DSB repair in the variable region of the 45S rDNA
To determine the mechanism of DSB repair in the 45S rDNA in somatic cells, we induced DSBs within the 45S rDNA repeats using Staphylococcus aureus Cas9 (SaCas9)14. The target site was selected in the variable IGS region, so that only a subset of repeats was cleaved (Fig. 1a; Supplementary Table 1). As a control, we designed a gRNA targeting the non-essential ADH1 gene (Ubi-SaCas9-ADH1) which induces a unique DSB. Col-0 plants were transformed with the expression vector, Ubi-SaCas9-IGS, as well as the control vector. After two weeks of growth, 40 transgenic plants per line were selected and pooled for extraction of genomic DNA. Subsequently, the 45S rDNA target site was amplified and subjected to next-generation sequencing (NGS) to determine the repair pattern. The deep sequencing analysis revealed that the majority of mutations at the break site were deletions, most of them smaller than 10 bp, and that the repair occurred without MH (Fig. 1b and c). Only a tiny fraction of reads showed insertions or InDels at the break site. These results strongly imply that the cNHEJ is the major DSB repair pathway within the 45S rDNA. To verify this, DSBs were induced in the 45S rDNA of DNA repair mutants at the same target site. The cNHEJ mutants, ku70-1 and lig4-5, as well as the HR mutant, rad54-1, were transformed with the expression and the control construct. The RAD54 mutant was chosen because it is, in contrast to other HR factor mutants, not sterile but has a strong deficiency in somatic HR15. After two weeks of growth, the survival rates of all transformed mutant lines were determined and normalized to Col-0. We expected that an insufficient 45S rDNA repair, due to loss of DNA repair factors, should result in a reduction of the survival rate. Indeed, both cNHEJ mutant lines, ku70-1 and lig4-5, showed a dramatic reduction in survival rate, when the 45S rDNA was cleaved (Fig. 1d). In contrast, no reduction of the survival rate could be observed in the rad54-1 HR mutant line. Thus, cNHEJ plays a key role in 45S rDNA DSB repair in somatic cells.
a The 45S rDNA in Arabidopsis is arranged in up to 750 repeats, in a head-to-tail-fashion. Each repeat consists of a unit of three rRNA genes (18S, 5.8S, and 25S), separated by a variable intergenic spacer (IGS) region. The individual rRNA genes are separated by two highly conserved internal transcribed spacers (ITS). The triangles represent the cleavage sites for SaCas9 within the repeats. By targeting the variable IGS region (petrol), only a subset of repeats was cleaved. b The break site was analyzed by deep sequencing. Only a very small fraction of mutations was identified as InDels (yellow) or insertions (blue), whereas over 95% were deletions (green). Further analysis of the reads, including the deletions, showed that 94% of these were small deletions with a size of 10 bp or smaller. The deletion pattern of these small deletions revealed that no microhomologies were used for the repair (red). c Detailed illustration of exemplary junctions with deletions in Col-0 plants on the sequence level. d Cleavage of the IGS region in DNA repair mutants revealed that only the cNHEJ mutants, ku70-1 and lig4-5, but not the HR mutant, rad54-1, responded highly sensitive, resulting in a drastically reduced survival rate (n = 3 biologically independent experiments). Data are presented as normalized mean values ± SD. Statistical differences were calculated using the two-tailed t-test with unequal variances: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source data file.
DSB induction in the conserved region of the 45S rDNA
Next, we used the two intergenic transcribed spacer (ITS) regions as target sites, which are conserved between all repeats. Two gRNAs were designed for ITS1 (Ubi-SaCas9-ITS1-1 and Ubi-SaCas9-ITS1-2) and one guide for ITS2 (Ubi-SaCas9-ITS2-1) (Fig. 1a; Supplementary Table 1). Again, the gRNA targeting the non-essential ADH1 gene (Ubi-SaCas9-ADH1) was used as a control, which showed an editing efficiency of 60% (Supplementary Fig. 1a). Transformation with the T-DNA targeting ADH1 proved to be efficient: 3.2% of the progeny were gentamycin-resistant. In contrast, a dramatic decrease in transformants was observed in the ITS target lines, with a reduction of 96% in case of Ubi-SaCas9-ITS1-1 as well as Ubi-SaCas9-ITS1-2 and a reduction of 99% in case of Ubi-SaCas9-ITS2-1 (Fig. 2a). The fact that quantitative real-time PCR showed no alteration in 45S rDNA copy number in the surviving seedlings suggests that in these few cases, the T-DNA was either truncated or silenced, resulting in insufficient DSB induction (Supplementary Fig. 1b and c). To prove that the nuclease activity of the SaCas9, and not just simple binding of the protein to the rDNA, is required, an additional set of control lines was transformed with catalytically inactive SaCas9 (dCas9) targeting ADH1 (Ubi-dCas9-ADH1) and ITS2 (Ubi-dCas9-ITS2-1). The survival rate determined in both dCas9 approaches was similar and comparable to that of the Ubi-SaCas9-ADH1 control line (Fig. 2b). Thus, the sgRNA ITS2-1 was only able to induce a lethal effect in combination with an active nuclease, demonstrating that massive DSB induction is responsible for the lethal effect. This result was surprising, considering that a recent study reported a dramatic reduction of copy number, but not lethality, when another Cas9 orthologue, namely Streptococcus pyogenes Cas9 (SpCas9), was used for DSB induction16. This difference might have been caused by the fact that SaCas9 is superior in DSB induction in comparison to SpCas9 in Arabidopsis14,17. Our results clearly demonstrate that targeting the conserved part of the 45S rDNA with a highly efficient nuclease is lethal for the transformed cells (Fig. 2c). Based on this observation, we speculated that massive DSB induction could be used as a tool to induce cell death during plant development. Thus, if Cas9 expression was confined to specific growth phases or specific organs, inducing cell death in a highly specific and cell-autonomous manner, a novel form of tissue engineering could be developed (Fig. 2c). By using this system, named CRISPR-Kill, we expected to be able to obtain plants devoid of individual organs or cell types.
a Cleavage of the conserved ITS region of the 45S rDNA using SaCas9 in the wild type resulted in a dramatic decrease of the number of transformants in all three ITS-targeting lines (Ubi-SaCas9-ITS1-1, Ubi-SaCas9-ITS1-2, and Ubi-SaCas9-ITS2-1) compared to the control line, Ubi-SaCas9-ADH1 (n = 3 biologically independent experiments). Data are presented as normalized mean values ± SD. P = 0.00043 for ITS1-1; P = 0.00085 for ITS1-2; P < 0.0001 for ITS2-1. b The survival rate of lines transformed with a catalytically inactive SaCas9 (dCas9) targeting ADH1 (Ubi-dCas9-ADH1) and ITS2 (Ubi-dCas9-ITS2-1) was analyzed. No reduction in survival rate could be observed in the ITS2-1 line compared to the ADH1 line (n = 3 biologically independent experiments). Data are presented as normalized mean values ± SD. c SaCas9 (blue) is under the expression of an exchangeable promoter (light pink). Utilization of SaCas9 under expression of a constitutive results in cleavage of the 45S rDNA in all cells (red), leading to cell death. By exchanging the constitutive promoter with a tissue-specific promoter, it is possible to obtain plants devoid of individual organs. Statistical differences were calculated using the two-tailed t-test with unequal variances: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Source data are provided as a Source data file.
CRISPR-Kill for elimination of floral organs
As a proof of concept, we decided to test Cas9 expression under control of the promoter of the floral homeotic gene APETALA1 (AP1, At1g69120). AP1 is a key player for flower meristem identity. It is highly expressed at the early stages of flower development and later gets restricted to specific tissue where it is required for sepal and petal development18 (Fig. 3a). The ap1 mutant shows a disturbed flower architecture where, instead of petals, carpel and stamen are formed (Fig. 3b). We were eager to test if the use of CRISPR-Kill would enable us to specifically eliminate petal tissue. For this approach, the most efficient guide, ITS2-1, was chosen, while the ADH1 guide served as a control. We expected that the utilization of the AP1 promoter would restrict SaCas9 expression exclusively to the early floral meristem and, later on, to sepals and petals. After determination of the transformation rate, no significant reduction could be observed in pAP1-SaCas9-ITS2-1 in comparison to pAP1-SaCas9-ADH1 T1 plants (Fig. 3c). After 4 weeks, 133 independent pAP1-SaCas9-ITS2-1 T1 plants were analyzed for phenotypical deviations in comparison to the control lines. While the control plants showed normal growth, a spectrum of different phenotypes was observed for the CRISPR-Kill plants (Table 1 and Fig. 3d). A very small fraction of CRISPR-Kill T1 lines had died, while one-fifth showed a strong phenotype, characterized by a lack of flowers or lack of sepals and petals. Another fifth showed a weak phenotype, characterized by a full set of sepals, but an incomplete set of petals. In almost half of all lines, an intermediate phenotype could be observed, described by the presence of, at least some, sepals in combination with varying numbers of petals. This intermediate phenotype also consists of the ap1 mutant phenotype (19.8%), characterized by a disturbed floral architecture. To determine whether there is a correlation between the strength of the phenotype and the amount of induced DSBs in the respective line, we isolated floral tissue and performed deep sequencing analysis of the ITS2 target site. In total, two ADH1 control lines (#81 and #82), as well as ten ITS-2-1 lines were analyzed for InDels. The latter ones were subdivided based on their phenotype in ‘no phenotype’ (#5 and #3), ‘intermediate phenotype’ (#1, #4, and #2), and ‘strong phenotype’ (#26 to #30) (Fig. 4). Due to the fact that only a minor portion of respective cell types survived elimination and DNA from adjacent tissues was co-purified, we expected to detect InDels as indications of DSB repair in only a small fraction of reads. The reads of the two control lines showed no InDels at the break site. Occasionally, 1 bp deletions (0.15% for #81 and 0.16% for #82) could be observed, reflecting the natural variances within the repeats. In contrast, all CRISPR-Kill lines that showed a distinguishable phenotype revealed InDels at the break site, with mutation frequencies proportional to the severeness of the phenotype. In the CRISPR-Kill lines with the ‘intermediate phenotype’ mutation, frequencies between 0.31% and 0.99% could be observed. In turn, the CRISPR-Kill lines with the ‘strong phenotype’ showed mutation frequencies between 1.60% and 2.14% (Supplementary Figs. 2 and 3). These findings demonstrate that efficient organ elimination can be achieved by using the CRISPR-Kill system and that the strength of the phenotype is correlated with the efficiency of DSB induction.
a The floral homeotic protein, APETALA1 (AP1), is highly expressed in the early floral meristem at flower development stages 1 and 2 (red). At development stage 3, the expression is restricted to specific tissue, which is determined to form sepals and petals. b Schematic representation of the ap1 mutant phenotype: a disturbed flower architecture with an extra set of carpel and stamen instead of sepals and petals. c A comparison of survival rates between the T1 control line (pAP1-SaCas9-ADH1) and the T1 flower-specific CRISPR-Kill line (pAP1-SaCas9-ITS2-1) revealed no significant differences (n = 3 biologically independent experiments). Data are presented as normalized mean values ± SD. Statistical differences were calculated using the two-tailed t-test with unequal variances. d Representative floral phenotypes of pAP1-SaCas9-ADH1 control line and pAP1-SaCas9-ITS2-1 CRISPR-Kill line. Bar = 1 mm. Statistical differences were calculated using the two-tailed t-test with unequal variances. Source data are provided as a Source data file.
The target site of ITS2-1 was analyzed via deep sequencing in floral tissue of two ADH1 control lines (#81 and #82) and ten ITS2-1 lines. The CRISPR-Kill lines were subdivided according to phenotype in ‘no phenotype’ (#5 and #3), ‘intermediate phenotype’ (#1, #4, and #2), and ‘strong phenotype’ (#26 to #30). The reads of the two ADH1 lines showed no repair events (gray), only also naturally occurring 1 bp-deletions (blue). In contrast, analysis of the CRISPR-Kill lines with a distinguishable phenotype revealed repair events like deletions (blue), insertions (red), and InDels (yellow), at the break site, with mutation frequencies linked to the severity of the phenotype. Source data have been deposited in SRA [accession: PRJNA726366].
Extension of CRISPR-Kill to target centromeric repeats
As the cleavage of 45S rDNA tandem repeats for tissue elimination was highly efficient, we speculated whether the cleavage of other repetitive elements would lead to a comparable result. The 178 bp centromeric sequences represent another highly repetitive cluster of functional tandem repeats19,20. Therefore, the gRNA of the CRISPR-Kill system was programmed to target the conserved 178 bp repeat for DSB induction (Ubi-SaCas9-Cen) (Fig. 5a). Transformation of Col-0 with this construct resulted in a reduction of resistant seedlings by 80% in comparison to the ADH1 control (Fig. 5b). In contrast, AP1-specific expression (pAP1-SaCas9-Cen) did not reduce the number of transformed seedlings (Fig. 5c). The phenotypical analysis revealed the ap1 mutant phenotype as the most abundant, confirming the successful organ elimination (Fig. 5c). Interestingly, in these 40 independent lines the phenotype was much more uniform than in pAP1-SaCas9-ITS2-1.
a The core centromere is characterized by a highly repetitive cluster of functional tandem repeats. The yellow triangle represents the cleavage site for SaCas9 within the centromere repeats (SaCas9-Cen). b Transformation of Col-0 plants with Ubi-SaCas9-Cen resulted in a reduction of the survival rate of T1 plants by 80% (n = 3 biologically independent experiments). Data are presented as normalized mean values ± SD. P = 0.00064. c Floral-specific expression of SaCas9-Cen using the AP1-promoter did not result in a reduction of the survival rate (n = 3 biologically independent experiments). Data are presented as normalized mean values ± SD. The phenotypical analysis revealed an ap1-mutant phenotype with disturbed floral architecture in most cases. Bar = 1 mm. Statistical differences were calculated using the two-tailed t-test with unequal variances: ***P < 0.001. Source data are provided as a Source data file.
CRISPR-Kill for elimination of lateral roots
To demonstrate the versatility of the system, we used CRISPR-Kill to modify root development. In A. thaliana, the pericycle cells form lateral roots (Fig. 6a). Thus, elimination of this cell type during root development was expected to hinder lateral root formation. Using the XPP (xylem pole pericycle) promoter, which is exclusively expressed in the pericycle21,22, we transformed Col-0 plants with XPP-SaCas9-ITS2-1, XPP-SaCas9-Cen constructs, and pXPP-SaCas9-ADH1 as a control (Supplementary Fig. 4a). Due to their early expression, a slight reduction of transformation efficiency could be observed in case of the two pXPP-CRISPR-Kill constructs compared to the control (Supplementary Fig. 4b). In all three approaches, 20-day old transgenic T1 plants were checked for lateral root growth. In the 45S rDNA CRISPR-Kill line, a three-fold decrease was observed in lateral root numbers, whereas the centromere CRISPR-Kill line showed a two-fold decrease (Fig. 6b and d). In addition, the relative lateral root length was determined (Supplementary Fig. 4c). Again, a three-fold decrease compared to the control line was observed in the rDNA-targeting line and a two-fold decrease in the centromere-targeting line (Fig. 6c and d). A detailed analysis of individual seedlings showed that, by using CRISPR-Kill, a larger portion of plants hardly developed any lateral roots (Supplementary Fig. 4d and e). This indicates that the severeness of the induced phenotype might vary depending on the T-DNA integration site, as seen in the case of pAP1-SaCas9-ITS2-1 in flower development. As we were interested in whether the phenotypical effects would be reproducible in the following generation, we analyzed root development in presence of XPP-SaCas9-ITS2-1 in the T2 generation. Progeny of T1 plants were selected for presence of the CRISPR-Kill construct and respective plants were analyzed as described above. For both lateral root number and relative lateral root length, again, a 3-fold decrease could be observed compared to the control (Fig. 7 and Supplementary Fig. 5).
a Schematic overview of different root cell types in the A. thaliana root tip. The root tip consists of the lateral root cap and columella (pink). The central cells of the root (stele, gray) are framed by pericycle cells (red), followed by endodermis (yellow), cortex (purple), and epidermis (blue), which arise from the stem cell niche (white). Lateral roots arise from the pericycle cells (red). b Representative pictures of lateral roots in both CRISPR-Kill lines (pXPP-SaCas9-ITS2-1 and pXPP-SaCas9-Cen) in comparison to the control line (pXPP-SaCas9-ADH1). Bar = 1 cm. c Total number of lateral roots in control line and CRISPR-Kill lines. The control line, pXPP-SaCas9-ADH1, showed an average of 13 lateral roots per plant, whereas in the CRISPR-Kill lines an average of 5 (ITS2-1-CRISPR-Kill line) and 7 (Cen-CRISPR-Kill line) lateral roots per plant was observed, respectively. P < 0.0001 for ITS2-1; P < 0.0001 for Cen. d Lateral root length in relation to main root length. In the control line, the sum of the length of all lateral roots was on average 4.4 times longer than the main root. In contrast, the sum of the length of all lateral roots in the ITS2-1- and Cen-CRISPR-Kill lines was on average only 1.4 and 2.2 times longer than the main root, respectively. P < 0.0001 for ITS2-1; P < 0.0001 for Cen. Data are presented as box plots (n = 108), boxes show the first to third quartile with median, whiskers encompass 1.5x the interquartile range reaching to minimum and maximum, and data beyond that threshold is indicated as outliers. P values were calculated using the one-way ANOVA-test: ****P < 0.0001. Source data are provided as a Source data file.
a Total number of lateral roots in control line and rDNA CRISPR-Kill lines. The control line, pXPP-SaCas9-ADH1, showed an average of 11 lateral roots per plant, whereas the ITS2-1-CRISPR-Kill line showed a 3-fold decrease and only an average of 3 lateral roots per plant. P < 0.0001. b Lateral root length in relation to main root length. A 3-fold decrease could also be observed in the relative lateral root length of the ITS2-1-CRISPR-Kill line. While the sum of the length of all lateral roots in the control line was on average 3.6 times longer than the main root, the sum of the length of all lateral roots in the ITS2-1 CRISPR-Kill line was on average only 1.1 times longer than the main root. P < 0.0001. Data is presented as box plots (n = 30), boxes show the first to third quartile with median, whiskers encompass 1.5x the interquartile range reaching to minimum and maximum, and data beyond that threshold is indicated as outliers. P values were calculated using the one-way ANOVA-test: ****P < 0.0001. Source data are provided as a Source data file.
