Cas9 cassette optimisation
Our pilot work in barley (14) utilised a Cas9 coding sequence (CDS) which was codon optimised for expression in human cells (HsCas9) (15) and whilst functional in barley resulted in rather low efficiency of mutagenesis (23% of plants for one construct and 10% for a second construct). More recently, other versions of Cas9 CDS’s were reported to work efficiently in other plant species. An Arabidopsis codon optimised version with one intron (AtCas9+1int) (16) outperformed HsCas9 in Arabidopsis. In another report a Zea mays codon optimised version with thirteen introns (ZmCas9+13int) outperformed HsCas9 in Arabidopsis and was also very effective in N.benthamiana and C. roseus (17). To compare these three coding sequence variants in barley, we targeted five genes. Each gene was targeted by four gene specific guides, which were paired with either HsCas9, AtCas9+1int or ZmCas9+13int in three separate constructs. In total fifteen constructs were made and transformed into the barley cultivar Golden Promise, enabling CDS comparison over five different target genes (Additional file 1). Between 17 and 24 independent T0 lines were made for each construct, which were subsequently screened by PCR and Sanger amplicon sequencing. Lines were scored positive for mutagenesis where indels or larger fragment insertions/deletions were detected by alignment to wild type reference sequence. Both intron containing versions (AtCas9+1int & ZmCas9+13int) outperformed HsCas9, but overall ZmCas9+13int performed best (figure 2a & additional file 2). The average mutagenesis efficiency for the three CDS variants over the five genes was 33% (HsCas9), 88% (AtCas9+1int), 96% (ZmCas9+13int).
Cas9 guide architecture optimisation
Most frequently guide RNAs for Cas9 are expressed using a polymerase 3 promoter, typically with one promoter per guide. Polymerase 2 promoters can be used to successfully drive guide expression in plants, where a single promoter can drive a multiple guide array. In this situation all guides are transcribed on a single transcript and are separated by sequences which bring about precise cleavage of the transcript to release individual guide RNAs, enabling Cas9/guide ribonucleoprotein formation. One way to release guides from a single transcript has been to use self-cleaving ribozyme sequences, which has been shown in Cas12a systems (18). Another way to bring about transcript processing is to separate each guide with the recognition/cleavage sequence for CSY4, a CRISPR associated protein, which is co-expressed with the Cas nuclease and is therefore able to release individual guides. This system is reported to work very effectively in some dicotyledonous species (19-21). We decided to compare these three different guide architectures for expressing Cas9 guides in barley, this time by targeting three genes simultaneously. Three constructs were made, differing only in the guide architecture used (Additional file 3). The protospacers used in each architecture were identical and in the same sequential order for the three constructs. Each of the three genes was targeted by 4 different gene specific guides, meaning each construct contained 12 guides. In the first construct (A), each of the 12 guides was driven by its own polymerase 3 promoter, either TaU6, TaU3 or HvU3. In the second construct (B) the 12 guides were driven by a single OsUbiquitin polymerase 2 promoter and separated by self-cleaving hammerhead (HH) and hepatitis delta virus (HDV) ribozyme sequences. In the third construct (C) the 12 guides were again driven by a single OsUbiquitin polymerase 2 promoter but this time separated by CSY4 recognition sequences. The CSY4 protein was also co-expressed in this construct to bring about transcript processing. The Cas9 variant used in all 3 constructs was ZmCas9+13int. We produced 20 T0 lines for constructs A and B, but despite multiple repeat transformations, were only able to produce 4 T0 lines for construct C making comparisons with A and B of limited value. Architecture A was the most efficient where 100% of T0 plants were mutagenised in all three of the target genes (figure 2b & additional file 4). Architecture B was almost as efficient where 100% of T0 plants were mutagenised in two of the target genes and 90% were mutagenised in the third gene. Architecture C was the least efficient, with 75% of T0 plants mutagenised in one of the target genes and 50% mutagenised in two target genes. The mean mutagenesis efficiencies of Cas9 guide architectures A,B & C were 100%, 97% and 58% respectively.
Cas9 in wheat
We took the best performing Cas9 parts from the barley experiments above (ZmCas9+13int & guide architecture A) and used them to target a single receptor like kinase gene (RLK) in hexaploid wheat (cv. Fielder) using a single construct (Additional file 5). Four guides each targeting the A,B & D copies (TraesCS7A02G264400, TraesCS7B02G162500, TraesCS7D02G265400) were used. In addition, a GRF-GIF overexpression cassette was included (13) to boost transformation efficiency. Forty-eight transgenic lines were screened by PCR/Sanger sequencing at each of the A,B & D target sites in conjunction with alignment to wild type reference sequences to identify mutagenic lines. The mutagenesis efficiency was 98%, 94% & 92% in the A,B & D subgenomes respectively with an overall mean of 95% (figure 2c & additional file 6).
Cas12a CDS optimisation
The first report of LbCas12a use in plants was in rice where the CDS used (OsCas12a) was codon optimised for that species (22). Later a CDS which was originally used in human cells and codon optimised accordingly (HsCas12a) (23) was shown to work in N.benthamiana, tomato and Arabidopsis(24). A third version was codon optimised for Arabidopsis and was shown to be considerably more efficient for mutagenesis at the relatively low temperatures plants grow at (tt=temperature tolerant), when a D156R mutation was included (ttAtCas12a) (25). We decided to compare these CDS variants in barley in addition to three further versions, firstly ttAtCas12a without the temperature tolerant mutation (AtCas12a), secondly HsCas12a containing the D156R mutation (ttHsCas12a), and thirdly ttAtCas12a into which we inserted eight Arabidopsis introns (ttAtCas12a+8int). To compare these 6 Cas12a CDS’s, we made six constructs (Additional file 7) differing only in the CDS used. Each construct contained a 3 guide array targeting the barley genes chr6Hg0653951, chr7Hg0684671 and chr2Hg0138701 using a guide architecture shown to function in Arabidopsis(26). Twenty T0 plants were made for each of the six constructs which were screened at the target loci by Illumina amplicon sequencing. The percentage of mutant reads was determined at each of the three target loci for each T0 plant produced (Additional file 8). Figure 2d shows the overall mutagenesis efficiencies of the Cas12a CDS variants over the three target genes. Surprisingly the OsCas12a was not functional in barley with no increase in the mutations detected compared to the wild type control (mean overall background artifact mutations of 3.6% and 3.5% respectively). AtCas12a had a mean efficiency of 26% although chr7Hg0684671 was not hit much above the baseline of the WT control (AtCas12a=5%, WT =4.5%) (Additional files 8 & 9). HsCas12a was more effective overall, having a mean efficiency of 44% although still ineffective at targeting chr7Hg0684671. Addition of the D156R temperature tolerant mutation in both Arabidopsis and Human optimised CDSs increased the overall mean mutagenesis efficiency to 53% and 72% for ttAtCas12a and ttHsCas12a respectively. The D156R addition also enabled the mutagenesis of chr7Hg0684671 with a mean efficiency of 17% and 33% respectively. Overall ttAtCas12a+8int performed best with a mean overall efficiency of 87%. Even the difficult to target chr7Hg0684671 was mutagenised at a mean efficiency of 68%.
Dissecting the role of introns in Cas12a
We decided to dissect the role that the eight introns inserted into ttAtCas12a+int played in boosting the performance of ttAtCas12a. To do this we made a further four constructs which retained some of the introns contained in ttAtCas12a+8int. ttAtCas12a+1int retained just the first intron, ttAtCas12a+S1int the first 3 introns, ttAtCas12a+S2int the second 3 introns and ttAtCas12a+S3int the last 2 introns (Additional file 10). These four additional Cas12a variants were coupled with the guide cassettes targeting the same three barley genes as before (chr6Hg0653951, chr7Hg0684671 and chr2Hg0138701) and once again twenty T0 lines were produced and screened using Illumina amplicon sequencing. Inclusion of the first intron in ttAtCas12a+1int had no great boosting effect with a similar overall mean efficiency to the no intron version ttAtCas12a of 54% & 53% respectively (Additional files 10 & 11). The first three introns in ttAtCas12a+S1int enhanced overall efficiency by 14%, the second three introns in ttAtCas12a+S2int by 31% and the final 2 introns in ttAtCas12a+S3int by 13%. The majority of intron mediated enhanced mutagenesis was because of the 3 introns in ttAtCas12a+S2int which has a very similar overall percentage enhancement compared to the 8 introns in ttAtCas12a+8int (31% & 34% respectively). However, comparing the percentage enhancement of ttAtCas12a+8int and ttAtCas12a+S2int at the most recalcitrant target chr7Hg0684671 (51% & 43% respectively) shows that there is a modest additional enhancement of 8% with all eight introns compared to just the three in ttAtCas12a+S2int (Additional files 10 & 11).
Combinatorial effect of D156R and introns in Cas12a
We also wanted to see if there was any combined advantage with the inclusion of introns and the D156R mutation which was greater than the sum of the individual parts. To do this we required one further construct targeting the three barley genes chr6Hg0653951, chr7Hg0684671 and chr2Hg0138701. This construct had a wild type D156 residue and all eight introns, being designated AtCas12a+8int. By comparing the efficiency of AtCas12a to ttAtCas12a, AtCas12a+8int and ttAtCas12a+8int the separate and combined enhancing roles of D156R and introns could be determined. Overall mean efficiencies (Additional files 12 & 13) were improved by the inclusion of singular D156R (53%) and introns (55%) in ttAtCas12a & AtCas12a+8int relative to AtCas12a (26%), the CDS which contained neither. The overall mean efficiency when both D156R and introns were combined in ttAtCas12a+8int was highest at 87%. For the target genes chr6Hg0653951 and chr2Hg0138701 the mean efficiencies of singular DI56R (ttAtCas12a) and introns (AtCas12+8int) were too high to allow determination of the sum of the individual parts (D156R enhancement + intron enhancement +base AtCas12a efficiency) because these values were more than 100% (Additional file 13). However, the summed efficiencies at the recalcitrant target chr7Hg0684671 were just 20%, allowing the gain from the combinatorial D156R/intron effect to be determined as 48%, considerably more than the sum of the individual parts (Additional file 12).
Cas12a guide architecture optimisation
We already achieved efficient Cas12a mutagenesis in the CDS comparison above using a ribozyme-based guide architecture we designate as version 2 (V2). Previously we tested a version 1 (V1) architecture which relies on the innate processing activity of Cas12a itself to release individual guides from a single transcript and found V2 superior to V1 (27). Later, it was reported that a Cas12a guide architecture containing tRNA sequence instead of ribozymes worked best in wheat (28) and so we decided to test this in barley, designating the tRNA version V3. For this we used two constructs each targeting the same three genes that were used in the Cas12a CDS comparison (chr6Hg0653951, chr7Hg0684671 and chr2Hg0138701). One construct contained guides in a V2 array and the other a V3 array (Additional file 14). Twenty T0 plants were made for each construct which were screened by Illumina amplicon sequencing to determine the percentage mutation at each locus. Overall, there was little difference in the mean editing efficiency between V2 and V3 over the three targets which recorded 87% and 90% respectively (Additional file 15). However, the mean efficiencies at the recalcitrant chr7Hg0684671 locus had V2 & V3 values of 68% & 90% respectively (figure 2e), showing a 22% improvement when using V3.
Cas12a in wheat
We trialled two Cas12a CDS’s which performed well in barley (ttHsCas12a & ttAtCas12a+8int) in wheat. To do this we selected three guides which were previously shown to be effective at targeting two genes in hexaploid wheat, GW2 and GW7 (29). Guides GW7T14 and GW7T13 target the three subgenome copies of GW7 and GW2T6 targets the three subgenome copies of GW2. Two constructs were made differing only in the Cas12a CDS used which contained these three guides in a V2 guide architecture (Additional file 16). A third construct was also made in response to the recent report (28) which identified a tRNA based guide architecture (V3) as being superior to a V2 architecture in wheat. Additionally, all three constructs contained a GRF-GIF overexpression cassette to increase the transformation efficiency (13). Constructs were transformed into the spring variety Fielder and 48 lines per construct were screened at the target loci by Sanger amplicon sequencing and alignment to the wild type reference. As was seen in barley, ttAtCas12a+8int performed better than ttHsCas12a (figure 2f & additional file 17) in wheat, giving at all subgenome GW2 and GW7 loci, a greater percentage of mutagenised T0 plants. The average percentage of mutagenised T0 plants across all subgenomes of both GW2 and GW7 was 46% for ttHsCas12a and 72% for ttAtCas12a+8int. Comparison of the mutagenesis efficiency resulting from the two guide architectures shows that V3 outperforms V2 in 5 of the 6 subgenome loci and raises the average percentage of mutagenised T0 plants from 72% to 90%.
Cas9 and Cas12a co-mutagenesis in barley
We tested whether it was possible to introduce both Cas9 and Cas12a components on the same T-DNA to edit simultaneously in barley. To do this, we made constructs expressing both ZmCas9+13int and ttAtCas12a+8int in addition to guide expression cassettes of type A for Cas9 and V2 for Cas12a (Additional file 18). Four constructs were made, two targeting two genes (constructs 1 & 2), one targeting three genes (construct 3) and one targeting four genes (construct 4). Each gene was targeted by four different gene specific guides, either four Cas9 guides, four Cas12a guides, or a combination of Cas9 and Cas12a guides. This made the T-DNA inserts rather large (construct 4 targeting 4 genes had a T-DNA of 28.1 Kb) which likely impacted transformation efficiency. However, we still managed to produce between 11 and 19 T0 plants for each construct, which were screened by Sanger amplicon sequencing and alignment to the wildtype reference. Table 1 shows that with all four constructs Cas9 and Cas12a were able to effectively mutagenise all target genes simultaneously. In T0 lines where Cas9 was effectively editing, it was apparent that Cas12a was also editing. For example, with construct 1, target chr5Hg0454731 addressed by Cas12a was edited in 45% of T0 lines created. Target chr4Hg0361381 addressed by Cas9 in construct 1 was also effective in exactly the same T0 lines, giving the same 45% efficiency at this target. This explains the overall 45% efficiency at “all targets” i.e. the percentage of T0 plants where both targets were simultaneously edited. This pattern was repeated with the remaining constructs 2, 3 and 4 and although not every target gene was hit in each active T0 line, the percentage edited in all targets was always determined by the target with the lowest editing efficiency for that construct. For example, construct 4 targeted chr1Hg0000251 at 68%, chr5Hg0508401 at 90%, chrHg0829711 at 95% and chr3Hg0346691 at 90%. The lowest efficiency for these four targets was 68% for chr1Hg0000251, limiting the percentage of T0 lines where all four genes were successfully targeted to 68%.
Table 1 Barley T0 editing efficiency of constructs 1,2,3 & 4 which co-express Cas9 and Cas12a components.
Construct
|
T0 plants
|
Target 1
|
Target 2
|
Target 3
|
Target 4
|
% all targets edited
|
1
|
19
|
chr5Hg0454731
|
% T0 edited
|
chr4Hg0361381
|
% T0 edited
|
NA
|
% T0 edited
|
NA
|
% T0 edited
|
Guides
|
45
|
Guides
|
45
|
NA
|
NA
|
NA
|
NA
|
45
|
Cas12a
|
Cas9
|
2
|
11
|
chr6Hg0674791
|
100
|
chr6Hg0674101
|
100
|
NA
|
NA
|
NA
|
NA
|
100
|
Guides
|
Guides
|
Cas9
|
Cas12a
|
3
|
19
|
chr7Hg0693091
|
42
|
chr2Hg0110011
|
32
|
chr5Hg0460461
|
47
|
NA
|
NA
|
32
|
Guides
|
Guides
|
Guides
|
Cas9 & Cas12a
|
Cas9
|
Cas12a
|
4
|
19
|
chr1Hg0000251
|
68
|
chr5Hg0508401
|
90
|
chrHg0829711
|
95
|
chr3Hg0346691
|
90
|
68
|
Guides
|
Guides
|
Guides
|
Guides
|
Cas9
|
Cas9 & Cas12a
|
Cas9 & Cas12a
|
Cas9 & Cas12a
|
The toolkit
We tested various Cas9 and Cas12a CDS variants and guide architecture designs in barley then applied the best to wheat, achieving very efficient mutagenesis. These components are the basis for a complete CRISPR Cas9 & Cas12a toolkit we have made available via AddGene. Species specific level 2 binary vectors are provided in the toolkit (table 2) as well as level 1 guide accepters (figure 3).
Table 2 summarises the level 2 binary vectors available.
Requirement
|
Vector
|
EC64420
|
EC67842
|
EC64434
|
EC67844
|
EC70364
|
EC67907
|
EC67843
|
EC67841
|
EC67908
|
Barley
|
✓
|
✓
|
✓
|
×
|
×
|
×
|
×
|
×
|
×
|
Wheat
|
×
|
×
|
×
|
✓
|
✓
|
✓
|
✓
|
✓
|
✓
|
Cas9
|
✓
|
×
|
✓
|
✓
|
✓
|
✓
|
×
|
×
|
×
|
Cas12a
|
×
|
✓
|
✓
|
×
|
×
|
×
|
✓
|
✓
|
✓
|
GRF-GIF
|
×
|
×
|
×
|
×
|
✓
|
×
|
×
|
✓
|
×
|
GR-GRF-GIF
|
×
|
×
|
×
|
×
|
×
|
✓
|
×
|
×
|
✓
|
For barley we provide three level 2 binary vectors containing a hygromycin marker for plant selection and either one or two nuclease expression cassettes. EC64420 contains a single ZmCas9+13int expression cassette, EC67842 contains a single ttAtCas12a+8int expression cassette and EC64434 contains both ZmCas9+13int & ttAtCas12a+8int expression cassettes. For wheat we provide six level 2 binary vectors containing a hygromycin marker for plant selection and either a ZmCas9+13int (EC67844, EC70364, EC67907), or a ttAtCas12a+8int (EC67843, EC67841, EC67908) expression cassette. Each of the nuclease versions is also available with (EC70364 & EC67841) and without (EC67844 & EC67843) an additional GRF-GIF overexpression cassette which greatly increases the efficiency of transformation in many wheat cultivars, but also is associated with some pleiotropic effects including sterility in some varieties. To retain the benefits of GRF-GIF overexpression during the transformation process, whilst alleviating pleiotropic effects later in plant development, we developed two further Cas9 and Cas12a versions possessing GR-GRF-GIF fusions (EC67907, EC67908). These allow induction of GRF-GIF activity in the presence of dexamethasone (Dex), which can be provided in the tissue culture media. Later absence of Dex during growth on soil should eliminate or reduce any undesirable GRF-GIF phenotypes. To derive the GR-GRF-GIF fusion we tested both N-terminal and C-terminal fusions of the rat glucocorticoid receptor (GR) to GRF-GIF (GR-GRF-GIF, GRF-GIF-GR) in two Fielder transformation experiments (Additional file 19). We found that the N-terminal version showed a greater benefit to the number of wheat shoots regenerated in the presence of Dex than the C-terminal version and a greater ability to switch off in the absence of Dex. GR-GRF-GIF gave an average transformation efficiency over two experiments of 55% in the presence of Dex and 15% in the absence of Dex. GRF-GIF-GR gave an average transformation efficiency of 33% with Dex and 24% without Dex. Therefore, the N-terminal fusion gave a 22% higher transformation rate in the presence of the inducer than the C-terminal version (55%>33%) and a greater difference in transformation efficiency between the induced and non-induced states (40%>9%). Whilst it is important to have a high rate of transformation in the presence of the inducer, it is also important that the non-induced condition has a considerably lower rate of transformation, indicating that the GRF-GIF effect is indeed switchable and likely to reduce pleiotropic effects later in development where the inducer is not present. We grew ten of the regenerated Fielder wheat plants containing GR-GRF-GIF taken from the +Dex conditions to maturity and observed no obvious phenotypic effects. All ten had normal spike development and were completely fertile.
For the toolkit, cloning begins at level 1 (figure 4a) by the insertion of complementary oligonucleotide pairs representing the protospacer sequence into level 1 guide accepters. Once oligonucleotides have been inserted into the guide accepters the entire transcriptional units can then be cloned either directly into level 2 binary vectors (figure 4b), or into level M accepters as an intermediate step (figure 4c). We have found that cloning is most efficient when the number of fragments ligated in any one step are kept to a minimum. When between 1 and 4 guides are required, level 1 can proceed directly to level 2 with no intermediate step. When more than 4 guides are required in the final level 2 vector, then the intermediate level M step is used, with again up to 4 guides stacked in any level M vector. Level 1 guide accepters with TaU6, TaU3 and HvU3 promoters are provided in all seven Golden Gate positions, both for the Cas9 guide architecture A (Additional file 3) and Cas12a V2 architecture (Additional file 14). In addition, we also provide Cas12a V3 guide accepters (Additional file 14) in all three promoter & seven position formats. We have found that when stacking level 1 guide cassettes in level M and level 2 it is good practice to alternate the sequential order of U3 and U6 promoters to avoid issues with bacterial recombination which are caused during cloning by repetitive tandem arrays.
Additional file 20 details (in the appropriate tabs) the various level 1, level M, level 2 & end linker plasmids available via AddGene. Genbank (.gb) file downloads of these plasmids are available on the AddGene browser. Additional file 20 also indicates the appropriate vector choices when between 1 and 26 guides are required in the final level 2 assembly (vector choices tab).
Additional file 21 details the constructs, guides and PCR primers used in this study.
Additional files 22-61 are the gb files of the constructs listed in additional file 21.