Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Jul 15.
Published in final edited form as: Nat Biomed Eng. 2025 Mar 7;9(8):1309–1319. doi: 10.1038/s41551-025-01372-1

T cell gene targeting and selection using non-viral intron knockins

Theodore L Roth 1,2,3,4,5,*, Johnathan Lu 1,2,3,4,5, Alison McClellan 1,2,3,4,5, Courtney Kernick 1,2,3,4,5, Oliver Takacsi-Nagy 1,2,3,4,5, Ansuman T Satpathy 1,2,3,4,5,*
PMCID: PMC12261302  NIHMSID: NIHMS2093231  PMID: 40055580

Abstract

The ability to insert new DNA sequences into defined sites in the genome of human T cells has facilitated more powerful basic research technologies and the engineering of clinical cell therapies. However, current methods target exonic regions, which limit the choice of integration sites and require the use of complex or harsh cell selection strategies. We developed a suite of DNA templates to enable non-viral intron targeting, where synthetic exons are targeted to endogenous introns and incorporated into mRNA transcripts of primary human T cells, causing knockout of the targeted gene preferentially in successfully edited cells. Non-viral intron knockin of a chimeric antigen receptors (CAR) to the T cell receptor alpha constant locus (TRAC) enabled efficient purification of >90% CAR+ T cell populations via negative selection of T Cell Receptor negative cells. Intron knockin templates could be engineered to extend to large >5 kb synthetic exons, use 5’ and 3’ exon splicing architectures that enabled simultaneous expression of the knockin and endogenous gene transcripts via alternative splicing, and using synthetic promoters with either endogenous or user defined gene regulation. The utility of non-viral intron knockins was generalizable across genomic target sites, with successful integrations across multiple introns within four distinct endogenous surface receptor genes. Thus, intron knockins increase flexibility of target sites for genetic engineering and enable simplified and high-throughput negative selection of successfully edited primary human T cells.

INTRODUCTION

RNA guided nucleases such as Cas9 and Cas12a have dramatically expanded applications of endogenous gene editing1,2. Combined with the addition of an exogenous DNA template, new genetic sequences can be inserted into defined sites in the genomes of primary human cells through homology directed repair (HDR)3. These targeted gene integrations have opened diverse research and therapeutic applications, from correcting disease causing mutations, to integrating new synthetic genes such as knockin of a chimeric antigen receptor (CAR) or T cell receptor (TCR) into the endogenous TCR alpha locus (TRAC) in primary human T cells4,5. The majority of current gene targeting methods integrate new genetic material within exons, knocking out the endogenous gene by targeting the middle of the coding sequence. Endogenous gene expression can be maintained by targeting the N or C terminus, although this limits guide RNA (gRNA) target sites, resulting in effective editing in only a subset of potential target genes6.

After gene targeting, the ability to purify or select for successfully edited cells is essential, and several methods exist to separate cells with a desired gene edit from unedited or incorrectly edited cells7. Classical selection methods include positive selection using fluorescently conjugated antibodies and cell sorting (FACS), positive selection with antibodies conjugated to magnetic beads, or drug selection using resistance genes. Recently, these selection methods have been supplemented by selection strategies taking advantage of the ability to target transgenes to specific endogenous loci, such as by integration within essential genes8–12. However, each of these selection methods has significant drawbacks: (1) positive selection requires direct cellular manipulation (such as antibody binding), which can leave bulky reagents bound to the cell surface after selection, inhibiting subsequent cellular functions and creating potential immunogenic antigens, (2) drug selections expose cells to toxic compounds and require the expression of foreign resistance genes, (3) both positive and drug selections require the introduction of additional genetic material beyond the desired therapeutic genetic sequence13, and (4) knockins to essential genes lose the benefits of choosing optimal endogenous gene regulatory circuits for a particular transgene (e.g. CAR knockins to TRAC4,5,14,15). ‘Touchless’ negative selections offer an ideal alternative, with the desired cells for selection never “touched” by antibody binding, small molecule drugs, or other manipulations, but for gene editing applications would require only successfully edited cells to lose expression of a selectable marker that all unedited and incorrectly edited cells retain.

Traditional gene insertion methods integrate new genetic material with an exogenous promoter to random (lenti/retrovirus) or non-expressed sites (safe harbors) or use gene editing to target an exon of an expressed gene so that the new sequence can be integrated into an existing mRNA under endogenous regulatory control. However, while these approaches are compatible with positive selection (by including additional DNA encoding a selection marker), they do not create the needed antigenic difference to enable negative selection. In the case of targeted editing of an essential gene’s exon, successfully HDR edited cells may lose expression of the targeted gene, but the majority of non-HDR edited cells will acquire NHEJ mutations16–18 or large deletions19–21 which also results in target protein loss, as demonstrated by knockin of a CAR to the first exon of TRAC in primary human T cells (Fig. 1a)4,5.

Figure 1. Non-viral TRAC intron knockins and negative selection of successfully edited CAR T cells.

Figure 1.

Open in a new tab

a, Traditional endogenous gene targeting approaches integrate a new transgene into endogenous exons, causing knockout of the endogenous gene, as shown by knockin of a GFP-CAR multicistronic cassette into the first exon of TRAC, causing TCR loss in Homology Directed Repair (HDR) edited cells. The majority of cells lacking a gene knockin show knockout of the TCR, due to Non-Homologous End Joining (NHEJ) and large deletions. Unedited Wild Type (WT) cells remain TCR positive.

b, Knockin of a synthetic exon containing a GFP-CAR multicistronic cassette flanked by splice acceptor and donor sites within the first intron of TRAC shows the majority of non-HDR edited cells continue to express the targeted endogenous gene (TCR+ CAR−), while HDR edited cells have largely lost endogenous gene expression (TCR− CAR+). Successfully HDR edited cells can thus be negatively selected by depletion of TCR (CD3) expressing cells.

c, Efficient knockins across TRAC Intron 1. Knockin efficiency of a synthetic exon containing a GFP-CAR multicistronic cassette across TRAC Intron 1 (avoiding a highly repetitive region within the intron) using both Cas9 and Cas12a nucleases. n = 2-4 unique donors per target site.

d, CD3 depletion purifies successfully edited cells after Cas9 (>80% purity) and Cas12a (>90% purity) intron knockins but provides only marginal enrichment for exon knockins due to undesired NHEJ edits causing TCR knockout. n = 4 unique donors. ns = not significant, ** = P<0.01, Two-sided paired t test.

e, In Vitro target cell killing by intron vs exon TRAC knockin CAR T cells. TRAC intron knockin CD19-28z CAR T cells kill Nalm6 target cells at same efficiency as TRAC exon knockin CAR T cells 48 hours post mixing at a 1:2 E:T ratio. Four technical replicates of n = 2 unique donors. ns = not significant, Two sided Mann-Whitney test.

We thus sought a generalizable method to address both issues: (1) limitations in gRNA target sites using exon targeting, and (2) limitations in existing selection strategies, particularly to enable negative selection of successfully gene edited primary human cells. Inspired by previous methods which integrated new genetic sequences into intronic regions, such as in gene trap mutagenesis experiments, affinity tagging, and certain gene therapies, among others22–25, here we present non-viral intron knockins, where synthetic exons are knocked into intronic regions of endogenous genes in primary human T cells. Successful knockin to an endogenous intron results in splicing of the inserted sequence into the mature endogenous mRNA transcript, disrupting the endogenous gene and resulting in loss of endogenous protein expression (e.g. a surface protein for selection purposes), as used historically in gene-trap random mutagenesis screening methods26,27. But in contrast to exon targeting, NHEJ edits within introns are predominantly spliced out of mature mRNA transcripts24, resulting in continued protein expression (except for rarer large multi kb deletions28). Thus, non-viral intron knockins enabled negative selection of successfully gene edited primary human T cells. Non-viral intron knockins were possible with even large synthetic exons, and detailed engineering of 5’ and 3’ splicing architectures enabled user control of the splicing behavior of the integrated synthetic exons, allowing for the targeted endogenous gene to be either knocked out or expressed depending on desired applications. Intron knockins were successful across genomic target loci in primary human T cells and offer a generalized gene targeting method enabling both greater flexibility in target site choice while also allowing for simpler negative selections of successfully edited cells.

RESULTS

Non-viral intron knockins at the TRAC locus in primary T cells

We first tested non-viral intron knockins at the TRAC locus in primary human T cells, seeking to integrate a CAR along with a GFP reporter as a synthetic exon into the first TRAC intron (Fig. 1b). We electroporated a CRISPR/Cas9 RNP with a gRNA targeting TRAC intron 1 along with a non-viral DNA homology directed repair template containing a GFP-CAR synthetic exon flanked by synthetic splice acceptor and donor sequences. Indeed, knockin of the GFP-CAR synthetic exon into the first intron of TRAC resulted in CAR-positive, TCR-negative cells, but unlike exon knockins, most CAR-negative cells remained TCR-positive (Fig. 1b). Negative selection of TRAC intron knockin CAR T cells by depletion of TCR (CD3) positive cells resulted in a dramatically enriched population of successfully-edited CAR T cells, with CAR positive cells increasing from 29.8% to 87.2% post selection. The resulting purified TRAC intron knockin CAR T cell population was TCR negative, which is beneficial for clinical applications4,5,14.

Non-viral intron knockin of a CAR at the TRAC locus proved to be successful across 18 tested gRNAs targeted different regions of TRAC Intron 1, and efficient gene integrations were possible using both Cas9 and Cas12a RNPs (Fig. 1c and Extended Data Fig. 1). Across four unique human donors, TRAC intron knockins enabled negative selection of CAR T cells to greater than 90% purity (Fig. 1d). In vitro target cell killing assays showed that TRAC intron knockin CARs killed approximately 90% of target antigen positive Nalm6 leukemia cells within 24 hours post co-incubation, the same (p > 0.05) as TRAC exon CAR knockins (Fig. 1e), which previous work has demonstrated shows similar or improved functionality compared to traditional lentiviral or retroviral mediated engineered T cell production4,5,29. Importantly, the post-editing viability, knockin efficiency, donor variability, and T cell activation profile as measured by CD69 and CD25 expression for non-viral TRAC intron knockin T cells were similar to previously reported non-viral TRAC exon knockins (Extended Data Fig. 2), which have successfully been used in human clinical trials5,29–31.

Finally, to directly compare different commonly used selection methods for primary human T cells, we designed a single synthetic exon that enabled selection of edited CAR T cells with four selection methods: (1) negative selection by anti-CD3 depletion, (2) positive selection using magnetic beads, (3) fluorescent cell sorting, and (4) drug selection with puromycin (Extended Data Fig. 3). All four selection methods resulted in enriched and functional CAR T cell populations, although only negative selection resulted in a population without residual TCR+ cells, with less than 0.1% of cells post selection still expressing the TCR (Extended Data Fig. 3b). Negative selection did not require antibody binding to the surface of the non-viral TRAC intron edited T cells, compared to previous positive selection methods that require antibody binding which can induce TCR or CAR cross-linking depending on their target32. Negatively selected CAR T cells showed a slight increase in in vitro proliferation after stimulation with target cells, and successfully killed target cells across tested E:T ratios (Extended Data Fig. 3c–d). Overall, TRAC intron knockins with negative selection enabled negative selection of successfully edited CAR T cells without integration of any additional expressed exogenous DNA.

Intron knockins can introduce large functional synthetic exons

We next determined the extent to which intron knockins could accommodate integration of large new genetic sequences. Many past intron targeting approaches have focused on addition of short (<1 kb) introns coding for small purification tags for biochemical experiments in cell lines23,33. We designed a series of increasingly large synthetic exons comprised of increasing numbers of fluorescent proteins and detectable cell surface reporters flanked by synthetic splice acceptor (SA) and splice donor (SD) sites, starting with a 2.5 kb GFP-CAR cassette and increasing to a 5.3 kb tNGFR-GFP-tEGFR-mCherry-CAR cassette. In primary human T cells, we found that non-viral TRAC intron knockins were able to efficiently integrate functional synthetic exons of greater than 5 kb (Fig. 2a–b). Increasing the size of intron knockins did show decreasing protein expression levels, although this was not specific to intron knockins as equivalently sized large knockins to TRAC exon 1 showed statistically similar (p > 0.05) lower levels of protein expression (Fig. 2c). Even knockin of a 2.4 kb template at TRAC exon 1 showed lower resulting expression levels than with knockin of the same sized cassette into TRAC intron 1, potentially due to the larger resulting size of the endogenous TRAC exon, now containing both the endogenous sequence and new synthetic sequence (Fig. 1a). Negative selection by CD3 depletion enabled enrichment of over 5kb intron knockins to greater than 50% purity on average (p < 0.01) (Fig. 2d), and TRAC intron-edited CAR T cells bearing large synthetic exons maintained the same in vitro target cell killing capacity as TRAC exon-targeted CAR T cells, successfully killing approximately 80% of target Nalm6 leukemia cells at 24 hours after co-incubation (p < 0.05) (Fig. 2e).

Figure 2. Introduction of >5 kb large synthetic exons with TRAC intron knockins.

Figure 2.

Open in a new tab

a, Synthetic exons of increasingly large sizes from 2.5 kb to 5.3 kb targeted to TRAC intron 1.

b, Flow cytometric plots demonstrating knockin efficiency of large synthetic exons at TRAC intron 1.

c, Increasing synthetic exon size shows decreasing transgene expression levels. Mean fluorescence intensity of GFP expression measured by flow cytometry in knockin positive cells after integration of increasingly large synthetic exons to TRAC intron 1. As the size of the synthetic exons increased, observed GFP expression decreased, potentially due to decreasing splicing efficiency of larger synthetic exons. n = 3 unique donors. ns = not significant, Two-sided paired t test.

d, Efficient knockin of increasingly large synthetic exons at TRAC intron 1 across three donors before and after negative selection by CD3 depletion. Even a 5.3 kb integration could successfully be enriched to greater than 50% knockin efficiency on average. n = 3 unique donors. * = P<0.05, ** = P<0.01, Two-sided paired t test.

e, In Vitro target cell killing with large synthetic exons. TRAC intron CAR T cells expressing a CD19-28Z CAR as part of increasingly large synthetic exons killed target Nalm6 tumor cells at equivalent efficiencies to TRAC exon CAR T cells with similar efficiencies to equivalent TRAC exon CAR T cells. Killing measured 48 hours after co-culture with a 1:2 E:T ratio. n = 2 unique human donors with 2 (Intron) or 4 (Exon) technical replicates each. ns = not significant, * = P<0.05, Two-sided unpaired t test.

Engineered control of synthetic exon splicing behavior after intronic knockin

We reasoned that whereas exon knockins can only remove the endogenous gene (unless targeted to the N or C terminus), intron knockins could be: (1) alternatively spliced into a coding transcript, resulting in gene knockout, or (2) skipped, enabling the endogenous mRNA to be expressed along with the synthetic knockin mRNA. We thus sought to determine whether alternative splicing of intron knockins could be programmed through engineering of the 5’ and 3’ end sequences of integrated synthetic exons. By removing the 3’ splice donor site and replacing it with a polyA sequence, the splice sites present in the pre-mRNA become unbalanced, with the preceding endogenous exon’s splice donor able to splice with the integrated synthetic exon (“synthetic splicing”), or with the downstream endogenous exon (“endogenous splicing”; Fig. 3a and Extended Data Fig. 4a). Indeed, while intron knockins with balanced splice sites resulted predominantly in loss of the targeted endogenous gene (GFP+ / TCR−), intron knockins with unbalanced splice sites caused most knockin positive cells to express both the gene encoded in the synthetic exon, as well as the targeted endogenous gene (GFP+ / TCR+; Fig. 3b). Having observed that the splicing behavior of synthetic exons integrated through non-viral intron knockins could be controlled through engineering of splice sites, we designed a series of constructs testing different sequence architectures at synthetic exon’s 3’ ends to determine the degree to which endogenous vs synthetic splicing could be controlled. The observed amounts of alternative splicing between the synthetic and endogenous exons could be tuned by polyA choice, with a shorter SV40 polyA resulting in more endogenous splicing than the longer WPRE polyA (p < 0.01), potentially due to greater ability of the longer polyA to halt transcription (Fig. 3c and Extended Data Fig. 4b). Rebalancing splicing through addition of a splice donor sequences after the polyA sequence similarly decreased alternative splicing (p < 0.01) (Fig. 3c and Extended Data Fig. 4b).

Figure 3. Engineered control of synthetic exon splicing using intron knockins.

Figure 3.

Open in a new tab

a, After intron knockin, successful splicing of a synthetic exon into the endogenous mRNA transcript disrupts the endogenous gene (“synthetic splicing”). If the synthetic exon is not spliced into the mature endogenous mRNA transcript, the new gene contained within the synthetic exon will not be expressed but endogenous gene expression is maintained (“endogenous splicing”). See Extended Data Fig 3a.

b, Replacing the synthetic exon’s 3’ splice donor sequence with a polyA creates unbalanced splicing, resulting in the majority of cells expressing both the synthetic gene, e.g. GFP (“synthetic splicing”, Blue Box) while also expressing the targeted endogenous gene, e.g. TCR/CD3 (“endogenous splicing”, Orange Box).

c, Varying the 3’ splicing architecture of synthetic exons allows different balances of synthetic splicing only (GFP+ TCR−) compared to synthetic and endogenous splicing (GFP+ TCR+). See Extended Data Fig 3b. n = 6 unique donors. ns = not significant, ** = P<0.01, Two-sided paired t test.

d, Altering degenerate bases of the synthetic exon’s 5’ end (within the 2A multicistronic element necessary to separate the synthetic gene from the preceding endogenous exon’s translation) to include either Exonic Splice Silencer (ESS) or Exonic Splice Enhancer (ESE) elements similarly enables control over the degree of synthetic vs alternative splicing.

e, Frequency of cells expressing both the synthetic knockin gene and the endogenous gene can be controlled by inclusion of ESS or ESE elements with the 5’ splicing architecture of synthetic exons. Note, the SV40 polyA construct in b-c contained the ESS element and its data is represented again for comparison with alternative 5’ splicing architectures. n = 2-6 unique donors. See Extended Data Fig. 3c.

Alternative splicing could also be further induced through manipulation of the 5’ end of the synthetic exon (Fig. 3d). Exonic Splicing Silencers (ESS) and Enhancers (ESE) are short ~6-8 bp degenerate sequences that are bound by SR proteins to control the efficiency of mRNA splicing at adjacent splice acceptor sites34,35. We optimized the degenerate codons of the 2A multicistronic element at the 5’ end of the integrated synthetic exon (which separates the gene encoded by the synthetic exon from the preceding fragment of the targeted endogenous gene), systematically adding or removing predicted ESS and ESE sequences36. Across multiple constructs, the ESS-2A sequence induced alternative splicing, with both the synthetic and endogenous gene being expressed in approximately 75% of knockin positive cells, whereas the optimized ESE-2A sequence predominantly resulted synthetic splicing only, with less than 20% of edited cells showing evidence of alternative splicing (Fig. 3e and Extended Data Fig. 4c). Combinatorial assessment of the tested 5’ and 3’ splicing architectures nominated optimal splicing design to bias towards either synthetic or endogenous splicing outcomes (Extended Data Fig. 4d).

Biallelic intronic knockins confirm alternative splicing of synthetic exons

To confirm that a single targeted allele could generate both synthetically spliced and endogenously spliced mRNA transcripts, we hypothesized that biallelic intron knockins could be used to directly evaluate whether a single allele after intron knockin could alternatively splice to produce both the new synthetic gene as well as the unaltered endogenous gene (Fig. 3). Simultaneous intron knockin of two synthetic exons containing two separate fluorescent proteins (GFP and RFP/mCherry) enabled the identification of dual positive cells with biallelic edits. With intron knockin of synthetic exons with balanced splice sites, gating on dual GFP+ / RFP+ cells showed that less than 5% of the dual positive cells still expressed the targeted endogenous TCR gene (Fig. 4a). In contrast, intron knockin of synthetic exons containing optimized Exonic Splicing Silencing elements at the 5’ ends and a polyA at the 3’ end (creating unbalanced splice sites) showed that almost 70% of the GFP+ / RFP+ positive cells were TCR+ (Fig. 4b). As both TRAC alleles possess a knocked in synthetic exon in the dual positive cells, one or both alleles must be capable of alternative splicing to maintain expression of the endogenous TCR. Finally, intron knockin of synthetic exons containing Exonic Splicing Enhancer elements at the 5’ end did not show any evidence of alternative splicing, with almost all dual GFP+ / RFP+ positive cells negative for the endogenous gene (Fig. 4c). The ability of a handful of silent degenerate basepair changes at the 5’ end of the synthetic exon (ESS → ESE element) to drastically change the degree of dual positive cells expressing the endogenous gene (TCR) further supports that a single allele with an intron knockin is capable of either expressing only the gene encoded in the synthetic exon (using splicing architectures in Fig. 4a and Fig. 4c), or is capable of alternative splicing and expression of both the synthetic gene and the endogenous gene (Fig. 4d). Off-target integration analysis further confirmed that the biallelic intron knockin cells expressing the endogenous TCR (as in Fig. 4b) were due to alternative splicing rather than off-target effects (Fig. 4e).

Figure 4. Biallelic TRAC intron knockins demonstrate alternative splicing of synthetic exons.

Figure 4.

Open in a new tab

a, Biallelic knockin of synthetic exons containing either GFP or RFP with balanced splice sites. Gating on dual positive cells showed almost none of the dual GFP positive / RFP positive cells still expressed the endogenous gene (TCR, measured by CD3 staining).

b, Biallelic knockin of a synthetic exon containing Exonic Splicing Silencing elements at its 5’ end and a polyA at its 3’ end. Gating on dual positive cells showed the majority of dual GFP positive / RFP positive cells still expressed the endogenous gene (TCR, measured by CD3 staining).

c, Biallelic knockin of a synthetic exon containing Exonic Splicing Enhancer elements at its 5’ end. Gating on dual positive cells showed almost none of the dual GFP positive / RFP positive cells still expressed the endogenous gene (TCR, measured by CD3 staining).

d, Quantification of the percentage of dual GFP positive / RFP positive cells that continue to express the endogenous gene across the three tested splicing architectures in a-c. n = 3-4 unique donors.

e, The degree of off-target integrations is important for interpreting biallelic integration data, as dual positive cells also expressing the endogenous gene (GFP / RFP / TCR positive cells) could be due to one knockin allele, one unedited endogenous allele, and then an off-target integration of the second fluorescent protein. The degree of off-target vs on-target integrations was assessed by electroporation of the GFP DNA Homology Directed Repair Template used in b without its accompanying RNP. While a low amount of GFP expression was seen in ~1-2% cells with HDRT only electroporations, likely due to off-target integrations, this was markedly less than the knockin rate and expression level of GFP when the HDRT was electroporated with its on-target RNP, similar to prior studies5. The low frequency of off-target integrations could not account for the degree of endogenous gene (TCR) positivity in the dual GFP / RFP positive cells in b. n = 3 unique donors.

Non-viral intron knockins are effective across genomic loci and T cell types

Finally, to assess whether non-viral intron knockins can be used as a generalizable technique for targeting across the genome, we designed intron knockin templates for three highly expressed surface receptors, the TCR complex member, CD3E, Beta-2 Microglobulin (B2M), and CD47 (Fig. 5). Efficient knockin was observed across all three targets, and the majority of cells that acquired a knockout of the targeted surface receptor expressed the knockin reporter tNGFR (Fig. 5a–c). At CD3E introns, average knockin efficiency ranged from approximately 10% to 40% depending on the gRNA used, and the percentage of CD3 negative cells that were knockin positive relative to knockin negative ranged from approximately 75% to 97% (Fig. 5a). At the targeted B2M intron, average knockin efficiency ranged from approximately 10% to 50% depending on the gRNA used, and the percentage of B2M (HLA-A/B/C) negative cells that were knockin positive relative to knockin negative ranged from approximately 95% to 97% (Fig. 5b). At CD47 introns 2 and 4, knockin efficiency ranged from approximately 20% to 40% depending on the gRNA used, and the percentage of CD47 negative cells that were knockin positive relative to knockin negative ranged from approximately 84% to 94% (Fig. 5c). Moreover, efficient intron knockin and resulting knockout of the targeted surface receptor was possible across multiple introns of CD3E and CD47 and the single tested intron of B2M, and across both CD4 and CD8 T cell types (Fig. 5a–c). Knockin across multiple targets exemplified how intron knockins take advantage of endogenous regulatory circuits, with intron knockin to the highly expressed B2M locus showed greater expression levels than CD3E or CD47 (Fig. 5a–c). For applications desiring greater expression than that driven by an endogenous gene locus, TRAC intron knockin of a GFP cassette with expression driven by an exogenous promoter showed a twenty-fold increase in expression compared to expression from the endogenous TRAC promoter (p < 0.001), while maintaining the ability of negative selections to enrich for cells with successful intron knockins (Fig. 5d).

Figure 5. Efficient non-viral intronic knockins across three genomic loci and distinct T cell types.

Figure 5.

Open in a new tab

a, Non-viral intron knockin within Intron 4 and Intron 5 of CD3E, a member of the TCR signaling complex. Two gRNAs at each insertion site were tested, and efficient knockin was observed for all four tested gRNAs. Knockin at the tested CD3E Intron 4 site appeared to be more efficient that the tested CD3E Intron 5 site. Knockin efficiency was similar in both CD4+ and CD8+ primary human T cells. n = 2 unique donors.

b, Non-viral intron knockin at two different sites within Intron 1 of B2M. Observed expression of the knocked in tNGFR surface receptor was observably higher for B2M intron knockins than other tested loci. Due to extremely high endogenous expression levels of B2M, some successful intron knockin cells cluster at the maximum recorded fluorescence values. n = 2 unique donors.

c, Non-viral intron knockin within Intron 2 and Intron 4 of CD47. Similarly to other tested endogenous loci, efficient knockin was observed across both CD4 and CD8 T cells, with the vast majority of CD47 negative cells possessing the tNGFR knockin. n = 2 unique donors.

d, TRAC intron knockin of a GFP template with splicing architecture to integrate into an endogenous mRNA transcript, with expression driven by the endogenous TCR-α promoter in comparison to intron knockin of a template containing an exogenous EF1-α core promoter driving expression of GFP. Intron knockin of an exogenous promoter allowed for the expression level of the knocked in GFP to be significantly increased, while still maintaining the enrichment of GFP+ cells within the CD3 (TCR) negative population. n = 3 unique donors, *** = P<0.001, Two-sided paired t test.

DISCUSSION

Methods to integrate synthetic DNA elements into intronic sequences have previous been used for diverse applications, from gene trap mutagenesis22, to gene knockouts by disrupting splice sites37, to gene therapy and model and commercial organism engineering . We have shown how intron knockins with optimized splicing architectures and in combination with simple selection protocols expand the methodological toolbox for endogenous gene targeting in primary human T (Fig. 6). First, intron knockins offer significantly more potential gRNA sites (~30% of the genome is intronic compared to <3% coding) to target endogenous genes than conventional exonic targeting strategies. The greater targeting flexibility of intron knockins may expand the search space for potential gene therapy applications requiring targeting of a specific mutated locus3,40, although an appreciable portion of intronic regions are made up of repetitive elements that may not be conducive to targeted gene editing41 . Second, intron knockins offer greater flexibility in the control of endogenous gene expression through engineered control of alternative splicing of integrated synthetic exons. When expression of the targeted endogenous gene needs to be maintained, exon knockins can only target the N or C terminus, drastically limiting available gRNA options (Extended Data Fig. 5a). In contrast, Intron knockins with engineered alternative splicing allow for target site selection across a gene’s entire intronic region (Extended Data Fig. 5b).

Figure 6. Intron knockins enable flexible endogenous gene targeting with simplified selection.

Figure 6.

Open in a new tab

a, Intron knockins expand the toolbox for gene targeting methods. Intron knockins with 5’ and 3’ architectures that can alternatively splice between synthetic and endogenous exons allow more flexible targeting of synthetic genes under endogenous regulatory control, and intron knockins with splicing architectures resulting in only synthetic splicing enable negative selection of successfully gene edited cells.

Intron knockins uniquely enable successfully edited cells to be purified by negative selection. By using intron knockins to introduce a synthetic exon with engineered splicing architectures that prevent alternative splicing, the targeted endogenous gene is knocked out (Extended Data Fig. 5c–d). While exonic knockins also generally cause knockout of the targeted endogenous gene, unsuccessfully edited cells without gene knockins largely have NHEJ mediated knockout of the targeted endogenous gene (Extended Data Fig. 5c). In contrast, the target sites of intron knockins are spliced out of mature mRNA transcripts, allowing them to tolerant of small NHEJ mediated mutations while still maintaining endogenous gene expression. With appropriate splicing architectures, intron knockins can result in successfully edited cells losing expression of the targeted endogenous gene while unedited cells or NHEJ edited cells maintain expression (Extended Data Fig. 5d). When the targeted gene is a surface receptor, successfully edited intron knockin cells can be negatively selected based on this selective loss of surface receptor expression. Application of intron knockins in primary human T cells generated TRAC intron knockin CAR T cells that were negatively selected to >90% purity without the need for additional bulky transgenes or requiring disruption of essential endogenous genes, and with complete removal of TCR positive cells

As non-viral TRAC exon knockins have now entered clinical trials for a variety of engineered TCR and CAR T applications30,31,42–44, non-viral TRAC intron knockins may offer a simpler clinical manufacturing method to select for pure clinical cell therapy products. Intron knockins also offer potential advantages for future in vivo targeted editing applications. Unlike exonic knockins, intron knockins do not need to be inserted at an exact basepair target site, expanding the potential range of editors that can be used for integration to targeted but inexact transposases, integrases, and recombinases45–49. Overall, Intron knockins offer a simple gene targeting strategy that overcomes key problems with prior selection methods for gene edited cellular therapies, while also dramatically expanding the flexibility of endogenous genetic manipulations.

METHODS

Primary Human T Cell Isolation and Culture

PBMCs from healthy human blood donors were collected under an approved IRB protocol by the Stanford Blood Center and used to isolate human T cells. Briefly, leukoreduction chambers (LRS) from processing of platelet donations were used to isolate PBMCs using density centrifugation with Ficoll (Lymphoprep, StemCell) within SepMate tubes (StemCell) according to manufacturer’s instructions. Next, primary human CD3 positive T cells were isolated by negative selection using Human CD3 T Cell Enrichment kit (StemCell) according to manufacturer’s instructions. Isolated primary human CD3 T cells were counted using an automated cell counter (Countess, Thermo), and activated using anti-human CD3/CD28 dynabeads (Cell Therapy Systems, Thermo) at a 1:1 ratio in XVivo 15 media (Lonza) supplemented with 5% FBS (MilliporeSigma) and 50 U/mL of human IL-2 (Peprotech). T cells were activated at 1:1 ratio of cells to dynabeads, and initially cultured in standard tissue culture incubators at approximately 1e6 cells / mL media. After gene editing/electroporations, T cells were counted and reseeded at approximately 1e6 cells / mL XVivo 15 media with fresh IL-2 every 2-3 days.

Non-viral gene knockins

Two days after activation, human T cells were harvested, dynabeads were magnetically removed by incubating for two minutes at room temperature on a magnet (EasySep Magnet, StemCell), and cells were counted using an automated cytometer. For electroporations, one million T cells per editing condition were gently pelleted by centrifugation at 90G for 10 minutes, followed by careful aspiration of the supernatant. T cell pellets were resuspended in 20 uL per editing condition in P3 Buffer (Lonza) and then mixed with prepared RNP and DNA HDRT templates. For each Cas9 knockin condition, RNPs were prepared by first complexing the gRNA by mixing 0.375 uL of 200 uM tracrRNA (IDT) with 0.375 uL of 200 uM crRNA (IDT) and incubating for 15 minutes at room temperature. Next 0.25 uL of 100 mg/mL PGA (15–50 kDa poly(L-glutamic acid); MilliporeSigma) was then added to the complexed gRNA and mixed by pipetting up and down. Next 0.5 uL of 40 uM SpCas9 (UC Berkeley MacroLab) was then added, mixed by pipetting up and down, and incubated for 15 minutes at room temperature to form the final Cas9 RNP. For Cas9 knockins, 20 uL of T cells were mixed with 1.5 uL of RNP (20 pmols total RNP) and 4 uL of plasmid DNA HDR Template at 1 ug/uL (4 ugs total HDRT). For each Cas12a condition, RNPs were prepared by first mixing 0.4 uL of 200 uM Cas12a gRNA (IDT) with 0.2 uL of 100 mg/mL PGA and pipetting up and down. Next 0.4 uL of 60 uM AsUltraCas12a (UC Berkeley MacroLab) was added and mixed by pipetting up and down, followed by incubation at room temperature for 10 minutes. For Cas12 a knockins, 20 uL of T cells were mixed with 1 uL of Cas12a RNP (24 pmols total RNP) and 4 uL of plasmid DNA HDR Template at 1 ug/uL (4 ugs total HDRT). For both Cas9 and Cas12a knockins, T cells were electropoated on a Gen2 Lonza 4D electroporation/nucleofection system using 96 well plate attachment and 20 uL cuvettes, using pulse code EO-151. Immediately following electroporation, 75 uL of pre-warmed XVivo 15 media was added to each cuvette, and cells were rested within the cuvettes for 15 minutes in a standard 37C Tissue Culture Incubator prior to moving to culture plates or flasks. An annotated list of all gRNA sequences used in the study is available in Supplementary Table 1.

DNA Constructs and HDR Templates

All DNA constructs used in the study were generated using standard cloning methods, primarily using PCRs (Q5 ultra II HotStart Polymerase, NEB), Gibson Assemblies (NEBuilder HiFi DNA Assembly Master Mix, NEB) and bacterial transformations (Mix & Go! Competent Cells - DH5 Alpha, Zymo). DNA constructs were sequence confirmed by Sanger Sequencing (Elim Bio) or whole plasmid next-generation sequencing (Primordium). For gene editing, plasmid Homology Directed Repair Templates were produced by bacterial culture and standard plasmid preparation (Zymo Midiprep and Maxiprep kits) according to manufacturer instructions, including endotoxin removal steps. Final plasmid HDR templates were eluted in molecular grade water, quantified (Nanodrop), and diluted to final concentrations of 1 ug/uL for electroporations. An annotated list of all DNA constructs used in the study, as well as full DNA sequences for all constructs, is available in Supplementary Table 2. Annotated genebank files for all DNA constructs used in the study are available upon request.

Flow Cytometry

All flow cytometric data shown was acquired either within the Stanford Blood Center Flow Cytometry Core using a BD Fortessa analyzer, or using a Biorad ZE5 analyzer. All antibodies used for flow cytometric experiments and staining dilutions are listed in Supplementary Table 3. Briefly, for flow cytometric staining, samples of approximately 100,000 T cells per analyzed condition were placed into round bottom 96 well plates and centrifuged for 5 minutes at 300G. After discarding the supernatant, each well was resuspended in 20 uL of FACS Buffer (PBS + 2% FBS) mixed with desired antibodies at indicated dilutions (Supplementary Table 3), and incubated at 4C for 20 minutes in the dark. Cells were washed once with FACS Buffer and resuspended in FACS Buffer for analysis. Flow cytometric data was analyzed using FlowJo v10 software.

Negative Selection by CD3 Depletion

Negative selections for successfully intron edited T cells were performed by CD3 depletion. Briefly, six days after non-viral gene editing by electroporation, subsequently expanded T cells were harvested, counted, and anti-CD3 biotinylated antibodies were added, following manufacturer’s instructions (EasySep™ Human CD3 Positive Selection Kit II, StemCell). After addition of magnetic beads and incubation at room temperature on a magnet, the supernatant containing untouched, CD3 negative cells was poured off into a new tube. Magnetic beads were added a second time according to manufacturer’s instructions, and again the supernatant was poured off after incubation on a magnet at room temperature, resulting in a final population of negatively selected, untouched, CD3 negative T cells.

Positive Selection by ~tNGFR Magnetic Bead Purifications

Positive selection for successfully intron edited T cells were performed by antibody binding of the surface expressed selection marker tNGFR followed by conjugation to a magnetic bead for magnetic purification six days after non-viral gene editing. Briefly, an anti-tNGFR biotinylated antibody (Supplementary Table 3) was bound to the surface of T cells using the same staining procedure as described above for flow cytometric fluorescent antibody staining. After antibody binding, T cells were incubated with magnetic beads coated with streptavidin (Dynabeads MyOne Streptavidin T1, Thermo) at 1.25 uL of beads per 1e6 cells. T cells were then incubated at room temperature on a rotating mixture for 15 minutes. Bound T cells were then placed onto a magnet and incubated for 5 minutes before pouring off the supernatant. T cells were resuspended in FACS Buffer and placed on the magnet a second time. After pouring of the supernatant, T cells were resuspended in XVivo 15 culture media containing 1 uM free biotin (D-Biotin, Thermo) for 2 days to competitively compete for streptavidin binding sites and help to remove bound magnetic beads from the T cell surface before returning cells to standard culture conditions.

Fluorescent Cell Sorting

Sorting based selection of successfully intron edited T cells was performed using standard fluorescence activated cell sorting six days after non-viral gene editing. Briefly, cells were bound with an anti-tNGFR fluorescent antibody (Supplementary Table 3) for sorting as described for flow cytometric staining. A BD Aria Cell Sorter (Stanford Blood Center Flow Cytometry Core) was used for all cell sorting experiments. Cells were maintained at 4C throughout the duration of sorting, and sorted cells were collected into destination tubes of XVivo15 media mixed 1:1 with FBS. After sorting, cells were centrifuged for 5 min at 300G prior to resuspension in culture media.

Drug Selection with Puromycin

Drug based selection of successfully intron edited T cells was performed using puromycin selection for forty-eight hours beginning six days after non-viral gene editing. Briefly, T cells were cultured in XVivo15 based media as described above with the addition of 5 ug/mL of puromycin (Puromycin Dihydrochloride, stock concentration of 10 mg/mL in 20 mM HEPES buffer, ThermoFisher). Concentrations used in dose titration experiments ranged from 1.0 ug/mL to 10 ug/mL as indicated in Extended Data Fig. 2b. After two days of puromycin selection, cells were centrifuged for 5 minutes at 300g and the supernatant was removed before resuspension in standard culture media without puromycin.

In Vitro Cancer Target Cell Killing Assays

At eight days post non-viral gene editing, edited CAR T cells (either bulk populations without selection or selected cells as described) were mixed at indicated E:T ratios with target Nalm6 leukemia cells in 96 well plates, with 40,000 Nalm6 cells and varying numbers of T cells per well. Cell killing was assessed by flow cytometery at 48 hours, and the percentage of Nalm6 tumor cell killing was calculated by taking 1 – (# Nalm6 cells alive in experimental condition / # Nalm6 cells alive in no T cell conditions). Nalm6 leukemia cells (ATCC) were cultured in RPMI+10% FBS and passaged every 2-3 days to maintain cell densities of approximately 1e6 cells / mL. Nalm6 cells were cultured for up to ~20 passages before discarding cultures are returning to low passage number frozen aliquots of initial cell line stock acquired from ATCC.

In Vitro Proliferation Assays

After initial stimulation at a 1:4 E:T ratio with Nalm6 target cells eight days post non-viral editing, T cells were expanded in XVivo15 media + 5% FBS with 50 U/mL human IL-2 over the next two weeks. Every 2-3 days, T cells were counted by automated cell counter, and reseeded at concentrations of approximately 1e6 T cells / mL culture media. Total cumulative proliferation was calculated compared to the input number of CAR T cells for each of the four tested selection methods.

Extended Data

Extended Data Figure 1. Efficient knockin of a synthetic exon across TRAC intronic sites.

Extended Data Figure 1.

Open in a new tab

a, Observed efficiency of TCR knockout (measured by flow cytometric surface staining for CD3) after gene editing with Cas9 or Cas12a RNPs containing gRNAs targeting TRAC exon 15 or 18 distinct targets within TRAC intron 1 (avoiding a highly repetitive region within the intron). TCR knockout was highly efficient for the exon targeting gRNA, but significantly lower for intronic guides.

b, Percentage of total cells expressing the knockin gene cassette (GFP+) following CD3 depletion (removing TCR positive cells) after intron knockin at 18 unique sites throughout TRAC intron 1. n = 2-4 unique donors.

c, TCR knockout (measured by flow cytometric surface staining for CD3) after gene editing with Cas12a RNPs containing gRNAs targeting TRAC intron 2.

d, Knockin efficiency at four unique sites within TRAC intron 2.

e, Percentage of total cells expressing the knockin gene cassette (GFP+) following CD3 depletion (removing TCR positive cells) after intron targeting at 4 unique sites throughout TRAC intron 2.

f, Across four TRAC intron 1 target sites targeted with Cas9 RNPs, no major difference in knockin efficiency and successful protein expression from the integrated synthetic exon was observed when using the endogenous splice acceptor and donor sites from the adjacent endogenous TRAC exons compared to synthetic consensus splice acceptor and donor sites.

a-f, n = 2-4 unique primary human T cell donors.

Extended Data Figure 2. Viability, editing, and activation metrics with non-viral intron editing and negative selection in human T cells.

Extended Data Figure 2.

Open in a new tab

a, Timeline of primary human T cell editing using non-viral intron knockins, followed by negative selection. T cells are isolated and activated on Day 0, followed by electroporation based non-viral intron editing on Day 2. CD3 Negative selection was performed on Day 8, six days followed editing.

b, Total T cell counts on Day 1 and Day 4 post electroporation relative to no electroporation controls following non-viral intron knockins by electroporation of a TRAC intron targeting CRISPR-Cas9 RNP along with a DNA Homology Directed Repair Template. Approximately half of the cell loss observed in the intron knockin condition (“RNP+DNA”) is due to electroporation itself, and half due to DNA toxicity, as observed previously for non-viral exon knockins using electroporation in primary human T cells5,29. n = 3 donors with 3 technical replicates each.

c, The percentage of successfully edited intron knockin cells across timepoints, beginning three days after editing (Day 5 post activation). Across n = 3 donors, editing percentages were stable in culture during post-editing expansion. Error bars represent standard error of the mean.

d, At Day 8 post activation, TRAC intron knockin T cells were negatively selected by binding magnetic beads to the CD3 complex. Compared to the input population prior to selection, post-negative selection intron knockin T cells did not show any decrease in viability as measured by Tryphan Blue staining and Live/Dead dye flow cytometric staining. The observed increase in viability after selection is likely due to washing steps removing dying cells present in the input culture.

e, After either negative selection or no selection, the activation status of TRAC intron knockin CD19-28z CAR T cells was assessed through in vitro activation by Nalm6 target cells at a 1:4 Effector:Target cell ratio. 24 hours post co-incubation, CD69 and CD25 expression levels in CAR positive (tNGFR+) and CAR negative (tNGFR−) cells was assayed by flow cytometry. n = 2 unique donors with 3 technical replicates. ns = not significant, Two-sided paired t test.

Extended Data Figure 3. Direct comparison of common edited primary human T cell selection methods.

Extended Data Figure 3.

Open in a new tab

a, Intron knockin of a tNGFR-CAR-PuroR synthetic exon enabled a single population of edited cells to be compatible with four common selection methods. Representative flow cytometric plots of TRAC intron CAR T cells after purification with four distinct selection methods. Knockin of a synthetic exon containing a tNGFR-CAR-PuroR multicistronic cassette to TRAC intron 1 enabled successfully edited cells to be negatively selected by CD3 Depletion (removal of TCR positive cells), positively selected by streptavidin magnetic bead enrichment after binding of anti-tNGFR biotinylated antibodies, fluorescence-activated cell sorting after binding of anti-tNGFR fluorescent antibodies, or drug selection after culture in puromycin. Negative selection by CD3 depletion yields predominantly a successfully edited CAR+ T cell population without the endogenous TCR, although rarer TCR negative / knockin negative cells are present (likely due to the RNP induced double stranded break within the intron causing a large deletion that included a portion of one or both adjacent exons, instead of the more common NHEJ repair outcome of smaller indels). Positive selection, sorting, and drug selection in contrast remove all knockin negative cells, but retain a population of TCR positive / knockin positive cells (likely due to successful HDR mediated knockin to one TRAC allele with either no editing or a small indel removed during mRNA splicing on the second TRAC allele; while in some T cells one TCRα loci is silenced, numerous T cells express functional TCRα chains from both alleles50). Only negative selections do not require additional genetic material (e.g. no selection marker or resistance gene necessary) and leave cells untouched post-selection.

b, Percentage of residual TCR positive cells following four different selection methods. TRAC intron knockin followed by CD3 Negative Selection (Blue) leaves almost no detectable TCR+ cells remaining.

c, In vitro killing of Nalm6 target cells by TRAC intron knockin CD19-28z CAR T cells following negative, positive, sorting, or drug selection, measured 48 hours post co-incubation. n = 4 unique donors. ns = not significant, * = P<0.05, Two-sided paired t test.

d, In vitro proliferation of TRAC intron knockin CD19-28z CAR T cells following negative, positive, sorting, or drug selection after co-culture with Nalm6 target cells. Error bars represent standard error of the mean from n = 4 unique donors. ns = not significant, * = P<0.05, Two-sided Mann-Whitney test.

e, Puromycin dose titrations reveal a tradeoff between purity and cell yield when performing drug selections. Increasing doses of puromycin yielded greater T cell purity, but overall edited cell yield began to decline with increasing puromycin concentrations. Beginning 6 days post editing, 100,000 bulk edited T cells were treated with indicated doses of puromycin for 48 hours. A dose of 5 ug/mL was used for experiments in Fig. 2b–c to balance purity and yield. n = 2 donors with 4 technical replicates each.

Extended Data Figure 4. Variable 5’ and 3’ splicing architectures enable control of alternative splicing of synthetic exons.

Extended Data Figure 4.

Open in a new tab

a, Correlation between observed cellular protein phenotypes by flow cytometry with inferred splicing behavior. Cells expressing both the protein encoded by the synthetic exon (e.g. GFP) and the endogenous protein (e.g. TCR) must be undergoing both synthetic splicing and endogenous splicing. These two splicing outcomes could be occurring from the same pre-mRNA transcript from a single allele by alternative splicing (single pre-mRNA transcript spliced into two different mature mRNA transcripts, one encoding the synthetic gene and a second encoding the endogenous gene), or the two splicing outcomes could be occurring within two separate pre-mRNA transcripts expressed from two different alleles (e.g. one allele with successful intron knockin of the synthetic exon, with the second allele being either unedited or possessing small indels from NHEJ repair that do not interfere with endogenous gene expression). Biallelic intron knockin experiments (Fig. 5) support that the majority, but not all, dual positive cells result from alternative splicing of the same pre-mRNA from a single allele.

b, Variable 3’ synthetic exon splicing architectures lead to controllable degrees of alternative splicing. Balanced splicing (2A-SD, far left) results in the majority of knockin (GFP) cells being negative for the endogenous gene (TCR−), with the residual GFP+ TCR+ cells likely due to one knockin allele and one WT or NHEJ edited allele. Creation of unbalanced splicing by replacement of the splice donor with a polyA (either SV40, WPRE, or TRAC’s endogenous polyA51) resulted in an increase in dual positive cells, with varying levels of endogenous gene expression depending on the polyA sequence used (SV40 > WPRE or TRAC). The differing levels of endogenous gene expression observed were potentially due to variable efficiency of RNA transcription termination by the polyA sequences, with the shorter SV40 polyA sequence less efficient at stopping transcription, enabling more mRNA transcripts to continue to the endogenous exons downstream from the knocked in synthetic exon and thus be capable of alternative splicing. Indeed, addition of a splice donor sequence following the polyA tails decreased the expression levels of the endogenous gene for all three polyA sequences, returning completely to baseline (2A-SD construct, far left) with the WPRE and TRAC polyA, and lowering the TCR expression level for the SV40 polyA, supporting the interpretation that unbalanced splice sites in the polyA only constructs lead to greater alternative splicing. SV40 polyA flow plot reproduced from Fig. 3b for comparisons.

c, Alternative splicing can also be controlled by varying the synthetic exon’s 5’ splicing architecture. In intron knockin constructs containing both an SV40 or TRAC endogenous polyA sequences at their 3’ end, inclusion of Exonic Splicing Silencer (ESS) DNA sequences adjacent to the synthetic exon’s 5’ splice acceptor resulted in increased numbers of dual positive cells showing evidence of alternative splicing. The short 6-8 bp ESS sequences were introduced into the 2A multicistronic element at the 5’ end of the synthetic exon (necessary to separate the new synthetic gene’s protein translation from the translation of the preceding endogenous exon) using degenerate bases to maximize the number and strength of the ESS sequences present36. In pre-mRNA transcripts containing unbalanced splice sites due to the 3’ polyA, inclusion of ESS sequences at the synthetic exon’s 5’ end reduces binding by the SR proteins that mediate splicing and increases the chance that the preceding endogenous exon will splice with the downstream endogenous exon rather than the synthetic exon. Both splicing outcomes occur with alternative splicing, resulting in expression of both the endogenous gene and the new synthetic gene. In contrast, similar optimization of the 2A elements degenerate bases to include Exonic Splicing Enhancer elements (or insertion of in-frame ESE elements prior to the unaltered 2A sequence) largely prevented alternative splicing, returning the levels of dual positive cells back to amount seen with synthetic exons containing balanced splice sites (e.g. the amount of dual positive cells seen due to one knockin allele and one unedited/NHEJ allele). ESS-2A – SV40 polyA flow plot reproduced from Fig. 3d for comparisons.

d, Summary heatmap of tested combinations of 5’ and 3’ splicing architectures. For applications requiring both endogenous and synthetic splicing of a synthetic exon integrated into an endogenous intronic region, an Exonic Splicing Silencer 5’ architecture paired with a short SV40 polyA 3’ architecture proved optimal. For applications requiring predominantly synthetic splicing, an Exonic Splicing Enhancer (ESE) 5’ architecture paired with any of the tested polyA 3’ architectures showed approximately equivalent degrees of synthetic splicing.

Extended Data Figure 5. Generalized endogenous gene targeting with exon and intron knockins.

Extended Data Figure 5.

Open in a new tab

a, A new synthetic gene can be introduced under endogenous regulatory control of an existing gene without also loosing expression of the existing gene only be integration at the N terminus (immediately before the Start codon) or C terminus (immediately before the Stop codon) of the targeted gene. This limitation in target sites means that efficient gRNAs for gene knockin may not be present for many genes, as observed in previous studies6.

b, Intron knockins offer greater flexibility for placing a synthetic gene under endogenous regulatory control without disrupting the endogenous gene. Using optimized 5’ and 3’ splicing architectures to induce alternative splicing, a synthetic exon can be introduced throughout the intronic regions of the endogenous gene, with alternative splicing resulting in two separate mature mRNA transcripts, one encoding the new synthetic protein and a second encoding the endogenous gene. Orders of magnitude more gRNA target sites are available within intronic regions than only at the very N or C terminus of a gene.

c, Exonic targeting of a new synthetic gene within the coding sequence of an endogenous gene disrupts the endogenous gene’s sequence, resulting in a multicistronic mRNA transcript encoding the new synthetic protein and two partial fragments of the endogenous protein, causing knockout of the endogenous gene.

d, Intron targeting with a synthetic exon containing optimized 5’ and 3’ splicing architectures to induce synthetic splicing only also results in endogenous gene knockout. However unlike in exonic targeting, where alleles with unsuccessful knockins largely have NHEJ induced indels resulting in frameshift mutations and protein knockout, with intron targeting NHEJ induced indels reside within an intronic sequence that is largely tolerant of short DNA changes which will be spliced out of the final mRNA transcript (more rare but detectable larger deletions that include parts of the surrounding exons induced by double stranded breaks can still result in endogenous protein knockout). If the targeted endogenous gene is a surface receptor, then intron knockins offer the unique ability to perform negative selections to purify successfully edited cells.

Supplementary Material

Supplementary Tables

ACKNOWLEDGEMENTS

We thank the members of the Satpathy lab for stimulating discussions. This work was supported by a Career Award for Medical Scientists from the Burroughs Wellcome Fund (A.T.S.), a Lloyd J. Old STAR Award from the Cancer Research Institute (A.T.S.), the Parker Institute for Cancer Immunotherapy (A.T.S), a CRISPR Cures for Cancer Award (A.T.S.), and the California Institute for Regenerative Medicine (DISC2–13212, A.T.S.).

COMPETING INTERESTS

T.L.R. is a founder of Arsenal Biosciences. A.T.S. is a founder of Immunai, Cartography Biosciences, Santa Ana Bio, and Prox Biosciences, an advisor to Zafrens and Wing Venture Capital, and receives research funding from Merck Research Laboratories and Astellas Pharma. The remaining authors declare no competing interests.

Data Availability

All data supporting the findings of this study are available within the paper and its Supplementary Tables. An annotated list of all gRNA sequences used in the study is available in Supplementary Table 1. An annotated list of all DNA constructs used in the study, as well as full DNA sequences for all constructs, is available in Supplementary Table 2. Annotated genebank files for all DNA constructs used in the study are available upon request.

REFERENCES

  • 1.Wang JY & Doudna JA CRISPR technology: A decade of genome editing is only the beginning. Science 379, eadd8643 (2023). [DOI] [PubMed] [Google Scholar]
  • 2.Makarova KS et al. Evolutionary classification of CRISPR–Cas systems: a burst of class 2 and derived variants. Nat. Rev. Microbiol 18, 67–83 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Maeder ML & Gersbach CA Genome-editing Technologies for Gene and Cell Therapy. Mol. Ther 24, 430–446 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Eyquem J et al. Targeting a CAR to the TRAC locus with CRISPR/Cas9 enhances tumour rejection. Nature 543, 113–117 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Roth TL et al. Reprogramming human T cell function and specificity with non-viral genome targeting. Nature 559, 405–409 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Shy BR et al. High-yield genome engineering in primary cells using a hybrid ssDNA repair template and small-molecule cocktails. Nat. Biotechnol (2022) doi: 10.1038/s41587-022-01418-8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Mikkelsen NS & Bak RO Enrichment strategies to enhance genome editing. Journal of Biomedical Science vol. 30 Preprint at 10.1186/s12929-023-00943-1 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Allen AG et al. A highly efficient transgene knock-in technology in clinically relevant cell types. Nat. Biotechnol (2023) doi: 10.1038/s41587-023-01779-8. [DOI] [PubMed] [Google Scholar]
  • 9.Wiebking V et al. Metabolic engineering generates a transgene-free safety switch for cell therapy. Nat. Biotechnol 38, (2020). [DOI] [PubMed] [Google Scholar]
  • 10.Levesque S, Carleton G, Duque V & Goupil C Pharmacological control of CAR T cells through CRISPR- driven rapamycin resistance. (2024). [Google Scholar]
  • 11.Chavez M, Rane DA, Chen X & Qi LS Stable expression of large transgenes via the knock-in of an integrase-deficient lentivirus. Nat. Biomed. Eng 7, (2023). [DOI] [PubMed] [Google Scholar]
  • 12.Mosti L et al. Targeted multi-epitope switching enables straightforward positive/negative selection of CAR T cells. Gene Ther. 28, 602–612 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Yin H, Kauffman KJ & Anderson DG Delivery technologies for genome editing. Nature Reviews Drug Discovery vol. 16 Preprint at 10.1038/nrd.2016.280 (2017). [DOI] [PubMed] [Google Scholar]
  • 14.Schober K et al. Orthotopic replacement of T-cell receptor α- and β-chains with preservation of near-physiological T-cell function. Nat. Biomed. Eng 3, 974–984 (2019). [DOI] [PubMed] [Google Scholar]
  • 15.Odé Z, Condori J, Peterson N, Zhou S & Krenciute G CRISPR-Mediated Non-Viral Site-Specific Gene Integration and Expression in T Cells: Protocol and Application for T-Cell Therapy. Cancers 12, 1704 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Bennardo N, Cheng A, Huang N & Stark JM Alternative-NHEJ Is a Mechanistically Distinct Pathway of Mammalian Chromosome Break Repair. PLOS Genet. 4, e1000110 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Chang HHY, Pannunzio NR, Adachi N & Lieber MR Non-homologous DNA end joining and alternative pathways to double-strand break repair. Nat. Rev. Mol. Cell Biol 18, 495–506 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Burma S, Chen BPC & Chen DJ Role of non-homologous end joining (NHEJ) in maintaining genomic integrity. DNA Repair 5, 1042–1048 (2006). [DOI] [PubMed] [Google Scholar]
  • 19.Simeonov DR et al. A large CRISPR-induced bystander mutation causes immune dysregulation. Commun. Biol 2, 1–6 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Kosicki M, Tomberg K & Bradley A Repair of double-strand breaks induced by CRISPR–Cas9 leads to large deletions and complex rearrangements. Nat. Biotechnol 36, 765–771 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Adikusuma F et al. Large deletions induced by Cas9 cleavage. Nature 560, E8–E9 (2018). [DOI] [PubMed] [Google Scholar]
  • 22.Stanford WL, Cohn JB & Cordes SP Gene-trap mutagenesis: past, present and beyond. Nat. Rev. Genet 2, 756–768 (2001). [DOI] [PubMed] [Google Scholar]
  • 23.Serebrenik YV, Sansbury SE, Kumar SS, Henao-Mejia J & Shalem O Efficient and flexible tagging of endogenous genes by homology-independent intron targeting. Genome Res. 29, (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Fu J et al. Improved and Flexible HDR Editing by Targeting Introns in iPSCs. Stem Cell Rev. Rep 18, (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Li J et al. Gene replacements and insertions in rice by intron targeting using CRISPR–Cas9. Nat. Plants 2, 1–6 (2016). [DOI] [PubMed] [Google Scholar]
  • 26.Stanford WL, Cohn JB & Cordes SP Gene-trap mutagenesis: Past, present and beyond. Nature Reviews Genetics vol. 2 Preprint at 10.1038/35093548 (2001). [DOI] [PubMed] [Google Scholar]
  • 27.Sigal A et al. Dynamic proteomics in individual human cells uncovers widespread cell-cycle dependence of nuclear proteins. Nat. Methods 3, (2006). [DOI] [PubMed] [Google Scholar]
  • 28.Simeonov DR et al. A large CRISPR-induced bystander mutation causes immune dysregulation. Commun. Biol 2, (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Schober K et al. Orthotopic replacement of T-cell receptor α- and β-chains with preservation of near-physiological T-cell function. Nat. Biomed. Eng 3, 974–984 (2019). [DOI] [PubMed] [Google Scholar]
  • 30.Foy SP et al. Non-viral precision T cell receptor replacement for personalized cell therapy. Nature 615, 687–696 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Zhang J et al. Non-viral, specifically targeted CAR-T cells achieve high safety and efficacy in B-NHL. Nature 609, 369–374 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 32.Orentas RJ A positive take on negative selection for CAR-T manufacturing. Mol. Ther. Methods Clin. Dev 32, 101218 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Zhong H et al. High-fidelity, efficient, and reversible labeling of endogenous proteins using crispr-based designer exon insertion. eLife 10, (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Liu HX, Zhang M & Krainer AR Identification of functional exonic splicing enhancer motifs recognized by individual SR proteins. Genes Dev. 12, (1998). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Cáceres EF & Hurst LD The evolution, impact and properties of exonic splice enhancers. Genome Biol. 14, (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Ke S et al. Quantitative evaluation of all hexamers as exonic splicing elements. Genome Res. 21, (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Kluesner MG et al. CRISPR-Cas9 cytidine and adenosine base editing of splice-sites mediates highly-efficient disruption of proteins in primary and immortalized cells. Nat. Commun 12, 2437 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Tsunekawa Y et al. Developing a de novo targeted knock-in method based on in utero electroporation into the mammalian brain. Development 143, 3216–3222 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Auer TO, Duroure K, De Cian A, Concordet J-P & Del Bene F Highly efficient CRISPR/Cas9-mediated knock-in in zebrafish by homology-independent DNA repair. Genome Res. 24, 142–153 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Dunbar CE et al. Gene therapy comes of age. Science 359, eaan4672 (2018). [DOI] [PubMed] [Google Scholar]
  • 41.Smith CJ et al. Enabling large-scale genome editing at repetitive elements by reducing DNA nicking. Nucleic Acids Res. 48, 5183–5195 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Labanieh L & Mackall CL CAR immune cells: design principles, resistance and the next generation. Nature 614, 635–648 (2023). [DOI] [PubMed] [Google Scholar]
  • 43.Larson RC & Maus MV Recent advances and discoveries in the mechanisms and functions of CAR T cells. Nat. Rev. Cancer 21, 145–161 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.June CH, O’Connor RS, Kawalekar OU, Ghassemi S & Milone MC CAR T cell immunotherapy for human cancer. Science 359, 1361–1365 (2018). [DOI] [PubMed] [Google Scholar]
  • 45.Durrant MG et al. Bridge RNAs direct programmable recombination of target and donor DNA. Nature 630, 984–993 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Durrant MG et al. Systematic discovery of recombinases for efficient integration of large DNA sequences into the human genome. Nat. Biotechnol. 41, 488–499 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Vo PLH et al. CRISPR RNA-guided integrases for high-efficiency, multiplexed bacterial genome engineering. Nat. Biotechnol 39, 480–489 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Rezazade Bazaz M, Ghahramani Seno MM & Dehghani H Transposase-CRISPR mediated targeted integration (TransCRISTI) in the human genome. Sci. Rep 12, 3390 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Tou CJ, Orr B & Kleinstiver BP Precise cut-and-paste DNA insertion using engineered type V-K CRISPR-associated transposases. Nat. Biotechnol 41, 968–979 (2023). [DOI] [PubMed] [Google Scholar]
  • 50.Padovan E et al. Expression of Two T Cell Receptor α Chains: Dual Receptor T Cells. Science 262, (1993). [DOI] [PubMed] [Google Scholar]
  • 51.De Silva D et al. Robust T cell activation requires an eIF3-driven burst in T cell receptor translation. eLife 10, (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Tables
NIHMS2093231-supplement-Supplementary_Tables.xlsx (70.3KB, xlsx)

Data Availability Statement

All data supporting the findings of this study are available within the paper and its Supplementary Tables. An annotated list of all gRNA sequences used in the study is available in Supplementary Table 1. An annotated list of all DNA constructs used in the study, as well as full DNA sequences for all constructs, is available in Supplementary Table 2. Annotated genebank files for all DNA constructs used in the study are available upon request.

RESOURCES