De novo PAM generation to reach initially inaccessible target sites for base editing

Abstract

Base editing by CRISPR crucially depends on the presence of a protospacer adjacent motif (PAM) at the correct distance from the editing site. Here, we present and validate an efficient one-shot approach termed 'inception' that expands the editing range. This is achieved by sequential, combinatorial base editing: de novo generated synonymous, non-synonymous or intronic PAM sites facilitate subsequent base editing at nucleotide positions that were initially inaccessible, further opening the targeting range of highly precise editing approaches. We demonstrate the applicability of the inception concept in medaka (Oryzias latipes) in three settings: loss of function, by introducing a pre-termination STOP codon in the open reading frame of oca2; locally confined multi-codon changes to generate allelic variants with different phenotypic severity in kcnh6a; and the removal of a splice acceptor site by targeting intronic sequences of rx3. Using sequentially acting base editors in the described combinatorial approach expands the number of accessible target sites by 65% on average. This allows the use of well-established tools with NGG PAM recognition for the establishment of thus far unreachable disease models, for hypomorphic allele studies and for efficient targeted mechanistic investigations in a precise and predictable manner.

Keywords: De novo PAM; Base editing; CRISPR; Medaka; Targeted mutagenesis.

Fig. 1.

Inception increases the base editing scope through de novo PAM generation. (A) Schematic representation of the one-shot inception approach (simultaneous application of two base editors and two guide RNAs). If the adenosine(s) of NGA, NAA and NAG motifs are contained within the canonical base editing window, an A-to-G edit event leads to the generation of (up to six) new PAM(s) (green, step 1), rendering a new guide RNA target site available for a second base editing event (step 2) 27-36 nucleotides upstream of the canonical PAM (orange). As the second base editing relies on the first, the inception approach depicts a locally confined sequential editing event. (B) Abundance analysis of canonical base editor target sites with NGG PAM (orange) and target site increase upon inception approach (green) normalized to gene locus length. Base editor target sites contain A or C nucleotide(s) in the respective editing window. Comparison of the top ten studied human genes (AKT1, APOE, EGFR, ESR1, IL6, MTHFR, TGFBI, TNF, TP53 and VEGFA) and their orthologs in commonly used model organisms (Table S1). Across all loci and organisms, inception increases accessible target sites by 64.8±6.0%. Boxplot shows median with boundaries representing the 25th and 75th percentiles. Whiskers extend to a maximum of 1.5 times the interquartile range. PAM, protospacer adjacent motif.

Fig. 2.

Inception efficiently introduces loss-of-pigmentation mutations in oca2. (A) Loss-of-function mutations via inception at the oca2 locus by synonymous de novo PAM generation in step 1 (ABE8e, CAG>CGG=p.A337A) and subsequent non-synonymous editing (ancBE4max, CAG>TAG=p.Q333* and ACC>ATT=p.T332I) to introduce non-synonymous codon changes, including a pre-termination STOP codon (PTC) at p.T332 and p.Q333 in step 2. Besides anticipated edits (black arrows), potential further edits (white arrows) may occur due to combined injection of A-to-G and C-to-T base editors (compare with Fig. S1). (B) At 4.5 days post-fertilization, full pigmentation of wild-type Oryzias latipes eyes (left) is lost in oca2 inception editants (right), as quantified by analysis of mean gray values per eye. The oca2-inception injection mix contained ABE8e and ancBE4max mRNA, and oca2-step1 and oca2-step2 guide RNAs. Each control mix lacked one component and injections did not show spurious activity. In the case of the ABE control, mild effects through editing T332A matched our previous report (Cornean et al., 2022), compare with Fig. S2. Boxplot shows median with boundaries representing the 25th and 75th percentiles. Whiskers extend to a maximum of 1.5 times the interquartile range. Scale bar: 200 μm. (C) Illumina amplicon sequencing reveals a high abundance of anticipated edits. The nucleotide abundance is displayed as the mean±s.d. of three replicates (pools of eight phenotypic embryos each; 246,045 reads total; Fig. S2). (D) Frequency analysis of resulting alleles (translated) highlights anticipated edits to reach 48.4±6.7% abundance. Analysis is based on CRISPResso2 Alleles frequency table with cut-off at >0.2% Illumina sequence read abundance per replicate (Fig. S3). AA pos, amino acid position; nt, nucleotide; PAM, protospacer adjacent motif; WT, wild type. Orange indicates canonical PAM; green indicates de novo PAM.

Fig. 3.

Inception efficiently introduces locally confined multi-codon changes in kcnh6a. (A) Local multi-codon editing via inception (both non-synonymous; anticipated edits are indicated by black arrows) at the kcnh6a locus to create hypomorphic alleles with ABE8e and two guide RNAs [kcnh6a-step1 and sequence adjusted (red outlined box) kcnh6a-step2-adjusted]. (B) At 4 days post-fertilization, the two-chambered embryonic wild-type Oryzias latipes heart displays strong diastole/systole periodicity [regularly patterned kymograph; the line used spans the ventricle (v) and atrium (a), indicated in left panel]. kcnh6a-inception editants displayed heart phenotypes (26.9%), including heart morphology, reduced ventricular contractility and the exemplary displayed 2:1 atrioventricular block [white brackets in kymograph (line indicated in left panel; Movie 1)]. Scale bar: 200 μm. Control injections were wild type-like (barplot; data are mean±s.d. of three replicates, n=number of embryos). (C) Illumina amplicon sequencing reveals high abundance of anticipated edits. The nucleotide abundance is displayed as the mean±s.d. of three replicates (pools of five to ten embryos each; 164,108 reads total; Fig. S6). (D) Frequency analysis of resulting alleles (translated) comparing the anticipated single, double and triple codon changes. Analysis based on CRISPResso2 Alleles frequency table with cut-off at >0.2% Illumina sequence read abundance per replicate (Fig. S7). AA pos, amino acid position; nt, nucleotide; PAM, protospacer adjacent motif; WT, wild type. Red box indicates A>G adjustment in kcnh6a-step2-adjusted guide RNA; orange indicates canonical PAM; green indicates de novo PAM.

Fig. 4.

Efficient splice site acceptor removal via inception in rx3. (A) A splice acceptor site mutation (CAG>CAA, black outlined box) in rx3 via intronic de novo PAM generation (rx3-step1 guide RNA, ABE8e; GAG>GGG) and subsequent C-to-T splice acceptor site editing [sequence adjusted rx3-step2-adjusted guide RNA (red outlined box), evoBE4max; CAG>CAA]. (B) At stage 26, Oryzias latipes eye development is apparent in wild-type embryos (left), but drastically affected in rx3-inception editants (right, 23.9±13.4%) and rarely found in control injections (barplot; data are mean±s.d. of three replicates; n=number of embryos). Scale bar: 200 μm. (C) Illumina amplicon sequencing reveals the high rates of the anticipated splice acceptor mutation (32.0±6.1%). The nucleotide abundance is displayed as the mean±sd of three replicates (pools of five to nine embryos each; 136,642 reads total; Fig. S6). AA pos, amino acid position; nt, nucleotide; PAM, protospacer adjacent motif; SA, splice acceptor; WT, wild type. Red box indicates sequence adjustment in rx3-step2-adjusted guide RNA; orange indicates canonical PAM; green indicates de novo PAM.