Correction of the NIPBL c.5483G > A mutation in a HEK293-CdLS cell model
To set up the correction strategy for the c.5483G > A mutation we generated HEK293 cell clones carrying the NIPBL coding sequence either wild-type or mutated. The NIBPL cDNA comprising exons 27–30 carrying the c.5483G > A mutation in exon 29, was stably integrated into HEK293 cells and two clones were isolated (HEK293/CdLS-cl1 and HEK293/CdLS-cl2). Sanger sequencing confirmed the presence of the c.5483G > A mutation in both clones (Fig. 1A, Supplementary Fig. 1A).
We initially tested the correction of the c.5483G > A mutation by using CRISPR base-editors, which have been developed to modify genomes in the absence of DSB [17, 36, 37]. We analyzed the protospacer adjacent motive (PAM) sequences 30bp downstream from the mutated A, to select the best base-editor candidates to induce specific A to G transition. The PAM search was performed by taking into consideration the deaminase editing window which has been reported ranging between specific nucleotide positions with respect to the PAM17. We found no optimal PAM sequences (-NGG) for Streptococcus Pyogenes Cas9 (SpCas9), while we found a PAM (-NNGRRT) for Staphylococcus Aureus Cas9 with a compatible deaminase activity window to specifically modify the mutated A in NIPBL . We thus designed a sgRNA (gRNA + 9) targeting a nickase SaCas9 adenine base-editor (ABEsa). We evaluated diverse versions of adenine deaminases (ABEmax, ABE8 and ABE8.20m[38–40]) combined with nickase SaCas9 and gRNA + 9, by measuring the A to G transition. The mutated A nucleotide in position 12 with respect to the PAM (A12), which was located in the optimal predicted position for A to G transition, was minimally modified (almost 5% with the most efficient ABEmax), while we detected higher modifications (up to 14,5%) of a non-target A in position 5 (A5) from the PAM (Fig. 1B).
Since the base-editing approach did not produce substantial A to G reversion and bystander modification was significant on a non-target nucleotide (A5), we then tested the most recent CRISPR technology, prime-editing (PE), which similarly to base-editing allows to modify the genome without DSBs. To this aim we designed four prime editing guide RNAs (pegRNAs, Supplementary Table 1) to apply both the PE2 and PE3 strategy as described by Anzalone et al. Even though the editing efficacy was higher than the one achieved with ABEs (up to 10.5% using the PE3 approach, Fig. 1C), the overall efficacy of c.5483G > A correction was not compatible with an application in primary cells which requires superior editing efficacy.
We then turned to the gene substitution approach using HDR induced by CRISPR-Cas9 nuclease activity in combination with a donor DNA sequence. To identify the most efficient strategy we tested a variety of Cas9 orthologs having compatible PAMs which should be sufficiently close to the targeted mutation, ideally less then 10 bp distant[42, 43]. We found PAMs usable with SpCas9-NG, SpCas9-VQR, enAsCas12a and SaCas9 (Supplementary Fig. 1B) and compared their editing efficiency through formation of small insertions and deletions (InDels) in the HEK293/CdLS clones. The SpCas9-NG and Sp-Cas9 in combination with gRNA + 1 and gRNA + 4 targeting sequences near the c.5483G > A mutation respectively showed the highest editing rates with up to 42.5% InDels (Supplementary Fig. 1B-C and Supplementary Table 1). As donor we used a single strand oligonucleotide (ssODN-CdLS) carrying the correct NIPBL sequence and two additional silent mutations located in the seed region of the complementary sgRNA to prevent recutting after correction (Fig, 1A and Supplementary Table 1). We initially attempted gene substitution through HDR by transfecting both the donor ssODN-CdLS and plasmids expressing the Cas9 nucleases. The HDR analysis revealed a higher correction efficiency by using SpCas9 in comparison to SpCas9-NG (7.2% and 2.5% respectively), while the amounts of InDels generated by non-homologous end-joining (NHEJ) were 49.2% and 47.8% respectively (Supplementary Fig. 1D).
Since HDR efficiency is improved by ribonucleoprotein (RNP) delivery of CRISPR-Cas, we electroporated a high-fidelity version of SpCas9 recombinant protein, SpHiFiCas9, along with a chemically synthesized gRNA. Strikingly, compared with plasmid transfection we obtained 2.6-fold improvement in HDR (18.7%) (Fig. 1D and E); as expected the InDels produced by NHEJ are higher (around four folds) than specific sequence substitution (Fig. 1D and E). To further enhance HDR, we tested NU7441, a compound that by blocking the NHEJ pathway through inhibition of the DNA-PK, favors HDR repair. The HDR was further enhanced by the NU7441 resulting in 43.8% sequence substitution, thus at least two folds more than untreated cells, while InDels generated by NHEJ decreased at similar levels as HDR (Fig. 1D and E).
Overall, these results suggest that the HDR strategy with SpHiFiCas9 and gRNA + 4 delivered as RNP together with ssODN-CdLS and the NU7441 treatment, is the most efficient method to correct the c.5483G > A mutation in the NIPBL gene in a HEK293-CdLS cellular models.
Correction of the NIPBL c.5483G > A mutation in patient-derived hiPSCs
Patient-derived hiPSCs are extensively used for disease modeling, drug screenings and somatic cell therapy[16, 47, 48]. Therefore, we generated hiPSCs from a CdLS patient carrying a c.5483G > A mutation in the NIPBL gene and then corrected the locus to generate isogenic wild-type and mutated cells. Gene correction was performed by electroporation of the mutated hiPSCs (hiPSCs-c.5483G > A) with SpHiFiCas9 and gRNA + 4 RNPs together with ssODN-CdLS and treated or not with NU7441. Editing efficiency in the bulk population was assessed after 3, 7 and 10 days, obtaining up to 16.4% of editing efficiency without NU7441, and up to 30.8% of editing efficiency with NU7441 after 10 days (Fig. 2A). Interestingly, HDR/InDels ratio increased with time, from 0.53 at day 3 to 1.9 and day 10 without NU7448, and from 1.1 at day 3 to 2.87 at day 10 with NU7448 (Fig. 2B).
To generate monoclonal edited derivatives, fourteen days after electroporation cells were sorted by flow cytometry using forward scatter and side scatter as parameters for the sorting. Single clones were expanded and Sanger sequencing analysis confirmed the presence of three fully corrected clones (hiPSCs-cl1/cl2/cl3) (Fig. 2C and Supplementary Fig. 2A).
Characterization of edited hiPSC clones
Fully corrected and unmodified hiPSCs were expanded and analyzed for expression of pluripotency-associated markers. Immunofluorescence analysis confirms the expression of endogenous pluripotency markers, including NANOG and OCT4 in all three corrected clones and in control non-edited cells (Fig. 3A and Supplementary Fig. 2B). Moreover, we detected by flow cytometry high levels of surface stem cell markers, including EpCAM, TRA-1-81 and SSEA-4 further confirming the staminal status of the cells preserved during the editing treatments and clonal selection (Fig. 3B and Supplementary Fig. 2C).
To functionally evaluate the pluripotency competence of the corrected hiPSCs, we performed an Embryoid Body (EB) assay and checked for the expression of germ layers markers 14–21 days following their formation. Specifically, we found the presence of cells positive for βIII-Tubulin (ectodermal marker), αSMA (mesodermal marker) and GATA4 (endodermal marker), thus indicating the stemness of the hiPSCs (Fig. 5). The pluripotency potential was further quantitatively confirmed by assessing the level of expression of NESTIN (ectoderm), αSMA (mesoderm) and AFP (endoderm) transcripts (Fig. 3C-D, Supplementary Fig. 3A-D). The marker profiles of the hiPSCs generated in this study were controlled using commercially available hiPSCs (see Material and Methods).
Editing precision and genomic integrity of hiPSCs corrected via CRISPR-Cas9 technology
To verify whether major genomic alterations may have occurred during editing of NIPBL locus and the expansion of the clones, DNA-seq of the entire genome was performed through shallow Whole Genome Sequencing (sWGS) in both hiPSCs-c.5483G > A and edited hiPSC clones. No substantial abnormalities could be observed in the karyotype, as well as in the copy number profile, between the hiPSCs-c.5483G > A and the edited clones (Fig. 4A-D and Supplementary Fig. 4A-D).
The potential off-target generated by the SpHiFiCas9-gRNA + 4 cleavages was assessed through the in silico analysis (Cas-OFFinder) and a genome wide assay, the GUIDE-seq method. The off-target prediction performed through the Cas-OFFinder software revealed 188 potential off-targets from 1 to 4 mismatches (Supplementary Table 2). The experimental off-target genome-wide analysis was performed through GUIDE-seq in HEK293 treated with SpCas9-gRNA + 4 and showed mainly on-target cleavages and near background levels of unpredicted cuts (5 sites) (Fig. 4E). The five unpredicted cleaved sites showed near back-ground cleavages (sequence reads below 10) apart for one that even though appeared with much less sequence reads than the on target, was higher than the other sites (62 sequence reads) (Fig. 4E). To verify the potential modification of the most represented off-target site we performed Sanger sequencing in the corrected hiPSC clones. The sequencing results showed that the potential off-target is not altered in the modified hiPSCs thus indicating a precise editing protocol (Fig. 4E-F and Supplementary Fig. 4E-F).