High-yield genome engineering in primary cells using a hybrid ssDNA repair template and small-molecule cocktails

Abstract

Enhancing CRISPR-mediated site-specific transgene insertion efficiency by homology-directed repair (HDR) using high concentrations of double-stranded DNA (dsDNA) with Cas9 target sequences (CTSs) can be toxic to primary cells. Here, we develop single-stranded DNA (ssDNA) HDR templates (HDRTs) incorporating CTSs with reduced toxicity that boost knock-in efficiency and yield by an average of around two- to threefold relative to dsDNA CTSs. Using small-molecule combinations that enhance HDR, we could further increase knock-in efficiencies by an additional roughly two- to threefold on average. Our method works across a variety of target loci, knock-in constructs and primary human cell types, reaching HDR efficiencies of >80-90%. We demonstrate application of this approach for both pathogenic gene variant modeling and gene-replacement strategies for IL2RA and CTLA4 mutations associated with Mendelian disorders. Finally, we develop a good manufacturing practice (GMP)-compatible process for nonviral chimeric antigen receptor-T cell manufacturing, with knock-in efficiencies (46-62%) and yields (>1.5 × 10⁹ modified cells) exceeding those of conventional approaches.

Extended Data Fig. 1 |. Illustration of ssCTS designs.

Detailed illustrations of CTS designs evaluated in the manuscript highlighting the location and orientation of gRNA target sequences (red), 4 bp mismatch (pink), PAM interaction site (yellow), and transgene (blue). (a-j) Illustration of short CD5-HA HDRT designs evaluated in Fig. 1a–c and Extended Data Fig. 2c. (k) Representative illustration of optimal ssCTS design used for large HDRTs throughout the manuscript. CTS = Cas9 Target Site, ssCTS = ssDNA HDRT + CTS sites, HDRT = homology-directed-repair template.

Extended Data Fig. 2 |. Comparison of CTS template designs.

(a) Diagram of CD5-HA knock-in strategy and control ssDNA HDRTs. (b) Representative flow cytometry plots demonstrating CD5-HA knock-in. (c) Live cell counts for each ssCTS design using a CD5-HA knock-in construct at 160 nM – 4uM concentration. (d-e) Comparison of knock-in efficiency, live cell counts, and knockin cell counts with insertions of increasing size using (d) dsCTS or (e) ssCTS HDRTs targeting the IL2RA gene. Evaluated transgenes encode either GFP (~1.4 kb total HDRT length), an IL2RA-GFP fusion (~2.3 kb total HDRT length), or an IL2RA-GFP fusion plus separate EF1a-mCherry expression cassette (~3.5 kb total HDRT length). Each experiment was performed with T cells from 2 independent healthy human blood donors represented by individual dots + mean. RNP = Ribonucleoprotein, CTS = Cas9 Target Site, ssCTS = ssDNA, HDRT + CTS sites, HDRT = homology-directed-repair template.

Extended Data Fig. 3 |. Evaluation of ssCTS design and mechanism with large HDRTs.

(a-f) Comparison of different CTS designs with a large ~2.7 kb CD5-HA knock-in construct. (a) Diagram of long CD5-HA knock-in strategy, representative flow cytometry plot, percent knock-in, live cell counts, and knock-in cell yield counts. (b) Comparison of CTS with a gRNA target sequence that is specific for the cognate RNP ( + CD5 CTS), an alternative gRNA sequence (+ IL2RA CTS), a CTS incorporating a PAM site and scrambled gRNA sequence (+ scramble CTS), or an equivalent amount of dsDNA within the 5’ end of the homology arm (+ end protection). (c) Comparison of complementary oligos covering different regions of the CTS and surrounding sequences. Constructs with CTS sites on both 5’ and 3’ end (green bars), 5’ end only (blue bars), or 3’ end only (red bars) are shown on the right panel. (d) Evaluation of varied 5’ ends including different length of buffer sequence upstream of the CTS site. *indicates no data available for the marked column. (e) Comparison of CTS with different numbers of scrambled bases at the 5’ end of the gRNA target sequence using WT or SpyFi Cas9. (f) Length of homology arm that is covered by the complementary oligonucleotide. (g) Evaluation with and without (‘−’) CTS sites on the 5’ and 3’ end of long ssDNA IL2RA-GFP HDRTs with PAM facing inwards toward the homology arm (‘In’) or outwards away from the homology arm (‘Out’). (h-i) Comparison of knockout and knockin with large CD5-HA (h) or IL2RA-GFP (i) ssDNA and ssCTS HDRTs using RNPs formulated with Cas9 +/− NLS sequences. Each experiment was performed with T cells from 2 independent healthy human blood donors represented by individual dots + mean. RNP = Ribonucleoprotein, CTS = Cas9 Target Site, ssCTS = ssDNA HDRT + CTS sites, PAM = Protospacer Adjacent Motif, HDRT = homology-directed-repair template.

Extended Data Fig. 4 |. Additional parameters affecting ssCTS knockin and biophysical analysis of RNP interactions with HDRTs.

(a-b) Comparison of knockin efficiency (top) and live cell counts (bottom) +/− PGA using (a) large IL2RA-GFP ssCTS templates or (b) large CD5-HA ssCTS templates (~1.3 kb homology arms). (c) Comparison of knockin efficiency (top) and live cell counts (bottom) with PGA, ssDNA enhancer, or no anionic polymer using a BCMA-CAR ssCTS templates. (d-f) evaluation of (d) indel formation by amplicon sequencing, (e) knockin efficiency with short ssDNA CD5-HA HDRTs, or (f) knockin efficiency with short ssCTS CD5-HA HDRTs (40 nucleotide homology arms) using varied molar amounts of RNP and HDRT. (g) Representative AFM images of gel purified dsCTS or ssCTS templates +/− Cas9 RNPs. Brightness shows the relative height as indicated in by scale bars to right of figure. Background circular forms in all panels are likely residual agarose. Experiments in panels a-f were performed with T cells from 2 independent healthy human blood donors represented by individual dots + mean (a, b, c, e, f) or mean alone (d). RNP = Ribonucleoprotein, CTS = Cas9 Target Site, ssCTS = ssDNA HDRT + CTS sites, HDRT = homology-directed-repair template, AFM = atomic force microscopy.

Extended Data Fig. 5 |. Arrayed knockin analysis and target locus characteristics.

(a) Comparison of HDRT variations for knock-in constructs targeting a tNGFR marker across 22 different target loci. Shown for each construct are live cell counts, knock-in cell count yields, relative %knock-in and relative knock-in counts compared to dsDNA templates. Data show mean and individual values from 2 independent healthy human blood donors (b-d) Evaluation target locus characteristics in comparison to ssCTS knockin efficiency by (b) Amplicon-Seq, (c) RNA-Seq, and (d) ATAC-Seq methodologies. Knock-in efficiency for panels b-d shows mean from 2 independent healthy human blood donors. Amplicon-Seq, RNA-Seq, and ATAC-seq data in panels b-d were generated 6 independent healthy human blood donors presented as mean +/− SD. Line-of-best-fit and R squared from a simple linear regression are shown for normally distributed data. RNA-seq data was log-transformed prior to linear regression. Spearman r is shown for non-linear correlations as determined by Shapiro-Wilk test (‘distance from cut site’ and ATAC-seq evaluations). Y axis for panels c-d is log 10. In c and d, separate analyses were performed for CD4 + (top) and CD8 + T cells (bottom) for RNA-Seq and ATAC-Seq comparisons. tNGFR = truncated nerve growth factor, MMEJ = microhomology mediated end joining, NHEJ = non-homologous end joining, ATAC = Assay for Transposase-Accessible Chromatin.

Extended Data Fig. 6 |. Application of ssCTS to diverse primary human hematopoietic cell types.

(a-c) Evaluation of CLTA-mCherry knock-in efficiency and live cell counts 0–10 days postelectroporation in primary human T cells. (a) Live cell counts represented as a percentage of the no electroporation control on day 4 post-electroporation. (b) Knock-in efficiency on day 2–10 post-electroporation. (c) Growth curves for control cells (no electroporation, electroporation only, and Cas9 RNP only) and cells edited with dsCTS or ssCTS HDRTs on day 0–10 post-electroporation. (d) Comparison of knock-in efficiency (top) and live cell counts (bottom) using ssCTS and dsCTS HDRTs (blue line) across a variety of primary human hematopoietic cell types using knockin constructs encoding a CLTA locus mCherry fusion protein. Each experiment was performed with cells from 2 independent healthy human blood donors. Each experiment was performed with T cells from 2 independent healthy human blood donors represented by individual dots + mean (c, d, f) or mean alone (e). CTS = Cas9 target site, dsCTS = dsDNA HDRT + CTS sites, ssCTS = ssDNA HDRT + CTS sites, HDRT = homology-directed repair template, RNP = ribonucleoprotein, CLTA = Clathrin, Treg = regulatory T cells, HSC = hematopoietic stem cell, NK cells = natural killer cells, γδ T cells = gamma delta T cells.

Extended Data Fig. 7 |. Evaluation of small molecule inhibitor cocktails in primary human T cells.

(a) Evaluation of relative increase in percent knock-in using an ssDNA CD5-HA knock-in construct over varied concentrations of 5 different small molecule inhibitors assessed by flow cytometry. Red bars indicate concentrations chosen for subsequent experiments. (b) Comparison of relative percent knock-in (top), live cell counts (middle), and viability with Ghost Dye 780 (Tonbo) (bottom) with small molecule inhibitor combinations. Cocktails chosen for subsequent experiments are highlighted in red (M3814), blue (MT) and yellow (MTX). (c-d) Evaluation of Novobiocin effects on (c) live cell counts and (d) knock-in efficiency using a small CD5-HA ssDNA HDRT. (e) Evaluation of Novobiocin effects on knockin efficiency at varied concentration using a small CD5-HA ssDNA HDRT in combination with M3814, MT, and MTX inhibitors. Each experiment was performed with T cells from 2 independent healthy human blood donors represented by individual dots + mean. M = M3814, MT = M3814 + Trichostatin A, MTX = M3814 + Trichostatin A + XL413, NVB = Novobiocin.

Extended Data Fig. 8 |. Analysis of genome editing outcomes with CTS templates and small molecule inhibitors.

(a-c) Evaluation of genome editing outcomes by either (a) flow cytometry or (b) amplicon sequencing using small dsDNA, dsCTS, ssDNA, or ssCTS CD5-HA HDRTs at non-toxic concentrations (800 nM) with and without M, MT, and MTX inhibitor combinations. (c) Ratio of perfect:imperfect HDR events with each combination. (d) Comparison of dsCTS and ssCTS templates in combination with small molecular inhibitors in 5 different knock-in constructs using a large CD5-HA HDRT (−2.7 kb, n = 4 donors), a tNGFR knock-in to the IL2RA gene (~1.5 kb, n = 4 donors), an mCherry fusion in the clathrin gene (~1.5 kb, n = 4 donors), a near full length CTLA-4-GFP fusion to the CTLA4 gene (~2.1 kb, n = 6 donors), and a full length IL2RA-GFP fusion to the IL2RA gene (~2.3 kb, n = 6 donors). (e-f) Evaluation of (e) live cell counts and (f) viability +/− MT and MTX inhibitor combinations using 44 different knock-in constructs targeting a tNGFR marker across 22 different target loci with 2 gRNA per gene (g1 and g2). Panel a shows mean and individual values from two healthy blood donors. Panels b, c, e, and f show mean values from two healthy blood donors. Panel d shows mean +/− SD. CTS = Cas9 target site, HDRT = homology-directed repair template, dsCTS = dsDNA + CTS HDRT, ssCTS = ssDNA + CTS HDRT, M = M3814, MT = M3814 + TSA, MTX = M3814 + TSA + XL413.

Extended Data Fig. 9 |. IL2RA and CTLA4 ORF replacement strategies.

(a) Gating for GFP + cells are shown for WT and S166N IL2RA-GFP knock-in constructs. (b) Diagram of the CTLA4 gene (top), CTLA4 protein levels (bottom), and cutting efficiency (bottom) illustrating a screening panel of 12 gRNAs examined within exon 1 and intron 1. gRNAs were assessed in activated CD4 + T cells for protein disruption by CTLA4 flow cytometric analysis (flow plots and top row of numbers demonstrate the % of CTLA4-negative cells for each donor), and for cutting efficiency as determined by TIDE indel analysis (bottom row of numbers indicate the %indel at target locus). (c) CTLA4 expression levels assessed by flow cytometry with endogenous protein (black) and WT CTLA4-GFP knock-in protein (red) are shown for CD4- T cells, CD4 + T cells, and regulatory T cells with (dotted line) and without (solid line) stimulation. (d) Gating for GFPhi cells is shown for WT, R70W, R75W, and T124P CTLA4-GFP knock-in cells. Each experiment was performed with T cells from 2 independent healthy human blood donors. WT = Wild-Type, Treg = regulatory T cell.

Extended Data Fig. 10 |. Evaluation of a nonviral strategy for anti-BCMA CAR-T cell manufacturing.

(a-c) Comparison of (a) knockin efficiency (mean +/− SD), (b) flow cytometric immunophenotypes, and (c) tumor burden of MM1S-bearing NSG mice treated with TRAC anti-BCMA CAR-T cells generated using either AAV or non-viral ssCTS HDRTs (mean +/− SD). (d) Live cell counts for large-scale GMP-compatible manufacturing process at Day 7 and Day 10 post-activation. (e) Tumor burden (average radiance) of individual MM1S-bearing NSG mice treated with Unmodified T cells and TRAC anti-BCMA CAR-T cells generated in GMP-compatible anti-BCMA-CAR T cell scaleup experiment. (f) Kaplan–Meier analysis showing overall survival of MM1S xenotransplant NSG mice treated with anti-BCMA-CAR or unmodified T cells. (g-j) Targeted Locus Amplification (TLA) analysis for anti-BCMA TRAC CAR-T cell products generated in GMP-compatible scaleup experiments. (g) Integration site analysis based on TLA sequencing demonstrating targeted insertion at the expected TRAC locus on chromosome 14. (h) Mean percentage of perfect and imperfect HDR events by TLA sequencing from 2 independent healthy human blood donors. (i) Table of perfect HDR, imperfect HDR, and off-target events for individual donors by TLA sequencing. (j) TLA sequence coverage aligned on the TRAC anti-BCMA CAR ssCTS reference construct. Grey bars on Y axis indicate sequence coverage. Low coverage across the CTS indicates relatively rare non-HDR events incorporating the indicated bases. Panel a was performed with 3 independent healthy human blood donors. Open circles represent use of serum-free media post-electroporation, closed circles represent use of serum-containing media post-electroporation. Panel c performed with the indicated number of mice using T cell products generated from a matched single healthy blood donor. Panel d performed with 2 independent healthy human blood donors. Panel e-f performed with unmodified T cells (n = 4 mice) and BCMA-CAR T cells (n = 5 mice) generated from one healthy human blood donor. A second cohort of mice treated with cells from a second donor was excluded because tumor failed to efficiently engraft in control group. TLA Analyses performed in 2 independent healthy human blood donors. **P < 0.05; ns, not significant. P values obtained by (a) unpaired two tailed t-test, (c) two tailed Mann-Whitney test, or (f) log-rank Mantel–Cox test (survival). rAAV = recombinant adeno-associated virus, HDRT = homology-directed repair templates, RNP = ribonucleoprotein, TCR = T cell receptor, CTS = Cas9 target site, ssCTS = ssDNA + CTS HDRT, CAR = chimeric antigen receptor, GMP = good manufacturing practice, TLA = targeted locus amplification, LHA = left homology arm.

Fig. 1 |. Development of ssCTS templates for high yield knock-in.

a, Diagram of hybrid ssDNA HDRT designs incorporating CTS sites. b, Panel of ssCTS designs tested. c, Knock-in efficiency for each ssCTS design using a CD5-HA knock-in construct at 160 nM–4 μM concentration assessed by flow cytometry. Dotted line represents mean knock-in percentage for control ssDNA HDRTs without CTS (construct a, gray). d–f, Knock-in strategy, gating, knock-in efficiency, live cell counts and knock-in cell counts are shown for large ssCTS templates including a tNGFR knock-in at the IL2RA locus (d), a IL2RA-GFP fusion protein knock-in to the IL2RA locus (e) or two different HDRTs inserting a BCMA-CAR construct at TRAC locus via two different gRNAs (g526 and g527) (f). Each experiment was performed with T cells from two independent healthy human blood donors represented by individual dots plus mean. CTS, Cas9 target site; FITC, fluorescein isothiocyanate; ssCTS, ssDNA HDRT + CTS sites. This figure was generated in part using graphics created by Biorender.com.

Fig. 2 |. Evaluation of ssCTS design requirements.

a–e, Comparison of different CTS designs with an IL2RA-GFP knock-in construct targeting IL2RA locus assessed by flow cytometry. a, Comparison of CTS with a gRNA target sequence that is specific for the cognate RNP (+IL2RA CTS), an alternative gRNA sequence (+CD5 CTS), a CTS incorporating a PAM site and scrambled gRNA sequence (+scramble CTS) or an equivalent amount of dsDNA within the 5′ end of the homology arm (+end protection). b, Comparison of complementary oligos covering varying regions of the CTS and surrounding sequences (design schematics left, knock-in results right). Constructs with CTS sites on both 5′ and 3′ end (green bars), 5′ end only (blue bars) or 3′ end only (red bars) are shown on the right panel with two best performing designs indicated (right). c, Evaluation of varied 5′ ends including different length of buffer sequence upstream of the CTS site. d, Comparison of CTS designs with varying numbers of scrambled bases at the 5′ end of the gRNA target sequence using WT or SpyFi Cas9. e, Knock-in percentages are shown with varying length of homology arm covered by the complementary oligonucleotide. Each experiment was performed with T cells from two independent healthy human blood donors represented by individual dots + mean. All comparisons except for b include complementary oligos covering the entire 5′ buffer + gRNA + PAM + homology arms. HA, homology arms.

Fig. 3 |. Application of ssCTS knock-in templates across diverse target loci, knock-in constructs and primary human hematopoietic cell types.

a–c, Comparison of knock-in cell yields using ssDNA (red) and dsDNA HDRTs (blue) with CTS sites across a variety of primary human hematopoietic cell types. Note that different cell type comparisons were performed with different blood donors. All comparisons were performed using a knock-in construct generating an CLTA-mCherry fusion at the CLTA locus. Shown for each cell type using HDRT concentrations from 5 to 160 nM are knock-in cell count yields (a), maximum fold-change in knock-in count yields (relative to dsCTS templates) (b) and maximum percentage knock-in (c). d, Knock-in efficiencies for constructs targeting a tNGFR marker to 22 different target genome loci. e–g, Comparison of large ssDNA and dsDNA HDRTs with CTS sites for knock-in of a pooled library of polycistronic constructs targeted to the TRAC locus (2.6–3.6 kb total size). Shown for each HDRT variation is relative percentage knock-in in comparison to maximum for dsDNA + CTS templates (e), relative knock-in cell count yields in comparison to maximum for dsDNA + CTS templates (f) and representation of each library member in knock-in cells postelectroporation in comparison to construct representation in the input plasmid pool (g). h, Evaluation of ssCTS templates ± MT or MTX inhibitor combinations with a panel of 44 different knock-in constructs targeting a tNGFR marker across 22 different target loci including genes implicated in PID or with potential importance for T cell engineering. Two gRNAs and corresponding ssCTS templates were used for each gene (g1 and g2). All experiments were performed with T cells from two independent healthy human blood donors represented by individual dots + mean (a–g) or mean alone (h). dsCTS, dsDNA HDRT + CTS sites.

Fig. 4 |. Whole ORF replacement at target genes for therapeutic and diagnostic human T cell editing. a–d, IL2RA exon 1–8 ORF replacement strategy.

a, Diagram of the IL2RA gene with reported patient coding mutations and knock-in strategy using an IL2RA-GFP fusion protein targeted to exon 1. The S166N mutation examined in c and d is noted. b, IL2RA and GFP expression in CD4⁺ T cells electroporated with IL2RA-GFP ssCTS templates and cognate RNP followed by MTX inhibitor combination (green), in comparison to RNP only (red) or no electroporation control cells (blue). c, Comparison of extracellular (surface staining in nonpermeabilized cells) or total IL2RA expression (staining in permeabilized cells includes total surface and intracellular protein) with WT and S166N IL2RA-GFP knock-ins assessed by flow cytometry. Percentage IL2RA⁺ is shown for each panel. d, Subcellular localization of WT and S166N IL2RA-GFP protein by fluorescence microscopy. Scale bars, 20 μm. e–i, CTLA4 exon 2–4 ORF replacement strategy. e, Diagram of the CTLA4 gene with reported patient mutations and knock-in strategy using a CTLA4-GFP fusion protein targeted to intron 1. The R70W, R75W, T124P mutations examined in g–i are noted. f, CTLA4 and GFP expression in CD4⁺ T cells electroporated with CTLA4-GFP ssCTS templates and cognate RNP followed by MTX inhibitor combination (green), in comparison to RNP only (red) or no electroporation control cells (blue). g, Quantification of percentage knock-in for WT, R70W, R75W and T124P constructs electroporated with ssCTS templates and treated with the MTX inhibitor combination assess by flow cytometry. Mean and individual values are shown for two independent healthy human blood donors. h, Structure of CTLA4 dimer with CD80/86 interaction domain highlighted (yellow) along with location of R70W (blue), R75W (orange) and T124P (green) mutations. i, Comparison of extracellular CTLA4 (surface staining), total CTLA4 (staining in permeabilized cells, which includes total surface and intracellular protein) and biotinylated recombinant CD80 ligand interaction stained with Streptavidin-APC in WT, R70W, R75W and T124P knock-in CD4⁺ T cells. DAPI, 4′,6-diamidino-2-phenylindole nuclear stain; rCD80, recombinant CD80.

Fig. 5 |. GMP-compatible process for nonviral CAR-T cell manufacturing.

a, Diagram of nonviral CAR-T cell manufacturing process. T cells are isolated from peripheral blood and activated on day 0 with anti-CD3/anti-CD28 Dynabeads, IL-7 and IL-15. Cells are electroporated using the Maxcyte GTx electroporator on day 2 with Cas9 RNPs+ssCTS HDRTs and then expanded for a total of 7–10 days using G-Rex 100M culture vessels supplemented with IL-7+IL-15. b, Representative day 10 flow plots showing BCMA-CAR knock-in for control (no inhibitors), M3814 and MT conditions. c, BCMA-CAR knock-in rates on days 7 and 10 for each condition. d, Absolute number of CAR⁻ cells on days 7 and 10. Gray box highlights anticipated patient doses of 50–400 × 10⁶ CAR⁻T cells. e, T cell immunophenotypes on day 10 based on CD45RA and CD62L expression. f, In vitro killing of BCMA+MM1S multiple myeloma cell lines in comparison to unmodified T cells from same blood donors. Experiments performed with T cells from two independent healthy human blood donors represented by individual dots + mean (c,d,f) or mean alone (e); a was generated in part using graphics created by Biorender.com. M, M3814; Tscm, T stem cell memory; Tcm, T central memory; Tm, T effector memory; Teff, T effector.