Using CRISPR-Cas12a for precise gene editing, gene disruption, and efficient gene insertion in human immune cells

In mammalian cells, transfection of endonuclease and crRNA RNP complexes enables high-efficiency genome editing with reduced off-target editing, compared to plasmid-based systems that constitutively express the CRISPR protein and crRNA.^23,24 To enable genome editing in human cells, a codon-optimized MAD7 sequence²⁵ was flanked with either a single or quadruple nuclear localization signal (NLS) (Figure S1a↗). The in-cellulo editing activity of the modified MAD7 nuclease was determined by nucleofection of RNPs in a T-cell leukemia cell line (Jurkat), followed by amplification of the edited part of the locus, targeted next-generation sequencing for identification of the edits, and finally computational analysis of modification frequency using the CRISPResso2 algorithm.²⁶

Firstly, using a crRNA targeting the DNMT1 locus, we compared the editing efficiency of different RNP amounts with either 1 × NLS or 4 × NLS MAD7. We observed RNP concentration-dependent modification efficiency, as evidenced by an increased fraction of modified amplicons with increased amounts of RNP complexes (Figurea). Notably, editing was markedly enhanced by use of 4 × NLS compared to 1 × NLS tagged MAD7, which indicates that optimization of the NLS is important to achieve maximal editing efficiency. Moreover, we observed that the fraction of modified amplicons increased along with the concentration of RNP, while the viability of cells concurrently decreased with the increasing concentration of RNP complexes (Figurea). In summary, the highest modification frequency and cell viability were achieved using 100 pmol 4 × NLS-MAD7 per reaction (2:3 MAD7:crRNA).

To optimize in cellulo editing activity in our model system, we tested 96 different transfection conditions—using 32 nucleofection programs in combination with three buffers—on the Lonza Nucleofector 96-well Shuttle System. The majority of the buffer-program transfection combinations resulted in suboptimal viability and modification frequency; however, our analysis revealed seven out of 96 conditions that supported substantial rates of both cell viability and editing efficiency (both ≥70%; Figure S1b↗). The two best transfection conditions observed in the screen, namely, SF-CA-137 and SG-CA-138, were then validated and compared to the Lonza-recommended nucleofection programs for Jurkat cell lines, namely, SE-CL-120 and SE-CK-116. We established that the optimal modification frequency and cell viability are achieved using the SF-CA-137 transfection condition (Figureb).

In this study, we used the Jurkat cell line as a model system to screen crRNAs for high-efficiency editing sites. The screen included 298 unique crRNAs targeting the immune checkpoint receptors PDCD1, TIM3, LAG3, TIGIT, and CTLA4; the checkpoint phosphatases PTPN6 (SHP-1) and PTPN11 (SHP-2); and the TCR signaling subunit CD247 (CD3ζ) (Table S1↗). The activity of the crRNAs was determined by targeted NGS and CRISPResso2 data analyses.

CRISPResso2 software reports the frequency of modifications (insertions, deletions, and substitutions) within a quantification window flanking the position of MAD7-induced cleavage in the amplicon sequence. To better understand the detection of editing events, we first compared the type of modifications detected in 230 amplicons that were sequenced in both crRNA-treated and MOCK’ samples (no MAD7). We observed relatively high modification frequencies (median 1%) in MOCK’ reactions as a result of high frequency of substitutions (Figurec); substitutions were detected at a median frequency of 0.96%, likely due to the errors in NGS base calling or substitutions arising during DNA amplification, while insertions and deletions were found at much lower median frequencies of 0.003 and 0.042%, respectively. Thus, we used the frequency of both insertions and deletions (INDEL) to quantify the editing activity of the CRISPR-MAD7 system. Moreover, we observed that low INDEL frequencies in MOCK’ reactions enabled the sensitive detection of editing events at a significantly greater fraction of sites (P = 3e^–12). Analysis of crRNAs with low INDEL frequencies showed statistically significant editing in crRNA-treated samples compared to MOCK’ samples at INDEL frequencies as low as 0.5% (P = 4e^–8; Figured). This indicates the sensitivity of the assay to detect modifications in the sub-1% range. The low INDEL frequency in MOCK’ amplicons also suggests that the controls are largely superfluous when screening for active crRNAs with ≥10% INDEL frequency and can be reserved for validation experiments.

Since MAD7 is described to target a wide range of PAM sequences, we screened crRNAs adjacent to all YTTN PAM variants, which enabled the analysis of MAD7 specificity in mammalian cells. Our data indicate that MAD7 edited sites with all eight combinations of YTTN PAM; however, editing was significantly higher at the YTTV and TTTV consensus sequences (P = 2 × 10^–3 and P = 2 × 10^–4, respectively). While the majority of highly active crRNAs (≥50% INDEL frequency) were found at sites with YTTV and TTTV PAMs, moderately active crRNAs (≥10% INDEL frequency) were found to target every PAM sequence with the exception of CTTT. This indicates that MAD7 can edit a wide range of target PAMs, albeit at reduced frequencies (Figuree).

Given the large number of crRNAs analyzed, we next determined if the targeted DNA sequence biases editing efficiency. Sequence logos were made to compare the DNA–complementary crRNA sequences of inactive (≤1% INDELs), active (1–10% INDELs), moderately active (10–50% INDELs), and highly active (≥50% INDELs) crRNAs (Figure S2a↗). While there were no strong biases for ribonucleotides at specific positions, guanine appeared to be overrepresented and uracil underrepresented on moderately active and highly active crRNAs. Next, we examined the frequency of ribonucleotide bases within the same four classes of crRNAs (Figure S2b↗), and the analysis confirmed significant enrichment of guanine and depletion of uracil on highly active crRNAs. Moreover, the data showed that nearly 40% of inactive crRNAs had runs of three or more adenine or uracil ribonucleotides, while none of the highly active and ≤20% of moderately active crRNAs contained such runs (Figuref). While these sequence features cannot determine crRNA activity, they may be useful for selecting putative high-activity crRNAs during initial rounds of screening and could reduce the overall cost of identifying crRNAs for various genes of interest or increase the overall scale of genes studied.

We next validated high-efficiency crRNAs identified in our initial analysis by assaying INDEL frequency for the top three or five crRNAs for each of the selected immunologically relevant genes (Figurea). In the validation experiment, the INDEL frequency was significantly correlated with the measurements from the initial screen, highlighting the reproducibility of the INDEL assay (Figure S3↗). Using the CRISPResso2 software, we then estimated the degree of open reading frame (ORF) disruption for each of the validated crRNAs (Figurea). In addition, for four highly active crRNAs targeting three different exons at the PDCD1 locus, we measured the surface expression of the PDCD1 protein by flow cytometry 4, 7, and 11 days post-transfection. The data revealed that the protein surface expression after transfection with crPDCD1_1, a crRNA targeting the PDCD1 gene at the extracellular domain of the protein, was as low as 4% 4 days post-transfection and remained at this level even at day 11 post-transfection. The surface expression after transfection with the remaining three crRNAs was significantly higher, 19% and 46% after transfection with crPDCD1_2, and both crPDCD1_3 and crPDCD1_4, respectively. It is important to emphasize that induced alterations in the proximity of the transmembrane domain (crPDCD1_2) or the cytoplasmic domain (crPDCD1_3 and crPDCD1_4) might not be reflected in the expression measurements of the surface protein (Figure S4↗). This is in line with the ORF data analysis, which showed that for most of the crRNAs including the highly active crPDCD1s, the predicted number of INDELs leading to frameshifts was similar to what we would expect from an unbiased DNA repair process, with frameshifts in two-thirds of the edited loci (Figureb). However, several of the crRNAs had a markedly different degree of ORF disruption; crCD247_4 resulted in frameshifts with 97% frequency, while crTIM3_1 and crTIM3_3 resulted in frameshifts with 23% and 44% frequencies, respectively (Figureb). The analysis of repair products indicates that in the case of crTIM3_1, and to some extent crTIM3_3, the bias arose from directly repeated sequences at the DNA cleavage site, which possibly promoted microhomology-mediated end joining (MMEJ) repair following DNA cleavage. These data help inform the selection of crRNAs for gene KO since some crRNAs, such as crTIM3_1, have a much lower frequency of gene disruption than would be predicted based on the frequency of INDEL formation.

Another consideration for selecting crRNAs is the potential for off-target cleavage events. We analyzed the list of validated crRNAs using the Cas-OFFinder software²⁷ to predict potential off-target editing sites in the genome with up to four mismatches between the crRNA and the target DNA sequence. Using the Bioconductor R packages,^{28 −31} we then matched the predicted off-target sites with the human gene database and extracted those sites that targeted exons and introns within the genes. Afterward, we examined the degree of editing activity at these sites by targeted next-generation sequencing, more specifically at 25 predicted off-target sites for the top-two PDCD1 crRNAs, i.e., crPDCD1_1 and crPDCD1_2. The analysis revealed low-level off-target activity at crPDCD1_2_13 and crPDCD1_2_15 sites; however, INDEL formation at these two sites was statistically insignificant compared to MOCK samples (non-targeting crRNA) (P ≥ 0.05; Figure). We further assayed the INDEL frequency at 43 putative off-target sites with up to three mismatches between crRNA and target DNA sequences for the top-two crRNAs targeting seven remaining genes (i.e., TIM3, LAG3, TIGIT, CTLA4, PTPN6, PTPN11, and CD247). The analysis revealed no detectable activity at any of the putative off-target sites (Figure), which reinforces the high cleavage fidelity of MAD7.

Insertion of exogenous transgenes is an important aspect of mammalian cell engineering. Gene insertion with CRISPR-Cas is achieved by HDR of CRISPR-induced DNA breaks using HDR-donor templates to integrate exogenous genetic sequences into targeted DNA loci. Several studies indicate that HDR templates, composed of linear double-stranded DNA, provide the most robust and efficient method of transgene insertion using CRISPR-Cas genome editing systems.^18,22

Here, we used the Jurkat cell line to evaluate the transgene insertion and expression efficiency using CRISPR-MAD7. We used a highly active crRNA targeting the AAVS1 safe-harbor locus (Figure S5↗; Vlassis et al. under review) and eight different HDR-repair templates flanked with symmetric homology arms (HAs) of 500 base pairs (bp) in the amount of 0.5 μg reaction^–1. The HDR inserts consisted of eight promoters differing in size and promoter strength to drive GFP expression (Figure S6↗). When the transient GFP expression diminished at day 14 post-transfection, we observed comparable insertion efficiencies with stable GFP expressions of up to 30% using four out of eight promoters, that is, JET (195 bp), PGK (511 bp), EF-1α (1195 bp), and CAG (1723 bp) (Figure S6↗), suggesting that the insert size had not affected the integration efficiency at AAVS1 in the Jurkat cell line. Subsequently, keeping the RNP amounts per reaction constant (100:150 pmol MAD7:crRNA), we evaluated the effect of various HA lengths (100 vs 500 bp) and HDR template amounts (0.125, 0.25, 0.5, and 1 μg) on the insertion efficiency using two selected promoters, namely, JET and EF-1α. We observed up to 30% higher integration efficiency with HDR templates flanked with HA of 500 compared to 100 bp. Moreover, the data showed improved insertion efficiencies with increasing amounts of HDR templates flanked with either 100 or 500 bp HA but at the same time somewhat reduced cell viability (Figurea).

To assess the potential of CRISPR-MAD7 as an editing platform for ex vivo cell therapies, we tested its performance in human primary cells. Using Pan-T cells isolated from the peripheral blood from three donors and the best procedure from the experiments above, i.e., 100:150 pmol MAD7:crRNA together with the 1 μg HDR template, in combination with 160 μg of poly-l-glutamic acid (PGA)²¹ per reaction, we evaluated the integration efficiency of a clinically relevant CAR transgene containing the JET or EF-1α promoter flanked with HA of 100 or 500 bp and a bovine growth hormone-derived polyadenylation sequence. We used an anti-CD19 CAR with fully human variable regions (Hu19CAR), CD8α hinge and transmembrane domains, a CD28 costimulatory domain, and a CD3ζ activation domain.³² We observed moderate insertion efficiency at AAVS1 but a stable CAR expression of up to 14% and 16% using HDR templates flanked with 100 and 500 bp HA, respectively. However, the normalized cell viability (normalized to MOCK) measured 24 h post-transfection ranged from 23% with JET-500-CAR, 65% with JET-100-CAR, 82% with EF-1α-100-CAR, to 95% with EF-1α-500-CAR. Mean cell viabilities (non-normalized) measured at day 7 and day 11 post-transfection exceeded 70% in all treatments (Figureb). It is important to emphasize that both CAR insertion efficiency and normalized cell viability significantly increased in the samples treated with PGA compared to the treatment without PGA (P ≤ 0.05; Figure S7↗).

We then re-evaluated multiple parameters of our procedure to further optimize primary T-cell post-transfection viability and CAR insertion efficiencies at AAVS1. Using Pan-T cells isolated from the blood from two donors, we systematically tested the effect of RNP amounts (with 160 μg of PGA) and EF-1α-500-CAR HDRT amounts on CAR insertion efficiency and cell viability (Figure S8↗). We observed that by reducing the RNP amount to 50:75 pmol MAD7:crRNA while increasing the HDRT amount to 1.5 μg reaction^–1, this resulted in improved CAR insertion efficiencies without substantially affecting cell viability (Figure S8↗). In addition, we observed that by using the abovementioned transfection conditions in combination with the cell recovery in a post-transfection cultivation medium pretreated with 2 μM M3814, this resulted in nearly five-times more efficient CAR insertion (Figure; patent application US63/315483) compared to the protocol above (see Figureb). In summary, utilizing our optimized CRISPR-MAD7 transfection procedure resulted in a CAR insertion efficiency of up to 85% 13 days post-transfection (mean of 50% across five donors), together with the normalized cell viability (normalized to MOCK) as high as 62% 24 h post-transfection (mean 52%). Mean cell viabilities (non-normalized) measured at days 7 and 13 post-transfection surpassed 90% (Figure).

The next generation of allogeneic T-cell therapies will involve multiplex genome editing to enhance the specificity, stealth, and persistence of the engineered T cells in vivo.³³ Here, we evaluated the efficiency of multiplex genome editing with the CRISPR-MAD7 system. We used Pan-T cells isolated from human peripheral blood (one donor) and a modified procedure for transgene insertion from Figure: that is, 50:75 pmol MAD7:crRNA (160 μg PGA) targeting AAVS1 for CAR insertion (knock-in, KI) and 50:75 pmol MAD7:crRNA (160 μg PGA) targeting either CTLA4, LAG3, PDCD1, TIGIT, CD247, PTPN6, or PTPN11 (Figure) for gene disruption (knockout, KO), together with 1.5 μg of HDRT and 2 μM M3814. We observed similar CAR insertion efficiencies of approximately 30% across all treatments, except for the double-edited LAG3 treatment (Figure, left panel). Moreover, we observed a significantly reduced expression of surface proteins CTLA4, LAG3, PDCD1, and TIGIT relative to their respective MOCK (Figure, middle panel); these induced alterations were further confirmed by the INDEL assay (Figure, right panel). Finally, while the expression of intracellular proteins CD247, PTPN6, and PTPN11 was not measured in this study, gene modification frequencies were evaluated with the INDEL assay, which revealed INDEL frequencies of up to 30% at the CD247 and PTPN6 loci and up to 50% at the PTPN11 locus (Figure, right panel). The non-normalized viability of double-edited cells at day 7 post-transfection exceeded 55% across all treatments, with cell viabilities of 59, 56, 64, 57, 64, 63, and 68% in the CTLA4, LAG3, PDCD1, TIGIT, CD247, PTPN6, and PTPN11 treatments, respectively (Figure, right panel).

The Jurkat human T-cell leukemia cell line (Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures GmbH (ACC 282)) was grown in RPMI 1640 medium (Thermo Fisher Scientific) with 10% heat-inactivated fetal bovine serum (FBS) (Thermo Fisher Scientific) supplemented with a 1% penicillin–streptomycin antibiotic mix (Thermo Fisher Scientific). Cells were cultured at 37 °C in 5% CO₂ incubators and maintained at a density of 0.5–1.5 × 10⁶ cells mL^–1; 24 h before transfection, the cells were passaged at 0.1 × 10⁶ cell mL^–1. A cell culture media supernatant was periodically tested for mycoplasma contamination using the MycoAlert PLUS Mycoplasma Detection Kit (Lonza).

This research was performed in accordance with the Declaration of Helsinki. Human peripheral blood was obtained from healthy adults after obtaining informed consent (Technical University of Denmark-Rigshospitalet National Hospital approval BC-40). T cells were isolated from the PBMC population by immune-magnetic negative selection using the EasySep Human T Cell Isolation Kit (STEMCELL Technologies). After isolation, T cells were activated in 25 μL ImmunoCult Human CD3/CD28/CD2 T-Cell Activator (STEMCELL Technologies) per 1 mL ImmunoCult-XF T Cell Expansion Medium (STEMCELL Technologies) containing 12.5 ng mL^–1 of Human Recombinant IL-2, 5 ng mL^–1 of IL-7, and 5 ng mL^–1 of IL-15 (STEMCELL Technologies) and seeded at 1.0 × 10⁶ cells mL^–1 until transfection 48 h later. The cells were maintained at 37 °C in 5% CO₂ incubators.

crRNAs were designed using the Benchling CRISPR Guide RNA design tool (https://benchling.com↗). Briefly, the software takes a target species, Ensembl transcript ID, and a base editor PAM variant as an input. Selecting the exon sequences, the software extracts the region around splice sites based on the specified window. Afterward, the pattern of N₂₀-NTTN also matched the reverse strand of the sequence. Finally, matched patterns are assigned off-target scores (lowest 0, highest 100) as well as on-target scores; however, due to lack of data, the on-target scores are currently not available for Cas12a PAMs.

MAD7 nuclease expression and purification followed Jedrzejczyk et al. (2022).⁴⁸

RNP complexes were generated by incubating relevant crRNAs with MAD7 in the molar ratio of 2:3 MAD7:crRNA for 15 min at RT immediately before transfection. For Jurkat experiments, the RNP complexes were generated by mixing the specific crRNA (150 pmol), MAD7 (100 pmol), and nuclease-free water up to 5 μL, unless otherwise stated. For T-cell experiments, 1.6 μL of an aqueous solution of 15–50 kDa PGA (100 μg μL^–1, Alamanda Polymers; modified from Nguyen et al. [²¹]) was added to crRNAs, followed by the addition of MAD7. For multiple editing T-cell experiments, the RNP complexes for KO and KI were prepared separately and as described above, followed by the addition of nuclease-free water to the final RNP volume of 9 μL.

HDR templates containing site-specific HAs, a promoter gene, and GFP or Hu19 scFv-CD8α-CD28-CD3ζ CAR gene were amplified from corresponding pTwist Ampicillin high-copy plasmids (Twist Bioscience) using HA-specific PCR primers. CMV- (pEGFP-C1, Addgene), SCP-,⁴¹ CMVe-SCP-,^41,42 PGK-,⁴³ CMVmax- (pmaxGFP, Lonza), JET-,⁴⁴ and EF-1α⁴⁵-driven templates were amplified in a two-step PCR program: initial denaturation at 98 °C for 30 s, cycle denaturation at 98 °C for 10 s, and extension at 72 °C for 30 s per 1 kb amplicon for 40 cycles with a hold at 72 °C for 10 min. Each 50 μL PCR reaction contained a 10 ng amplification template (plasmid DNA), 0.5 μM of HA-specific forward and reverse primers, nuclease-free water (IDT), 3% DMSO, and 1× Phusion High-Fidelity PCR Master Mix with HF Buffer (Thermo Fisher Scientific). PCR products were purified using the NucleoSpin Gel and PCR Clean-up Kit (Macherey-Nagel) with double 20 μL elution. Plasmids containing CAG⁴⁵-driven templates were propagated overnight in One Shot TOP10 Chemically Competent E. coli (Invitrogen, Thermo Fisher Scientific), and plasmid DNA was isolated and purified using anion exchange columns (HiSpeed Plasmid Maxi Prep, Qiagen). HDR templates were cut from the plasmid using restriction digestion, followed by purification on and isolation from agarose gel, and finally purified using NucleoSpin Gel (Macherey-Nagel). Purified HDR templates were collected and concentration measured on a NanoDrop One Microvolume UV–Vis Spectrophotometer (Thermo Fisher Scientific). Templates were up-concentrated using Amicon Ultra-0.5 mL 30 K centrifugal filters: 100 μg DNA per unit was transferred, filled with nuclease-free water to 500 μL, and centrifuged at 10,000 × g for 10 min to reduce the volume to 50 μL. DNA was washed twice with nuclease-free water and recovered into a fresh tube by unit inversion and centrifugation at 10,000 × g for 15 s. HDR templates were collected and diluted and concentrations quantified using Qubit dsDNA HS Assay Kit (Thermo Fisher Scientific). HDR donor templates of 0.5–1.5 μg reaction^–1 were used for cellular studies, unless otherwise stated.

A Lonza 4D-Nucleofector with Shuttle unit (V4SC-2960 Nucleocuvette Strips) was used for transfection, following the manufacturer’s instructions. For transfection, cells were harvested by centrifugation (200 × g, RT, 5 min) and re-suspended at 10 × 10⁶ cells mL^–1 (2 × 10⁵ cells 20 μL^–1) in the SF Cell Line Nucleofector X Kit buffer (Lonza), unless stated otherwise. The cell suspension was mixed with the RNPs, immediately transferred to the nucleocuvette, and transfected using the CA-137 Nucleofector program, except where indicated otherwise. After transfection, the cells were immediately re-suspended in the pre-warmed cultivation medium and plated onto 96-well, flat-bottom, non-treated plates (Falcon) and cultured at 37 °C in 5% CO₂ incubators and maintained at a density of 0.5–1.0 × 10⁶ cells mL^–1. After 48 h, the cells were harvested for the viability assay and genomic DNA, as described below. For the HDR template insertion, the HDR template was added to the cells and the suspension transferred to the RNPs immediately before transfection. The transfection parameters, cell recovery step, and proliferation conditions were as described above. The cells were harvested 48 h post-transfection for the viability assessment and after 7, 14, and 21 days for GFP insertion efficiency.

Forty-eight hours after isolation, the cells were harvested by centrifugation (300 × g, RT, 5 min) and re-suspended at 50 × 10⁶ cells mL^–1 (1 × 10⁶ cells 20 μL^–1) in the supplemented P3 Primary Cell Nucleofector Kit buffer (Lonza). The cells were mixed with HDRT, and the suspension was transferred to the RNPs immediately before transfection (Nucleofection program EH-115). After transfection, 80 μL of pre-warmed cultivation medium without IL-2 was added to the electroporation cuvettes. In the experiments with M3814 (Selleckchem), 80 μL of pre-warmed cultivation medium containing 2 μM of M3814 final concentration without IL-2 was added to the electroporation cuvettes. After a 10 min incubation at 37 °C, cells were transferred onto 96-well, flat-bottom, non-treated plates (Falcon) containing a pre-warmed cultivation medium pretreated with 2 μM of M3814 final concentration and 12.5 ng mL^–1 of IL-2. The cells were seeded at a density of 2.5 × 10⁶ cells mL^–1, or 1.3 × 10⁶ cells mL^–1 in the experiments with M3814, and kept at 37 °C in 5% CO₂ incubators. The viability assay was carried out 24 h post-transfection, after which the cells were reseeded in the fresh cultivation medium containing 12.5 ng mL^–1 of IL-2. The insertion efficiency of CAR was measured after 7 days, and 11 or 13 days post-transfection.

Flow cytometric assessments were carried out on a CytoFLEX S instrument (Beckman Coulter) using a 96-well plate format. Measurements of cell viability, exhaustion marker expression, GFP expression, and CAR expression were performed on 10,000 or 20,000 single-cell events in Jurkat or T cells, respectively.

For the cell viability and GFP knock-in measurements, approximately 250,000 cells per sample were transferred onto 96-well V-bottom cell culture plates and assessed following a series of consecutive washing and staining steps. The first step included centrifuging the cells at 300 × g for 5 min at RT, discarding the supernatant, and washing Jurkat cells or T cells in 150 μL Dulbecco’s PBS/2% FBS (STEMCELL Technologies) or Cell Staining Buffer (BioLegend), respectively, followed by the second centrifugation and removal of supernatant. The final step included viability staining of Jurkat cells or T cells using 150 μL Dulbecco’s PBS/2% FBS with 7-amino-actinomycin D (7-AAD, 1:1000; Thermo Fisher) or 50 μL Cell Staining Buffer with Zombie Violet dye (1:200; BioLegend), respectively. The fluorescence excitation of 7-AAD is at 561 nm (yellow-green laser) of Zombie Violet at 405 nm (violet laser) and of GFP at 488 nm (blue laser).

For detection of CAR or exhaustion marker expression in primary T cells, approximately 250,000 cells per sample were transferred onto 96-well V-bottom plates, centrifuged at 300 × g for 5 min at RT, and re-suspended in 50 μL Cell Staining Buffer (BioLegend) with either PE Anti-Myc tag antibody [9E10] (1:30; Abcam) and Zombie Violet dye (1:200; BioLegend) for 30 min for CAR expression and viability or PD-1 Antibody (J116) [PE/Cy5.5] (1:200; Novus Biologicals), CD152 (CTLA-4) Monoclonal Antibody (14D3), PE-eFluor 610 (1:40; eBioscience), TIGIT Monoclonal Antibody (MBSA43), APC, (1:40; eBioscience), CD223 (LAG-3) Monoclonal Antibody (3DS223H) or eFluor 506 (1:60; eBioscience), and Zombie Violet dye (1:200; BioLegend) for 30 min for exhaustion marker expression and viability. Afterward, the cells were washed in two subsequent washing steps using 150 μL Cell Staining Buffer and finally re-suspended in 100 μL Cell Staining Buffer for the flow cytometry measurements. See Figure S9↗ for the gating strategy.

For detection of PDCD1 knockout efficiency, approximately 250,000 Jurkat cells per sample were transferred onto 96-well V-bottom plates, centrifuged at 300 × g for 5 min at 4 °C, re-suspended in 100 μL Cell Staining Buffer (BioLegend) with APC/Cyanine7 anti-human CD279 (PD-1) Antibody (1:100; BioLegend), and incubated for 30 min at 4 °C in the dark. The next step included two repeats of centrifugation at 300 × g for 5 min at 4 °C, supernatant removal, and cell washing in 150 μL ice-cold Cell Staining Buffer (BioLegend). In the final step, the cells were re-suspended in 100 μL Cell Staining Buffer for the flow cytometry measurements using red laser (638 nm).

Cells were harvested 48 h post-transfection by centrifugation (1000 × g, 10 min) in 96-well, V-bottom plates (Greiner), washed with PBS (Sigma Aldrich), and lysed in 20 μL QuickExtract DNA Extraction Solution (Epicenter, Lucigen). DNA was extracted following the manufacturer’s protocol of 15 min at 65 °C, 15 min at 68 °C, and 10 min at 95 °C, cooled down, and stored at 4 °C. Genomic DNA was diluted 20× in nuclease-free water before amplicon PCR reactions.

Extracted genomic DNA was quantified using the NanoDrop (Thermo Fisher Scientific). Amplicons were constructed in two PCR steps: in the first PCR, regions of interest (150–400 bp) were amplified from 10 to 30 ng of genomic DNA with primers containing Illumina forward and reverse adapters on both ends, using Phusion High-Fidelity PCR Master Mix (Thermo Fisher Scientific) supplemented with 0.4 U Phusion Hot Start II High-Fidelity PCR Master Mix (Thermo Fisher Scientific) per reaction. Amplification products were purified with Agencourt AMPure XP beads (Ramcon), using the sample-to-bead ratio of 1:1.8. The DNA was eluted from the beads with nuclease-free water and the size of the purified amplicons analyzed on a 2% agarose E-gel using the E-gel electrophoresis system (Thermo Fisher Scientific). In the second PCR, unique pairs of Illumina-compatible indexes (Nextera XT Index Kit v2) were added to the amplicons using the KAPA HiFi HotStart Ready Mix (Roche). The amplified products were purified with Agencourt AMPure XP beads (Ramcon), using the sample-to-bead ratio of 1:1.8. The DNA was eluted from the beads with 10 mM Tris–HCl pH = 8.5 + 0.1% Tween 20. Sizes of the purified DNA fragments were validated on a 2% agarose gel using the E-gel electrophoresis system (Thermo Fisher Scientific), quantified using Qubit dsDNA HS Assay Kit (Thermo Fisher), and then pooled in equimolar concentrations. The quality of the amplicon library was validated using the Bioanalyzer High-Sensitivity DNA Kit (Agilent) before sequencing. The final library was sequenced on the Illumina MiSeq System using the MiSeq Reagent Kit v.2 (300 cycles, 2 × 250 bp, paired-end reads). De-multiplexed FASTQ files were obtained from BaseSpace (Illumina).

An initial quality assessment of the obtained reads was performed with FastQC36. The sequencing data were aligned and analyzed with the CRISPResso2 software,²⁶ using the CRISPRessoBatch command with the parameters --cleavage_offset 1 --quantification_window_size 10 --quantification_window_center 1 --expand_ambiguous_alignments for the INDEL frequency analysis. For the ORF disruption analysis, the CRISPRessoBatch command with the parameters --cleavage_offset 1 −coding_seq < EXON_SEQ > −-quantification_window_size 0 --quantification_window_center 1 --expand_ambiguous_alignments was used. Modification rates from the CRISPResso2 software output were analyzed in MS Excel.

Analysis was performed in RStudio, using either two-tailed T-test or Fisher’s exact test. The level of significance used was 0.05.