Neuronal DNA repair reveals strategies to influence CRISPR editing outcomes

Abstract

Genome editing is poised to revolutionize treatment of genetic diseases, but poor understanding and control of DNA repair outcomes hinders its therapeutic potential. DNA repair is especially understudied in nondividing cells like neurons, which must withstand decades of DNA damage without replicating. This lack of knowledge limits the efficiency and precision of genome editing in clinically relevant cells. To address this, we used induced pluripotent stem cells (iPSCs) and iPSC-derived neurons to examine how postmitotic human neurons repair Cas9-induced DNA damage. We discovered that neurons can take weeks to fully resolve this damage, compared to just days in isogenic iPSCs. Furthermore, Cas9-treated neurons upregulated unexpected DNA repair genes, including factors canonically associated with replication. Manipulating this response with chemical or genetic perturbations allowed us to direct neuronal repair toward desired editing outcomes. By studying DNA repair in postmitotic human cells, we uncovered unforeseen challenges and opportunities for precise therapeutic editing.

Extended Data Figure S1:. Schematic of how DSB repair pathways determine the CRISPR editing outcome.

Cas9 induces a blunt or staggered DSB, depending on where the RuvC domain cleaves (Shou et al, Mol Cell, 2018. PMID: 30033371). The exposed DNA ends are then subjected to either end protection or end resection (or other processing). End protection generally leads to cNHEJ. If the protected ends are still chemically compatible for ligation, cNHEJ often ligates them fidelitously, yielding an unchanged sequence which can be re-cut by any remaining Cas9 RNP. If the protected ends are not compatible for ligation, or if end protection was outcompeted by processing machinery such as polymerases and nucleases, then NHEJ processing can occur (Stinson et al, Mol Cell, 2020. PMID: 31862156). This processing sometimes introduces indels. In dividing cells, end resection often outcompetes end protection, leading to resection-dependent pathways such as MMEJ, HDR and SSA. Resection-dependent pathways can cause indels (MMEJ/SSA) or templated repair (HDR).

Extended Data Figure S2:. Characterizing the purity of the neuronal differentiation.

a) By Day 7 of differentiation, less than 1% of cells are proliferative (Ki67+). Bars show what percentage of DAPI+ nuclei were Ki67+, averaged across 3 replicate wells. HCS Studio SpotDetector. b) By Day 4 of differentiation, 95% of cells express a neuron-specific marker (NeuN+). Bars show what percentage of DAPI+ nuclei were NeuN+, averaged across 3 replicate wells. CellProfiler. For a-b: Each dot is one replicate well, totaled across 13 non-overlapping fields per well. Error bars show SEM. c-e) Representative ICC images showing DAPI and Ki67 staining from Days 4/7/10 of differentiation; quantified in a. f-h) Representative ICC images showing DAPI, NeuN, and TUBB3 staining from Days 4/7/10 of differentiation; quantified in b. TUBB3 is another marker of mature neurons. Note for c-h: HCS Studio software baked the scale bar annotations into the output montages, and only the “merged” panel of each montage is shown here. Differences in font size and bar thickness are simply due to differences in dimensions between two-panel (DAPI+Ki67) and three-panel (DAPI+NeuN+TUBB3) montages. Scale bar lengths remain accurate for each panel.

Extended Data Figure S3:. Establishing VLP delivery of Cas9 to human postmitotic neurons.

a) Schematic depicting the components of virus-like particles. Matrix, capsid, and nucleocapsid are part of the Gag polyprotein. VSV-G and BRL are envelope (env) proteins for pseudotyping to mediate broad and efficient cellular transduction. b) Maps and nomenclature of optimized FMLV VLP vectors. c) Vectors used to produce HIV VLPs, also known as enveloped delivery vehicles (EDVs), were previously described in Hamilton et al., Nat Biotechnol, 2024. PMID: 38212493. d) Assessment of editing efficiency with optimized FMLV VLPs in glioblastoma cells. Monoclonal mCherry-expressing glioblastoma cells (LN229-LC11) were transduced with the indicated VLPs harvested in regular growth media, serum-free growth media, or Opti-MEM. Target cells were transduced at the indicated VLP dilution, with addition of polybrene (PB, 5 μg/ml). At day six post-transduction, mCherry editing efficiency (mCherry-) was assessed by flow cytometry. Non-transduced cells were used for normalization. VLP sgCherry-1: CRISPR-Cas9 VLP containing a previously validated mCherry-targeting sgRNA (Knott et al, eLife, 2019. PMID: 31397669). VLP sgNT-1: CRISPR-Cas9 VLP containing a non-targeting control sgRNA. Error bars indicate standard deviation. e) Assessment of editing efficiency with optimized FMLV VLPs in astrocytes. Normal human astrocytes expressing ZsGreen (NHA-PC5), and previously treated with puromycin-targeting VLPs (Tan et al, Cell Reports, 2023. PMID: 37917583), were transduced with sgZsGreen-targeting VLPs (harvested in regular growth media) at the indicated dilution, with addition of polybrene (PB). Cells were either transduced with a single VLP to generate indels or with a mixture of two VLPs to induce a deletion in ZsGreen. At day six post-transduction, ZsGreen editing efficiency (ZsGreen-) was assessed by flow cytometry. Non-transduced cells were used for normalization. VLP sgZs-1/2/3: CRISPR-Cas9 VLPs containing ZsGreen-targeting sgRNAs. Error bars indicate standard deviation. f) Optimized FMLV VLPs efficiently transduced human iPSC-derived neurons. Representative microscopy image of neurons after transduction with Cas9 VLPs co-encapsulating an mNeonGreen transgene. g-j) Our optimized FMLV VLPs and HIV VLPs (EDVs) both transduced human iPSC-derived neurons efficiently. Flow cytometry 1-week post-transduction with no VLP (untransduced), HIV VLPs, or optimized FMLV VLPs shows up to 97% transduction efficiency. Dosage: 2 μL VLP per 100 μL media.

Extended Data Figure S4:. Cas9-VLPs induce DSBs in human postmitotic neurons.

a) Unmerged panels from Figure 1c, showing DSBs induced by Cas9-VLPs in human iPSC-derived neurons, compared to age-matched untransduced neurons. For a-b: Neurons transduced 2 weeks into differentiation, and imaged 3 days post-transduction. DSBs are co-labeled by markers γH2AX (red) and 53BP1 (green). Dose: 1 μL FMLV VLP per 100 μL media. Scale bar is 20 μm. b) Additional representative ICC images showing DSBs induced by Cas9-VLPs in human iPSC-derived neurons, compared to age-matched untransduced neurons.

Extended Data Figure S5:. CRISPR editing outcomes differ in postmitotic neurons compared to isogenic dividing cells.

a-d) For each of four separate sgRNAs, CRISPR editing outcomes differ between nondividing neurons and dividing iPSCs. Despite differences in which indel outcomes each sgRNA was amenable to overall, in each case, the MMEJ-like deletions were predominant in iPSCs whereas the NHEJ-like small indels were predominant in neurons. Dose: 2 μL VLP (FMLV) per 100 μL media. Average of 6 replicate wells, transduced in parallel. Genomic DNA was harvested 5 days post-transduction, processed for amplicon-NGS, then analyzed by CRISPResso2.

Extended Data Figure S6:. Cas9-VLP-induced indels accumulate for weeks post-transduction in neurons.

a) Even in experiments where iPSCs plateaued at a lower editing efficiency, they still reached that plateau sooner than neurons. Regraphed data from Figure 2a, and overlayed data from a separate experiment with a less-efficient batch of VLPs. In this experiment, iPSCs plateaued at 60–70% indels instead of 90%+, but still reached that plateau within ~4 days. b-d) For three additional sgRNAs, despite differences in which indel outcomes each sgRNA was amenable to overall, all available indel outcomes at the 4 day timepoint increased in prevalence by the 30 day timepoint. Dose: 1 μL VLP (FMLV) per 100 μL media. e-h) The time course of indel accumulation was reproduced very comparably between FMLV and HIV based Cas9-VLPs. With either delivery particle, indels plateaued within 4 days post-transduction in iPSCs (e), but continued to increase for up to 16 days post-transduction in neurons (f). The overlaid time courses look very similar with FMLV VLPs (g-h) compared to HIV VLPs (Figure 2a, 2c). Dose: 2 μL VLP per 100 μL media.

Extended Data Figure S7:. Testing additional hypotheses about the prolonged editing time course in neurons.

a) The prolonged indel accumulation in neurons is not driven by residual VLP in the media. Replacing the media after 2 days post-transduction (as is required for iPSCs) did not significantly affect neuron editing efficiency at 4 days post-transduction. Notable because the steepest increase in indels in neurons typically occurs over the first 4 days post-transduction. Unpaired t test, ns = not significant (p>0.05). 6 replicate wells per condition, transduced in parallel. Dose: 1 μL VLP (FMLV) per 100 μL media. b) ABE-VLPs confirm that the slow indel accumulation we observed is not a product of deficient VLP delivery to neurons. When the same HIV VLPs are used to deliver adenine base editors (ABEs) instead of Cas9, neurons can match and even exceed the editing efficiency of iPSCs, within only 3 days post-transduction. Error bars show SEM; 3 replicate wells per condition, transduced in parallel. Dose: 4 μL VLP (HIV) per 100 μL media – but these VLPs were half as concentrated as normal, since VLPs were harvested from 3 10 cm plates per batch instead of 6 but still resuspended to the same volume. Therefore, equivalent to a 2 μL VLP dose from normal batches. ABE-VLP cloning protocol is described in the fourth tab of Supplemental Table 3. c) Postmitotic iPSC-derived cardiomyocytes (CMs) also show a weeks-long accumulation of indels, for two different sgRNAs. At day 30+ of differentiation, after lactate purification to select for postmitotic CMs, CMs were transduced with 1 μL FMLV VLP per 100 μL media. 3 replicate wells per condition, transduced in parallel. CRISPResso2 analysis of amplicon-NGS. CMs were generated from WTC background iPSCs using the protocol described in Lian et al, Nat Protoc, 2013 (PMID: 23257984).

Extended Data Figure S8:. Cas9-induced DSB repair signals persist in neurons for at least one week post-transduction.

a-c) DSB repair markers over time in untransduced (a), B2Mg1-transduced (b), and NEFLg1-transduced (c) neurons. DSBs are co-labeled by ICC markers γH2AX (red) and 53BP1 (green). Dose: 1 μL FMLV VLP per 100 μL media. Neurons were fixed at 1,4, or 7 days post-transduction as labeled. One representative image from each condition is shown. Transduction was 2 weeks into differentiation. Scale bar is 20 μm. Same experiment as Figure 2d–e, but now showing unmerged panels individually, and including additional conditions (timepoints, sgRNAs). Therefore, the merged panels for untransduced and B2Mg1-transduced at 1 day and 7 days are the same as in Figure 2d–e, but uncropped here.

Extended Data Figure S8:. Cas9-induced DSB repair signals persist in neurons for at least one week post-transduction.

a-c) DSB repair markers over time in untransduced (a), B2Mg1-transduced (b), and NEFLg1-transduced (c) neurons. DSBs are co-labeled by ICC markers γH2AX (red) and 53BP1 (green). Dose: 1 μL FMLV VLP per 100 μL media. Neurons were fixed at 1,4, or 7 days post-transduction as labeled. One representative image from each condition is shown. Transduction was 2 weeks into differentiation. Scale bar is 20 μm. Same experiment as Figure 2d–e, but now showing unmerged panels individually, and including additional conditions (timepoints, sgRNAs). Therefore, the merged panels for untransduced and B2Mg1-transduced at 1 day and 7 days are the same as in Figure 2d–e, but uncropped here.

Extended Data Figure S9:. DSB repair signals remain detectable at the cut site for at least 8 days post-transduction.

a) ChIP-qPCR for γH2AX binding at various distances from the cut site over time. Same conditions as Figure 2f, but with γH2AX antibody instead of Mre11. b) Schematic illustrating our strategy to detect cut-and-resealed loci by using a ChIP-qPCR amplicon that spans across the cut site. Repair protein binding suggests that the locus had been cut, and successful PCR amplification suggests that the cut has since been resealed. Note: however, it remains ambiguous whether these loci were sealed with or without an indel. c-d) Some loci have been resealed as early as 2 days post-transduction. ChIP-qPCR using the spanning amplicon to detect cut-and-resealed loci, with both Mre11 (c) and gH2AX (d). Same procedures as Figure 2f and S9a, but using different amplicons (cut site spanning, and different chromosome control). e) Cas9 protein in iPSCs gets quickly diluted and/or degraded to background levels within 2 days post-transduction; therefore, these neuron ChIP-qPCR data cannot be compared to iPSCs. Pulse-chase to track degradation of Halo-tagged Cas9 in iPSCs. First, iPSCs (with/without lentivirally integrated Halo-Cas9 and B2Mg1) were seeded onto glass-bottom 96-well plates with ~2,000 cells per well. iPSCs were pulsed with 40 μM fluorescent Halo ligand (Promega HaloTag-JF549, cat. #GA1110) for 1 hour, then washed with fresh media 3 times to prevent newly translated Cas9 protein from being labeled. iPSCs were then chased with 2 μM of an unlabeled Halo ligand (Promega ent-HaloPROTAC3, cat. #GA4110) as a binding competitor. Nuclei were labeled with NucBlue (ThermoFisher, cat. #R37605) 20 min before live cell imaging on the Image Xpress Confocal Microscope. Halo fluorescence signal was measured at several timepoints to track the degradation/dilution of the pulse-labeled Cas9 molecules over time. Analyzed in CellProfiler. 8 replicate wells; error bars show standard deviation.

Extended Data Figure S10:. Neuronal transcriptional response to Cas9-VLP is very consistent across three different sgRNAs.

a-d) The neuron-specific transcriptional response to Cas9-VLPs was replicated by two additional sgRNAs, NEFLg1 (a-b) and NTg1 (c-d), besides B2Mg1 shown in Figure 3a–b. Neurons have more DEGs overall upon transduction, and the most significant of these DEGs are enriched for DNA repair genes. Same parameters as Figure 3a–b, but with different sgRNAs. Note: NEFL is not expressed in iPSCs, so its expression is not expected to decrease upon NEFLg1 editing in iPSCs. e) The only two DEGs between B2Mg1-edited and NEFLg1-edited neurons are B2M and NEFL respectively. This reinforces the consistency of the neuronal transcriptional response across different sgRNAs. f) The transcriptional profile of NTg1-treated neurons is more similar to B2Mg1-edited and NEFLg1-edited neurons than to untransduced neurons. Therefore, at least some component of the neuronal response to Cas9-VLPs may be DSB-independent. Multidimensional scaling (MDS) plot visualizing similarity between the various RNAseq samples. As indicated in this plot, all RNA-seq analysis in this study was conducted on 3 replicate samples per condition, transduced in parallel; 24 total RNA-seq samples.

Extended Data Figure S11:. DNA repair genes are not enriched in the DEGs of transduced iPSCs.

a-c) Gene ontology (GO) analysis shows no enrichment for DNA repair genes in the DEGs of B2Mg1-transduced (a), NEFLg1-transduced (b), or NTg1-transduced (c) iPSCs, relative to untransduced iPSCs. Showing the top 20 GO terms in each comparison. Bar length indicates number of DEGs that fall into each GO category. Color indicates significance of adjusted p-value.

Extended Data Figure S12:. Transcriptional response of RNR subunits in neurons compared to iPSCs.

a-c) In iPSCs, non-targeting Cas9 (a) does not affect transcription of any RNR subunits. However, both of the cutting Cas9-VLPs (b-c) significantly induce transcription of RRM2B, the canonically DSB-responsive subunit of RNR. The other two subunits are unaffected. d-f) In neurons, the canonically S-phase-restricted RRM2 is one of the most significantly upregulated genes in all 3 transduced conditions, including with non-targeting Cas9.

Extended Data Figure S13:. Inhibiting RNR changes editing outcomes in neurons.

a) Toxicity of escalating doses of RRM2 inhibitor 3AP in neurons. Maximum tolerable dose was 3.75 μM. For a/c/f: Tolerability threshold was arbitrarily set to 0.75 or above, corresponding to less than a 25% reduction in viability. PrestoBlue viability assay at 8 days post-transduction, normalized to age-matched untreated neurons on the same plate. 3 replicate wells per condition, treated in parallel; error bars show SEM. b) Toxicity of escalating 3AP doses in neurons with or without Cas9-VLPs. Optimal 3.75 μM dose remains nontoxic even with Cas9-VLPs (1 μL FMLV) inducing DNA damage. Error bars show SEM. Two-factor ANOVA; ns = not significant (p>0.05). c) Toxicity of escalating doses of RRM2 inhibitor GW8510 in neurons, alongside Cas9-VLP treatment. Maximum tolerable dose was 2 μM. d-e) GW8510 co-treatment of B2Mg1-edited neurons increases indels overall (d), boosting deletions specifically, and roughly doubles the frequency of single-base deletions (e). Replicated the effects of RRM2 inhibitor 3AP from Figure 3g–h. Dose: 1 μL FMLV VLP per 100 μL media, and maximum tolerable dose of GW8510. Indels measured 8 days post-transduction. For d, error bars show SEM. One-Factor ANOVA, ** p<0.005. For d-e, 6 replicate wells per condition treated in parallel. f) Toxicity of escalating doses of RRM1 inhibitor gemcitabine in neurons, alongside Cas9-VLP treatment. Maximum tested dose was 300 nM, and still tolerable. g-h) Gemcitabine co-treatment of B2Mg1-edited neurons increases indels overall (g), boosting deletions specifically (h). Replicated the effect of RRM2 inhibitor 3AP from Figure 3g–h. Dose: 1 μL FMLV VLP per 100 μL media, and maximum tolerable dose of gemcitabine. Indels measured 8 days post-transduction. For g, error bars show SEM. One-Factor ANOVA, *** p<0.0005. For g-h, 6 replicate wells per condition treated in parallel.

Extended Data Figure S14:. RNR inhibition affects neuron editing outcomes in an sgRNA-dependent manner.

a-d) For B2Mg2 (a-b) and NEFLg1 (c-d), 3AP treatment significantly increases indels, but without the same selectivity for deletions as seen for B2Mg1. 6 replicate wells per condition, treated in parallel. e-f) For HSPB1g2, 3AP treatment significantly decreases indels instead of increasing them. This is consistent with its intrinsic indel distribution, which appears relatively impermissible to deletions. 6 replicate wells per condition, treated in parallel. For a/c/e: One-Factor ANOVA, ** p<0.005, * p<0.05. For a-f: 1 μL dose, FMLV VLPs.

Extended Data Figure S15:. NHEJ pathway exemplifies that many DNA repair factors are not reliably targetable with small molecule inhibitors.

Two-thirds of the factors in the NHEJ pathway (Stinson et al, Annu Rev Biochem, 2021. PMID: 33556282) are not reliably druggable by small molecule inhibitors. Determined by searching for availability of inhibitors on Tocris, Selleckchem, Sigma, and PubMed, as of 2023. This simply demonstrates how many DNA repair factors are not readily druggable at the protein level.

Extended Data Figure S16:. Lipid nanoparticles allow all-in-one delivery of Cas9, sgRNA, and siRNAs to manipulate editing outcomes.

a-c) Lipid nanoparticles transfect neurons with almost 90% efficiency. Neurons were transfected with LNPs encapsulating GFP mRNA at Day 17+ of differentiation. GFP fluorescence was measured by flow cytometry one week later. d) Toxicity of escalating doses of DNA-PKcs inhibitor AZD7648 in neurons, alongside Cas9-VLP treatment. Maximum tested dose was 2 μM, and still tolerable (arbitrary viability threshold of 0.75 as per Extended Data Figure S13). PrestoBlue viability assay at 8 days post-transduction, normalized to age-matched untreated neurons on the same plate. e-f) DNA-PKcs inhibitor AZD7648 reduces indels overall in B2Mg1 VLP-treated neurons. 6 replicate wells per condition, treated in parallel. For e: One-Factor ANOVA, *** p<0.0005. g-h) All-in-one LNPs co-encapsulating siRNAs against PRKDC reduce indels overall in B2Mg1 edited neurons. The effect of PRKDC RNA-level inhibition on all-in-one LNP-mediated editing mirrors the effect of PRKDC protein-level inhibition on VLP-mediated editing. 6 replicate wells per condition, treated in parallel. For g: One-Factor ANOVA, *** p<0.0005.

Figure 1:. Modeling CRISPR repair outcomes in postmitotic human neurons.

a) Schematic: Genome editing proteins can perturb DNA, but cellular DNA repair determines the editing outcome. b) Timeline of differentiating iPSCs (blue) into neurons (green). After at least 2 weeks of differentiation/maturation, postmitotic neurons are treated with VLPs delivering Cas9 protein (yellow) and sgRNA (orange). c) Cas9 VLPs induce DSBs in human iPSC-derived neurons. Representative ICC images of neurons 3 days post-transduction with B2Mg1 VLPs, and age-matched untransduced neurons. Scale bar is 20 μm. Arrows denote examples of DSB foci: yellow puncta co-labeled by γH2AX (red) and 53BP1 (green). Dose: 1 μL VLP (FMLV) per 100 μL media. d) Genome editing outcomes differ between iPSCs and isogenic neurons. CRISPResso2 analysis of amplicon-NGS, from cells 4 days post-transduction with B2Mg1 VLPs. Dose: 2 μL VLP (HIV) per 100 μL media. Data are averaged across 6 replicate wells per cell type transduced in parallel, and expressed as a percentage of total reads. Thick blue background bars are from iPSCs; thin green foreground bars are from neurons.

Figure 2:. Cas9-induced indels accumulate over a prolonged time span in neurons.

a) Cas9-induced indels accumulate more slowly in neurons than in iPSCs. Dose: 2 μL B2Mg1 VLP (HIV) per 100 μL media. For a-b: points represent individual replicates (some obscured by overlap); curves connect means at each timepoint. For a-c: 6 replicate wells per condition transduced in parallel. CRISPResso2 analysis of amplicon-NGS. b) Several sgRNAs show weeks-long accumulation of indels in neurons. Dose: 1 μL VLP (FMLV) per 100 μL media. c) Insertions and deletions both increase over time in neurons. Dose: 2 μL B2Mg1 VLP (HIV) per 100 μL media. Histogram: thick gray background bars are from 4 d timepoint, and thin green foreground bars are from 30 d. d-e) Cas9-induced DSBs remain detectable in neurons at least 7 days post-transduction. Representative ICC images of neurons 1 day (d) and 7 days (e) post-transduction with B2Mg1 VLPs, and age-matched untransduced neurons. Dose: 2 μL B2Mg1 VLP (FMLV) per 100 μL media. Arrows denote examples of DSB foci. See Extended Data Figure S8 for unmerged/uncropped panels. f) MRE11 is bound near the cut site in neurons for at least 8 days post-transduction. Dose: 2 μL B2Mg1 VLP (FMLV) per 100 μL media. Binding events quantified by ChIP-qPCR for each amplicon, normalized for amplification efficiency and input chromatin. Average of 3 replicate ChIP-qPCR reactions, normalized to untransduced control for each amplicon. Error bars show SD. g) Schematic of two possible models for prolonged indel accumulation in neurons.

Figure 3:. Neuronal response to Cas9-VLP reveals unexpected factors that influence editing outcomes.

a-b) Neurons (b), but not iPSCs (a), dramatically upregulate transcription of DNA repair factors upon Cas9-VLP transduction. Volcano plots show differential expression transcriptome-wide in transduced cells relative to untransduced. Dashed lines show cutoffs for significance (padj<0.05) and effect size (fold-change >2 or <0.5). For a-f: differential expression was calculated from bulk RNA-seq across 3 replicate samples per condition transduced in parallel. Methods section details the statistical tests used to define significant DEGs. Dose: 1 μL HIV VLP per 20,000 cells (1.25 μL VLP per 100 μL media). c) Transduced neurons consistently have more DEGs than transduced iPSCs for 3 different sgRNAs, and <10% of DEGs are shared between the cell types. d) Over 75% of the DEGs in either B2Mg1- or NEFLg1-transduced neurons are shared with NTg1-transduced neurons. e) The most significantly altered DEGs in transduced neurons are highly enriched for DNA repair factors. f) Transduced neurons significantly upregulate many DNA repair genes, including factors canonically associated with replication. Top 40 DNA repair DEGs are shown, rank-ordered by averaging the adjusted p-values from each transduced condition. Bold denotes repair genes ranked in the top 50 DEGs genome-wide. g) Inhibiting RRM2 yields a 50% increase in neuron editing efficiency, within 4 days post-transduction. Error bars show SEM. One-Factor ANOVA with Tukey’s multiple comparison test. Each condition vs No Drug, *** p<0.0005, ns = not significant. For g-h: CRISPResso2 analysis of amplicon-NGS, averaged across 6 wells per condition transduced in parallel. Dose: 1 μL B2Mg1 VLP (FMLV) per 20,000 cells in 100 μL media. h) RRM2 inhibition tripled the frequency of 1-base deletions at 4 days post-transduction. Thick gray bars are DMSO condition; thin green bars are 3AP.

Figure 4:. All-in-one particles deliver Cas9 and sgRNA while simultaneously manipulating DNA repair factors.

a) Schematic illustrating all-in-one LNPs that encapsulate Cas9 mRNA (yellow) and sgRNA (orange), along with siRNAs (blue) against a repair gene of interest. b) Multiple sgRNAs show days-long accumulation of neuron indels following LNP transfection. Individual points represent 6 replicate wells per condition transfected in parallel (some obscured by overlap). Curves connect means at each timepoint. For b-e: CRISPResso2 analysis of amplicon-NGS from neurons. c) RNA inhibition of RRM1 during editing phenocopies small molecule inhibition of RRM1/2, shifting B2Mg1 editing outcomes toward deletions. Two weeks post-transfection. Thick gray background bars are from siNT condition; thin green foreground bars are from siRRM1. For c-d: averaged across 6 replicate wells per condition transfected in parallel. d) Co-encapsulating siRRM1 increases total B2Mg1 editing at two weeks post-transfection. Error bars show SEM. One-Factor ANOVA, ** p<0.005. e) All-in-one LNPs reveal additional targets that increase B2Mg1 editing efficiency at 4 days post-transfection. Averaged across 8 replicate wells per condition, transfected in parallel. Error bars show SEM. One-Factor ANOVA with Tukey’s multiple comparison test. Each condition vs siNT, * p<0.05, ns = not significant. f) A model for why the hits from e accelerated editing in neurons. Inhibiting indel-free repair may have directed outcomes toward indels instead of repeatedly resealing and recutting.