Strategy to Engineer a tRNA-processing System for Producing Numerous gRNAs.

To simultaneously produce multiple gRNAs from one primary transcript, we aimed to compact a cluster of gRNAs with different spacers in one polycistronic gene and hijack an endogenous RNA-processing system for cleavage of this transcript into individual gRNAs in a nucleus. In search of RNA and endo-RNases that potentially meet this requirement, tRNA attracted our attentions for its characteristics as follows. (i) In nucleoplasm, the tRNA precursors (pre-tRNAs) are cleaved at specific sites in eukaryotes by RNase P and RNase Z (or RNase E in bacterium) to remove 5′ and 3′ extra sequences (Fig. 1A), respectively (14–18). (ii) RNase P and RNase Z recognize the tRNA structure, regardless of the pre-tRNA sequence (17, 18). Previous studies revealed that only the acceptor stem, D-loop arm, and TψC-loop arm of tRNA are necessary for RNase P and RNase Z cleavage of pre-tRNA (16–19). (iii) tRNA was found in the polycistronic transcription unit in bacteria (20) and occasionally in eukaryotes (19, 21), suggesting that the tRNA-processing system is likely used as an intrinsic mechanism to produce different small RNAs [e.g., small nucleolar RNA (snoRNA)] with tRNA from a single polycistronic gene (19). (iv) Because tRNA is one of the most abundant cellular components, the endogenous tRNA-processing system should be robust to process a large number of substrates. (v) The tRNA genes contain internal promoter elements (box A and B) to recruit the RNA polymerase III (Pol III) complex (Fig. 1B) (22, 23). Therefore, the tRNA sequence may also work as a potential transcriptional enhancer for Pol III. Based on these features, we hypothesized that the endogenous tRNA system could be engineered as a general platform for precise processing of gRNAs.

To use the endogenous tRNA system for multiplex genome editing with Cas9/gRNA, we designed a polycistronic tRNA-gRNA (PTG) gene for simultaneous production of numerous gRNAs. As shown in Fig. 1C, this PTG gene consists of tandem repeats of tRNA-gRNA and would be transcribed as a normal snoRNA or sgRNA gene under the control of the Pol III promoter. The endogenous tRNA-processing RNases (e.g., RNase P and RNase Z in plants) would recognize the tRNA components and excise individual gRNAs from the PTG transcript. The resulting gRNAs would then direct Cas9 to multiple target sites for genome editing (Fig. 1C).

Precise Processing of PTG to Produce Functional gRNAs with Desired Targeting Sequences.

To explore whether the synthetic PTG gene would be transcribed, would be processed, and would function as we predicted (Fig. 1C), we synthesized four PTG genes with a structure of tRNA-gRNA (PTG1 and PTG2) or tRNA-gRNA-tRNA (PTG1.1 and PTG2.1) (Fig. 2A). These four PTG genes were designed to produce gRNA1 (PTG1 and PTG1.1) and gRNA2 (PTG2 and PTG2.1), which have been tested previously (24) with sgRNA genes (sgRNA1 and sgRNA2) in Cas9-mediated genome editing (SI Appendix, Fig. S1A; and see gene sequence in SI Appendix, Table S1). Both gRNAs targeted the rice MPK5 gene, which encodes a mitogen-activated protein kinase (MAPK) involved in biotic and abiotic signaling pathways. These PTGs were constructed with a 77-bp long pre-tRNA^Gly gene, which recognizes the GGC codon and is widely present among various genomes (25). The chosen pre-tRNA^Gly sequence consists of a 5′ leader (5′-AACAAA-3′, 6 bp) and mature tRNA (71 bp) (Fig. 2B). Such tRNA–gRNA fusion in PTGs mimics the native tRNA–snoRNA43 dicistron in plants (19).

In this study, we used a plasmid vector (SI Appendix, Fig. S2) in which sgRNA or PTG is expressed with the rice U3 snoRNA promoter (U3p) and Cas9 is expressed with a rice ubiquitin promoter plus the complete 5′ untranslated region (UBIp). After transfecting rice protoplasts with plasmids containing U3p:sgRNA or U3p:PTG, circularized reverse transcription PCR (cRT-PCR) (26, 27) was performed to map both 5′ and 3′ ends of mature gRNAs (see SI Appendix, Fig. S3 and Table S2 for the principle of cRT-PCR). Mature gRNAs with the expected size were detected from the protoplasts expressing PTGs or sgRNAs (Fig. 2 C–F). However, the tRNA–gRNA fused transcript from PTGs was not detectable by cRT-PCR, probably due to the highly robust tRNA-processing system that cleaved most (if not all) primary transcripts of PTGs. Sequence analysis of the cRT-PCR products revealed that all four PTG transcripts were precisely cleaved at the tRNA–gRNA junction and produced mature gRNAs carrying the desired 5′ spacer sequences without extra nucleotides (Fig. 2G and SI Appendix, Figs. S4 and S5). On the other hand, the 3′ end of mature gRNAs carried a 1- to 7-nt-long poly (U) tail if it preceded the Pol III terminator (sgRNA, PTG1, and PTG2) (SI Appendix, Fig. S4), or 1- to 4-nt ribonucleotides from the tRNA-leader when it preceded the tRNA (PTG1.1 and PTG2.1) (SI Appendix, Fig. S5). The cRT-PCR data also confirmed that gRNAs transcribed from U3p:sgRNA were obligated to start with nucleotide A whereas U3p:PTGs produced gRNAs without constraint of the first nucleotide (SI Appendix, Figs. S4 and S5). In summary, the gRNAs produced from PTG are the same as the sgRNAs but are not obligated to start with a specific nucleotide (Fig. 2 G and H).

The functionality of PTG1 and PTG2 was confirmed by examining the insertion/deletion (indel) mutations introduced by nonhomologous end joining (NHEJ) repairing at the predicted Cas9:gRNA cleavage site. Because gRNA1 and gRNA2 targets contain the KpnI and SacI restriction enzyme (RE) sites (SI Appendix, Fig. S1A), respectively, the mutations induced by PTG1/Cas9 and PTG2/Cas9 could be readily analyzed by the digestion of PCR products encompassing the targeted sites with corresponding RE (PCR/RE assay). In rice protoplasts transfected with PTG1/Cas9 and PTG2/Cas9, 15% and 9% of the target sites were found to carry indels, respectively (Fig. 2I), which are slightly higher than the mutation rate of the sgRNA:Cas9 constructs we used previously (24). Consistent with our hypothesis that tRNA may function as a transcriptional enhancer for Pol III, the quantitative RT-PCR with gRNA-specific primers revealed that the transcript levels of PTG1 and PTG2 were about 3 and 31 times higher than those of sgRNA1 and sgRNA2 in protoplasts (Fig. 2J), respectively. Taken together, our results demonstrated that the endogenous tRNA system could be used as a precise and robust tool to produce gRNAs from PTG for Cas9-mediated genome editing.

Efficient Multiplex Genome Editing in Rice Protoplasts via PTG/Cas9.

To test the targeting capacity and efficiency of the PTG system, up to eight gRNAs (gRNA1–gRNA8) (SI Appendix, Table S3 and Fig. S1) were tandemly arrayed to construct PTGs for targeting four homologous rice MAPKs (MPK1/2/5/6), which may function redundantly in diverse cellular signaling pathways. These gRNAs were divided into four pairs and each pair targeted two genomic sites within a gene locus with a distance of 350–750 bp between them (SI Appendix, Fig. S1). By combining different gRNA pairs, we designed PTG genes encoding two (PTG3/4/5/6), four (PTG7/8), and eight (PTG9) gRNAs to simultaneously target one, two, and four MPK loci, correspondingly (Fig. 3A and SI Appendix, Table S1). We expected that such a design might result in a deletion of a short chromosomal fragment between two Cas9 cut sites and allow us to readily examine the efficacy of PTGs.

To synthesize PTG with repetitive tRNA—gRNA architecture, we designed a scalable and flexible approach to assemble PTGs from PCR components based on the principle of Golden Gate assembly strategy (28) (see SI Appendix, SI Methods and Figs. S6 and S7 for details). Our gene-assembly approach allowed fast synthesis of PTGs with different combination of gRNAs from common oligonucleotide primers. Using a one- or two-step assembly, we synthesized PTG3-PTG9 genes and cloned them into the CRISPR/Cas9-expressing vectors (SI Appendix, Fig. S2 and Tables S1–S3).

We then transfected rice protoplasts with these plasmids containing both U3p:PTG3-PTG9 and UBIp:Cas9. Chromosomal fragment deletions at four MAPK loci, which were revealed by amplification of truncated PCR products with target specific primers, were detected in protoplasts expressing respective PTGs with 4–45% efficiency (Fig. 3B). Even though eight gRNAs were simultaneously produced from PTG9, they still guided Cas9 to efficiently excise chromosomal fragments at all four targeted loci with 4–20% frequency (Fig. 3B). The fragment deletion efficiency at MPK5 loci was further confirmed by quantitative PCR using primers encompassing the gRNA2 cut site (SI Appendix, Table S4). Sequencing of these truncated PCR products confirmed that fragments between two gRNA/Cas9 cut sites were excised from targeted loci with or without additional indels (Fig. 3C). In general, the fragment deletion efficiency was negatively correlated to the distance between two paired cut sites (correlation coefficient r = −0.95) despite the fact that different gRNAs may have variable efficiencies. We noticed that the total number of gRNAs in one PTG affected the deletion efficiency with variable extent at different targets in protoplasts (Fig. 3B). At three targeting loci (MPK2/5/6), the PTGs (PTG4/5/6) containing two gRNAs exhibited only slightly higher efficiency than the PTGs with four (PTG7/8) and eight (PTG9) gRNAs. However, at the MPK1 locus, PTG3 (two gRNAs) showed ∼2 times higher deletion frequency than PTG7 (four gRNA) or PTG9 (eight gRNAs). Such a reduction of efficiency of PTGs with a higher number of gRNAs was likely due to the competition for Cas9 among gRNAs. Nevertheless, PTG with tandemly arrayed tRNA–gRNA architecture is capable of simultaneously producing numerous gRNAs and guiding Cas9 to multiple chromosomal targets with a high efficiency.

Improving Multiplex Genome Editing in Stable Transgenic Plants with PTG/Cas9.

Because many plants, including important crops like rice, could not be readily regenerated from protoplasts, we used the conventional Agrobacterium-mediated transformation to produce stable transgenic lines and evaluated the efficacy of PTG/Cas9 system for multiplex genome editing in intact rice plants. Mutagenesis frequency at gRNA1 and gRNA2 targets in transgenic plants expressing sgRNA1:Cas9, sgRNA2:Cas9, PTG6:Cas9, or PTG7:Cas9 was examined at T₀ generation. Among T₀ generation of sgRNA1:Cas9 (n = 32) and sgRNA2:Cas9 plants (n = 20), 44% and 60% of them carried indels whereas 13% and 20% of them had biallelic mutations, respectively (Table 1 and SI Appendix, Fig. S8). By contrast, indels at both targets were detected in all PTG6:Cas9 plants (100%, n = 17) including 35% (gRNA1) and 76% (gRNA2) biallelic mutations (Table 1 and SI Appendix, Fig. S9). However, the chromosomal-fragment deletion between gRNA1 and gRNA2 target sites was not detected in PTG6:Cas9 plants with PCR, which may occur at a lower frequency in regenerated calli/plants than in protoplasts. Nevertheless, the results showed that PTG6 not only expressed two gRNAs simultaneously, but also achieved a higher mutagenesis efficiency at individual targets than sgRNA did (Student’s t test, P < 0.01) (Table 1 and SI Appendix, Figs. S8 and S9).

When the total number of targets increased to four in PTG7:Cas9 plants, a comparable indel frequency to sgRNA1:Cas9 plants was observed at the gRNA1 site, but a significantly higher frequency of indels than sgRNA2:Cas9 was found at the gRNA2 site (Table 1 and SI Appendix, Fig. S10). Interestingly, we obtained one PTG7:Cas9 line (6%) carrying biallelic deletion of ∼350 bp in MPK1 and monoallelic deletion of ∼750 bp in MPK5 (Fig. 4 A and B). This result prompted us to further examine the efficiency of all eight gRNAs in PTG9:Cas9 lines. At five gRNA targets (gRNA1/2/3/5/7) whose mutation would destroy RE sites, 50% (n = 14) of PTG9:Cas9 T₀ lines carried biallelic mutations at all five targets (SI Appendix, Fig. S11). Interestingly, PCR/RE assays suggested that these five gRNAs exhibited comparable efficiency and that all of them showed significantly higher mutagenesis activities than what we observed in sgRNA1/2:Cas9 lines (Table 1). Sanger sequencing of the fragments from four MPK loci of PTG9:Cas9 plants confirmed that mutations were introduced at all eight sites (SI Appendix, Fig. S12). However, the fragment deletion was detected only at MPK2 and MPK5 loci in two PTG9:Cas9 lines (SI Appendix, Fig. S11). In comparison with protoplasts, the efficiency of targeted chromosomal fragment deletion between paired gRNA/Cas9 cut sites is lower in transgenic plants, which might be due to a relatively lower expression of gRNAs and Cas9 in rice calli/regenerated plants (typically only a single copy of a transgene is integrated into the rice genome after the Agrobacterium-mediated transformation). Our data demonstrate that the PTG method not only increases the gRNA number and targeting sites, but also likely enhances mutagenesis efficiency at individual sites, especially when multiple gRNAs are expressed using PTG.

To further substantiate the high efficiency of mutagenesis with PTG/Cas9, we synthesized PTG10 with two gRNAs (gRNA9 and gRNA10) to target the rice phytoene desaturase (PDS) gene (SI Appendix, Fig. S13A). PDS is frequently used as a convenient gene target to examine knock-out efficiency because plants with null functional PDS would lead to a visible photo-bleached phenotype. Strikingly, all PTG10:Cas9 transgenic seedlings (T₀ generation, n = 15) regenerated from calli showed photo-bleaching phenotype, and 13 of them were completely albino, indicating that these plants carried null functional PDS (Fig. 4C). Although we identified only one line carrying chromosomal deletions between two gRNA-targeted sites, sequencing of the PDS targets from all of the lines confirmed that indels were introduced at the target sites (SI Appendix, Fig. S13). By comparing these data with the previously reported efficiency of the sgRNA/Cas9 system (9, 11, 29–32), to our knowledge, the PTG/Cas9 approach yielded the highest efficiency of targeted mutagenesis (up to 100%) (Fig. 4 and Table 1) in plants. Our results also demonstrate that targeting one gene with two gRNAs using PTG will greatly increase the efficiency of complete gene knock-out.

Plant Materials.

Rice (Oryza sativa L. ssp) cultivars Kitaake and Nipponbare were used in this study. Rice plants were grown in a green house or growth chamber at 28 °C day/23 °C night with 12 h of light.

Plasmid Vector Construction.

The plasmid vector pRGE32 was used to transiently express U3p:sgRNA or U3p:PTG along with UBIp:Cas9 in rice protoplasts, and pRGEB32 was used for the Agrobacterium-mediated rice transformation (SI Appendix, Fig. S2). See SI Appendix, SI Methods for details about plasmid vector construction.

sgRNA:Cas9 and PTG:Cas9 Expression Constructs.

The U3p:sgRNA1 and U3p:sgRNA2 constructs were made as described previously (24). The specific spacer sequences for gRNA3-gRNA8 (SI Appendix, Table S3) were selected using the CRISPR-PLANT database (www.genome.arizona.edu/crispr/) (38). The PTG genes and U3p:PTG constructs were generated as described in SI Appendix, SI Methods, Figs. S6 and S7, and Table S3. The synthesized PTGs were inserted into the BsaI-digested pRGE32 or pRGEB32 for transient protoplast expression or stable rice transformation, respectively. The sequences of sgRNA1, sgRNA2, and PTG genes used in this study are listed in SI Appendix, Table S1.

Rice Protoplast Transfection.

The rice protoplast preparation and transfection were performed as described previously (24). Briefly, 20 µg of plasmid DNA was used to transfect 2 × 10⁵ protoplasts with a transfection efficiency of ∼40–50%. Total rice genomic DNAs were extracted from the protoplast samples at 36 h after transfection and then used for PCR and sequence analysis.

Agrobacterium-Mediated Rice Transformation.

Transgenic rice plants were generated by the Agrobacterium tumefaciens-mediated transformation using rice mature seed-derived calli according to a conventional protocol (39).

Genomic DNA Extraction and PCR/RE Assay.

Rice genomic DNA was extracted as described previously with the hexadecyltrimethylammonium bromide method (24). To detect mutagenesis at desired sites, target regions were amplified with specific primers (see SI Appendix, Table S2 for primer sequences) using GoTaq DNA polymerase (Promega). The PCR product was separated in 1% agarose gel and stained with ethidium bromide to detect chromosomal-fragment deletion. To detect indels at specific sites with PCR/RE, the PCR products encompassing the target sites were digested with appropriate restriction enzymes (RE) for 5 h and then were analyzed with 1% agarose gel electrophoresis and ethidium bromide staining. The stained gels were imaged using the Gel Doc XRS system (Bio-Rad) and quantified with the Quantity One 4.6 program (Bio-Rad). Selected PCR products were cloned into pGEM-T Easy Vector (Promega) for DNA sequencing.

RNA Extraction, cRT-PCR, and Quantitative PCR.

See SI Appendix, SI Methods for details.

Gene Accession Number in GenBank.

Genes and their GenBank RefSeq accession numbers are as follow: MPK1, Os06g0154500; MPK2, Os08g0157000; MPK5, Os03g0285800; MPK6, Os10g0533600; UBI, Os02g0161900; and PDS, Os03g0184000.