Development of a CRISPR–Cas9 transgene assay to evaluate knock-in fidelity

We have developed a novel fluorescence-based transgene integration assay that allows the quantification of integration events involving both the transgene and the plasmid backbone. This assay provides a platform for the subsequent testing and development of improvements to knock-in strategies.

The assay is plasmid-based, as this allows the flexibility to rapidly test different strategies by using available plasmid vectors for Cas9, guide RNAs and transgene donors (Fig. 1). The Cas9 expressing plasmid contains a CMV-driven Cas9 coding sequence followed by a T2A self-cleavage peptide and the mAmetrine reporter gene. The T2A sequence encodes a ‘self-cleaving’ amino sequence, adapted from an insect virus, Thosea asigna [43]. These 2A peptide sequences have been widely used for co-expressing multiple proteins from a single mRNA. The mechanism of ‘cleavage’ involves ribosome skipping at the C-terminus of the 2A producing two separate proteins. This design enables the evaluation of transfection efficiency, where co-expression of mAmetrine would be a marker for in vivo Cas9 expression.

Fig. 1

Development of the CRISPR–Cas9 knock-in assay. A Donor plasmid design for unmatched end joining strategies. The AAVS1 locus is targeted by CRISPR–Cas9 using the AAVS1-A1 guide. Cas9-targeted cleavage will form DNA DSBs at which the transgene is to be inserted. The unmatched end donor plasmid contains the transgene, which is a splice acceptor—driven EGFP coupled with the pac puromycin resistance gene for further selection. On the backbone, BFP expression is driven by a CAG promoter so as to report backbone integration. The donor plasmid contains Tet2 guide cleavage sites on either side of the transgene but does not have any sequences that are homologous to the AAVS integration site. B Overview of the knock-in assay. Three plasmids are transfected into HEK293 cells (i) Cas9 (wt or mutant) and mAmetrine expression and (ii) expression of sgRNAs for AAVS1 and Tet2, (iii) the unmatched end donor plasmid. Initial assessment by flow cytometry quantifies mAmetrine and BFP expression for transfection efficiency, and GFP to evaluate knock-in efficiency. Puromycin selection is then applied to enrich for cells in which the transgene is integrated and expressed. Final flow cytometry assessment is performed to collect GFP and BFP expression data to evaluate knock-in fidelity. C, D Data from the knock-in assay using wtCas9 and the unmatched end donor plasmid. C Flow cytometry quantification of GFP expressing cells 2 days after transfection. The donor plasmid only control was not transfected with the Cas9 and sgRNA plasmids. D Fluorescence profile of puromycin resistant cells after transfection and selection according to the unmatched end joining knock-in strategy. Error bars in C and D represent the standard error mean from three independent replicates. Students’ T tests indicate no significant differences between the donor only control and the unmatched end joining experiment. p > 0.05

To ensure stable expression of integration we selected a guide targeting the intron 1 of the adeno-associated virus locus 1 (AAVS1) genomic safe harbour which is often used as a target for CRISPR–Cas9 editing site [2] (Fig. 1A).

Evaluation of transgene integration was done by assessing the integration of donor fragments containing a GFP gene (Fig. 1A). The donor plasmid contains the transgene, which encodes a splice acceptor (SA)-driven EGFP followed by a T2A peptide and the pac puromycin resistance gene for antibiotic selection. This promoter-less reporter cassette ensures the transgene is only expressed if integrated in the correct location at the correct orientation. On the backbone of the plasmid, a CAG promoter-driven BFP fluorescence gene is used to report transfection efficiency and unwanted exogenous DNA integration. Cas9-mediated linearization of the donor plasmid can be induced using sgRNAs targeted to either end of the transgene.

To perform the assay, HEK293T cells are transfected with three plasmids; Cas9 expressing plasmid, sgRNA expressing plasmid, and the donor plasmid (Fig. 1B). Two days after transfection, cells are analysed by flow cytometry to evaluate transfection efficiency using plasmid-driven mAmetrine and BFP expression. The efficiency of initial transgene integration is assayed using GFP expression.

Cells were then selected with puromycin, to allow time for the transient, plasmid-based BFP expression to be lost while maintaining cells in which the GFP transgene has integrated. Subsequent flow cytometry analysis of the selected cells was used to quantify any unwanted integration of the donor plasmid by detecting BFP expression while faithful transgene integration measured by GFP expression does not contain BFP (Additional file 1: Fig. S1).

In the unmatched ends knock-in approach, Cas9 cleavage is directed at the AAVS1 locus and at either ends of the transgene in the donor plasmid, generating two plasmid DNA fragments. There is no sequence homology between the donor plasmid and the integration site. Initial transgene integration efficiency was 6.1% as indicated by GFP expression 2 days after transfection (Fig. 1C). Puromycin selection resulted in cells that were mainly GFP+BFP+ (Fig. 1D), and quantification revealed that 88.6% of the puromycin-resistant GFP expressing cells also express BFP. Only 8% of the puromycin enriched cells are GFP+BFP− (Fig. 1E). This knock-in profile of CRISPR–Cas9 edited cells was not significantly different compared to the donor plasmid control that did not include the CRISPR system.

Additionally, the data reveal that puromycin-resistant cell cultures are a polyclonal mix that are dominantly composed of cells that co-express GFP and BFP. This provides strong evidence that integration of unwanted exogenous DNA is highly pervasive in edited cells.

High occurrence of backbone integration reduces the fidelity of CRISPR–Cas9 mediated transgene integration

Our initial evaluation of CRISPR–Cas9 strategy involved the linearization of donor plasmid without any homologous sequences to guide on-targeted integration (Fig. 1). Various strategies have previously been developed to increase the efficiency of targeted integrations [5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34]. We adopted these strategies and evaluated their performances with our assay (Fig. 2).

Fig. 2

Donor plasmid design for matched end joining and HDR-based strategies. A AAVS1 locus is targeted by CRISPR–Cas9 using AAVS1-A1 guide. Cas9-targeted cleavage will create DNA DSBs where the transgene of interest is intended to be inserted. The left and right arms either side of the DSB may be incorporated into the donor template to direct on-target integration (see C and D). B–D The donor plasmids contain a splice acceptor—driven GFP transgene coupled with the puromycin resistance gene, and BFP driven by a CAG promoter on the plasmid backbone. B For the matched end joining strategy, the donor plasmid contains a copy of the AAVS1 sgRNA target site at both flanks of the transgene. Cas9 cleavage in vivo using the AAVS1-A1 guide would cleave the endogenous locus while also generating two DNA fragments from the plasmid, the transgene and the backbone. C, D For homology-directed repair strategies, the transgene is flanked by sequences corresponding to the left and right arms of the endogenous locus. However, the donor plasmids have a mutated PAM sequence of the AAVS1-A1 sequence target, preventing Cas9 cleavage. This ensures the HR-based plasmids remain circular to initiate the HR mechanism. Homology arms are 795 bp and 50 bp in the long and short HA donor plasmids, respectively (C, D). Orange lines indicate left arm homology. Blue lines indicate right arm homology

The “matched ends” strategy still involves the cleavage of the donor fragment at either side of the transgene, with 200 bp of sequence that is homologous to the AAVS1 target included on either side of the transgene fragment (Fig. 2B).

Two other donor plasmids were designed to explore homology directed repair (HR)-based pathways (Fig. 2C, D). Some reports have shown that the length of homology sequences at both ends of the transgene (from herein referred to as homology arms or HA) determines the efficiency of integration [11, 15, 16]. To evaluate the impact of the length of homology arms in our KI assay, we designed short (50 bp) or long (795 bp) HAs flanking the target site, to be referred to as short HA and long HA, respectively. To ensure that the donor plasmids are processed with HR machinery and not non-homologous end joining (NHEJ) or microhomology-mediated end joining (MMEJ), we modified the targeting sequence on the homology arm by deleting the PAM sequence of the CRISPR target site ensuring no Cas9 targeted cleavage on the donor plasmid. This modification will ensure the donor plasmids remain circular and must rely on strand invasion for on-target.

The initial transgene integration efficiencies of each strategy indicate that the matched end joining and long homology arms strategies both had more efficient transgene integration than the unmatched end joining strategy and donor control, although these differences did not reach statistical significance (Fig. 3A).

Fig. 3

High occurrence of backbone integration in all CRISPR–Cas9 mediated KI strategies. A Quantification of GFP expressing cells 2 days after transfection. GFP can only be expressed when the donor plasmid is integrated into the genome downstream of a transcribed RNA splice donor site. HEK293T cells were transfected with Cas9 and sgRNA plasmids along with the donor plasmid for either the unmatched ends, matched ends, long HA, or short HA strategy. Plasmid donor only transfections were performed as a control for random plasmid integration into the genome. Error bars indicate standard error from 3 biological replicates. B Fluorescence profile of puromycin-selected cells. Transfected cells were selected with puromycin for 3 weeks prior to Flow cytometry analysis. Any BFP expression arising from the backbone of the transiently transfected plasmid is lost by this time, so the relative proportions of cells expressing GFP and/or BFP can be calculated. The percentages of green, blue, and double fluorescent cells are plotted as stacked graphs. Error bars indicates standard error for GFP+ cells from three biological replicates. *p < 0.05 in student T-test

Interestingly, flow cytometry of selected cells produced different outcomes (Fig. 3B). The KI profile of the matched end joining strategy was similar to the unmatched end joining strategy. However, the long HA strategy generated a significant increase in KI fidelity compared to unmatched and matched end joining strategies (p = 0.05 and p = 0.03, respectively).

Our assay system also allows the direct comparison of different Cas9 variants. We compared wild type SpCas9 with espCas9 (1.1), a rationally engineered variant of Cas9 that has reduced off-target cleavage [21]. However, the use of espCas9 (1.1), from herein referred to Cas9es, did not significantly alter the KI fidelity of any of the strategies tested.

This data indicates that use of a homology-directed repair strategy for inserting transgenes leads to a significant reduction in unwanted integration of exogenous DNA.

The type of DNA DSBs influence KI outcomes

We next evaluate whether the type of CRISPR–Cas9 mediated DNA DSBs would induce distinct integration outcomes. Targeting the AAVS1 locus, we adopted six guides complement to the sense and antisense strand of the target locus [2] (Fig. 4A). Targeting guides are paired with a tail-to-tail orientation, as upon utilizing paired Cas9 D10A and H840A variants will generate 5ʹ or 3ʹ overhangs single stranded DNA (ssDNA) overhangs, respectively (Fig. 4B, C). As it has been reported that the length between Cas9 binding site and the orientation of the PAM is critical for efficient targeting [2, 4, 44, 45] we paired these AAVS1 guides to generate DNA DSBs with various lengths of overhangs (Additional file 1: Table S1). The different lengths of overhangs would enable us to evaluate the impact of ssDNA homologous sequences on transgene integration.

Fig. 4

Dual Cas9 nickase strategy mediates DNA DSBs with single stranded overhangs. A 6 different guides were used for the dual Cas9 nickase strategy. Three guides are complement to the sense strand (Guide A1, A2, and A3) and three guides target the anti-sense strand (Guide B1, B2, and B3). Two different Cas9 nickases were applied, Cas9 D10A and Cas9 H840A. B Two Cas9 D10A utilizing guides at a tail-to-tail orientation, will generate a DNA DSB with 5ʹ overhang ssDNA. As for C dual Cas9 H840A strategy will generate a 3ʹ overhang DNA DSB

Interestingly, efficiency of transgene integration was not affected by the type of DNA DSBs, the length of overhangs, and the use of enhanced specificity Cas9 variants (Fig. 5). Evaluation of stable puromycin selected cells however showed distinct integration outcome profiles for each strategy (Fig. 6). Furthermore, statistical analysis revealed that revealed that long 5ʹ overhangs, induced by dual Cas9 D10A using pair 5 guides, produced a statistically significant outcome compared to other pairs (Fig. 6A). Significant difference was also seen when comparing Cas9 and Cas9es version with pair 5 (two-sample t-test p = 0.028) indicating higher editing efficiency using the normal Cas9 D10A variant. However, no differences were observed among strategies employing 3ʹ overhang intermediates (Fig. 6B).

Fig. 5

No significant differences in the efficiency of transgene integration between different lengths of overhangs. HEK293T cells were transfected with plasmids expressing either Cas9 D10A nickases or H840A nickases, alongside plasmid containing AAVS1 pair guides and the donor plasmid. Cells were analysed for GFP expression by flow cytometry 2 days after transfections. A Strategies utilizing 5ʹ overhang DNA DSBs with various lengths of overhangs; B strategies utilizing 5ʹ overhang DNA DSBs with various lengths of overhangs. Error bars indicate standard error from three biological replicates. *p < 0.005

Fig. 6

Distinct integration profiles of stable puromycin selected cells from overhang mediated KI strategies. Transfected cells were grown under puromycin selection for 3 weeks before final flow cytometry analysis. Stable GFP+BFP− cells were quantified using flow cytometry and compared to non-Geminin results. A Strategies utilizing 5′ overhang DNA DSBs with various lengths of overhangs; B Strategies utilizing 5′ overhang DNA DSBs with various lengths of overhangs. Error bars indicates standard error from ≥ three biological replicates. *p < 0.05

The combination between Cas9es and geminin fusion increases faithful transgene integrations in long and short HA strategies

We hypothesised that restricting the formation of DSBs to S phase might have a positive effect on knock-in efficiency and fidelity, as error prone insertion through NHEJ would be minimised and HR would be promoted. Indeed, cell synchronization studies have revealed an increase of faithful editing in S-phase cells [7, 46]. Previously, two independent labs reported the development of Cas9-Geminin fusion, geminin being a DNA replication licensing protein, to enhance HR-mediated gene editing [32, 33].

To investigate whether restricting Cas9 expression to later phases of the cell cycle increases knock-in fidelity, we fused a 110 amino acid fragment of geminin to the C terminus of Cas9. We then applied this fused Cas9-geminin variant with our various KI strategies.

Surprisingly, the geminin fusion to Cas9 directed a significant increase in initial transgene integration strategies when using the enhanced specificity Cas9es with the blunt end HR strategies (Fig. 7A). The geminin fusion led to a 4.6-fold increase in GFP+ cells with the long HA strategy and 5.8-fold increase in the short HA strategy (p < 0.005). The percentage of GFP+BFP− cells after selection was 77.7%, which is substantially higher than in previous strategies (p < 0.005).

Fig. 7

Geminin-restricted Cas9es expression increases effective transgene integration in overhang DSB-mediated strategies. HEK293T cells were transfected with plasmids expressing either Cas9, Cas9es, Cas9-geminin or Cas9es-geminin, alongside the sgAAVS1-A1 plasmid and the donor plasmid for either the matched ends, long HA, or short HA strategy. Cells were analysed for GFP expression by flow cytometry 2 days after transfections. A Strategies utilizing blunt end DNA DSBs; B strategies utilizing overhang DNA DSBs. Error bars indicate standard error from three biological replicates. *p < 0.005

The effect on geminin on Cas9es experiments was also substantial in both 5ʹ and 3ʹ overhang strategies (Fig. 7B). Both short and long 5ʹ overhang DSB intermediates displayed a significant increase of transient GFP+ cells upon using the Cas9es-geminin fusion (p < 0.01). However, this fusion combination produces lower selected GFP+BFP− cells compared to blunt end strategy (Fig. 8). Integration fidelity was not affected by the length of overhang nor the use of enhanced specificity Cas9 variant. Taken together, the data suggest that strategies using overhang DSB intermediates are repaired and integrated more rapidly in the S phase of the cell cycle, but that faithful integration is independent of the timing of the DSB formation.

Fig. 8

Geminin-restricted Cas9es expression increases faithful transgene integration in HA strategies. Transfected cells were grown under puromycin selection for 3 weeks before final flow cytometry analysis. Stable GFP+BFP− cells were quantified using flow cytometry and compared to non-Geminin results. A Strategies utilizing blunt end DNA DSBs; B strategies utilizing overhang DNA DSBs. Error bars indicates standard error from ≥ three biological replicates. *p < 0.05

Reducing Cas9 expression increase faithful transgene integration

Reducing Cas9 activity has been shown to reduce off-target activity [47, 48]. Various inducible expression systems have been developed to restrict Cas9 expression [30, 49, 50]. However, these approaches suffer from decreased targeting activity and the requirement for an induction step. One alternative strategy that would alleviate these disadvantages is to apply a self-cleaving system [24, 28].

The concept of the CRISPR Cas9 self-cleaving system is to introduce a guide RNA that targets the coding sequence of the expressing vector. Cas9 targeting and cleavage of the expressing vector initiates DNA degradation and limits Cas9 expression, thus reducing unwanted off-target activity [28, 31].

The benefits of reduced off-targeting activity have been demonstrated in genetic knock-out (KO) purposes. We hypothesized that reduction in Cas9 expression would also aid CRISPR–Cas9 mediated KI strategies. We, therefore applied a self-limiting system to our assay to evaluate the effects on KI efficiency and fidelity (Fig. 9A).

Fig. 9

The self-cleaving Cas9 system. A The plasmid based self-limiting system incorporates a self-targeting guide into the Cas9 expressing plasmids. Upon expression, Cas9 assembly will generate two Cas9 complexes; one targeting and cleaving the AAVS1 genomic locus and the other cleaving the Cas9 expressing plasmid. B, C HEK293T cells were transfected with Cas9 expressing plasmids with their appropriate self-targeting guides. Cells were collected at each time points (8, 16, 24, 48, and 72 h after transfection) and nuclear extracts were obtained. Western blotting was used to detect Cas9 protein (161 kDa) or the TBP loading control (37.6 kDa) in B control and C cells transfected with Cas9 self-cleaving guides

Four self-cleaving guides targeting the N terminus of the Cas9 coding sequence were evaluated for their ability to restrict Cas9 expression following plasmid transfection. All self-cleaving guides reduced Cas9 protein levels during the 72-h evaluation (Fig. 9B). Self-cleaving guide A74 generated the greatest reduction in Cas9 expression and was used in subsequent experiments.

The use of the self-cleaving system significantly increased transgene integration efficiency in all the strategies tested (Table 2). Strikingly, an increase in KI fidelity was also seen in almost all the strategies, with all the long HA strategies achieving over 78% GFP+BFP− cells.

Table 2 Overall strategies that produced high intended transgene integration outcomes

To have an overview of the performance of all tested strategies in HEK293T cells, we plotted the average GFP+ expression to an efficiency-to-fidelity plot (Fig. 10). Four quadrants were determined with the 50% of either axis generating four outcome categories: (i) high efficiency and fidelity, (ii) high efficiency with low fidelity, (iii) low efficiency with high fidelity, and (iv) low efficiency and fidelity (Table 2).

Fig. 10

Self-cleaving Cas9 leads to substantial increases in knock-in fidelity with the long homology arm strategy. Each strategy was plotted into the dot graph for unselected GFP+ and selected GFP+BFP− results. The graph is divided into four quadrants at the 50% scale on each axis. Grouped samples are generated from the same strategy which consist of Cas9, Cas9es, Cas9-Gem, and Cas9es-Gem variants. Colour code was appointed to indicate high efficiency and fidelity (green), high efficiency with low fidelity (yellow), low efficiency with high fidelity (red), and low efficiency and fidelity outcomes (blue)

Use of DNA inhibitory molecules leads to small but significant increases in faithful integration of short HA donors

Integration of exogenous DNA can be driven by several cellular DNA repair pathways, including NHEJ, MMEJ and HR. NHEJ is active throughout the cell cycle but is error prone. HR relies on homologous sequences in the transgene and the integration site is only active at later stages of the cell cycle. In order to investigate whether selective inhibition of different repair pathways can improve the fidelity of transgene integration, we tested three DNA repair inhibitors in our assay; NU7441 and BO2 which inhibits DNA-PK and Rad51