Systematic optimization of prime editing for the efficient functional correction of CFTR F508del in human airway epithelial cells

Abstract

Prime editing (PE) enables precise and versatile genome editing without requiring double-stranded DNA breaks. Here we describe the systematic optimization of PE systems to efficiently correct human cystic fibrosis (CF) transmembrane conductance regulator (CFTR) F508del, a three-nucleotide deletion that is the predominant cause of CF. By combining six efficiency optimizations for PE-engineered PE guide RNAs, the PEmax architecture, the transient expression of a dominant-negative mismatch repair protein, strategic silent edits, PE6 variants and proximal 'dead' single-guide RNAs-we increased correction efficiencies for CFTR F508del from less than 0.5% in HEK293T cells to 58% in immortalized bronchial epithelial cells (a 140-fold improvement) and to 25% in patient-derived airway epithelial cells. The optimizations also resulted in minimal off-target editing, in edit-to-indel ratios 3.5-fold greater than those achieved by nuclease-mediated homology-directed repair, and in the functional restoration of CFTR ion channels to over 50% of wild-type levels (similar to those achieved via combination treatment with elexacaftor, tezacaftor and ivacaftor) in primary airway cells. Our findings support the feasibility of a durable one-time treatment for CF.

Fig. 1. Basic PE2 and PE3 systems inefficiently correct CFTR F508del.

a, Schematic of PE2 and PE3 systems. The factors that influence PE2 and PE3 editing efficiency include (1) pegRNA spacer sequence, (2) pegRNA PBS length, (3) pegRNA RTT length and (4) ngRNA spacer sequence. b, Quantification of PE2 correction of CFTR F508del in HEK293T cells using NGG2 pegRNAs with different combinations of PBS and RTT lengths, with the PBS and RTT lengths shown in nucleotides. c,d, PE3 correction of CFTR F508del in HEK293T cells using NGG2 PBS13 RTT29 pegRNA (c) or NGG2 PBS14 RTT41 pegRNA (d) in combination with several ngRNAs. The x-axis labels identify different ngRNAs by their nicking position relative to the pegRNA nick (in base pairs). The PE2 x-axis label specifies an editing condition with no ngRNA. e, Adenine base editing in HEK293T cells at adenines in the NGG1 and NGG2 protospacers. Adenines (A) are numbered 5′ to 3′ starting from the PAM-distal end of the protospacers. For b–e, the data and error bars represent the means and standard deviations, respectively, of the three independent biological replicates (shown as black dots). Source data

Fig. 2. PE enhancements synergistically enhance correction of CFTR F508del.

a, Schematic of the PE system with enhancements that improve F508del correction. Enhancements include (5) epegRNA 3′ structured RNA motifs, (6) co-expression of MLH1dn, (7) translationally silent edits to evade cellular mismatch repair and (8) engineered and evolved prime editor proteins (PEmax and PE6). b, Heatmap of F508del correction in HEK293T cells using NGG2 epegRNAs with variable combinations of PBS and RTT lengths, in nucleotides. The edits were completed with PE4max. c, Silent edit installation strategies SE0–SE4 are shown. The correction of the F508del CTT deletion alone is shown as SE0. d, PE5max correction of F508del in HEK293T cells with the NGG2 PBS13 RTT41 epegRNA is encoded with SE0–SE4. e, Comparison of F508del correction with PEmax and PE6 variants a–g in HEK293T cells. All conditions use the NGG2 PBS13 RTT41 SE2 epegRNA, MLH1dn and the +104 ngRNA. For b, d and e, data and error bars represent mean and standard deviation, respectively, of three independent biological replicates (shown as black dots). Source data

Fig. 3. Enhanced PE systems enable F508del correction in human immortalized airway epithelial cells.

a, ABE of 16HBEge-F508del cells at NGG2 guided by an sgRNA, pegRNA or epegRNA. Adenines (A) are numbered 5′ to 3′ starting from the PAM-distal end of the protospacers. b, Schematic of the PE system with dsgRNA (9) added to modulate the chromatin state of the target locus. c, Comparison of F508del PE correction in 16HBEge-F508del cells with NGG2-proximal dsgRNAs. dsgRNA x-axis label indicates the distance (in nucleotides) between the nicking site of NGG2 and the putative nicking site of the dsgRNA. The strand x-axis label indicates the genomic DNA strand to which the dsgRNA binds (NGG2 targets the (–) strand). All conditions use PE6c, the NGG2 PBS13 RTT41 SE2 epegRNA, MLH1dn and the +104 ngRNA. d, Combinatorial improvements in F508del correction with epegRNAs, silent edits, PE6c and dsgRNA. For epegRNAs, (–) denotes the use of a pegRNA and (+) denotes the use of an epegRNA; for silent edits, (–) denotes the use of SE0 and (+) denotes the use of SE2; for PE6c, (–) denotes the use of PEmax and (+) denotes the use of PE6c; and for dsgRNA, (–) denotes the use of a non-targeting (n.t.) dsgRNA and (+) denotes the use of the −40 dsgRNA. All conditions use MLH1dn co-expression and the +104 ngRNA. e, Effect of MLH1dn co-expression on F508del correction efficiency and edit-to-indel ratio. All conditions use PE6c, the NGG2 PBS13 RTT41 SE2 epegRNA, the −40 dsgRNA and the +104 ngRNA. For a–e, data and error bars represent mean and standard deviation, respectively, collected from three independent biological replicates (shown as black dots). n/a, not applicable. Source data

Fig. 4. PE corrects CFTR F508del in primary airway epithelial cells from patients with CF and rescues ion channel function.

a, Quantification of F508del correction in CFTR F508del homozygous primary airway epithelial cells from patients with CF using HTS. The untreated condition indicates primary cells that were not electroporated. Data and error bars represent mean and standard deviation, respectively, and were collected from three independent donors (shown as black dots). b, The representative Isc recordings are shown with the transepithelial voltage held at 0 mV. A total of 3 weeks after electroporation of PE reagents, the transepithelial I_sc of fully differentiated airway epithelial cells was quantified in response to F&I and CFTR(inh)-172 treatment. All the cells were pre-incubated with 10 µM forskolin and 100 µM IBMX for 24 h before recording Cl⁻ transport. Non-CF cultures and donor-matched F508/F508 cultures with ETI pretreatment served as positive controls. c, Summary of the change in short-circuit current (ΔI_sc) in response to F&I and CFTR(inh)-172 treatment. Data and error bars represent mean and standard error of the mean, respectively, and were collected from three independent donors (shown as black dots). P values, determined by two-way ANOVA, are shown. For a–c, mock condition indicates primary cells that were electroporated without RNA. d, Ratio of editing in treated versus untreated primary airway epithelial cells at off-target loci nominated by CIRCLE-seq. Indels and substitutions (Sub) are shown for sites cleaved in vitro by the epegRNA, ngRNA and dsgRNA used in our PE strategy. For the epegRNA and dsgRNA, the top 31 CIRCLE-seq nominated off-target sites are shown as one of the top 32 identified sites was CFTR F508del. For the ngRNA, the top 32 CIRCLE-seq nominated off-target sites are shown. All data points are the ratios of the mean of three replicates for treated and untreated samples. The red line indicates a ratio of editing in treated versus untreated cells of 1.5. The characteristics of off-target sites for which this ratio exceeds 1.5 are shown in the table to the right. Observed editing frequency is the mean of three independently edited primary airway epithelial cell lines. lncRNA, long non-coding RNA; n/a, not available. Source data

Extended Data Fig. 1. PE2 correction of CFTR F508del is undetectable with NGG1 pegRNAs.

A panel of 24 pegRNAs with variable PBS and RTT lengths was transfected into CFTR F508del HEK293T cells. PBS and RTT lengths are listed in nucleotides. Data and error bars represent mean and standard deviation, respectively, collected from three independent biological replicates (shown as black dots). Source data

Extended Data Fig. 2. PE4max correction of CFTR F508del in HEK293T cells with epegRNAs at several targets with NGA PAMs.

a, A schematic of the targeted NGA1-3 protospacers around the CFTR F508del CTT deletion. b–d, PE4max CFTR F508del correction using epegRNAs with variable PBS lengths and variable RTT lengths targeted to the protospacers NGA1 (b), NGA2 (c), NGA3 (d). PBS and RTT lengths are listed in nucleotides. For (b–d), data and error bars represent mean and standard deviation, respectively, collected from three independent biological replicates (shown as black dots). Source data

Extended Data Fig. 3. PE4max correction of CFTR F508del in HEK293T cells with epegRNAs at several targets with NGG PAMs.

a, A schematic of the targeted NGG1-4 protospacers around the CFTR F508del CTT deletion. b-e, PE4max CFTR F508del correction using epegRNAs with variable PBS lengths and variable RTT lengths targeted to the protospacers NGG1 (b), NGG3 (c), NGG4 (d), NGG2 (e). PBS and RTT lengths are listed in nucleotides. For (b–e), data and error bars represent mean and standard deviation, respectively, collected from three independent biological replicates (shown as black dots). Source data

Extended Data Fig. 4. Correction of CFTR F508del in HEK293T cells with several ngRNAs using PE5max and an epegRNA.

Using the epegRNA NGG2 PBS13 RTT41, a PE5max experiment was performed against a panel of ngRNAs to identify the most efficient strategy to correct CFTR F508del. X-axis labels identify different ngRNAs by their nicking position relative to the epegRNA nick (in base pairs). The PE4max x-axis label specifies an editing condition with no ngRNA. Data and error bars represent mean and standard deviation, respectively, collected from three independent biological replicates (shown as black dots). Source data

Extended Data Fig. 5. Silent edits co-installed with the F508del corrective CTT insertion disrupt NGG2’s PAM and recode several F508del mutation proximal codons.

Schematics of the silent edits (shown as blue base pairs) co-installed with the F508del corrective CTT insertion (shown as green base pairs) as part of silent edit strategies SE1-SE4. The undisrupted PAM of NGG2 is shown as orange base pairs in the SE0 schematic. Source data

Extended Data Fig. 6. Prime and base editing in 16HBEge-F508del cells via mRNA electroporation.

a, Effect of epegRNA 3′ modifications on prime editing efficiency at CFTR F508del. Both epegRNAs consist of NGG2 PBS13 RTT41 SE2, and both have three phosphorothioate (mS) modifications in the place of phosphodiester bonds on their 5′ end. One epegRNA incorporates 3 mS modifications on its 3′ end and the other incorporates 2 phosphonoacetate (mP) modifications on its 3′ end. b, Base editing at CFTR F508del proximal protospacers. Deamination efficiencies for all target bases within the first 12 bases of the guide RNA protospacer sequence (indexed from the PAM-distal end of the protospacer) are shown. For (a–b), data and error bars represent mean and standard deviation, respectively, collected from three independent biological replicates (shown as black dots). Source data

Extended Data Fig. 7. pegRNA, ngRNA, and petRNA optimization in HEK293T cells in the context of other prime editing enhancements.

a, epegRNA PBS and RTT length screen using SE2 silent edits, PE6c, MLH1dn, the +104 ngRNA, and the -40 dsgRNA. b, ngRNA screen using PE6c, MLH1dn, epegRNA NGG2 PBS13 RTT41 SE2, and the -40 dsgRNA. c, petRNA screen using SE2 silent edits, PE6c, MLH1dn, the +104 ngRNA, and the -40 dsgRNA. d, Editing efficiencies using petRNAs at previously reported positive control target sites. For (c) and (d), a nCas9 (based on PEmax) and a MCP-RT (using a PE6c-based reverse transcriptase) were used for petRNA editing, as previously described. For (a–d), data and error bars represent mean and standard deviation, respectively, collected from three independent biological replicates (shown as black dots). Source data

Extended Data Fig. 8. CFTR F508del correction efficiency in HEK293T cells by epegRNAs designed using computational tools.

CFTR F508del correction mediated by the top 24 ranked epegRNAs designed by: a, DeepPrime without the co-installation of silent edits (SE0); b, PRIDICT without the co-installation of silent edits (SE0); c, DeepPrime with the co-installation of SE2 silent edits; and d, PRIDICT with the co-installation of SE2 silent edits. All conditions use PE6c, MLH1dn, the +104 ngRNA, and the -40 dsgRNA. For (a–d) data and error bars represent mean and standard deviation, respectively, collected from three independent biological replicates (shown as black dots). n.d., no data; this epegRNA was not cloned due to DNA synthesis errors. Source data

Extended Data Fig. 9. Scaffold integration and partial edit incorporation in prime edited cells.

a, Scaffold integration in prime edited primary airway epithelial cells from patients with CF, edited in Fig. 4a. The frequency of HTS reads containing CTT insertion and a defined number of scaffold-incorporated bases, indexed from the 3′ end of the pegRNA scaffold, is shown. b, Scaffold integration in prime edited 16HBEge-F508del cells. Scaffold integration frequencies are shown as described in (a) for both PEmax and PE6c edited cells. Data collected from Fig. 3d condition (+) epegRNA, (+) SE2, (- and +) PE6c, and (+) dsgRNA -40. c, Occurrences of all possible combinations of SE2 edit incorporation, both with and without concomitant CTT insertion, in PE-treated primary airway epithelial cells from Fig. 4a. d, Schematic showing the location of Cas9 PAM and protospacer edits within SE2. For (a–c), the epegRNA used consists of NGG2 PBS13 RTT41 SE2 with 2 × 3′ phosphonoacetate modifications. Source data

Extended Data Fig. 10. Analysis of off-target editing in primary airway epithelial cells from patients with CF.

a, Indel and substitution quantification at the top 32 human genomic sites identified by CIRCLE-seq for epegRNA NGG2 PBS13 RTT41 SE2. b, Indel and substitution quantification at the top 32 human genomic sites identified by CIRCLE-seq for ngRNA +104. c, Indel and substitution quantification at the top 32 human genomic sites identified by CIRCLE-seq for dsgRNA –40. For (a–c) data represents mean collected from three independent biological replicates (shown as black dots). Source data