gRNA15, 66% versus 74%; gRNA17, 58% versus 68%) and much lower knockdown on WT Myh6 (gRNA15, 11% versus 65%; gRNA17, 10% versus 60%; Figure 1D through 1F). Fluorescence-activated cell sorting assays further validated that allelic discrimination of TNR was superior to WT (Figure S3D). Therefore, we named TNR high-precision Cas13d (hpCas13d) in the following experiments.

hpCas13d Has Low Enzymatic Kinetics on Single-Mismatched Target RNA

To gain insight into the mechanism by which variants introduced into hpCas13d enhance allelic discrimination, we purified wtCas13d and hpCas13d protein from Escherichia coli and performed the in vitro dose-dependent cleavage assay (Figure 2A through 2D; Figure S4A). With gRNA targeting Myh6-R872H, hpCas13d almost completely cleaved the Myh6-R872H transcript at 225 nM, whereas 600 nM was needed to cleave the Myh6-wt transcript comparably. In contrast, wtCas13d exhibited similar cleavage of both Myh6-R872H and Myh6-wt at 225 nM (Figure 2A through 2C; Figure S4B). A kinetic assay further illustrated that hpCas13d exhibited cleavage kinetics similar to wtCas13d when gRNAs and targets matched. However, a mismatch of gRNA against Myh6-R872H with the Myh6-wt RNA target dramatically reduced the rate of cleavage by hpCas13d but not wtCas13d (Figure 2D; Figure S4C). These results suggest that the TNR substitutions enhance SNV discrimination by dampening the cleavage kinetics of Cas13d when gRNA and target differ by a single nucleotide.

Figure 2.

In vitro cleavage and discrimination window exploration. A and B, Representative gels depicts in vitro cleavage reactions on Myh6-R872H (RH; A) and Myh6-wt (B) with increased concentration of Cas13d variants. C, Dose-responsive cleavage activity of wtCas13d or hpCas13d. A total of 300 nM target RNA and 75 to 600 nM Cas13d/guide RNA (gRNA) were used for reaction. Exponential fits are shown as solid lines. D, Cleavage kinetics of wtCas13d or hpCas13d in a time course. Exponential fits are shown as solid lines. The pseudo-first-order rate constants (k_obs; mean) representing the kinetics are shown on the right. E through G, A mismatched gRNA array targeting variant Myh6 revealed a discrimination region of hpCas13d. Knockdown efficiencies of these titling gRNAs (E) with wtCas13d (F) and hpCas13d (G) were measured with the dual reporter system. Data are mean±SD; n=3 independent experiments.

Identification of the Seed Region in gRNA

To uncover the effect of mismatch location on allelic discrimination, we used the molecular dynamic (MD) model to calculate the root mean square fluctuation (the average fluctuation of the RNA duplex) of each nucleotide in gRNA and in target RNA complexed with wtCas13d and Cas13d variants. As anticipated, TNR mutagenesis diminished the interactions between Cas13d and the crRNA-target RNA duplex (Figure S5A through S5D), and might alter the stability of the RNA duplex as a consequence. Compared with WT, N641A and R648A decreased the stability of gRNA spacer central region nucleotides 10 to 14, which was further intensified by the TNR triple sequence variation (Figure S5D). N641A and R648A did not change the root mean square fluctuation of target RNA, but TNR continued to increase it at 14 to 18 nt of target RNA, matching to the same spacer central region 10 to 14 nt (Figure S5E). Elevated root mean square fluctuation suggests RNA duplex instability and possible mismatch intolerance around this spacer central region. We tested this hypothesis by measuring the knockdown efficiency of gRNAs containing a single mismatch at nucleotides 1 to 23 of the spacer (Figure 2E). The location of the mismatch did not substantially affect wtCas13d knockdown efficiency (Figure 2F). In contrast, hpCas13d knockdown efficiency was substantially compromised when the mismatch was located at spacer nucleotides 15 to 21 (Figure 2G), immediately next to the central region with highest root mean square fluctuation change (Figure S5E). Unlike the central region, which contacts Cas13d protein, this portion of the spacer is free of protein binding and exposed to solvents, which matches a key feature of the seed region of other CRISPR gRNAs. The elevated instability at nucleotides 10 to 14 caused by TNR may convey proximally to nucleotides 15 to 21 and thereby increase their mismatch intolerance.

Reduced Off-Target Effects of hpCas13d

The intrinsic collateral cleavage activity of Cas13d, which refers to the gRNA-independent degradation of bystander RNA, substantially limits its therapeutic potential. We assessed the collateral activity of hpCas13d in a new dual fluorescence assay, wherein a GFP-labeled Cas13d variant was cotransfected with an Myh6-dmCherry reporter and a gRNA targeting either the mCherry transcript or an unrelated transcript (Figure 3A). In this assay, we compared hpCas13d, wtCas13d, dCas13d (an enzymatic dead form of Cas13d carrying R239A, H244A, R858A, and H863A sequence variations in the HEPN domains), and recently reported low collateral variance hfCas13d (carrying A134V, A140V, A141V, and A143V). When targeting the dmCherry transcript, wtCas13d induced substantial GFP collateral cleavage compared with dCas13d, as previously reported.^26,27 However, this collateral cleavage was markedly reduced by hpCas13d (Figure 3B). We also evaluated gRNAs targeting unrelated, endogenous transcripts for PPIA and RPL4, which are known to exhibit substantial collateral cleavage activity. wtCas13d displayed high collateral cleavage activity on GFP and dmCherry, whereas hpCas13d exhibited a substantial reduction in collateral cleavage (Figure 3C and 3D). The minimal collateral cleavage activity of hpCas13d was comparable to that of hfCas13d, yet hpCas13d demonstrated substantially higher on-target cleavage activity. These findings were further corroborated by real-time reverse transcription quantitative polymerase chain reaction (Figure 3E through 3G).

Figure 3.

Off-target assessment of hpCas13d. A, Schematic diagram of collateral off-target analysis in HEK293T cells. Cas13d/GFP plasmids with mCherry-targeting or PPIA-targeting guide RNA (gRNA) were transfected with variant Myh6-mCherry plasmids. A total of 48 hours later, GFP⁺ cells were sorted by fluorescence-activated cell sorting (FACS). Quantitative polymerase chain reaction (qPCR) was used to evaluate on-target cleavage and collateral cleavage activity. B, FACS analyzing the collateral cleavage of GFP with mCherry-specific gRNA in 293T cells transfected with dCas13d, wtCas13d, hpCas13d, and hfCas13d. C and D, FACS analyzing the collateral cleavage of GFP and mCherry with gRNAs specific to endogenous PPIA (C) or RPL4 (D) in 293T cells transfected with dCas13d, wtCas13d, hpCas13d, or hfCas13d. E through G, Reverse transcription qPCR validated the compromised collateral activity of hpCas13d and hfCas13d compared with wtCas13d with gRNAs specific to mCherry (E), PPIA (F), or RPL4 (G). Data are mean±SD; n=3 independent experiments. Two-way ANOVA post hoc with Sidak multiple comparison test. ***P<0.001.

To further evaluate the collateral effects of hpCas13d genome-wide, we conducted RNA sequencing analysis of the transcriptome in HEK293T cells after the transfections with dCas13d, wtCas13d, hpCas13d, and hfCas13d (Figure S6). Compared with dCas13d, wtCas13d with a PPIA gRNA (reported to induce substantial collateral cleavage) resulted in considerable downregulation of 1085 genes. In sharp contrast, hpCas13d and hfCas13d showed much less downregulated genes (185 and 165, respectively; Figure S6). These findings collectively demonstrate that hpCas13d exhibits reduced collateral activity, rendering it suitable for in vivo applications.

hpCas13d Prevented HCM in Myh6^R872H/+ Mice

R870 is located in the rod domain of human MYH7 and conserved across multiple species (Figure S7A and S7B). Heterozygous p.R870H variant has been linked to HCM in multiple HCM families, but its pathogenicity has not been evaluated in animal models. We introduced this missense variant into exon 21 of mouse Myh6 (c.2615 G>A, p.R872H), the major murine cardiac MHC analogous to human MYH7 (Figure S7C through S7E). Myh6^R872H heterozygous (Myh6^RH/+) mice were grossly normal and only developed mild hypertrophy without fibrosis by age 12 months (Figure S8A through S8C; Table S2). However, oral administration of 1 mg/g cyclosporine A for 3 weeks increased left ventricular hypertrophy (Figure S8D through S8G). These results suggest that Myh6 p.R872H causes HCM in mice.

To test whether hpCas13d selectively suppresses the variant allele in Myh6^RH/+ mice in vivo, we treated mice with adeno-associated virus type 9 (AAV9) that expresses hpCas13d under control of the cardiac-specific troponin T promoter and gRNA targeting the RH allele under the U6 promoter (Figure 4A). A total of 1×10¹¹ vg AAV-hpCas13d[RH] was injected subcutaneously to each neonatal mouse, and cyclosporine A was administered at weeks 5 through 8 (Figure 4B). Digital droplet PCR indicated that hpCas13d specifically downregulated the Myh6^R872H variant allele but not the WT allele (knockdown 27.14±3.74% versus −3.954±2.82%; Figure S8E). Hyperdynamic systolic function observed in Myh6^RH/+ mice was normalized by hpCas13d (ejection fraction 76.8±3.3% [PBS] versus 56.9±7.3% [hpCas13d]; P<0.001), as were left ventricular posterior wall thickness (in diastole, 0.64±0.02 versus 0.83±0.07 mm [P<0.001]; in systole, 0.695±0.10 versus 1.35±0.11 mm [P<0.001], respectively) and left ventricular systolic internal diameters (end-diastolic, 2.94±0.19 versus 2.74±0.29 mm [P=0.3307]; end-systolic, 2.09±0.20 versus 1.54±0.20 mm [P=0.0014]; Figure 4C; Figure S8D; Table S2).

Figure 4.

Therapeutics of hpCas13d on Myh6^R872H/+ hypertrophic cardiomyopathy mice. A and B, Schematic diagram of AAV-Cas13d-guide RNA (gRNA) structure and in vivo experimental design. Myh6^R872H/+ (RH/+) mice were injected with 1×10¹¹ vg AAV-hpCas13d on postnatal day 3. Wild-type mice and RH/+ mice were injected with PBS as controls. When the mice grew to 5 weeks of age, 1 mg/g cyclosporine A (CsA) in chow was administered for 3 weeks to induce hypertrophy; the mice were evaluated with echocardiogram, ECG, and pathology thereafter. C, M-mode echocardiograms showed CsA increased ejection fraction (EF) and posterior left ventricular diameter (left ventricular posterior wall thickness in diastole [LVPWd], left ventricular posterior wall thickness in systole [LVPWs]) and decreased left ventricular internal diameter (end-diastolic left ventricular systolic internal diameter [LVIDd], end-systolic left ventricular systolic internal diameter [LVIDs]) in RH/+ mice, which was reversed by hpCas13d. Data are mean±SD; n=6 or 7 each group. One-way ANOVA post hoc with Tukey multiple comparison test. **P<0.01, ***P<0.001. D, Dissected heart images after treatment. CsA induced the hypertrophy of RH/+ heart that was blocked by hpCas13d injection. E, Heart cross-sections of hematoxylin & eosin staining revealed a thinner heart wall in hpCas13d-treated RH/+ mice compared with RH/+ mice. F and G, ECG recordings (F) and quantification (G) showed hpCas13d[RH] alleviated the arrhythmogenesis. Data are mean±SD; n=6 or 7. One-way ANOVA post hoc with Tukey multiple comparison test. **P<0.01, ***P<0.001.

AAV-hpCas13d[RH]–treated Myh6^RH/+ hearts appeared smaller than their untreated counterparts (Figure 4D). Histological analysis indicated reduced myocardial wall thickening and smaller cardiomyocyte size (Figure 4E; Figure S8F and S8G). Arrhythmia is one of the major clinical manifestations of HCM, with MYH7 variants predisposing to cardiac arrest and sudden death. Surface ECG revealed that Myh6^RH/+ mice developed prolonged QTc and smaller S waves after 3 weeks of cyclosporine A. These abnormalities were prevented by AAV-hpCas13d (Figure 4F and 4G). All AAV-hpCas13d[RH]–treated mice survived to the experimental end points, indicating the in vivo safety of hpCas13d.

Furthermore, hpCas13d[RH] treatment did not alter cardiac function or structural remodeling in Myh6^RH/+ mice without cyclosporine A, which indicated that hpCas13d[RH] had minimal adverse effects on cardiac function at a physiological level (Figure S9 and Table S2).

Myh6^R872H/R404Q Compound Heterozygous Mice Developed Early-Onset and Spontaneous HCM

The heterozygous p.R870H sequence variation causes mild HCM in mice and humans, but was also identified in pediatric-onset HCM⁹ cases with a more severe clinical phenotype than adult-onset HCM.²⁸ Compounding mutagenesis with other MYH7 pathological variants might be the genetic mechanism underlying the phenotypic heterogeneity. To investigate this hypothesis, we crossed Myh6^R872H/R872H (RH/RH) mice with an established Myh6^R404Q/+ (RQ/+) mouse,²⁹ which is analogous to human MYH7 p.R403Q, a well-characterized malignant HCM variant located in the myosin head (Figure S10A through S10C). The resulting Myh6^R872H/R404Q (RH/RQ) compound heterozygous mice spontaneously developed left ventricular hypertrophy by age 2 months and worsened by age 4 months. In contrast, at this age, Myh6^R404Q/+ and Myh6^R872H/+ heterozygous mice did not exhibit substantial HCM (Figure S11A; Table S2). These results depict that p.R870H compound with other MYH7 variants leads to early-onset spontaneous HCM with more severe phenotype than any single variant.

hpCas13d Inhibited HCM Development in Myh6^R872H/R404Q Mice

To further benchmark the therapeutic potential of hpCas13d for HCM with other MYH7 variants, we applied hpCas13d to Myh6^R872H/R404Q mice. Using our dual-fluorescence single nucleotide discrimination assay, we selected a gRNA that specifically and efficiently interfered with the mouse Myh6^R404Q reporter expression (Figure S11B). We packaged hpCas13d and gRNA expression cassettes into AAV9 (AAV9-hpCas13d[RQ]) and administered it to neonatal Myh6^R872H/R404Q mice (Figure 5A and 5B). AAV-hpCas13d[RQ] decreased the Myh6^R404Q transcript by ≈32% and 23.5% compared with untreated levels 2 and 4 months after administration, whereas the Myh6^R872H transcript level was largely preserved (Figure S11C and S11D). Four months after injection, AAV-hpCas13d[RQ] markedly prevented increase in left ventricular posterior wall thickness and reduction of left ventricular internal diameter, and restored heart function almost to normal levels (hpCas13d[RQ] versus PBS hearts: left ventricular posterior wall thickness in diastole 0.92±0.15 versus 1.28±0.18 mm [P=0.001]; left ventricular posterior wall thickness in systole 1.17±0.12 versus 1.70±0.19 mm [P<0.001]; end-diastolic left ventricular systolic internal diameter 3.31±0.31 versus 2.84±0.33 mm [P=0.0375]; end-systolic left ventricular systolic internal diameter 2.38±0.27 versus 1.62 ±0.18 mm [P<0.001]; ejection fraction 55.7±3.4% versus 74.77±7.86% [P<0.001]; Figure 5C and 5D; Table S2).

Figure 5.

Therapeutics of hpCas13d on Myh6^R872H/R404Q compound heterozygous mice. A and B, Experimental design. 1×10¹¹ vg of AAV-hpCas13d targeting Myh6 R404Q sequence variation (A) was subcutaneously injected into Myh6^R872H/R404Q (RH/RQ) mice at P3 (B). C and D, M-mode echocardiograms (C) and quantification (D) implicated mitigated left ventricular diameter (left ventricular posterior wall thickness in diastole [LVPWd], left ventricular posterior wall thickness in systole [LVPWs]) and EF and increased left ventricular internal diameter (end-diastolic left ventricular systolic internal diameter [LVIDd], end-systolic left ventricular systolic internal diameter [LVIDs]) in hpCas13d[RQ]-treated hearts. Data are mean±SD; n=6 mice per group. One-way ANOVA post hoc with Tukey multiple comparison test. *P<0.05, **P<0.01, ***P<0.001. E and F, ECG recordings (E) and quantification (F) illustrated that hpCas13d normalized the aberrant QRS amplitude and QRS intervals of RH/RQ hearts. Red arrows show a dramatic decrease of both Q amplitude and R amplitude. Data are mean±SD; n=6 mice. One-way ANOVA post hoc with Tukey multiple comparison test. **P<0.01, ***P<0.001.

Consistent with severe ventricular hypertrophy observed in echocardiograms, ECG analysis revealed abnormal cardiac impulse conduction in Myh6^R872H/R404Q mice, characterized by a reduced PR interval, prolonged QRS interval, increased corrected QT interval, and altered QRS complex morphology (reduced Q and R wave amplitudes; Figure 5E and 5F). hpCas13d[RQ] administration significantly improved electrical conduction and normalized these electrocardiographic measurements (hpCas13d[RQ] versus PBS mice: PR interval 34.65±5.39 versus 21.99±7.197 ms [P=0.002]; QRS interval 9.43±0.89 versus 14.15±2.62 ms [P=0.0001]; corrected QT interval 47.74±3.55 versus 69.73±20.68 ms [P=0.0161]; R amplitude 0.56±0.11 versus 0.27±0.07 mV [P<0.001]; Q amplitude 0.023±0.010 versus −0.163±0.148 mV [P=0.0017]; Figure 5E and 5F).

Furthermore, postmortem analysis illuminated that AAV-hpCas13d[RQ] prevented decrease in heart size and weight (hpCas13d[RQ] versus PBS mice: heart weight/body weight 4.86±0.21 versus 6.19±0.44 mg/g [P<0.001]; Figure 6A; Figure S11E) in Myh6^R872H/R404Q mice. Consistent with the echocardiogram and ECG results, histological analysis of cross-sections demonstrated reduced ventricular thickness (Figure 6B), a lower cardiomyocyte cross-sectional area (hpCas13d[RQ] versus PBS cardiomyocytes 267.9±112.6 versus 402.3±150.7 μm²; n=262–293 cells [P<0.001]; Figure 6C and 6D), and better alignment of myofibrils in hpCas13d[RQ]-treated hearts (Figure 6E). Reduced fibrosis and cellular apoptosis (hpCas13d[RQ] versus PBS hearts: fibrosis 3.29±1.45% versus 5.76±0.81% [P=0.002]; TUNEL⁺ cells, 4.90±0.43% versus 17.03±2.95% [P<0.001]; Figure 6F through 6I) were also observed in hpCas13d[RQ]-treated hearts compared with untreated Myh6^R872H/R404Q controls. All AAV-hpCas13d[RQ]-treated mice survived to pilots, further supporting the safety of hpCas13d.

Figure 6.

Histological analysis of hpCas13d[RQ]-treated hearts. A, Heart images show severe hypertrophy of Myh6^R872H/R404Q heart, which was alleviated by hpCas13d[RQ]. B, Hematoxylin & eosin cross-sections revealed that hpCas13d[RQ] markedly reduced the wall thickness and increased the ventricular chamber of RH/RQ hearts. C and D, Wheat germ agglutinin staining (C) and quantification (D) showed that the enlarged cardiomyocytes of RH/RQ hearts were normalized by hpCas13d[RQ] (n=262 to 293 cardiomyocytes from 3 individual hearts). Nonparametric Kruskal-Wallis test with Dunn multiple-comparisons test. ***P<0.001. E, Hematoxylin & eosin staining showed that hpCas13d[RQ] alleviated the myofibril disarray in RH/RQ hearts. F and G, Masson trichrome staining (F) and quantification (G) revealed that hpCas13d[RQ] alleviated the collagen deposition in 6-month-old RH/RQ hearts. Data are mean±SD; n=5 or 6 mice. One-way ANOVA post hoc with the Tukey test. ***P<0.001. H and I, TUNEL staining (H) and quantification (I) showed that hpCas13d[RQ] decreased cell apoptosis in RH/RQ hearts (n=3). One-way ANOVA post hoc with Tukey test. ***P<0.001.

Moreover, we applied the hpCas13d[RQ] to 2-month-old Myh6^RH/RQ mice that had already developed HCM and found that hpCas13d[RQ] partially restored the hypercontractile phenotype (Figure S12; Table S2). These results demonstrate that hpCas13d is able to suppress distinct allelic sequence variations in vivo and can be used as an efficient gene therapy tool to treat HCM.

hpCas13d Resolved Diverse Human Pathological MYH7 Variants

To further assess the ability of hpCas13d to discriminate human MYH7 pathogenic SNVs, we expanded our dual-fluorescence single nucleotide discrimination assay to encompass 45 MYH7 variants (32 pathogenic sequence variations from ClinVar³⁰ and 13 variants designed to assess detection of diverse sequence variations) within one human MYH7 coding sequence, fused with dmCherry (Figure S13A). For each SNV, we designed 3 gRNAs positioned within nucleotides 15 to 17 of the spacer discrimination region (Figure 2). Because of their incompatibility with U6 transcripts, gRNAs with 4 consecutive Ts were excluded. We ultimately constructed 130 gRNAs targeting 43 MYH7 SNVs in expression vectors with either wtCas13d or hpCas13d (Figure 7A). The allelic discrimination ratio for wtCas13d ranged from −24% to 42.29% and averaged 5.99% (Figure 7B and 7C). In contrast, the allelic discrimination ratio for hpCas13d ranged from −6.95% to 71.57% and averaged 29.03%, which is significantly higher than wtCas13d (Mann-Whitney test; P<0.001). Moreover, 42.31% of gRNAs (55 of 130) targeting 55.81% sequence variation sites (24 of 43) had a discrimination efficiency >30% (Figure 7D and 7E), which was therapeutically effective in our in vivo studies (Figures 4 and 5). The discrimination ratio of hpCas13d but not wtCas13d had a positive correlation (Pearson correlation R=0.55 [P=1.3e-11]) with hybridization minimum free energy calculated from mismatched gRNA/target RNA duplex, further supporting the high precision profiles of hpCas13d (Figure S13B and S13C).

Figure 7.

hpCas13d resolved multiple single-base mismatches in a human MYH7 sequence variation library. A, A total of 32 pathogenic sequence variations, along with an additional 13 sequence variations, were introduced in human MYH7 plasmid. Guide RNAs (gRNAs) perfectly matched with variant MYH7 and with a single mismatch to wtMYH7 at spacer 15 to 17 nt were designed. Human MHY7 dual-fluorescence plasmids cotransfected with AAV-Cas13d-gRNA plasmids into HEK293T cells. B, Scatterplots show mismatch resolving efficiencies of hpCas13d versus wtCas13d with 130 gRNAs targeting 43 human variants (n=3 gRNAs for each MYH7 sequence variation). C, Average resolving efficiencies of hpCas13d and wtCas13d were calculated. Mann-Whitney test, 2-tailed. ***P<0.001. D, Pie chart shows the single-base mismatch resolving efficiencies of wtCas13d and hpCas13d with 130 gRNAs differently distributed in 3 quartiles. E, The quantile of discrimination of hpCas13d and wtCas13d on 43 human MYH7 sequence variation sites. F, The quantile of discrimination ratios of hpCas13d and wtCas13d on 18 MYH7 sequence variation sites unavailable for base editors.

Among the 43 variants are 18 A/T, C/A, and G/T transitions that cannot be edited by current base editors. Thirteen of these variants (35 gRNAs) were efficiently resolved by hpCas13d with a discrimination ratio >30% (Figure 7F; Figure S13D). The average discrimination ratio of hpCas13d at these sites was 38.78% (Figure S13E). These results demonstrate that hpCas13d has the capacity to discriminate many SNVs from WT, suggesting that it will be a potent tool for human gene therapy of disorders caused by dominant SNVs.

DISCUSSION

Leveraging an allele-specific reporter platform and mutagenesis screen, we identified a high-precision Cas13d that efficiently cleaves point mutant allele while preserving the WT allele. This single nucleotide discrimination was robust across a wide range of tested SNVs. hpCas13d also exhibits high fidelity with minimized collateral cleavage. In vivo administration of hpCas13d specifically suppressed 2 different SNV alleles that cause HCM and resulted in a profound mitigation of the HCM phenotype. Further in vitro screening provided a substantial list of efficient gRNAs for treating human MYH7 HCM. Our study uncovered great potential for CRISPR-Cas13 RNA nuclease in treating genetic cardiomyopathy. The enhanced programmable RNA nuclease developed in this study, characterized by high precision and fidelity, could be broadly used for RNA biology and gene therapy.

Gene therapies, especially base editor–based therapy, have recently showed great potential in treating HCM and other cardiac diseases.^11,31–35 However, available base editors have inherent limitations, including protospacer adjacent motif constraints, editing window constraints, and inefficient editing for A/C, A/T transversion.¹² As a consequence, at least 30% of genetic variants cannot be efficiently edited. Moreover, for efficient delivery to cardiomyocyte, these base editor–based gene therapies often use AAV. The long-term expression trait of AAV raises substantial safety concerns because of nonnegligible off-target and bystander editing, despite the evolution of plentiful base editor variants to minimize these effects.^36,37

Instead of manipulating the genetic material of DNA, RNA nuclease CRISPR-Cas13d cleaves the RNA, which is reversible and safe for long-term AAV delivery. In this regard, CRISPR-Cas13d has been applied to treat several genetic disorders, such as Huntington disease,¹⁸ but allelic-specific knocking down of RNAs differing by a single nucleotide was previously unavailable until our studies. Autosomal dominant gain-of-function sequence variations account for ≈36% of all human genetic diseases, according to the Online Mendelian Inheritance in Man database.³⁸ Approximately 30% of HCM cases result from dominant sarcomere gene SNVs. Traditional RNA interference¹⁰ and WT Cas13d (as shown in Figure 1 and Figure S2) have limited applicability to this class of genetic diseases because of their inability to discriminate between WT and SNV transcripts. Indistinguishable suppression of WT and variant allele not only fails to remedy but also exacerbates disease progression.³⁹ Our hpCas13d overcomes this barrier and enables efficient SNV-specific knockdown, offering a new therapeutic approach with great translational potential for these allele-specific and dominant-negative diseases.

In addition to improved precision, our strategy also enhances the fidelity of Cas13d. Both fluorescent reporter assays and RNA sequencing analysis demonstrated that hpCas13d had reduced collateral effects. Mutating the RNA binding region proximal to the enzymatic HEPN domain was reported to reduce the collateral cleavage.²⁶ However, our hpCas13d sequence variations are not located in this region and do not overlap with the sequence variations previously reported to reduce collateral cleavage.²⁶ Collateral cleavage depends on gRNA-induced HEPN activation.¹⁵ Therefore, one possible mechanism by which hpCas13d reduces collateral cleavage is to blunt the off-target RNA-induced HEPN activation. HEPN domain activation requires conformational rearrangements of multiple domains.²³ Therefore, another possible mechanism for collateral activity reduction is that the TNR substitutions may change the confirmational kinetics of RfxCas13d regarding their positions at the confirmational-executive domains of HEPN1, Helical-2, and HEPN2. Elucidating these changes is beyond the scope of MDs and necessitates a structural analysis of hpCas13d with mismatched RNA duplexes.

Residual collateral cleavage activity of hpCas13d was still observed, although substantially reduced, which could potentially hinder its clinical application. To circumvent this curb, the fidelity of hpCas13d needs to be improved. One strategy is to search for other amino acids related to reducing collateral cleavage or to include the variants from hfCas13d. We also found different gRNA sequences were associated with varying degrees of collateral cleavage. Therefore, for each target, selection of gRNA with minimized off-target before in vivo usage becomes critical. It is requisite to include multiple reliable and sensitive assays such as dual reporter assays and RNA sequencing for sake of complete evaluation of the residual RNase activity. With these refinements, hpCas13d holds great potential for clinical gene therapy in the near term.

Many genetic disorders are not monogenic. With respect to HCM, recent studies show that many HCM cases are caused by multiple pathological variants.^3,40 Patients with multiple variants are also considerably younger, commonly presenting with childhood-onset hypertrophy.⁴¹ In agreement with these clinical observations, we found that compound heterozygosity for 2 MHC variants causes spontaneous and severe HCM not observed for each variant alone. Suppression of one variant was sufficient to prevent the severe HCM phenotype. Thus, our study provides an important guide for designing gene therapy in oligogenic diseases.

This study opens a new avenue for using high-precision and fidelity CRISPR-Cas13 tools to treat familial cardiomyopathy, and this approach is also expected to have broader applications for other genetic disorders, such as transthyretin amyloidosis,⁴² amyotrophic lateral sclerosis,¹⁷ and Huntington disease.¹⁸

ARTICLE INFORMATION

Acknowledgments

The authors thank Dr Hui Yang (Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, China) for providing hfCas13d plasmids; Drs Xuanming Yang and Wanting Wang (Sheng Yushou Center of Cell Biology and Immunology, Shanghai Jiao Tong University) for assistance with flow cytometry analysis; and Mei Gao and Xi Liu (Instrumental Analysis Center, Shanghai Jiao Tong University) for echocardiography.

Sources of Funding

This work was supported by grants from the National Key Research and Development Program of China (2020YFA0803800, 2020YFA0803802, 2023YFA1800700, and 2023YFA1800702), National Foundation of Distinguished Young Scholars of China (grant 82225006), Innovation Program of the Shanghai Municipal Education Commission (grant 2021-01-07-00-02-E00088), WLA Program of Shanghai Science and Technology Commission (grants 21dz2210202 and 21dz2210200), and SJTU STAR Award (awards YG2021ZD11 and YG2022ZD023; Dr B. Zhang); National Key Research and Development Program of China (grants 2023YFA1800700 and 2023YFA1800702), Startup Fund for Young Faculty of SJTU (grant 21X010500929), and the Youth Program of National Natural Science Foundation of China (81901515; Dr Yang); the National Key Research and Development Program of China (grants 2023YFA1800700 and 2023YFA1800703), National Natural Science Foundation of China (grant 32200926), and SJTU STAR Award (AF4150057; Dr S. Zhang); and the Ministry of Science and Technology of China (grants 2021YFA0804803 and 2022YFA1104204), the Major Project of Natural Science Foundation of Hunan Province (Open Competition 2021JC0002), and the National Science Foundation of China (grants 81930012 and 82241027; Dr Chen).

Disclosures

None.

Supplemental Material

Methods

Tables S1–S2

Figures S1–S13

References 43–49