Boosting genome editing efficiency in human cells and plants with novel LbCas12a variants

Directed evolution of LbCas12a with enhanced activity

To enhance the on-target activity of LbCas12a, we first attempted to mimic the mutations of AsCas12a Ultra (AsCas12a: M537R/F870L) in LbCas12a based on sequence alignment (Additional file 1: Fig S1A) [5]. However, incorporating both mutations in LbCas12a (i.e., N527R/E795L) did not increase the on-target activity when delivered as RNP (Additional file 1: Fig S1B). Specifically, we found E795L enhanced the editing efficiency at HPRT 38,330 site using 0.4 μM RNP, but outweighed by the detrimental effect of N527R (Additional file 1: Fig S1B). These results indicate the impact of point mutations on closely related Cas12a orthologs at aligned positions can be highly variable. As a result, previously identified beneficial mutations from AsCas12a Ultra cannot simply be used to improve LbCas12a function. We therefore proceeded to conduct a de novo screen to identify mutations in LbCas12a that improve intrinsic enzymatic activity.

To accomplish this, we designed an unbiased screen that systematically evaluates the effect of all single point mutations of LbCas12a on DNA cleavage in E. coli (Fig. 1A, B). We first established an E. coli-based activity assay for LbCas12a, where the degree of cell survival upon ccdB (which encodes a toxin) induction is directly linked to the cleavage activity of LbCas12a (Fig. 1A). We next generated a saturation mutagenesis library of LbCas12a spanning the entire coding sequence, with most plasmid clones in the library only containing one codon change (Fig. 1B) [10]. This library was then iteratively selected for four rounds and the enrichment of LbCas12a variants in the last round of selection was quantified by next generation sequencing (NGS). This strategy allowed us to measure the phenotype of over 9000 point mutations with high reproducibility (Fig. 1C). As expected, nearly all synonymous mutations (847) displayed no significant enrichment, as their scores tightly clustered around 0. There were 1977 point mutants with phenotype scores (i.e., natural logarithm of relative enrichment) greater than zero over three biological replicates (Additional file 2: Table S1).

Fig. 1figure 1

High-throughput characterization of DNA cleavage activities of LbCas12a point mutations in E. coli. A Schematic representation of bacterium-based selection assay to isolate LbCas12a mutants with enhanced activity. E.coli BW25141:DE3 cells containing an inducible ccdB expression plasmid were transformed with LbCas12a variant library, which was programmed to cleave the reporter plasmid through a target site with TTTT PAM sequence. Active LbCas12a mutants with enhanced activity enable the clearance of reporter plasmid, thus avoiding the cell death upon the induction of ccdB expression with arabinose. LbCas12a plasmids from survived cells were extracted and used for subsequent round of selection. Four rounds of sequential selection were performed. Round 3 and 4 libraries were sequenced and used to calculate the enrichment of LbCas12a variants. B Design of a saturation mutagenesis library for LbCas12a. Every codon of LbCas12a was randomized with NNK degenerate primers by nicking mutagenesis [10]. Importantly, each member of plasmid library only contains one codon change at a time, which was verified by Sanger sequencing of 24 individual plasmids from the library. C Enrichment scores for ~ 9000 LbCas12a point mutations over the last round of selection (round 4). Synonymous variants without changes on the protein sequences were colored in red. Variants with a minimal of 50 counts were included in this analysis. D Enrichment scores for 5 selected positions of LbCas12a that have been evaluated prior to this study

The effect of point mutations on enzyme activity can depend on the experimental system and conditions. In the context of mutation screens for CRISPR enzymes in E. coli, false positives have been reported in the literatures [11,12,13] where positive responses in bacteria failed to translate in human cells. To evaluate the reliability of our high-throughput data generated in E. coli, we first checked the phenotype scores of mutations known to affect the intrinsic cleavage activity of LbCas12a. For those two positions (N527/E795) taken from AsCas12a Ultra, N527 suffered from significant reductions in enrichment scores, indicating mutations at this position are detrimental to DNA cleavage in E. coli (Fig. 1D). While the enrichment score of N527R is not available in this analysis, a lysine substitution at this site (i.e., N527K) showed measurable reduction of cleavage activity of LbCas12a in E. coli, suggesting a positive charged side chain at position 527 for LbCas12a is detrimental to function. On the other hand, the E795L mutation was enriched upon selection (Fig. 1D), which is consistent with our empirical observations when the corresponding LbCas12a variants were delivered as RNP in human cells (Additional file 1: Fig S1B).

Previous research with AsCas12a showed that both E174R and S542R mutations significantly enhanced on-target activity [14]. Moreover, K548R mutants were also altered in PAM preference with little effect on overall activity [14]. Therefore, we attempted to introduce similar mutations from enAsCas12a (E174R/S542R/K548R) into LbCas12a to generate enLbCas12a (D156R/G532R/K538R). However, only D156R exhibited any beneficial effect, whereas the other two mutations are detrimental to editing [15]. This observation is faithfully recapitulated by our E. coli screen (Fig. 1D). For position D156, nearly all amino acid changes were enriched over the selection process, suggesting the removal of negative charge at the side chain of position 156 is the underlying feature to improve on-target activity. In contrast to the D156 site, most point mutations with measurable enrichment scores at positions G532 or K538 are detrimental, including the G532R introduced into enLbCas12a. In addition, we found significant data dropouts at both positions, further indicating that both positions are not amendable for mutagenesis (Fig. 1D). Taken together, we provided examples where the enrichment scores of 5 different mutation sites correlated well with the known effects on the intrinsic activity of LbCas12a. While we cannot rule out the absence of false positives, we are confident that the mutations represent true positive hits with enhanced activity in our dataset.

LbCas12a-RVQ is the optimal nuclease with robust activity in human cells

To further validate the editing performance of some mutations in human cells, 1977 positive mutation were ranked by the phenotype (i.e., enrichment score; Additional file 2: Table S1) and then prioritized over residues on the DNA/R-Loop binding interface, which ended up with 24 point mutations. We characterized the on-target editing efficiency of 24 novel LbCas12a point mutations in human cells using RNP delivery. Purified LbCas12a proteins were programmed with 4 different synthetic crRNAs to edit the human HPRT gene in HEK293 cells, and the efficiency was measured by T7EI assay 2 days post-delivery (Additional file 1: Fig S2A). The relative activity of each mutant over WT LbCas12a was shown in Fig. 2A. Overall, we found 16 of the 24 mutations enhanced the on-target editing efficiency of LbCas12a. Mapping the validated hits on the existing crystal structure of the LbCas12a-crRNA-dsDNA complex (PDB: 5XUS [16]) revealed that most hits were in proximity to dsDNA or DNA/RNA duplexes (Fig. 2B). We then combined these validated hits in various double and triple mutation combinations and identified a specific variant with 3 mutations (RVQ: G146R/R182V/E795Q) that performed the best overall (Fig. 2A, B). We also generated and tested some higher-order quadruple and quintuple mutants which however did not show improved editing efficiency; Rather, editing efficiency was drastically reduced with some combinations (Fig. 2A). For example, combining more positive hits on RVQ, such as RVQVK (RVQ + P799V + T814K), reduced the overall efficiency of LbCas12a when delivered as RNP (Fig. 2A). Overall, our analysis identified RVQ as a top-performing LbCas12a variant for robust genome editing in human cells.

Fig. 2figure 2

LbCas12a-RVQ is a novel variant with robust editing efficiency in human cells. A The normalized editing efficiencies of 56 LbCas12a variants in human HEK293 cells over four target sites when delivered as RNP. See Additional file 1: Fig. S2A for the raw editing efficiency. B Mapping the validated positive hits with enhanced activity on the existing LbCas12a structure. Red: residues used in LbCas12a-RVQ constructs. Gold: other residues with improved activity but not included in the LbCas12a-RVQ. C Editing efficiency of LbCas12a-WT, LbCas12a-RVQ and LbCas12a-RRVQ in HEK293 cells over additional 11 target sites at a low RNP dosage (0.1 μM). NC indicates negative control. One-way analysis of variance (ANOVA) with Tukey’s multiple comparison test (p < 0.05) were analyzed using GraphPad Prism 9. D, E The DNA binding specificities of WT LbCas12a (D) and LbCas12a-RVQ (E) measured by Spec-seq over the same target site and presented as motif logo for visualization. F, G The DNA cleavage specificities of WT LbCas12a (F) and LbCas12a-RVQ (G) measured by SEAM-seq

We further expanded the evaluation of LbCas12a-RVQ over an additional 11 sites by targeting a variety of therapeutic relevant targets. Cas12a-RNPs were intentionally delivered to HEK293 cells at low dosage (0.1 μM) via nucleofection, to better reveal the activity differences between these nucleases. As the baseline, WT LbCas12a showed a median editing efficiency at 7.4%. while LbCas12a-RVQ significantly enhanced the editing efficiency (median efficiency: 18.8%) (Fig. 2C). However, combining RVQ with D156R (RRVQ) compromised the overall efficiency (Fig. 2C). We thus concluded that the RVQ variant is a robust LbCas12a nuclease for human genome editing via RNP delivery.

We next characterized the intrinsic sequence specificity of LbCas12a-RVQ in vitro by SEAM/Spec-seq [16]. The DNA cleavage and binding specificities of LbCas12a RNPs (WT or RVQ) were measured over a library with sequences with up to 4 mismatches to the crRNA (Additional file 1: Fig S2B). Highly consistent measurements of both binding and cleavage specificities were obtained from two biological replicates (Additional file 1: Fig S2C). Biophysical models that describe the quantitative penalties of mismatches at each position of the target site were generated by non-linear regression and presented as motif logos for visualization (Fig. 2D–G). In general, the overall profiles of binding and cleavage specificities are highly similar between WT and RVQ, where the first 18-bp of crRNA mediates most sequence specificities. Incorporating RVQ mutations did not alter any base preference at each position, including the PAM region. As the trade-off for the enhanced activity, we observed slightly reduced binding and cleavage specificities of RVQ over the entire target site, which is analogous to the difference between WT AsCas12a and AsCas12a Ultra under the identical experimental condition [5]. As the mutations in the AsCas12a Ultra marginally affected the off-target profile in human cells [5], we reasoned that LbCas12a-RVQ largely maintained the intrinsic sequence specificity of WT LbCas12a, but with enhanced activity.

LbCas12a-RV confers robust singular and multiplexed editing in rice and tomato protoplasts

Previous studies have shown that the editing efficiency of Cas12a varied from human cells to plant species and in general Cas12a nucleases are sensitive to temperature [3, 17,18,19]. Although various engineered Cas12a proteins have been reported to possess enhanced temperature tolerance and relaxed PAM requirements [2, 20], the editing efficiency is still relatively low in most plant species. There is a clear need to enhance the editing activity of Cas12a in plants at low temperatures that are more relevant to plant tissue culture. Therefore, we evaluated the editing efficiency of our engineered LbCas12a variants in the protoplasts of rice (a monocot crop) and tomato (a dicot crop) at both 32 °C and 25 °C.

Besides LbCas12a-RVQ and LbCas12a-RRVQ, we included another two new variants LbCas12a-RV (G146R/R182V) and LbCas12a-RRV (G146R/D156R/R182V) in the rice protoplast assay (Fig. 3A). LbCas12a-RV enabled highest editing efficiency at the 6 out of 8 target sites, while LbCas12a-RRV was comparable with LbCas12a-RVQ at all target sites (Fig. 3B). LbCas12a-RRVQ showed compromised editing efficiency as similarly observed in human cells (Figs. 2C and 3B). We further compared LbCas12a-RV with WT LbCas12a in rice protoplasts by editing four target sites with RNP delivery (Fig. 3C). At low RNP dosage (0.001 μM), LbCas12a-RV showed significantly higher editing efficiency than WT LbCas12a at all four target sites at 32 °C (Additional file 1: Fig S3A). The same trend was observed at two sites (Os03g52594-TTTA and GA1-TTTA) when performing the editing at a lower temperature (25 °C), although the efficiency was globally reduced for both WT-LbCas12a and LbCas12a-RV (Additional file 1: Fig S3A). As expected, increasing the dosage from 0.001 μM to 0.01 μM enhanced the editing efficiency for all the nucleases (Fig. 3D and Additional file 1: Fig S3A). At this higher concentration, LbCas12a-RV showed significantly higher editing efficiency than WT LbCas12a in rice protoplasts across all four target sites at both 25 °C and 32 °C (Fig. 3D).

Fig. 3figure 3

Enhanced editing efficiency of LbCas12a-RV in rice and tomato protoplasts using plasmid and RNP delivery. A Diagram of the plasmid expression system of Cas12a and crRNA. B Editing efficiency of LbCas12a variants at four TTTV target sites and four VTTV target sites in rice protoplasts. C Diagram of RNP delivery in rice protoplasts. Rice protoplasts were transfected with 0.01 μM RNP and incubated in two temperature conditions (25 °C and 32 °C) for 2 days. D Editing efficiency of LbCas12a and LbCas12a-RV at two TTTV target sites and two VTTV target sites in rice protoplasts with RNP delivery. E Diagram of RNP delivery in tomato protoplasts for multiplexed genome editing. Tomato protoplasts were transfected with 0.06 μM RNP (0.01 μM RNP per target) and incubated in two temperature conditions (25 °C and 32 °C). F Editing efficiency of LbCas12a and LbCas12a-RV at six target sites in tomato protoplasts with RNP delivery. The editing efficiency was measured by NGS. One-way analysis of variance (ANOVA) with Tukey’s multiple comparison test (p < 0.05) were analyzed using GraphPad Prism 9

We next evaluated LbCas12a variants in tomato protoplasts by simultaneously targeting 6 sites with RNP delivery (Fig. 3E). To keep the same RNP concentration for each target site in tomato protoplasts as in rice protoplasts, 0.006 μM and 0.06 μM RNP were used. At 0.006 μM RNP concentration, LbCas12a-RV resulted in significantly higher editing than WT LbCas12a at 5 out of 6 sites at 25 °C and at all 6 sites at 32 °C (Additional file 1: Fig. S3B). At this low RNP concentration, LbCas12a-RV displayed significantly higher editing efficiency at 32 °C than at 25 °C (Additional file 1: Fig. S3B). At 0.06 μM RNP concentration, genome editing efficiencies by WT LbCas12a and LbCas12a-RV were both high at 32 °C, about 60% at all target sites (Fig. 3F). At 25 °C, LbCas12a-RV sustained the same high editing efficiencies of 32 °C at all 6 sites (Fig. 3F). By contract, WT LbCas12a’s editing efficiencies were all dropped significantly at this relatively low temperature (Fig. 3F). Taken together, genome editing by LbCas12a-RV appears to be less sensitive to temperature than WT LbCas12a at all target sites in rice and tomato protoplasts with the higher RNP concentration (Fig. 3D, F) and at most target sites with the lower RNP concentration (Additional file 1: Fig S3). Thus, the LbCas12a-RV variant is robust for plant genome editing at variable temperatures. To further verify the editing efficiency of the new LbCas12a variants, they were subsequently tested in transgenic rice and poplar.

LbCas12a-RRV outperforms other variants in stable transgenic rice

Agrobacterium-mediated stable transformation of CRISPR reagents is the primary delivery method for obtaining genome-edited plants. To evaluate the performance of our LbCas12a variants in transformed plants generated by this common method, we first compared the editing efficiency of WT, D156R, RV, and RRV versions of LbCas12a in rice T0 plants using a multiplexed Cas12a/crRNA expression system (Fig. 4A) [21]. Four target sites were selected with each being duplicated in the rice genome with the identical protospacers but with different PAMs: canonical TTTV vs non-canonical VTTV PAMs (Fig. 4B). In total, eight target sites were targeted simultaneously by four crRNAs (Fig. 4A, B). Of the 4 canonical TTTV sites, 2 sties (GA1-TTTA and Os01g09810-TTTC) showed significantly greater indel frequencies with all three variants and one site (GA1-TTTA) showed significant difference between LbCas12a-RRV and LbCas12a-RV (Fig. 4B). LbCas12a-RRV generated greater biallelic editing compared to the other variants with 100% of the tested lines for the GA1-TTTA site showing biallelic editing (Fig. 4C and Additional file 1: Fig S4). LbCas12a-RRV consistently showed higher indel frequencies compared to the other variants and WT LbCas12a at the four non-canonical VTTV PAM sites (Fig. 4B, C and Additional file 1: Fig S4B, D). Impressively, LbCas12a-RRV achieved 100% biallelic editing at 4 out of 8 target sites (GA-TTTA, Os11g19880-TTC, Os11g20160-TTTC, and Os01g09810-TTTC), whereas other Cas12a variants only generated 100% biallelic editing at one site, Os11g20160-TTTC (Fig. 4C and Additional file 1: Fig S4C, E). These results suggest that LbCas12a-RRV is a promising and robust Cas12a nuclease for generating genome edited rice plants not only at canonical TTTV PAM sites but also non-canonical VTTV PAM sites.

Fig. 4figure 4

Robust genome editing by LbCas12a-RRV in rice T0 transgenic plants. A Diagram of multiplexing four crRNAs to simultaneously target eight sites using a dual ZmUbi promoter and tandem HH-crRNA-HDV system. B Indel frequencies of eight target sites in rice T0 plants with each dot representing one T0 plant. C Genotypes of T0 plants at GA1-TTTA, Os11g19880-TTC and Os12g12600-TTA sites. Wild type (WT) denoted as empty rectangles, biallelic edit denoted as fully filled rectangles, monoallelic edit denoted as half-filled rectangles, chimeric edit denoted as doted rectangles. One-way analysis of variance (ANOVA) with Tukey’s multiple comparison test (p < 0.05) were analyzed using GraphPad Prism 9. Bars without assigned letters indicate no significant differences among four LbCas12a nucleases

LbCas12a-RRV outperforms other variants in stable transgenic poplar

Poplar is a perennial tree that also serves as an important bioenergy and biomaterials feedstock in foundational and translational research. Biallelic genome editing in the first generation is necessary for functional analysis of the target genes in trees like poplar. To further evaluate the performance of LbCas12a-RRV, we used Agrobacterium to deliver LbCas12a/crRNAs into poplar. Six sites were simultaneously targeted using a multiplexed expression system (Fig. 5A). Since AsCas12a was previously reported to show higher editing efficiency than LbCas12a in poplar [22], AsCas12a was also included for comparison with LbCas12a variants. Based on the editing efficiency of WT-LbCas12a, there are 2 high-activity (SVP-crRNA1 and SVP-crRNA2) and 4 low-activity sites (4CL1-crRNA1, 4CL1-crRNA2, PII-crRNA1, and PII-crRNA2) (Fig. 5B). At the 2 high-activity sites in PtSVP (SVP-crRNA1 and SVP-crRNA2), LbCas12a-RV and LbCas12a-RRV showed 100% biallelic editing efficiency, significantly better than LbCas12a-D156R, LbCas12a-RVQ, LbCas12a-RRVQ, LbCas12a, and AsCas12a (Fig. 5B, C). At the 4 medium-to-low activity sites in Pt4CL1 and PtPII, LbCas12a-RRV showed significantly higher editing activity than all other Cas12a proteins tested (Fig. 5B). Although biallelic edits were detected for each of the LbCas12a variants, they were more frequent for LbCas12a-RRV (4 of 6 sites) than LbCas12a-D156R (2 of 6 sites) and LbCas12a-RVQ (2 of 6 sites) (Fig. 5C). Sequencing analysis of the two PtSVP target sites in T0 lines showed that LbCas12a-RRV resulted in 100% biallelic editing with deletions ranging from 3 to 14 bp and with homozygous mutation rates of 50% and 83.3%, respectively (Fig. 5D, E). Therefore, LbCas12a-RRV is more robust and efficient for poplar genome editing than other LbCas12a variants and AsCas12a. It is advantageous to use LbCas12a-RRV over other Cas12a nucleases/variants in poplar due to its high frequency biallelic and homozygous editing.

Fig. 5figure 5

Robust genome editing by LbCas12a-RRV in poplar T0 transgenic plants. A Diagram of multiplexing six crRNAs to simultaneously target six sites using a dual pAtUBQ10 promoter and tandem HH-crRNA-HDV system. B Indel frequencies of six target sites in poplar T0 plants with each dot representing one T0 plant. C Editing efficiencies at six target sites. D Genotypes of 12 T0 plants at SVP-crRNA1 by LbCas12a-RRV. E Genotypes of 12 T0 plants at SVP-crRNA2 by LbCas12a-RRV. One-way analysis of variance (ANOVA) with Tukey’s multiple comparison test (p < 0.05) were analyzed using GraphPad Prism 9

留言 (0)

沒有登入
gif