Improved specificity and efficiency of in vivo adenine base editing therapies with hybrid guide RNAs

Abstract

Phenylketonuria (PKU), pseudoxanthoma elasticum (PXE), and hereditary tyrosinemia type 1 (HT1) are autosomal recessive disorders linked to the PAH, ABCC6, and FAH and HPD genes, respectively. Here we evaluated the off-target editing profiles of clinical lead guide RNAs (gRNAs) that, when combined with adenine base editors (ABEs), correct the recurrent PAH P281L variant, PAH R408W variant, or ABCC6 R1164X variant or disrupt either of two sites in the HPD gene (a modifier gene of HT1) in human hepatocytes. To mitigate off-target mutagenesis, we systematically screened hybrid gRNAs with DNA nucleotide substitutions. Comprehensive and variant-aware specificity profiling of these hybrid gRNAs revealed dramatically reduced off-target editing and reduced bystander editing in cells. In humanized PAH P281L and ABCC6 R1164X mouse models of PKU and PXE, we showed that when formulated in lipid nanoparticles (LNPs) with ABE mRNA, selected hybrid gRNAs reverted disease phenotypes, reduced off-target editing, increased on-target editing, and reduced bystander editing in vivo. These studies highlight the utility of hybrid gRNAs to improve the safety and efficiency of adenine base editing therapies.

Extended Fig. 1 |. Reduction of variant-aware off-target editing by PAH1 hybrid guide RNAs.

Examples of variant sites nominated by ONE-seq, with reductions in ONE-seq scores observed with PAH1 hybrid gRNAs. Differences between reference and variant versions of nominated sites are in bold and underlined.

Extended Fig. 2 |. Reduction of blood phenylalanine levels by PAH1 hybrid guide RNAs.

Blood phenylalanine levels in homozygous PAH P281L mice (n = 4–5 animals per group) before and 48 hours after treatment with 2.5 mg/kg dose of LNPs with PAH1 standard or hybrid gRNA in combination with ABE8.8 mRNA (1 blood sample per timepoint). The dotted line marks the upper end of the normal range for blood phenylalanine, 125 μmol/L. Means are shown; P values were calculated with the Mann–Whitney U test.

Extended Fig. 3 |. Effects of PXE1 gRNA LNPs on blood pyrophosphate levels and ALT levels.

a, Blood pyrophosphate levels in wild-type mice, before and 7 days after treatment with PBS vehicle (n = 10 animals), or in homozygous ABCC6 R1164X mice, before and 7 days after treatment with 2.5 mg/kg dose of LNPs with PAH1 standard gRNA (n = 6 animals) or PAH1_hyb18 gRNA (n = 5 animals) in combination with ABE8.8 mRNA (1 blood sample per timepoint). Means are shown; P values comparing genotype groups were calculated with the Mann–Whitney U test, and the P value comparison for the aggregated R1164X groups (before versus after treatment) was calculated with the Wilcoxon signed-rank test. P values were not calculated within each individual R1164X gRNA group (before versus after treatment) because of inadequate sample sizes for the Wilcoxon signed-rank test. b, Blood alanine aminotransferase (ALT) levels in wild-type mice (n = 4 animals per group) before and 24 hours after treatment with PBS vehicle or with 2.5 mg/kg dose of LNPs with PAH1 standard or hybrid gRNA in combination with ABE8.8 mRNA (1 blood sample per timepoint). No effects on ALT levels by LNP treatment were evident.

Extended Figure 4.. Assessment of on-target and off-target editing by guide RNAs targeting the HPD gene.

a, Lentiviral screening in primary human hepatocytes for adenine base editing of heterologous HPD splice site target sequences by various gRNAs in combination with ABE8.20; target adenines are in bold and underlined. MIT score = off-target/specificity score indicated by UCSC Genome Browser. Four candidate gRNAs (HPD3, HPD4, HPD5, and HPD20) were taken forward for off-target analyses. b, Targeted amplicon sequencing of ONE-seq-nominated sites in plasmid-treated vs. control wild-type HuH-7 cells with HPD3, HPD4, HPD5, or HPD20 gRNA in combination with ABE8.8 (n = 2 treated and 2 untreated biological replicates). Sites with unsuccessful sequencing are omitted. Arrows indicate verified sites of off-target editing. Means ± standard deviations are shown.

Fig. 1 |. Assessment of on-target and off-target editing by candidate clinical lead guide RNAs.

a, Schematics of the genomic sites of the five loci that are evaluated in this study, adapted from the UCSC Genome Browser (GRCh38/hg38). Each vertical yellow bar indicates either the G altered to A (in red) by a pathogenic variant, or the target A (in red) at a splice site intended for disruption. Each horizontal grey bar indicates the protospacer (thick) and protospacer-adjacent motif (PAM) (thin) sequences targeted by the respective guide RNA (gRNA). b, Schematic of the ONE-seq methodology.

Fig. 2 |. Optimizing correction of the PAH P281L variant in cellulo and in vivo.

a, Hybrid capture sequencing of ONE-seq-nominated sites in treated vs. control PAH P281L HuH-7 cells with PAH1 gRNA in combination with ABE8.8 mRNA (n = 3 treated and 3 untreated biological replicates), with seven human sites verified to have off-target editing. b, Spacer sequences of PAH1 standard and hybrid gRNAs. “rN” = ribonucleotide; “dN” = deoxyribonucleotide; “m” = 2’-O-methylation; * = phosphorothioate linkage. DNA substitutions are in red bold and underlined. c, A-to-G editing of PAH on-target site (correction of P281L variant only, or correction plus unwanted nonsynonymous bystander editing) or verified PAH1 human off-target site (PAH1_OT3) in PAH P281L HuH-7 cells treated with PAH1 standard or hybrid gRNA in combination with ABE8.8 mRNA (n = 2–3 biological replicates per condition). d, Number of sites with ONE-seq score > 0.01 for each indicated PAH1 standard or hybrid gRNA with ABE8.8 protein. e, A-to-G editing of PAH on-target site or verified PAH1 human off-target site in PAH P281L HuH-7 cells treated with PAH1 standard or hybrid gRNA in combination with ABE8.8 mRNA (n = 11–12 biological replicates per condition). f, Number of sites with ONE-seq score > 0.01 for each indicated PAH1 standard or hybrid gRNA with ABE8.8 protein. g, Total A-to-G editing of PAH on-target site or any of seven verified PAH1 human off-target sites in untreated vs. treated PAH P281L HuH-7 cells with PAH1 standard or PAH1_hyb24 gRNA in combination with ABE8.8 mRNA (n = 11–12 biological replicates per condition). h, Number of variant (non-reference-genome) sites with ONE-seq score > 0.01 for each indicated PAH1 standard or hybrid gRNA with ABE8.8 protein. i, A-to-G editing of PAH on-target site or verified PAH1 mouse off-target site (PAH1_mOT3) in homozygous PAH P281L mice treated with 2.5 mg/kg dose of LNPs with PAH1 standard or hybrid gRNA in combination with ABE8.8 mRNA (n = 4–5 animals per group). Means are shown; P values were calculated with the Mann–Whitney U test.

Fig. 3 |. Optimizing correction of the ABCC6 R1164X variant in cellulo and in vivo.

a, Spacer sequences of PXE1 standard and hybrid gRNAs. “rN” = ribonucleotide; “dN” = deoxyribonucleotide; “m” = 2’-O-methylation; * = phosphorothioate linkage. DNA substitutions are in red bold and underlined. b, A-to-G editing of ABCC6 on-target site (correction of R1164X variant only, or correction plus unwanted nonsynonymous bystander editing) or either of two verified PXE1 human off-target sites (PXE1_OT1, PXE_OT2) in ABCC6 R1164X HuH-7 cells treated with PXE1 standard or hybrid gRNA in combination with ABE8.8 mRNA (n = 2–3 biological replicates per condition). c, Number of sites with ONE-seq score > 0.01 for each indicated PXE1 standard or hybrid gRNA with ABE8.8 protein. d, A-to-G editing of ABCC6 on-target site or either of two verified PXE1 human off-target sites in ABCC6 R1164X HuH-7 cells treated with PXE1 standard or PXE1_hyb18 gRNA in combination with ABE8.8 mRNA (n = 5–6 biological replicates per condition). e, A-to-G editing of ABCC6 on-target site or verified PXE1 mouse off-target site (PXE1_mOT24) in homozygous ABCC6 R1164X mice treated with 2.5 mg/kg dose of LNPs with PXE1 standard or PXE1_hyb18 gRNA in combination with ABE8.8 mRNA (n = 6 animals per group). Means are shown; P values were calculated with the Mann–Whitney U test.

Fig. 4 |. Optimizing targeting of two distinct sites in the HPD gene with hybrid gRNAs.

a, Spacer sequences of HPD20 standard and hybrid gRNAs. “rN” = ribonucleotide; “dN” = deoxyribonucleotide; “m” = 2’-O-methylation; * = phosphorothioate linkage. DNA substitutions are in red bold and underlined. b, A-to-G editing of HPD on-target site (target adenine only, or target adenine plus bystander editing) in wild-type HuH-7 cells treated with HPD20 standard or hybrid gRNA in combination with ABE8.8 mRNA (n = 3 biological replicates per condition). c, Number of sites with ONE-seq score > 0.01 for each indicated HPD20 standard or hybrid gRNA with ABE8.8 protein. d, Spacer sequences of HPD4 standard and hybrid gRNAs. e, A-to-G editing of HPD on-target site (target adenine only, or target adenine plus bystander editing) or verified HPD4 off-target site (HPD4_OT13) in wild-type HuH-7 cells treated with HPD4 standard or hybrid gRNA in combination with ABE8.8 mRNA (n = 3 biological replicates per condition). Means are shown.

Fig. 5 |. Optimizing correction of the PAH R408W variant with a PAM-relaxed ABE.

a, Spacer sequences of PAH4 standard and hybrid gRNAs. “rN” = ribonucleotide; “dN” = deoxyribonucleotide; “m” = 2’-O-methylation; * = phosphorothioate linkage. DNA substitutions are in red bold and underlined. b, A-to-G editing of PAH on-target site (correction of R408W variant with or without synonymous bystander editing, or correction plus unwanted nonsynonymous bystander editing) in PAH R408W HuH-7 cells treated with PAH4 standard or hybrid gRNA in combination with SpRY-ABE8.8 mRNA (n = 3 biological replicates per condition). c, Number of sites with ONE-seq score > 0.01 for each indicated PAH4 standard or hybrid gRNA with SpRY-ABE8.8 protein.