Conferring resistance to geminiviruses with the CRISPR–Cas prokaryotic immune system

Abstract

To reduce crop losses due to geminivirus infection, we targeted the bean yellow dwarf virus (BeYDV) genome for destruction with the CRISPR–Cas (clustered, regularly interspaced short palindromic repeats–CRISPR-associated proteins) system. Transient assays using BeYDV-based replicons revealed that CRISPR–Cas reagents introduced mutations within the viral genome and reduced virus copy number. Transgenic plants expressing CRISPR–Cas reagents and challenged with BeYDV had reduced virus load and symptoms, thereby demonstrating a novel strategy for engineering resistance to geminiviruses.

Geminiviruses are destructive plant pathogens that reduce grain, vegetable and fruit harvests worldwide¹. Geminiviruses contain circular, single-stranded DNA genomes that replicate through double-stranded DNA intermediates. Replication proceeds within the nucleus of plant cells by rolling-circle and recombination-dependent mechanisms². During the replication cycle, genomes are shuttled to neighbouring cells through plasmodesmata. Double-stranded genomes serve as templates for transcription and replication and are wrapped in histones to form minichromosomes³.

Transgenic approaches to achieve geminivirus resistance, including RNA-mediated interference-based strategies, have had limited success^4,5. To adopt the CRISPR–Cas system⁶ for reducing geminivirus replication in plants, single guide RNAs (sgRNAs) were designed that are complementary to sequences within the bean yellow dwarf virus (BeYDV) genome⁷. BeYDV encodes four proteins: replication initiator protein (Rep), RepA, movement protein and coat protein. It also harbours two cis-acting elements: the long intergenic region (LIR) and the short intergenic region (SIR). The LIR contains divergent promoters flanking the origin of replication, which consists of a Rep binding site (RBS), and a hairpin structure with an invariant nonanucleotide sequence within the loop, the initiation site of plus (viral) strand synthesis. The SIR is required for minus (complementary) strand synthesis. Targets were directed to sequences conserved at the nucleotide or amino acid level on either DNA strand (Fig. 1a). Six target regions within the BeYDV genome were chosen: RBS, hairpin, nonanucleotide sequence and three Rep motifs essential for rolling circle replication (motifs I, II and III).

Figure 1 |. Testing Cas9-sgRNA activity at targets within the BeYDV.

a, Illustration of the wild type BeYDV genome (reverse complemented) and sgRNA target sequences. Red nucleotides, sgRNA sequence; black nucleotides, BeYDV sequence; blue nucleotides, protospacer adjacent motif; lightning bolts, predicted sites for Cas9 cleavage; IR, inverted repeat; 9 nt, nanonucleotide sequence; M1, motif 1; M2, motif 2; M3, motif 3; +/−, DNA strand that pairs with the sgRNA; NOS, nopaline synthase. b, Approach to assess Cas9 and sgRNA activity within N. benthamiana leaves by measuring, c, GFP expression and d, CFUs. n, number of biological replicates; error bars represent standard deviation; *P < 0.05.

To assess the activity of Cas9 and sgRNAs, we developed a transient assay in Nicotiana benthamiana using BeYDV-based replicons (Fig. 1b). Two strains of Agrobacterium were co-infiltrated into leaves from 5-week-old plants. The first strain contained transfer DNA (T-DNA) harbouring CRISPR–Cas reagents: Cas9 was expressed from a double 35S promoter, and sgRNAs were expressed from either an AtU6 or At7SL RNA polymerase III promoter⁸. The second Agrobacterium strain contained T-DNA harbouring BeYDV replicon sequences⁹. To assess the activity of Cas9 and sgRNAs against BeYDV, the coat protein and movement protein genes were replaced with the enhanced green fluorescent protein (eGFP) gene. Five days after co-infiltration, average eGFP intensity was quantified. Results were normalized to a control expressing Cas9 and an sgRNA targeting Tomato golden mosaic virus motif III (gTM3⁺). From 11 sgRNAs, we observed a reduction in eGFP intensity ranging from ∼5% to ∼87% compared with the control (Fig. 1c). Some sgRNAs (gBHP1⁻, gBHP3⁻ and gB9nt⁺) targeting sequences near the hairpin were less active, suggesting that a secondary structure may influence Cas9-sgRNA cleavage or access. We also tested the combination of two sgRNAs (gBRBS⁺ and gBM3⁺) and observed a further reduction in eGFP intensity relative to each individual sgRNA, suggesting an additive effect when multiple sgRNAs target the same viral genome. Lastly, we tested different CRISPR–Cas architectures (Supplementary Fig. 1). Delivery of sgRNA alone (gBM3⁺), or delivery of catalytically dead Cas9 with gBM3⁺, did not reduce eGFP expression.

To assess whether the CRISPR–Cas system could reduce virus copy number, the BeYDV replicon was modified to harbour a ColE1 origin and β-lactamase gene. Five days after co-infiltration, eGFP intensity was quantified and total DNA was extracted from leaves. DNA was transformed into Escherichia coli and colony forming units (CFUs) were counted. To remove T-DNA plasmid from Agrobacterium, total DNA was digested with DpnI before transformation. With the sgRNAs tested, a correlation was observed between copy number and eGFP expression (Fig. 1d).

We predicted that double-strand breaks within geminivirus genomes created by Cas9 will result in repair by non-homologous end joining (NHEJ), leading to small insertions and deletions (INDELs) at the break site. To assess the frequency of NHEJ-induced mutations within the BeYDV genome, we used next-generation sequencing to survey the sgRNA targets in the DNA samples used for the CFU experiment. Four sgRNAs were tested, gBRBS⁺, gBM3⁺, gBM1⁻ and gB9nt⁺ and we observed INDELs at the predicted cleavage sites at frequencies of 70.01%, 22.37%, 0.03% and 7.85%, respectively (Fig. 2a). The majority of mutations were short (1–2 bp) INDELs (Supplementary Fig. 2). Although gBM1⁻ reduced replicon-based protein expression and copy number, no mutations were observed after deep sequencing. We speculate that inhibition by gBM1⁻ may occur by mechanisms other than cutting, possibly by blocking Rep transcription¹⁰.

Figure 2 |. Restricting BeYDV infection with the CRISPR–Cas system.

a, Illumina next-generation sequencing of four sgRNA targets. Dark bars, sgRNA gTM3⁺ (negative control). b, PCR detection of deletions within the genome of mobile BeYDV. The ∼657 bp band indicates unmodified sequences or sequences with small INDELs. The ∼349 bp band indicates sequences containing deletions; sequences of three clones for each biological replicate are shown. c, Representative images of transgenic plants challenged with BeYDV 35 days after inoculation. d, Quantitative PCR of virus levels within transgenic plants 37 days after inoculation.

Collectively, the data suggest that Cas9 can bind to and introduce double-strand breaks at sites within a geminivirus genome. However, the data do not distinguish between mutations introduced within the circular BeYDV genome or the input T-DNA. To confirm that Cas9 can cleave the genome of a geminivirus as it spreads throughout the plant, we developed a transient assay wherein virions must move through cells expressing CRISPR–Cas reagents. N. benthamiana leaf tips were infiltrated with Agrobacterium containing T-DNA harbouring the complete BeYDV genome flanked by LIR sequences. Two days later, a second, non-overlapping site on the same leaf was infiltrated with Agrobacterium containing T-DNA encoding CRISPR–Cas reagents. To facilitate detection of NHEJ-induced mutations, two sgRNAs (gBM3⁺ and gBRBS⁺) were co-expressed to delete intervening sequences. Five days after virus infiltration, total DNA was isolated from leaf tissue infiltrated with Agrobacterium harbouring the CRISPR–Cas reagents. DNA was digested with BamHI (a BamHI site resides between the sgRNA targets) to enrich for deletions, and used as a template for PCR with primers designed to amplify across the gBM3⁺ and gBRBS⁺ targets. From three biological replicates, we detected smaller PCR products consistent with the predicted deletion size (Fig. 2b). Deletions were confirmed by cloning and sequencing.

Two of the most active sgRNAs–gBM3⁺ and gBRBS⁺–were tested in whole plants for their ability to thwart BeYDV infection. Transgenic N. benthamiana constitutively expressing Cas9 and one of the sgRNAs were challenged with BeYDV by agroinfection and observed for symptom development. Transgenic plants showed reduced symptoms relative to control plants expressing Cas9 only (Fig. 2c and Supplementary Fig. 3). At 37 days after infection, total DNA was isolated from a systemic leaf on each of three plants and tested for virus copy number by quantitative, real-time PCR. Consistent with the reduced symptoms, plants expressing either the gBRBS⁺ or the gBM3⁺ contained significantly fewer copies of the BeYDV genome compared with plants expressing Cas9 alone (Fig. 2d).

In conclusion, we effectively interfered with geminivirus replication and systemic movement by transferring elements of the CRISPR–Cas prokaryotic immune system to plants. The flexibility of the CRISPR–Cas system makes it possible to target any sequenced geminivirus genome. The modularity and small size of the guide components enables sgRNA stacking in a single transgene, thereby directing several nucleases against a single virus to increase the potential for durability. Finally, with this approach it is possible to direct nucleases against multiple viruses and satellites, a key advantage in defeating the mixed infections of geminivirus disease complexes.

Methods

Vectors and Agrobacterium.

T-DNA plasmids (Cas9 (ref. 8) and BeYDV) were constructed in pCAMBIA vectors and transformed into Agrobacterium tumefaciens GV3101. Strains were grown overnight at 28 °C in Lysogeny Broth (LB) with selection. Cultures were transferred to LB with selection and 20 μM acetosyringone, and grown overnight at 28 °C. Strains were pelleted and resuspended in MES/MgCl₂ buffer with 150 μM acetosyringone at a final concentration of A_{600 nm} = 0.6 and 0.01 for Cas9/sgRNA and BeYDV T-DNAs, respectively.

Green fluorescent protein intensity assays.

Five days after agroinfiltration, N. benthamiana leaves were photographed and the average enhanced green fluorescent protein (eGFP) intensity was measured using ImageJ. Intensity values were normalized to Cas9/gTM3⁺ infiltrated on the same leaf.

CFU assays.

The β-lactamase gene and ColE1 origin were cloned downstream of eGFP. Five days after agroinfiltration, leaves were photographed, GFP intensity was measured and total DNA was harvested. DNA was digested with DpnI and transformed into NEB5^α high-efficiency chemically competent cells (New England Biolabs). Cells were plated on medium with carbenicillin. CFUs were normalized to the gTM3⁺ control on the same leaf.

Deep sequencing.

Sequencing libraries were prepared with the 16S Metagenomic Sequencing Library Preparation guide (Illumina) using the Nextera XT index kit. DNA isolated for the CFU assay was used for PCR amplification with Q5 polymerase (New England Biolabs). Thermocycling conditions were 30 s at 98 °C for denaturation, followed by 24 cycles of (7 s at 98 °C, 12 s at 69 °C, 15 s at 72 °C) and a final extension for 2 min at 72 °C. Amplifications were performed twice to generate technical replicates. Index PCR was performed with Q5 polymerase and the Nextera XT index kit before Illumina Next Generation Sequencing (MiSeq Nano V2, 250 bp paired end reads).

Next-generation sequencing (NGS) data analysis.

Sequence quality was verified in FastQC, and read trimming was performed with Trimmomatic-0.32 (ref. 11). Forward and reverse trimmed reads were merged by SeqPrep and mapped to the reference sequence using Geneious R7 mapper¹². Mapped reads were trimmed to include the nuclease target site and exported in bam format. Bed files with CIGAR (concise idiosyncratic gapped alignment report) string were generated from bam files using Bedtools v2.17.0 (ref. 13). A custom bash script was used to select all unique indel variants and determine their counts. All reads with deletions larger than 20 bp were removed from the RBS and 9nt control and treatment samples, and unique reads with deletions were manually verified to make sure they span the nuclease target site. Read variants that have a higher coverage than 20 were used for alignments and graph preparation.

Quantitative PCR.

Quantitative, real-time PCR was performed on total DNA from systemic leaves harvested at 37 days after infection. Primers were used to detect abundance of the Rep gene and normalized to the F-box and PP2a control genes as previously described^14,15.

Agroinfection.

Transgenic T1 plants were inoculated with Agrobacterium (A_{600 nm} = 0.95) carrying T-DNA with the complete BeYDV genomic sequence flanked by LIRs. Symptoms and quantitative PCR analysis were assessed as described in Fig. 2 and Supplementary Fig. 3.