Gene Editing Corrects In Vitro a G > A GLB1 Transition from a GM1 Gangliosidosis Patient

Exome sequencing and bioinformatics pipeline

Trio exome sequencing was performed at Rennes Hospital University (Molecular Genetics and Genomics Laboratory) using standard pipeline as previously described.³² Identified pathogenic variants were described according to the Human Genome Variation Society nomenclature guidelines (https://varnomen.hgvs.org/).

Patient-derived fibroblasts: primary culture and immortalization

The fresh skin punch biopsies were dilacerated with scalpels under a sterile area and then transferred into T25 flasks containing 5 mL of AmnioMAX C-100 (Gibco) medium. After 1–2 weeks, explant growing cells were washed with Dulbecco's phosphate-buffered saline (DPBS; Lonza), trypsinized with TrypLE™ Express Enzyme (1 × ; Gibco), and passaged in T75 flasks for fibroblast amplification. Primary cells were then immortalized by SV40 T antigen lentiviral transduction. Lentiviral particle production was performed in HEK293T cell line by following Trono Lab recommendations (http://tronolab.epfl.ch).

Briefly, psPAX2 (Addgene plasmid ID #12260), pVSVG (Addgene plasmid ID #8454), and pLOX-Ttag-iresTK plasmids (Addgene plasmid ID #12246) were transfected into HEK293T cells with lipofectamine 2000 (Thermo Fisher Scientific) according to the manufacturer's instructions. Cell supernatant containing lentiviral particles was collected after 2 days, centrifuged, and filtered (0.45 μm) to eliminate HEK293T-derived cell debris. Fibroblasts in primary culture (P6 well plate; Falcon^®) were infected with 500 μL of this infectious media in the presence of 8 μg/mL polybrene (Sigma) to achieve viral infection.

Cell culture

Immortalized patient-derived fibroblasts and HEK293T (ATCC) cells were cultured, respectively, in RPMI-1640 medium (Lonza) or Dulbecco's modified Eagle's medium (Gibco), each supplemented with 10% fetal bovine serum (Gibco), 1% glutamine (Lonza), and 1% penicillin-streptomycin (Sigma). When confluent, cells were washed with DPBS (Lonza), digested with TrypLE Express enzyme (1 × ; Gibco), and subsequently passaged at 1:10 ratio. All cells were maintained at 37°C in a humidified incubator with 5% CO₂. Cell lines were routinely tested for mycoplasma infection with PlasmoTest Mycoplasma Detection Kit (Invivogen) and negative results were obtained.

DNA extraction, polymerase chain reaction, and Sanger sequencing

DNA was extracted from frozen cell pellets with the Nucleospin tissue DNA extraction kit (Macherey-Nagel). DNA concentration was evaluated using the NanoDrop1000 spectrophotometer (Thermo Fisher Scientific). Polymerase chain reaction (PCR) was performed from 50 ng DNA with the Phusion™ High-Fidelity DNA Polymerase kit (Thermo Fisher Scientific). The list of primers is presented in Supplementary Table S1. Samples were subjected to thermal cycling as follows: 98°C for 30 s (initial denaturation step), 98°C for 10 s, primers' annealing temperature for 15 s, 72°C for 30 s (35 cycles), and 72°C for 5 min (final elongation step).

Amplicon size and purity were checked on 1.5% agarose gel electrophoresis before Sanger's reaction. Sanger sequencing was performed directly on 1 μL of PCR reaction in the presence of 0.5 μM primer and the Big Dye Terminator V3.1 (Applied Biosystems™). The following thermocycler program was used: 96°C for 5 min (initial denaturation step), then 96°C for 1 min, primers' annealing temperature for 1 min, and 62°C for 2 min (30 cycles). Sequence products were purified on Sephadex G50 beads (GE Healthcare) and directly loaded onto 3130xl Genetic Analyzer capillary electrophoresis laser coupled system (ABI PRISM).

RNA extraction, reverse transcription, and quantitative PCR

RNA extraction was performed from frozen tissue in culture plates (P6 well plate; Falcon) by using the NucleoSpin RNA kit (Macherey-Nagel). RNA concentration was measured with NanoDrop1000 spectrophotometer (Thermo Fisher Scientific). Reverse transcription was performed on 500 ng total RNA with the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems). Quantitative PCR (qPCR) was performed in 384-well plates on 2.5 ng of reverse-transcribed cDNA using the SYBR Green PCR Master Mix (Applied Biosystems) in the presence of 1 μM of forward and 1 μM of reverse primers, using the QuantStudio 5 quantitative PCR equipment (Applied Biosystems). The primer sequences used are available in Supplementary Table S1. Raw data were extracted with QuantStudio Design and Analysis Software (Applied Biosystems). Relative gene expression compared to control conditions was calculated by using the 2^−ΔΔ Ct method. GAPDH was used as a housekeeping gene for normalization.

Protein extraction

Cells were lysed and proteins were extracted from fresh fibroblast cells with RIPA buffer (Thermo Fisher Scientific) supplemented with protease and phosphatase inhibitor mini-tablet (Pierce, Thermo Fisher Scientific). Protein quantification was performed with Pierce BCA protein assay kit (Thermo Fisher Scientific) and absorbance was measured with spark microplate reader (TECAN) at 562 nm.

β-Gal enzymatic activity measurement

β-Gal activity was evaluated using a fluorogenic method initially described by Ho and O'Brien.³³ Briefly, 20 μL of fresh cell lysates were incubated in a white 96-well microplate (Grenier Bio-One) with 100 μL of buffer solution (pH 4.3) containing 0.5 mM of 4-methylumbelliferyl-β-d-galactopyranoside substrate (Sigma). A kinetic analysis was performed at 37°C for 1–2 h. Every 4 min, the fluorogenic substrate was excited at 366 nm and fluorescence emission, proportional to enzyme activity, was measured at 442 nm with a spark microplate reader (TECAN). Enzyme activity was assessed from the slope of the fluorescence = f (time) curve and normalized per microgram of protein.

Western blot

Protein samples were reduced and denatured with lithium dodecyl sulfate and reducing agent for 10 min at 70°C. Then, 20 μg of each sample was loaded in 4–12% Bis-Tris Gel (NuPAGE; Invitrogen) and migration was done with following parameters: 1 h 30 min, 200 V, and 400 mA. Proteins were then transferred onto a nitrocellulose membrane (Invitrogen™ iBlot™ Transfer Stack) with iBlot transfer device system (Invitrogen).

After transfer and incubation with a blocking solution (1 h at room temperature, TBS-Tween 0.1%; 5% bovine serum albumin; Eurobio), the membrane was probed with the primary antibody overnight at 4°C (rabbit anti-β-galactosidase antibody, 15518-1-AP Protein Tech, dilution 1/1000; and mouse anti-α-tubulin, dilution 1/10,000; T6199 Sigma-Aldrich) and then for 1 h at room temperature with the horseradish peroxidase (HRP)-linked secondary antibody (HRP-linked anti-rabbit IgG, dilution 1/2000; #7074 Santa Cruz; and HRP-linked anti-mouse IgG, dilution 1/2000; #7076 Santa Cruz). Signal was detected with Amersham ECL select substrate (Cytiva) under a luminescent image analyzer (ImageQuant LAS 4000). Uncropped western blots are available in Supplementary Figure S1.

Cloning

Single guide RNAs were designed according to the recommendations of Walton et al,¹⁹ Supplementary Table S1. Designed oligonucleotides were integrated into BPK1520 backbone (Addgene plasmid ID #65777) by golden gate assembly as previously described.³⁴ Plasmids were transformed into NEB^® Stable competent Escherichia coli C3040H (New England BioLabs) by following the manufacturer's recommendations. Bacteria were seeded onto ampicillin (100 μg/mL) Lysogeny broth (LB) agar plates and incubated overnight at 37°C. Isolated colonies were then amplified overnight at 37°C with LB growth medium and ampicillin (100 μg/mL) under constant agitation. Plasmid DNA was purified with NucleoBond^® Xtra Maxi kit (Macherey Nagel) according to the manufacturer's instructions. Constructs were verified by Sanger sequencing. The plasmid encoding the sgRNA number 8 used in this study is available on Addgene (BPK1520-sgRNA GLB1; Addgene #184378).

ABE transfection

Immortalized fibroblasts were seeded in 6-well plates (Falcon) (500,000 cells/well). Approximately 24 h after seeding (∼60% confluency), cells were transfected with 11.55 μL of Viafect reagent (Promega) according to the manufacturer's protocols with 2200 ng of ABEmax(7.10)-SpRY-P2A-EGFP plasmid (Addgene plasmid ID #140003) and 1100 ng of sgRNA plasmid (Addgene plasmid ID #65777). The medium was replaced 6 and 24 h after transfection to eliminate dead cells. Forty-eight hours after transfection, cells were washed with DPBS (Lonza), digested with TrypLE Express enzyme (Gibco), and resuspended into RPMI-1640 medium (Lonza). GFP positive cells were sorted by fluorescence-activated cell sorting (FACS, FACSAria FUSION; Becton Dickinson). Sorted cells were either lysed for DNA extraction and HTS or individualized by limit dilution for clonal amplification (96-well plates, 1 cell/well; Falcon).

Approximately 3 weeks after the limit dilution, pure clones were amplified in 12-well plates (Falcon). To easily identify the edited cells, β-galactosidase assays were performed and gene editing was confirmed by Sanger sequencing.

siRNA transfection

Immortalized fibroblasts were seeded in 6-well plates (500,000 cells/well; Falcon). Approximately 24 h after seeding (∼60% confluency), cells were transfected with 50 nm siRNA CTRL or siRNA GLB1 (IDT, sequence available in Supplementary Table S1) with lipofectamine RNAimax (Thermo Fisher Scientific) according to the manufacturer's instructions. Forty-eight hours after transfection, cells were stored at −80°C and proteins were then extracted.

Off-target identification (CRISPOR)

Off-target sites were predicted bioinformatically by using the CRISPOR online tool (http://crispor.tefor.net).³⁵ Putative off-target sites were ranked according to cutting frequency determination (CFD) score³⁶ and top 10 off-target hits were selected for deep sequencing (amplicons) (data are available in Supplementary Table S3). In addition, putative exonic off-target sites were checked by exome sequencing.

Amplicon HTS

Genomic regions of interest were amplified by PCR (Phusion High-Fidelity DNA Polymerase; Thermo Fisher Scientific or KAPA2G Robust HotStart PCR; Sigma) from genomic DNA samples. Amplicons were purified with AMPure XP magnetic beads (Beckman Coulter). The library was prepared with the SureSelect XT HS2 DNA kit (Agilent). We followed the manufacturer's recommendations (except that initial DNA fragmentation, hybridization, and capture steps were skipped). Pooled libraries were sequenced on a MiSeq instrument (Illumina) with a 2 × 250 bp paired-end running method (MiSeq Reagent Nano Kit v2 500 cycles; Illumina). The flow cell was loaded with 5 pM pooled libraries containing 5% PhiX control V3 (Illumina). Raw sequencing data were demultiplexed with Bcl2Fastq software (v2.19; Illumina) (data are available in Supplementary Table S3).

CRISPResso2

FastQ files were submitted to CRISPResso2³⁷ for precise editing quantification. We used the following parameters: -wc = −10, -q = 30, - -min-bp-quality-or-N = 30, --conversion-nuc-from = A, and - -conversion-nuc-to = G. The matrix “selected-nucleotide-percentage-table-around-sgRNA” generated by CRISPRESSO was used to calculate the potential editing percentage. For both on-target and off-target data, the line with nucleotide “N” (Q score <30) was removed with a Python script (Supplementary Fig. S2) to avoid potential editing underestimation (data are available in Supplementary Table S4).

Three-dimensional structure prediction

Maestro Software (Schrödinger Release 2021-3; Schrödinger, LLC, New York, NY, 2019) was used for visualization, basic molecular modeling steps (protein preparation, energy minimization, etc.), as well as for the molecular dynamic (MD) simulation trajectory using Desmond.³⁸ Crystal structure of the GLB1 protein in complex with galactose (PDB ID: 3THC) served as the starting point for molecular modeling. The protein preparation wizard was used to prepare the crystal structure as follows: hydrogen atoms were added and possible metal binding states generated.

The protonation and tautomeric states of Asp, Glu, Arg, Lys, and His were adjusted to match a pH of 4.3 and possible orientations of Asn and Gln residues were generated. Hydrogen bond sampling with adjustment of active site water molecule orientation was performed using PROPKA. Water molecules with fewer than three hydrogen bonds to nonwater molecules were deleted. Finally, the protein-ligand complexes were subjected to geometry refinements using the OPLS4 force field³⁹ in restrained minimizations. The S302G mutant was then generated using the “Residue and Loop mutation” tool included with maestro, followed by an energy minimization of the whole structure in implicit water.

The GLB1 wild-type (WT) and S302G structures were then prepared for MD with the system builder panel using the following parameters: each system was solvated with TIP3P water⁴⁰ under periodic boundary conditions with the minimum distance between any atom in the solute and the edge of the periodic box being 10.0 Å; Na⁺/Cl⁻ counterions were added as appropriate for the neutralization of the system; 0.154 M NaCl. MD simulations for each system were performed with Desmond as follows: 3 × 100 ns total time; OPLS4 force field; 100 ps trajectory recording intervals; system energy set to 1.2, NPT ensemble class; 300.0K; 1.01325 bar; and model relaxed before simulation. Quality control of our simulations was done using the “Simulation Quality Analysis” tool as well as the “Simulation Interactions Diagram” tool.

GLB1 genetic variants

First, we analyzed the genetic alterations of the GLB1 gene found in ClinVar to characterize the origin of GM1 gangliosidosis (Fig. 1 and Supplementary Table S2). The GLB1 variants are almost exclusively germline (>95%) SNVs (82%), the majority of which are missense (60%) and some frameshift (11%) (Fig. 1A–C). It is interesting to note that the GLB1 noncoding sequence is also impacted (untranslated region and splice sites; ∼20%).

Open in a separate window

FIG. 1.

Genetic alterations of GLB1. (A) Origin of the GLB1 variants (somatic, de novo, or germline), (B) types of GLB1 variants, (C) location and consequence of GLB1 variants, (D) number of GLB1 variants in the different classes, (E) scheme recapitulating base editor effects (ABE and CBE), (F) number of GLB1 variants (class 5) that might be edited by PE, ABE, or CBE. ABE, adenine base editor; CBE, cytosine base editor; PE, prime editing.

Next, we focused on the 93 known pathogenic variants (class 5 corresponding to 17% of all variants, N = 532) to estimate the ability of BE and PE to correct them (Fig. 1D, E). Importantly, PE, ABE, and CBE should, in theory, correct up to 100, 41, and 15%, respectively, of GBL1 pathogenic variants described in ClinVar (Fig. 1F). Of course, the genomic context of each variant must be taken on a case-by-case basis to consider compatibility with editing window and PAM of specific editors. Since BE is currently the most efficient approach for gene editing in humans,³ we evaluated the ability of ABE to correct a missense variant associated with GM1 gangliosidosis as a proof-of-concept experiment. We selected a patient with a missense variant targetable by ABE (Fig. 2).

Open in a separate window

FIG. 2.

Characterization of patient-derived fibroblasts. (A) Milestones of the patient with GM1 gangliosidosis, (B) GM1 gangliosidosis patient pedigree, indicating autosomal recessive inheritance (genealogical tree), (C) hypothetical effect of inherited variants on GLB1 mRNA and amino-acid sequence, (D) workflow describing patient-derived fibroblast purification and immortalization using SV40 T antigen (transduced by lentivirus), (E) β-galactosidase activity in patient-derived fibroblasts. The measured values of the patient were compared to the values obtained from his parents and a control case (control fibroblasts), n ≥ 3 biologically independent experiments, each histogram represents the mean ± SD. (F) GLB1 protein levels in the same fibroblasts. Specificity of the GLB1 antibody raised against GLB1 protein was evaluated using siRNA targeting GLB1 mRNA or a nontargeting siRNA (siCTR). Western blot results are representative of at least three experiments and α-tubulin serves as a loading control, (G) expression levels of GLB1 mRNA evaluated by RT-qPCR in patient-derived fibroblasts or control fibroblasts, (H) allele-specific qPCR quantification in patient-derived fibroblasts (ΔCt). Otherwise indicated, data were subjected to Student's t-test and p-values were deemed to be statistically significant as follows: *p < 0.05; **p < 0.01;***p < 0.001; ****p < 0.0001. Raw data are available in Supplementary Table S5. NMD, nonsense-mediated mRNA decay; RT-qPCR, real-time quantitative reverse transcription PCR; SD, standard deviation.

GLB1 characteristics of the patient

The patient is the second child of nonconsanguineous parents (Fig. 2A). He has one healthy brother. Pregnancy and delivery were uneventful. He was born at 40 weeks of gestation, and his birth weight was 4015 g. He achieved motor skills with slight delay, as he was able to sit at 9 months, to stand up at 19 months, and to make his first steps at 22 months of age (Fig. 2A). At 30 months of age, he began to lose previously acquired milestones as he could no longer walk independently. Regarding speech development, he could pronounce his first words at the end of his first year of age, but he did not progress further and progressively lost speech during the following months.

At 22 months of age, the parents noticed some episodes of eyelid shaking. The ophthalmologic examination was normal. A first brain magnetic resonance imaging (MRI) was performed at 23 months of age showing normal results. At 3 years of age, generalized epilepsy became obvious. Initial anticonvulsant therapy by Levetiracetam was switched to valproic acid due to side effects. Follow-up cerebral MRI at 3 years of age showed generalized brain atrophy and hypomyelination.

The first genetic investigations, which included screening for Steinert disease, Angelman and Prader Willi syndrome, and CGH array analysis, were normal (data not shown). Trio whole exome sequencing (WES) of the child and his parents identified two heterozygous probable pathogenic variants in the GLB1 gene (NM_000404.2): c.907G>A (p.Val303Met) and c.75 + 2 dup, inherited from his mother and father, respectively (Fig. 2B, C). The first variant is a missense variant (never described) and the second is located at a splice site, leading to a 20-nucleotide intron retention in GLB1 RNA.⁴¹

Since the GLB1 genetic alterations suggested an altered β-gal activity associated with GM1 gangliosidosis, we evaluated it in white blood cells. Analyses revealed a pathogenic drop of β-gal activity (8 nmol/h/mg of proteins in the patient vs. control: 197 nmol/h/mg of proteins, data not shown). The residual β-gal activity was estimated to be 4% of control. As a control, the neuraminidase activity was similar for the patient and the control sample (data not shown). Thus, the abnormal β-gal activity supported the pathogenicity of the two variants identified in GLB1 and led to the diagnosis of a late-onset form of infantile GM1 gangliosidosis for this patient.

To confirm this diagnosis, three skin biopsies were performed to generate primary fibroblast cultures (parents and patient) (Fig. 2D). To obtain enough cells, we immortalized these fibroblasts using T antigen from SV40 by a lentiviral transduction.

In accordance with results obtained with Peripheral Blood Mononuclear Cells, we confirmed that patient's fibroblasts displayed an undetectable β-gal activity in our experimental conditions (Fig. 2E). In contrast, the β-gal activity detected in fibroblasts isolated from the parents was reduced by 50% when compared to control fibroblasts. These results demonstrated that the combination of these two inherited mutations abrogated the β-gal activity and expression (Fig. 2F). Next, we confirmed a reduced amount of GLB1 mRNA in patient-derived fibroblasts when compared to control fibroblasts (Fig. 2G) and both alleles were detectable (Fig. 2H). Allele 1 (c:75 + 2dup) has reduced expression compared to allele 2, probably due to nonsense-mediated decay of transcript containing 20 bp of intronic sequence as already suggested.⁴¹ In conclusion, we showed that the combination of these two genetic alterations explains this case of GM1gangliosidosis.

Adenine base editing strategy for GM1 gangliosidosis

To conceive a therapeutic solution for this patient, we selected the ABE approach to correct only one allele; the GLB1 c.907G>A (p.Val303Met) mutation (maternally inherited), aiming to restore a β-gal activity (50%) as found in his asymptomatic parents (Fig. 3A). The ABE strategy is known to reach a greater efficiency than PE, and ABE is already used in clinical trials.³ Only a PE approach has the potential to repair the other genetic alteration inherited from the father.⁴²

Open in a separate window

FIG. 3.

Adenine base editing strategy for GM1 gangliosidosis. (A) Maternal inherited GLB1 variant is targetable by ABE, (B) scheme explaining the base editor window. Potential bystander edit might be obtained (proximal adenine to the on-target adenine located in the base editor window), (C) two sgRNA were designed to correct the pathogenic ATG (Met) into the GTG (Val) on exon 8 of GLB1 gene using ABE, (D) workflow used to evaluate the ABE efficiency and specificity for the patient-derived fibroblasts, (E) efficiency of ABE to introduce the on-target edit (A₈→G₈ or 907A→G) using the sgRNA8 in fibroblasts. Results were obtained by amplicon deep sequencing and analyzed by CRISPResso2. Bystander edits were identified. (F) Clones isolated after transfection sgRNA8+ABE or sgRNA CTR+ABE were compared to normal fibroblasts or untreated patient-derived fibroblasts. Each clone was characterized for β-galactosidase activity, β-galactosidase protein, and edits. n = 3 biologically independent experiments, each histogram represents the mean ± SD. α-Tubulin serves as a loading control. Data were subjected to Student's t-test and p-values were deemed to be statistically significant as follows: *p < 0.05; **p < 0.01;***p < 0.001; ****p < 0.0001. Raw data are available in Supplementary Table S5. MUT, mutated; WT, wild type.

To optimize our chance of designing effective sgRNA for ABE, we selected the human codon-optimized ABEmax(7.10): A-to-G base editor with nSpCas9 SpRY (D10A/L1111R/D1135V/G1218R/E1219F/A1322R/R1335V/T1337R).¹⁹ The main advantage of this Cas9 variant is that the classical NGG PAM evolves to NRN or NYN, with NRN PAM being better recognized than NYN PAM (NRN>NYN). Using this near-PAMless engineered CRISPR-Cas9 variant, we successfully designed and cloned two sgRNAs (Fig. 3A–C). Next, ABEmax(7.10) and one sgRNA were transfected into immortalized fibroblasts derived from the patient (Fig. 3D).

Transfected cells were selected using the enhanced green fluorescent protein (EGFP) signal encoded by the plasmid pCMV-T7-ABEmax(7.10)-SpCas9-NG-P2A-EGFP (Fig. 3D). Five days after the enrichment of transfected cells (EGFP⁺-sorted cells), the BE efficiency was estimated by DNA sequencing as previously described.^6,37 CRISPResso2 analyses indicated that A907 was converted into G907 (reference nucleotide) in ∼52% of reads from the maternal allele as expected with this BE approach (Fig. 3E). The sgRNA 7 was inefficient (Supplementary Table S4), confirming the requirement to evaluate several sgRNAs for BE.

Next, we performed a clonal selection to isolate edited G907 clones (on-target edit). β-Gal activity and β-gal protein levels were assessed (Fig. 3F). For the six clones exposed to the sgRNA8, ABE rescued a therapeutic level of β-gal protein and β-gal activity when compared to donor cells (wt GLB1). The restored β-gal activity reached a similar activity to that detected in the fibroblasts isolated from his asymptomatic parents (50% of the normal values) (Fig. 2E).

Window editing of ABE

Closer inspection of our sequencing results indicated that A904 (position A₅ from the 5′ sgRNA extremity) is also converted into G904 (∼31% of the reads) (Figs. 3F and 4A, B). This result could be explained by the active window of the ABEmax (7.10), which extends from nucleotides N₄ to N₉ (Fig. 3B).¹⁹ In parallel, we did not detect any significant editing event outside this editing window by NGS (raw data Supplementary Table S4).

Open in a separate window

FIG. 4.

Specificity assessment of ABE for GLB1 therapy. (A) Scheme recapitulating our workflow to identify the on-target and bystander edits on the GLB1 gene (base editor window), (B) workflow explaining the off-target effects of ABE in patient-derived fibroblasts. Potential off-target edits were predicted by CRISPOR and 41 potential sites were examined by combining whole exome sequencing and amplicon sequencing results. CFD score.^35,36 (C) Analysis of the top 10 off-target predicted sites for sgRNA8+ABE. Data were subjected to Student's t-test and p-values were deemed to be statistically significant as follows: ****p < 0.0001. Raw data are available in Supplementary Table S5. CFD, cutting frequency determination.

Since the clone#20 (G907-G904; on-target and bystander edits) displayed a similar β-gal activity than other edited clones (G907-A904; only on-target edit), we investigated the neutral role of this unwanted edit.

We predicted the 3D structure of GLB1 protein (Supplementary Fig. S3). In silico mutation followed by an energy minimization of the whole structure showed minimal impact on the surrounding residues and GLB1 structure (RMSD_all-atoms = 0.720 Å). To study the potential impact of this mutation more in depth, MD simulations were carried out (3 × 100 ns) on GLB1 WT and S302G (Supplementary Fig. S3A). Analysis of trajectories did not reveal any significant change in the secondary of tertiary structure of GLB1 (Supplementary Fig. S3B). Altogether, our results strongly suggest that this unwanted on-target event (bystander edit) might have no impact on GM1 gangliosidosis patients.

First, the authors would like to thank the family (parents and son) for their kind support.

We would like to thank their collaborators for helpful advices, protocols, and vectors: Pr. C. Bendavid, Dr. C. Moreau, Pr. D.R. Liu (vectors), Pr. B. Kleinstiver [pCMV-T7-ABEmax(7.10)-SpRY-P2A-EGFP (RTW5025)], the Biosit platforms, especially Dr. P. Gripon for BSL3 and L. Deleurme and A. Aimé for FACS analysis, the CRISP'edit platform (Bordeaux), especially Dr. B. Turcq and Dr. V. Prouzet-Mauleon, the head of the Inserm research unit Dr. E. Chevet, the technicians from the Cell Biology service CHU Rennes, and Pr. E. Tucker and Dr. A. Matvere for critical reading of this article.