Prime Editing in Mice with an Engineered pegRNA

Abstract

CRISPR editing involves double-strand breaks in DNA with attending insertions/deletions (indels) that may result in embryonic lethality in mice. The prime editing (PE) platform uses a prime editing guide RNA (pegRNA) and a Cas9 nickase fused to a modified reverse transcriptase to precisely introduce nucleotide substitutions or small indels without the unintended editing associated with DNA double-strand breaks. Recently, engineered pegRNAs (epegRNAs), with a 3’-extension that shields the primer-binding site of the pegRNA from nucleolytic attack, demonstrated superior activity over conventional pegRNAs in cultured cells. Here, we show the inability of three-component CRISPR or conventional PE to incorporate a nonsynonymous substitution in the Capn2 gene, expected to disrupt a phosphorylation site (S50A) in CAPN2. In contrast, an epegRNA with the same protospacer correctly installed the desired edit in two founder mice, as evidenced by robust genotyping assays for the detection of subtle nucleotide substitutions. Long-read sequencing demonstrated sequence fidelity around the edited site as well as top-ranked distal off-target sitesWestern blotting and histological analysis of lipopolysaccharide-treated lung tissue revealed a decrease in phosphorylation of CAPN2 and notable alleviation of inflammation, respectively.. These results demonstrate the first successful use of an epegRNA for germline transmission in an animal model and provide a solution to targeting essential developmental genes that otherwise may be challenging to edit.

Introduction

The ability to rapidly and precisely generate changes in the mouse genome has been democratized with the emergence of various editing platforms [1]. The clustered regularly interspaced short palindromic repeats (CRISPR), a three component editing system encompassing an endonuclease (Cas9), a programmable guide RNA, and an exogenous repair template, is often used to incorporate nonsynonymous substitutions in coding genes for purposes of elucidating the role of specific amino acids in protein function [2]. This platform is highly efficient at editing but produces mixed alleles during error-prone repair of double-strand breaks in DNA.

Prime editing (PE) is a two component-editing platform comprising a Cas9 nickase fused to a modified reverse transcriptase and a prime editing guide RNA (pegRNA) in which the 3’ end of the conventional guide RNA is extended with a reverse transcriptase repair template (RTT) and primer-binding site (PBS) [3]. The incorporation of desired edits in the 3’ extension of the pegRNA obviates the need for a repair template while the Cas9 nickase fusion minimizes unintended on/off-target indels [3]. The prime editors have efficiently introduced desired edits in zebrafish [4], drosophila [5], and mouse embryos [6, 7]. Recently, engineered pegRNAs (epegRNA), containing an RNA stabilizing motif at the 3’-terminus, were shown to confer higher editing efficiency in cells [8]. However, whether an epegRNA is effective in generating germline-transmitted edits in an animal model has yet to be reported.

Missense mutations in a protein can interfere with protein function. For example, mutations that disrupt protein phosphorylation may result in deregulation of a downstream signaling cascade [9]. Calpain-2 (CAPN2) is a ubiquitously expressed calcium-dependent cysteine protease [10] that is up-regulated in a variety of pathological disease states such as stroke, epilepsy, neurodegenerative diseases and aging [11–13]. Phosphorylation of CAPN2 at serine 50 is increased in vascular smooth muscle cells from human patients with pulmonary artery hypertension (PAH) as well as animal models of PAH. Further, the restricted loss of Capn2 in vascular smooth muscle halted the development and progression of PAH [14]. Here, we show the differential effectiveness of conventional three component CRISPR, prime editing, and engineered prime editing to substitute the Ser50 of Capn2 in mice.

Materials and Methods

Mouse models

The CRISPOR tool [15] was used to define a single guide RNA (sgRNA) close to the sequence of Capn2 to be edited. The sgRNA (GGATGACGGCAGGGCGGGGA) was in vitro assembled (below) and tested with purified Cas9 protein (Synthego) for in vitro activity using a PCR-based assay. Next, a single strand oligodeoxynucleotide (ssODN; IDT) was designed to carry substitutions that would disrupt the PAM sequence and generate the Ser50Ala replacement in the CAPN2 protein. The sequence of the ssODN was GGAGTAGGGCCCCAACTCCTTATAGCCCAAGGATGACGGCAGGGCGGGGAAAGCAGGATCCTGGAAGAGCGCCCCGGCCTCCAGGCACTCGTTCCGCAGCG (substitutions in bold red). The sgRNA (25 ng/μl) and Cas9 protein (50 ng/μl) were preincubated for 10 min in nuclease-free micro-injection buffer (M2, Sigma) and then combined with the ssODN (50 ng/μl) for injection into fertilized oocytes of strain C57BL/6J mice. Viable two-cell stage embryos were transferred to pseudopregnant ICR mice and founder pups genotyped either with allele-specific primers [16] or a restriction fragment length polymorphism (RFLP)-like assay. The allele-specific primers were 5’-CTACGAGACGCTGCGGAACGAGTGCCTGGAGGCCGGGGCGCTCTTCCAGGATCCTTCC-3’ (forward wild-type allele), 5’-GCGCTCTTCCAGGATCCTGCT-3’ (forward mutant allele), and 5’-CAATCGCGGAGCACTACTAA-3’ (common reverse). For RFLP-like assay, primers flanking the edited window were designed (IDT) and used to PCR genomic DNA that was subsequently digested with HpyAV (cuts CCTTCN₆ or N5GAAGG) generating bands of 81, 88, and 117 bp if wild-type sequence or 198 and 88 bp, if mutant. All PCR products were resolved in a 2.5% agarose gel and Sanger sequenced to confirm the fidelity of editing. Positively identified founder mice were bred to C57BL/6J mice for germline transmission and subsequent heterozygous inter-crossing to generate homozygous mutant mice.

For prime editing, the same protospacer sequence as above was used with an extended 3’ tracrRNA containing 10 and 15 nucleotides of RTT and PBS, respectively. The terminal 3’ end of the pegRNA contained UUU. The full sequence of the pegRNA was GGATGACGGCAGGGCGGGGAgttttagagctagaaatagcaagttaaaataaggctagtccgttatcaacttgaaaaagtggcaccgagtcggtgcaGGATCCTgCtTTCCCCGCCCTGCCGUUU (protospacer; RTT with two substitutions as above [lowercase], and PBS). The PE2 prime editor was prepared as described previously [7] and combined with the pegRNA for microinjection into fertilized oocytes and all subsequent steps were followed as described above.

For engineered pegRNA (epegRNA), we used the same pegRNA as above with addition of a degradation resistant pseudoknot at the 3′ end of the pegRNA plus modified prequeosine 1-1 riboswitch aptamer (evopreQ1; gift of Dr. David Liu) [8]. The same PE2 editor was used as above and the epegRNA and PE2 mRNA were microinjected into fertilized oocytes as above.

All guide RNAs were generated in-house through conventional cloning in Bsa1 digested Puc57-sgRNA expression vector (Addgene #51132); clones were verified by Sanger sequencing and then in vitro-transcribed using MEGAshortscript^™ Kit (Ambion, #AM1354) according to the manufacturer instructions. The PE2 editor was generated with the pCMV-PE2 (Addgene #132775) construct which was in vitro-transcribed using mMESSAGE mMACHINE^® T7 Ultra Kit (Ambion, #AM1345), as described previously [7].

Amplicon sequencing with Nanopore platform

PCR conditions: step 1, 95°C for 5 min; step 2, 95 °C for 30s, 55 °C for 30s, 72 °C for 40s for 30 cycles; step 3, 72 °C for 10 min. Amplicon sizes ranged from 305- to 513- bp. Amplicons were isolated by Qiagen QIAquick PCR purification kits (cat#QIA28106). To query the CRISPR-targeted locus and the ten off-target CRISPOR/ChopChop prediction loci, these eleven PCR amplicons from the brain, heart, and intestine from two founder mice (66 total amplicons) were barcoded with ONT-kits EXP-NBD196 and SQK-LSK109 and pooled to run on R9.4.1 flow cells with the ONT GridION platform following manufacturer’s instructions (www.nanoporetech.com). ONT software guppy (v4.2.2) converted fast5 files to fastq on MinKNOW (v20.10.3) MinKNOW Core (v4.1.2) with a fast base-calling option and minimum Q-score of 7 for read filtering. Data were demultiplexed by the ONT-MinKNOW software and then aligned to a custom reference using Qiagen CLC Genomics Workbench (www.qiagen.com) with Long-Read Support (beta) plugin that utilizes minimap2 [17] with default parameters (www.digitalinsights.qiagen.com). Reference sequences comprising the CRISPR targeted locus and the top-ten CRISPOR (http://crispor.tefor.net/) and ChopChop (https://chopchop.cbu.uib.no/) predicted CRISPR off-targets are listed in Supplementary Table 2. Library read counts ranged from ~8,000 to 415,000 (Suppl. Table 2). Sequence data can be found under BioProject number PRJNA979143 NCBI SRA (www.ncbi.nlm.nih.gov/sra).

Western blotting

Animal experiments were conducted under an approved institutional animal protocol. Mice were injected intraperitoneally with PBS (control) or 12.5 mg/kg lipopolysaccharide (LPS) for 16 hrs. Sections of lung (n=3 mice per arm of experiment) wer rapidly dissected and processed for hematoxylin and eosin staining. Samples of lung homogenates (10-20 μg of protein) were denatured and electrophoresed on SDS-PAGE gel. Separated proteins were electrotransferred to nitrocellulose membranes. The membranes were incubated with 5% fat-free milk or 5% BSA for 1 h, and then incubated with antibodies against phospho-calpain 2 (Ser50) (1:300 dilution, PAS-105976, Invitrogen) and total calpain 2 (1:1000 dilution, SC-373966, Santa Cruz) overnight at 4°C. The membrane was then washed with 0.1% Tween-20, 20 mM Tris-HCl (pH7.5) and 150 mM NaCl (TTBS) three times for 10 min. Secondary antibody IgG conjugated to alkaline phosphatase was incubated with the membranes at room temperature for 1 h and signals were detected by chemiluminescence (ECL kit, BioRad).

Results

The first exon of mouse Capn2 harbors the Ser50 residue that was targeted for editing (Fig, 1A). An initial attempt to generate a mouse with a Ser50Ala substitution utilized conventional three component CRISPR (Fig, 1Bi) with allele-specific primers [2] to distinguish correct editing from the wild-type sequence (Fig. 1Ci, 1Bi). A total of 317 zygotes were injected of which 237 (75%) 2-cell stage embryos were transferred to 7 surrogate Swiss Webster mice. The results revealed no live born pups. C-sectioning of several visibly pregnant mice revealed 33 fetuses, only 8 of which survived for 15 minutes ex utero. We genotyped these 8 fetuses and found no evidence of editing (data not shown). Collectively, these results suggested that sgRNA-mediated dsDNA breaks in the Capn2 gene resulted in embryonic arrest, consistent with the known embryonic lethality of global Capn2 knockout mice [18].

Figure 1. Strategy, design, and genotyping of S50A Capn2 mice.

A) UCSC Genome Browser screenshot of Capn2 locus with intended editing of nucleotides (in red) that destroy the PAM sequence (yellow CCT highlight) and generate the S50A substitution in exon 1 (blue highlighted sequence is the spacer). B) Different platforms used to generate the S50A Capn2 mouse; Bi), 3-Component CRISPR; Bii) pegRNA and prime editor (PE2); Biii) engineered prime editor (ePE2). C) Genotyping methods of the founder mice. Ci) Allele-specific PCR-based genotyping as indicated in (i, ii). Ciii) RFLP-like assay with HpyAV restriction sites shown. D) Allele-specific-based genotyping of founder mice from PE2. E) Allele-specific-based genotyping of founder mice from ePE2. F) RFLP-like assay confirming the founder E5 from ePE2. * Represents non-specific PCR bands. Pink object in panels Bii and Biii represents endogenous nuclease. Biorender was used for graphical illustration.

We next turned to the prime editing platform [3] (Fig. 1Bii) and engineered a pegRNA with the same protospacer sequence used in conventional three component CRISPR. A total of 250 zygotes were injected and 218 (87%) 2-cell stage embryos were transferred to 5 recipient female mice. A total of 47 pups were born, none of which exhibited the expected genotype based on an allele-specific PCR assay (Fig, 1Ci–ii, 1D). Next, we injected the same concentration of PE2 editor and an engineered pegRNA (Fig. 1Biii), shown recently to confer higher efficiencies in editing [8]; the protospacer of the epegRNA was the same as above. A total of 250 zygotes were injected and 219 (88%) 2-cell stage embryos were transferred to 5 recipient female mice. A total of 33 pups were born, two of which showed correct editing using both an allele-specific PCR assay (Fig. 1Ci–ii, 1E) and an RFLP-like PCR assay (Fig. 1Ciii, 1F).

The two founder pups that genotyped positive for the expected edit were bred to C57BL/6J mice for germline transmission. F₂ mice underwent Sanger sequencing and the results showed correct installation of the nucleotide substitutions in Capn2 (Fig. 2A). Amplicon sequencing of brain, heart, and intestinal tissues showed high on-target sequence fidelity around the targeted site (Fig. 2B–2D). The top 10 off-target sites were interrogated by long-read sequencing and the results revealed no evidence of off-targeting events in brain, heart and intestinal tissues (Suppl. Table 1).

Figure 2. Validation of S50A substitution.

A) Trace viewer sequences from wild type (WT) (i) and homozygous S50A (ii) mice showing the correct installation (TCC to GCT). B) Western blot analysis of lung lysate from LPS-treated mice shows a significant reduction of phosphorylated CAPN2 (i) with quantification (ii). C) On-target sequence fidelity assessed by amplicon sequencing of both + and − strands of DNA with the Nanopore sequencer, showing correct edit without additional indels within the surrounding 500 bp. D) Zoom-in of the violet rectangle in panel C shows the correct TCC to GCT edits. E) A summary table of percentage edits for the respective substitution in three different tissues (brain, heart, intestine) in each of the two positive founders (E5 and G12). F-G) Histomicrographs of lung tissue isolated from WT mouse (F) and homozygous S50A mouse (G) following 16 hrs of lipopolysaccharide injection (12.5 mg/kg, i.p.). Note the substantial reduction in inflammatory cell infiltration in the S50A mutant mouse. Scale bar is 100μm.

Heterozygous intercrossing resulted in homozygous Capn2^S50A mice that were viable with no overt phenotype. Isolated lung tissue from mice treated with LPS showed an attenuation in phosphorylated CAPN2 protein in Capn2^S50A suggesting the Ser50Ala substitution reduced agonist-induced phosphorylation of the protein (Fig. 2Bi, ii). Previous research has identified Ser50 CAPN2 is involved in PAH-mediated remodeling [14] To further investigate its potential role in these processes, we assessed the effect of Capn2^S50A on inflammatory cell infiltration in lung tissue. Mice treated with LPS, a well-known inflammatory mediator, exhibited significantly reduced leukocyte infiltration in their lung tissues compared to control mice Fig 2F–G. While these results are promising, further investigation will dissect the precise signaling pathways involved and explore the potential therapeutic efficacy of targeting Capn2^S50A in pre-clinical models of acute lung inflammation and PAH.

Discussion

Our findings demonstrate the inability of the conventional CRISPR-Cas9 and prime editor to yield any positive edited mice. On the other hand, an engineered prime editing guide RNA directed the Ser50Ala substitution in CAPN2 and provided a novel mouse model to study the loss of Ser50 phosphorylation of CAPN2 in its native context.

The genome editing toolbox has expanded considerably since the first reporting of genetic modification in the mouse using CRISPR-Cas9 [1, 3, 19–22]. The original CRISPR-Cas9 system of genome editing repurposed a natural bacterial system of adaptive immunity to create programmable double-strand breaks in DNA [23, 24]. In its simplest form, CRISPR-Cas9 is a two component system comprising a single guide RNA and the bacterial Cas9 protein [23]. Upon imperfect repair of the double-strand break, frameshifts in the targeted coding DNA can occur resulting in the inactivation of a protein-coding gene. This paved the way for a new, facile method of generating genetically-modified mouse models [25]. More precise edits were made possible upon introduction of a third CRISPR component, namely a DNA repair template that can bias the endogenous repair pathway to install user-defined substitutions, insertions, or deletions in the targeted sequence [2, 26]. Here, conventional three component CRISPR failed to introduce viable mice carrying the two nucleotide substitution at codon 50 of the Capn2 gene. In fact, very few mice were viable during embryogenesis and of those genotyped, none showed evidence of editing. The simplest explanation for this finding is that CRISPR-Cas9 introduced double-strand breaks resulting in embryonic lethality, a finding consistent with previous studies of Capn2 gene targeting [18]. Surprisingly, the original prime editing platform (PE2), with which we had previous success in the mouse [7], while yielding comparatively more live-born pups than CRISPR-Cas9, was also unsuccessful in generating mice with the desired Capn2 edit; the increase in yield of pups was likely a result of the nickase function of PE2. It is possible that injecting more embryos with the PE2 system may have revealed positive editing. Interestingly, an epegRNA with the same protospacer as CRISPR-Cas9 and PE2, yielded two pups with the desired S50A substitution. These results, while representing just one locus, support the original in vitro findings of superior editing efficiency with an epegRNA design [7].

Phosphorylation of CAPN2 at Ser50 by extracellular signal-regulated kinases (ERK1/2) accompanied by CAPN2 activation is involved in pulmonary vascular remodeling induced by pulmonary arterial hypertension (PAH) [14] and epidermal growth factor (EGF)-induced motility in fibroblasts [27]. Accordingly, we tested whether the Ser50Ala substitution minimized CAPN2 activation as measured by agonist-induced phosphorylation. Indeed, Western blotting showed a sharp reduction in phosphorylated CAPN2 induced by LPS, a major toxin of gram-negative bacteria that is commonly used to induce acute lung injury in rodents [28, 29]. Concomitant with the marked reduction of leukocyte infiltration, these findings suggest that phosphorylation of CAPN2 at Ser50 may be involved in LPS-induced acute lung injury. Further, our newly generated CAPN2 Ser50Ala knockin mice will be of utility for evaluating the role of activated CAPN2 in lung and vascular pathobiology.

An acknowledged limitation in this short communication is the reporting of epegRNA efficiency at a single locus with a single epegRNA design. Nevertheless, the results not only provide the first demonstration of an epegRNA design with a positive outcome in an animal model, they highlight a useful alternative to conventional CRISPR editing to avoid embryonic lethality when editing essential developmental genes such as Capn2 [18]. Future comparative studies, such as those described here, targeting other gene loci in animals could add further support of the superior nature of epegRNAs in vivo. We suggest that the epegRNA method represents a relatively simple method of generating genetically-modified mouse models.