CRISPR-Cas9 genome editing can cause large chromosome breaks

To confirm this hypothesis, we next hybridized the cells with one Chr10q sub-telomeric orange probe (Supplementary Figure 3a). We confirmed the complex pattern of Chr10 in HEK293T as revealed by a large analysis of a higher percentage of cells with only two Chr10q probes compared to non-transfected cells (9% in NT versus 22.6% and 20.1% in cells transfected with nuclease only or nuclease with ssODN, respectively, Supplementary Figure 4a). As previously, we then focused on HEK293T cells apparently trisomic for Chr10. To do so, we used another couple of probes: sub-telomeric probes at Chr10p-arm (labeled in green) and another one at Chr10q-arm (labeled in orange) (Fig. 3b) and focused on cells with a FISH profile with three green signals (3 G). The percentage of cells with a 3 O/3 G profile significantly decreased from 86% to 71.5 and 75% for nuclease without or with ssODN respectively in favor of cells with only two or one orange signals (from 14 to 28.5% and 25% for nuclease without or with ssODN respectively, p < 0.01). The use of sub-telomeric probes therefore allowed us to confirm that Chr10q truncation occurred in about 10% of HEK293T cells transfected with nuclease, without or with ssODN, resulting in a 7.5 Mb loss downstream of UROS (Fig. 3b). To check whether megabase-scale deletions could also occur upstream of UROS DSB, we compared the 3 O/2 G signal frequency between HEK293T cells non-transfected or transfected with nuclease. We did not observe any difference in 3 O/2 G signal frequency in two independent experiments (3.7 vs. 3.4% in first experiment and 7.6 vs. 8.5% in the second one respectively for NT cells and transfected with nuclease), suggesting that large deletions are mostly unidirectional downstream of UROS DSB.

To confirm the loss of 10q arm extremity and map the break point, we picked up 10 single clones of HEK293T cells previously transfected with nuclease. Using UROS-framing probes, FISH analysis revealed that 3 out of 10 clones lost one distal probe. We then performed high resolution array-Comparative Genomic Hybridization (CGH) (60 kb resolution) on two positive clones. Results demonstrated chromosomal truncation in both clones starting exactly from UROS DSB until the last probe near 10q chromosome telomere (50 kb from telomere) (Fig. 3c). Moreover, we observed in clone #8, upstream of UROS, a 31 Mb duplication with 50% of mosaicism. These data highlight the high genomic instability at the DNA break point with megabase–scale deletion or gain.

Altogether, these results strongly suggest that the editing strategy using a single DSB in one chromosome may induce megabase-scale chromosomal damage. To explore whether a similar effect occurs in other cell types, we transfected the K-562 hematopoietic cell line with nuclease and analysed outcomes using UROS-framing probes. Similarly, we found that nuclease transfection induced a significant increase in cells with only one signal of the probe distal to UROS (+ 11%), suggesting that chromosomal terminal deletion induced by nuclease-mediated DSB is not cell type-dependent (Supplementary Figure 4b).

HEK293T and K562 cell lines have a very complex and unstable karyotype and are deficient in p53 activity. To explore whether chromosome truncation could occur in primary cells, we repeated same experiments in primary human foreskin fibroblasts (hFF). Deep analysis of FISH assay using UROS-framing probes fortunately revealed that the percentage of 1 O/2 G cells does not increase after nuclease in fresh and hTERT (human telomerase reverse transcriptase) immortalized fibroblasts (Fig. 3d). Because FISH limit of detection is about 1%, we cannot exclude rare terminal truncations after nuclease transfection. We decided to use hTERT immortalized fibroblasts invalidated or not for the TP53 gene. In contrast to hTERT immortalization alone, TP53 knockout dramatically increased the risk, up to 10-fold (Fig. 3d). We demonstrate the strong involvement of p53 in Cas9 nuclease -induced chromosomal instability and highlight CRISPR-Cas9 nuclease genotoxicity in primary cells with inactivated p53.

We hypothesized that the ssODN template concentration was a critical limiting factor for HDR efficiency. For this reason, we carried out ssODN dose-scaling and found that HDR frequency reached 5.1% at the highest concentration (2-fold higher with 5 µM vs 0.5 µM, p < 0.01, Supplementary Figure 6a). We then retained 5 µM as the optimal ssODN concentration for further experiments. To evaluate whether a shorter ssODN would facilitate ssODN transfection, we designed an 80nt-ssODN template. The HDR rate was similar but not improved (Supplementary Figure 6b). To monitor ssODN transfection efficacy, we used a ssODN coupled to an Alexa-647 fluorochrome (80nt-ssODN-A647). Surprisingly, we observed an unexpected increase in HDR rate without sorting A647-positive cells (+382%, compared to non-optimized condition, Supplementary Figure 6b).

We hypothesized that fluorochrome could stabilize the ssODN template and improve HDR. To confirm this hypothesis, we added 3 locked nucleic acid (LNA) bound in 5’ of the same 80nt-ssODN template and observed a similar increase in the HDR rate to the 5’ Alexa-647 chemical graft (+325%, compared to non-optimized condition, Supplementary Figure 6b). With all these improvements (high concentration of stabilized ssODN), TIDER analysis validated the high frequency of SacI (12.2% of alleles) and with a very low level of indels (1.2%) (Supplementary Figure 2c). NGS confirmed the 18% HDR rate and, most importantly, still with a minimal rate of indels (0.7%, Fig. 4d). Moreover, due to the minimal indels, other alleles were not modified (80.8% are WT vs 1.3% with nuclease).

To test whether single nickase could be used to edit other cell types, we transfected the K-562 cell line with the same optimized tools targeting UROS exon 4. We obtained a high rate of HDR (9.47 ± 2%, n = 6) measured by RFLP and confirmed by NGS analysis (8.5%, Supplementary Figure 7a, b). Again, a high HDR/indels ratio up to 14.1 with only 0.6% of indels and no UROS dysfunction (absence of fluorocytes, Supplementary Figure 7c) were observed. To confirm that this was not specific to exon 4, we targeted exon 10 of UROS near to the frequent c.683 C > T point mutation (p.Thr228Met)¹⁸ and transfected a 75nt-ssODN-A647 template together with the nickase (Supplementary Figure 8a). Similarly, we confirmed the editing by RFLP (Supplementary Figure 8b) without any UROS dysfunction (absence of fluorocytes detected by flow cytometry (Supplementary Figure 8c)). These data demonstrate that the nickase strategy can be applied to other cell types and loci. Moreover, unlike nuclease, nickase allows editing and leaves the other alleles non-edited. The absence of indels made it possible to perform iterative editing to reach higher HDR rates. With successive transfections, we dramatically improved the HDR rate at two loci and in different cell types (Supplementary Figure 9).

To explore whether SSB can also avoid karyotype abnormalities, we first performed DNA-FISH using a centromeric probe. As expected, no chromosomal loss was observed (Supplementary Figure 3b). Then we used the two UROS-framing probes. To be highlighted, after nickase and ssODN template transfection, the percentage of cells with 2 O/3 G signal was significantly lower than with nuclease ssODN (5 and 17% respectively, Fig. 3a, p < 0.01) and similar to non-transfected cells. Sub-telomeric 10q FISH analysis of transfected cells with nickase showed significantly fewer cells with 1 O/3 G or 2 O/3 G (18% vs 25% Fig. 3b) than with nuclease. Importantly, FISH revealed the same pattern as non-transfected cells. Therefore, unlike nuclease, single nickase leads to HDR with only rare indels and no detectable chromosomal truncations using FISH.

Human embryonic kidney (HEK) cell line HEK293T (ATCC®, Manassas, VA, USA) was maintained in Dulbecco’s modified Eagle’s medium (DMEM), low glucose (1 g.L-1), L-Glutamine (1 g.L-1) and pyruvate (Gibco® by Lifetechnologies^TM, Carlsabad, CA, USA) supplemented with 10% fetal bovine serum, 100 U/mL penicillin, and 100μg/mL streptomycin (all from Eurobio, Courtaboeuf, France). K-562 cell line (ATCC®) was maintained in RPMI Medium 1640, l-Glutamine, 25 mM HEPES (Gibco®) supplemented with 20% fetal bovine serum, GlutaMAX (Gibco®), 100 U/mL penicillin, and 100 μg/mL streptomycin. Human foreskin fibroblasts (HFF) were from a healthy person in accordance with the ethical standards of the committee responsible for human experimentation (Centre Hospitalier Universitaire de Bordeaux). Skin fragments were treated with trypsin-EDTA for 3 h at 37 °C and fibroblasts were isolated. hTERT immortalized fibroblasts were from ATCC® (CRL 4001, BJ-5ta), invalidated or not for TP53 by CRISPR-Cas9. Fibroblasts were maintained in the same culture medium as HEK293T. Both cell lines were cultured at 37 °C, 5% CO₂ in a humidified chamber.

Cells were transfected by electroporation using the Nucleofector AMAXA electroporation system (Lonza®, Bale, Switzerland). In brief, 10⁶ HEK293T (or K-562) cells were nucleofected with 200 ng (or 4 µg) of nuclease- or nickase-containing plasmid and 0.05 µM of specified ssODN. ssODN optimized concentration was used when specified and corresponded to 1.7 µM. Cells were then seeded onto 6-well plates (Corning©, Tewksbury, MA, USA) and cultured as described above. Transfected cells were then positively selected 24 h after transfection either by addition of 1.25 μg mL⁻¹ puromycin (Sigma-Aldrich, Saint Louis MO,USA) for 36 h or by GFP-positive selection by Fluorescent activating cell sorting (FACS) using BD FACS Aria®.

For fibroblasts, 4D-Nucleofector system was used with P3 Primary Cell Line 4D-Nucleofector® in association with CZ-167 program. In brief, 200,000 hFF were nucleofected with 16.9 µg Cas9 RNP and 5 µM of Alt-R® Cas9 Electroporation Enhancer. To form RNP, Cas9 protein was complexed to crRNA:tracrRNA according to the manufacturer’s instructions. Then complexes were incubated for 20 min at room temperature before electroporation. Cas9 protein (Alt-R® S.p. Cas9 Nuclease V3), crRNA (Alt-R® CRISPR-Cas9 crRNA) were purchased from Integrated DNA Technologies.

Nuclease-containing plasmid was a modified version of pSpCas9n-2A-PuroR (pX462) obtained from Addgene (Cambridge, MA, USA) (#48141). First, to switch from Cas9D10A to Cas9, an insert containing the CBh promotor and 253 bp of Cas9 was obtained by XbaI and BgIII digestion of pX330 (Addgene, #42230), and then ligated in pX462 digested with the same enzymes. Then, the puromycin-resistance gene (pac) was replaced by a human codon-optimized version (PuroR-HO). sgRNA was cloned into pUC19 obtained from Addgene (#50005) using the BbsI restriction site and co-delivered at transfection with the nuclease-containing plasmid.

The nickase-containing plasmid was also a modified version of pX462 with a human codon-optimized version of the puromycin-resistance gene (PuroR-HO). The nickase-containing plasmid used for GFP-positive selection was pSpCas9n(BB)−2A-GFP (PX461), Addgene (#48140). In both nickase-containing plasmids, sgRNA was directly added using the BbsI restriction site according to the Feng Zhang protocol (Addgene). All sgRNAs were designed using the CHOPCHOP v2 algorithm⁶⁸ (chopchop.cbu.uib.no) and based on a unique sequence with 20 nucleotides. All ssODN templates used in the study were purchased from Integrated DNA Technologies (IDT, Coralville, IA, USA). For 80nt-ssODN-A647 and 75nt-ssODN-A647, an Alexa Fluor® 647 was chemically linked at the 5’ terminal end to ssODN by NHS Ester link. For 80nt-ssODN-LNA, three Locked Nucleic Acids were added at the 5’ terminal end.

Fluorocyte quantification and sorting for disease modeling and correction was performed by flow cytometry. UV-sensitive type-I porphyrins were excited at 496 nm and the emitted wavelength was approximately 667 nm, detected by the PE-Cy5A PMT channel (FACSCanto, BD, Franklin Lakes, NJ, USA). FL-1 is a control green fluorescent channel used to exclude autofluorescent cells. Cells were sorted by BD FACS Aria®.

Genomic DNA was extracted using Nucleospin® Tissue (Macherey-Nagel, Duren, Germany) according to the manufacturer’s protocol. The genomic region flanking UROS exon 4 (or exon 10) was amplified by PCR (HotStarTaq Plus DNA polymerase, Qiagen®, Venlo, Netherlands) with adequate primers (Supplementary Table 2). PCR products were purified with Nucleospin® Gel and PCR Clean-up (Macherey-Nagel) and digested with SacI (or ApaI for exon 10 UROS analysis) restriction enzyme (New England Biolabs, Ipswich, MA, USA) for 1 h at 37 °C. Then, 5 ng digestion products were loaded into the Agilent® 2200 TapeStation (Santa Clara, CA, USA) capillary electrophoresis using D1000 ScreenTape and D1000 reagents according to the manufacturer’s protocol. Quality control of enzymatic digestion efficiency is included in each assay.

For allelic analysis, purified PCR products were sub-cloned into a “TOPO TA Cloning” vector (Invitrogen, Life Technologies® (Carlsbad, CA, USA) according to the manufacturer’s protocol. Electro-competent bacteria (Stbl3™) (Thermo Fisher Scientific, Waltham, Massachusetts, USA) were transformed with the TOPO and colonies were analyzed by Sanger sequencing after PCR. For cellular clonal analysis, transfected cells were isolated by the limiting dilution method and each clone was analyzed by Sanger sequencing after PCR. Sanger sequencing was done on purified PCR products and sequenced by LIGHTRUN (GATC Biotech, Konstanz, Germany).

UROS activity was determined by an enzyme-coupled assay as described previously⁶⁹. Briefly UROS activity was determined by an enzyme-coupled assay. For that, porphobilinogen was first converted to hydroxymethylbilane, the natural substrate for UROS, by hydroxymethylbilane synthase. Then, the uroporphyrinogen reaction products were oxidized to their respective uroporphyrin isomers, which were then resolved and quantitated by reversed-phase high-pressure liquid chromatography. One unit was defined as the amount of enzyme that formed 1 nmol of uroporphyrinogen III per hour at 37 °C.

Tracking of Insertions, DEletions and Recombination events (TIDER)²⁸ was used to determine HDR and indels frequencies. 490pb PCR product (carrying c.217 T) from non-transfected HEK293T was provided as control chromatogram. For Fig. 6, 490pb PCR product carrying the c.217 C mutation was used as control. Reference chromatogram with either SacI-c.217 C (Fig. 5) or SacI-c.217 T (Fig. 6 and Supplementary Figure 2) were obtained from allelic analysis (Supplementary Figure 5 and Supplementary Figure 10). Defaults parameters were conserved for analysis. The limit of quantification (loq) was defined at 2%.

Genomic DNA was extracted using Nucleospin® Tissue (Macherey-Nagel®) according to the manufacturer’s protocol. The genomic region flanking UROS exon 4 was amplified by PCR (KAPA HiFi DNA polymerase, Kapa Biosystems®, Cape Town, South Africa) with adequate primers (Supplementary Table 2). PCR products were purified with Nucleospin® Gel and PCR Clean-up (Macherey-Nagel®). To prepare sequencing libraries, the Illumina Nextera XT Kit (Illumina®, San Diego, California, USA) was used and nested-PCR using Illumina primers was performed on purified PCR products. An Illumina MiSeq instrument (Illumina®) was used for high-throughput sequencing. The average depth of each genome analysis was 10,000. Quality of paired-end reads was checked with FastQC (Galaxy, https://usegalaxy.org/↗). Then, data were analyzed using the CRISPR Data Analysis and Visualization (CRISPR-DAV) pipeline⁷⁰.

FISH was performed on interphase nuclei of HEK cells or K-562 cell line, with probes targeting the following regions on chromosome 10: centromeric region (XCE 10 probe, labeled in orange) (MetaSystems Probes, Altlussheim, Germany), locus 10q26.11 (BAC RP11–79M19 probe, labeled in green) (Empire Genomics, Buffalo, NY, USA), locus 10q26.2 (BAC RP11–31M22 probe, labeled in orange) (Empire Genomics, Buffalo, NY, USA), sub-telomeric regions (p-arm and q-arm sub-telomere probes, respectively labeled in green and orange) (Cytocell Ltd, Cambridge, UK). Preparations were pre-treated as indicated below. Briefly, the slides were successively immersed in a 2xsaline-sodium citrate buffer for 10 min at 37 °C, in a 0.01% pepsin solution for 10 min at 37 °С, in a 1× phosphate-buffered saline (PBS) solution for 5 minutes, in a 3.7% formaldehyde solution for 5 min, and in a 1xPBS solution for 5 minutes. FISH probes and DNA were then co-denaturated according to the manufacturers’ protocols, and hybridization was performed overnight at 37 °C. The slides were then successively immersed in wash solutions and the nucleic acids were counterstained by 4,6-diamidino-2-phenylindole. The slides were then placed under an Axio Imager2 microscope with an epi-fluorescence source (Carl Zeiss AG, Oberkochen, Germany). The microscope was linked to the Metafer4 software for automated image acquisition and processing (MetaSystems GmbH, Altlussheim, Germany).

Genomic DNA was extracted with the Wizard Kit (Promega Corporation, Madison, USA) following the protocol validated in the laboratory.

For sgRNA targeting exon 4 UROS locus, the Top 10-predicted off-target sites, identified by CRISPOR software were amplified in genome-edited HEK293T and subjected to Sanger sequencing, followed by comparison to non-transfected cells by TIDER analysis.

Primers used for off-target analysis are in Supplementary Table. 2

Array-CGH was performed on 8 × 60k oligonucleotide microarrays (Agilent Technologies, CA). DNA was labeled (cyanine 3 or cyanine 5) using the Genomic DNA ULS Labeling Kit from Agilent Technologies and hybridized onto the microarrays according to the manufacturer’s instructions (Agilent). Scanning of the microarrays was performed using a G2565CA scanner (Agilent). Data analysis was carried out with Agilent Technologies software, namely Feature Extraction for Cytogenomics V5.0 for the fluorescence ratio calculation and Agilent CytoGenomics 3.0.1.1 to visualize chromosomal imbalances. Deletions and duplications in the heterozygous state were characterized by values of the log2 ratio of fluorescence intensities (cyanine5/cyanine3) below −0.5 and above + 0.3, respectively, with the statistical algorithm ADM2 used at a threshold of 5.

Statistical significance was inferred when necessary. Graph Pad Prism 6 software was used for statistical analysis. Results are presented as mean ± SEM (standard error of the mean). The two-tailed unpaired t test was done to compare means of two groups. One-way ANOVA, completed with unprotected Fisher’s Least Significant Difference test, was used to compare three groups. Two-tailed Khi2 tests were used to compare percentages. All comparisons are shown with black bars. Null hypothesis was rejected when p value < 0.05. *p < 0.05; **p < 0.01; ***p < 0.001; ns, non-significant.

Further information on experimental design is available in thelinked to this article. Nature Research Reporting Summary

Supplementary Information Reporting Summary Source Data