Deconstructed Viruses for the Delivery of Sequence-Specific Nucleases and Repair Templates

We chose to deconstruct the mild strain of the BeYDV (Halley-Stott et al., 2007) because it had previously been used to express heterologous proteins in plant cells (Regnard et al., 2010). BeYDV replication requires three viral elements: the cis-acting LIR, the short intergenic region (SIR), and the trans-acting replication-initiation protein (Rep) and RepA. Rep and RepA are expressed from the same precursor mRNA: Rep is produced from the spliced mRNA, whereas the unspliced mRNA yields RepA (Wright et al., 1997). Eliminating coding sequences for the coat and movement proteins abolishes cell-to-cell movement but effectively eliminates genome-size constraints and makes it possible to increase vector cargo capacity. To compensate for loss of cell-to-cell movement, we used Agrobacterium to deliver geminivirus constructs to cells. The Agrobacterium T-DNA was modified to harbor cis-acting sequences needed for replication in a LIR-SIR-LIR orientation (hereafter referred to as the LSL T-DNA; Figure 1A). Concomitant delivery of Rep protein should result in replicational release and amplification of circular geminivirus replicons (GVRs; Stenger et al., 1991).

Delivery of conventional T-DNA to plant cells by Agrobacterium can result in transient gene expression. By modifying T-DNA to harbor sequences for genome engineering reagents (sequence-specific nucleases and repair templates), both targeted mutagenesis and gene targeting can be performed (Reiss et al., 2000; Li et al., 2013). To better distinguish between the effects of GVRs and the input T-DNA, we sought to design the LSL T-DNA such that expression of heterologous sequence is minimized when virus proteins are not present. We chose to position heterologous sequence downstream of the LIR in the virion-sense direction. The LIR functions as a bidirectional promoter and promoter strength in the virion-sense direction is weak in the absence of Rep and RepA proteins (Hefferon et al., 2006). To promote high gene expression within circularized GVRs, we positioned a duplicated 35S promoter (P35S) upstream of the LIR in the virion-sense direction. Splice donor and acceptor sequences were positioned upstream and downstream, respectively, of the LIR sequence that is predicted to form after circularization. High-protein expression may also occur from the transactivation of the virion-sense LIR promoter by the viral Rep proteins (Hefferon et al., 2006).

To investigate whether our GVRs replicate and express protein in plant cells, we modified LSL T-DNA plasmid to encode green fluorescent protein (GFP; pLSLGFP) or β-glucuronidase (GUS; pLSLGUS; Figure 1B). GVRs were delivered to tobacco leaf tissue by infiltrating a mixture of Agrobacterium strains containing pLSLGFP or pLSLGUS and a separate T-DNA expression plasmid encoding Rep and RepA (pREP). Both GUS and GFP expression was enhanced when pREP T-DNA was delivered (Figure 1B; Supplemental Figure 5). Background GUS activity seen in the tissue transformed with pLSLGUS is most likely due to activity of the virion-sense LIR promoter within Agrobacterium and plant cells.

To correlate the enhanced protein expression with GVR production, replicational release was evaluated by PCR using primers that would only amplify a product if circularization occurred. Strong amplification of the LIR sequence was observed only from samples in which pREP T-DNA was delivered, confirming successful replicational release of replicons (Figure 1C; Supplemental Figure 6). In a separate experiment, we quantified the relative GVR gene copy number over a 2-week time course using quantitative PCR (Supplemental Figure 7). Five days after infiltration, we observed a maximum copy number of ∼6000 per single-copy gene. This is consistent with previous reports describing that mastreviruses replicate to very high copy number within plant cells (Timmermans et al., 1992). Together, these data illustrate that our GVRs replicate and express heterologous proteins in leaf cells. Furthermore, these results show that peak replicon and protein levels occur between 3 and 5 d after infiltration, which is consistent with previous reports using BeYDV-based replicons for protein expression (Zhang and Mason, 2006). In subsequent genome engineering experiments, we chose to sample tissues between 5 and 7 d after infiltration. We reasoned that sampling on these days would allow sufficient time for the GVRs to reach maximum copy number and for peak protein expression. Furthermore, as DNA modifications are stable events, a targeted change occurring at 3 d after infiltration, for example, should be detectable 7 d after infiltration.

Expression of ZFNs with GVRs

To demonstrate targeted mutagenesis using GVRs, an LSL T-DNA plasmid was modified to encode the Zif268:FokI ZFN (pLSLZ; Figure 2A). Target sequence for this ZFN is present within a reporter gene that is stably integrated in tobacco plants (Wright et al., 2005). This reporter is a nonfunctional, translational fusion between GUS and neomycin phosphotransferase (NPTII; Datla et al., 1991). Leaf tissue from transgenic tobacco plants carrying the gus:nptII reporter was syringe infiltrated with an Agrobacterium strain containing pLSLZ (one side of a leaf) or coinfiltrated with a mixture of Agrobacterium strains containing pLSLZ and pREP (the other side of the leaf). Total DNA was extracted 7 d after infiltration, replicational release was verified by PCR (Figure 2A), and the Zif268 target sequence was analyzed for mutations introduced by NHEJ. To this end, a 484-bp DNA fragment encompassing the Zif268 target sequence was amplified by PCR. Amplicons were digested with MseI (an MseI site is present in the spacer sequence between the two Zif268 binding sites), and bands were resolved by agarose gel electrophoresis (Figure 2B). Cleavage-resistant amplicons from leaf tissue transformed with both pLSLZ and pREP T-DNA were cloned and sequenced (Figure 2C). Six out of eight sequenced clones contained mutations at the ZFN target site. Five out of the six sequences had distinct NHEJ-induced mutations, indicating that DNA in multiple leaf cells had been modified. We observed significantly higher levels of targeted mutagenesis in cells transformed with pLSLZ and pREP compared with pLSLZ alone (Figure 2D). Cleavage-resistant amplicons observed from tissue delivered pLSLZ T-DNA are most likely due to expression of Zif268:FokI from the virion-sense LIR promoter. These results indicate that GVRs can express ZFNs for creating targeted DNA double-strand breaks.

Expression of TALENs and Components of the CRISPR/Cas System with GVRs

Having demonstrated that GVRs can deliver ZFNs to plant cells, we next tested if they were effective vehicles for delivering TALENs and CRISPR/Cas reagents. We constructed a pLSL vector that encodes the T30 TALEN pair (pLSLT; Supplemental Figure 8A; Zhang et al., 2013), which targets a site within the tobacco acetolactate synthase (ALS) genes. A second vector was constructed that encodes Cas9 and synthetic guide RNA (sgRNA), also targeting ALS (pLSLC; Supplemental Figure 8D). After delivery to plant cells, mutagenesis of endogenous targets was monitored by loss of a restriction enzyme site due to NHEJ-induced mutation at the site of cleavage. We observed an enhancement in mutagenesis when pREP was cotransformed, suggesting GVR-mediated expression of TALENs and CRISPR/Cas elements can be achieved (Supplemental Figures 8B, 8C, 8E, and 8F).

GVRs Enhance Frequencies of Gene Targeting

GVRs were next assessed for their ability to achieve gene targeting through the delivery of sequence-specific nucleases and repair templates. The target for modification was the defective gus:nptII gene, which can be repaired by correcting a 600-bp deletion that removes part of the coding sequences for both GUS and NPTII (Wright et al., 2005). An LSL T-DNA plasmid was constructed that encodes the Zif268:FokI ZFN followed immediately by repair template sequence (designated us:NPTII) designed to restore gus:nptII function (pLSLZ.D; Figure 3A). Cells having undergone gene targeting will express a functional GUS protein and, consequently, will stain blue when incubated in a solution with the GUS substrate X-Gluc. Random integration of the repair template or read-through transcription from virus promoters should not produce functional GUS protein due to 500 bp that is missing from the 5′ end of the coding sequence. This was experimentally confirmed by transforming nontransgenic leaf tissue with pLSLZ.D and pREP T-DNA; no GUS activity was observed 5 d after infiltration (Figure 3B, Neg. Ctrl.). To compare the performance of GVRs with conventional T-DNA, we engineered a T-DNA vector with Zif268:FokI sequence followed immediately by the us:NPTII repair template (p35SZ.D). Five days after infiltration, leaf tissue was stained in X-Gluc and assessed for cells having undergone gene targeting. We observed an enhancement in GUS activity in leaf tissue transformed with pLSLZ.D and pREP T-DNA (Figure 3B) relative to the p35SZ.D and pLSLZ.D controls. To verify repair of the reporter gene, PCR was performed on total DNA extracted from leaf tissue transformed with pLSLZ.D and pREP T-DNA. Primers were used that recognize sequence within the repaired reporter gene and ∼1 kb downstream of the homology carried by the repair template (Supplemental Figure 9). Amplification of a 2.3-kb product from 9 of the 11 leaf samples indicates the presence of the repaired GUS:NPTII transgene. Sequence analysis confirmed the presence of an NPTII repair template-transgene junction, consistent with repair of the reporter by gene targeting.

The relative enhancement in gene targeting was calculated by quantifying the density of blue sectors within transformed leaf tissue. A significant enhancement in gene targeting was observed in tissue transformed with pLSLZ.D and pREP T-DNA across multiple transgenic plant lines where the reporter gene is integrated at different chromosomal positions (Figure 3C; Wright et al., 2005). Notably, in plant line 1, we observed an enhancement in gene targeting greater than two orders of magnitude. Taken together, these results demonstrate that GVRs can deliver sequence-specific nucleases and repair templates for gene targeting and that GVRs contribute features that result in the enhancement of gene targeting in somatic cells relative to conventional T-DNA delivery.

Double-Strand Breaks Enhance Gene Targeting

There are several features of GVRs that may promote gene targeting, including replication to high copy numbers, high protein expression and pleiotropic activities of Rep and RepA. To test these features, we transformed single leaves with two Agrobacterium samples and directly compared GUS activity resulting from gene targeting (Figure 4A). Pairing two samples on a single leaf reduced the variation in results due to environmental conditions (e.g., differences in leaf age, size, and health). As T-DNA size affects transfer efficiency, we ensured that vectors being directly compared had similar T-DNA sizes. To determine the contribution of double-strand breaks to gene targeting and to test the effectiveness of this assay, Zif268:FokI (897 bp) was replaced with GFP (912 bp). We observed a significant decrease (100-fold) in GUS activity when Zif268:FokI was removed (Figure 4B), which is consistent with previous reports describing a stimulatory effect of double-strand breaks on recombination (Puchta et al., 1996).

Replication of Repair Templates Enhances Gene Targeting

To determine if replication of GVRs contributes to gene targeting, we compared our GVR system (pLSLZ.D and pREP) to a replication-inactive system (p35SZ.D and pREP). We observed a significant decrease in blue sectors when the cis-acting replication elements were removed (Figure 4C, left graph). These results indicate that replication enhances gene targeting; however, we could not conclude whether replicating sequence-specific nucleases or repair templates contributed more toward gene targeting. Therefore, to address this issue, we explored the role of sequence-specific nuclease expression on GVR-mediated gene targeting.

One explanation for the observed high frequency of gene targeting is that GVRs mediate high expression of sequence-specific nucleases, and, as a result, double-strand breaks are more frequent. To test this hypothesis, we compared the level of NHEJ-induced mutations in leaf tissue delivered Zif268:FokI using GVRs (pLSLZ and pREP) or conventional T-DNA (p35SZ). We predicted that if expression of sequence-specific nuclease was enhanced using GVRs, we would observe higher levels of NHEJ-induced mutations. Our results refute this hypothesis (Supplemental Figure 10A): We observed a lower level of NHEJ-induced mutations with GVRs, suggesting that replication of sequence-specific nucleases does not augment levels of gene targeting. As an alternative approach to assess the impact of nuclease amplification, we modified Rep and RepA to carry a mutation in the conserved ATPase domain (K219H for BeYDV). This mutation impairs replication (Desbiez et al., 1995; Supplemental Figure 10B); however, other functions of Rep and RepA should remain active (e.g., nicking activity, ligation activity, and transactivation of virion-sense transcription). As expected, pREPK219H T-DNA delivered to leaf tissue was not replicated (Supplemental Figure 10C), and we observed similar levels of targeted mutagenesis with pREPK219H as with pREP (∼10%; Supplemental Figures 10D and 10E). Collectively, the results suggest that replication of Zif268:FokI coding sequence does not significantly contribute to enhancing targeted mutagenesis.

To test the hypothesis that gene targeting is enhanced by replication of repair templates and not sequence-specific nucleases, we compared gene targeting frequencies of two GVR systems that either replicate sequence-specific nucleases or repair templates (Figure 4C, right graph). Here, we observed significantly higher levels of gene targeting when repair templates were replicated. Similarly, when we translocated repair templates from conventional T-DNA to GVRs, we observed substantially higher levels of gene targeting (Supplemental Figure 11). These results indicate that although expression of sequence-specific nucleases is critical for gene targeting, replication of repair template sequence contributes more toward enhancing gene targeting.

Pleiotropic Activity of Rep and RepA Enhances Gene Targeting

Geminivirus Rep and RepA proteins interact with a multitude of host proteins. Specifically, RepA protein from mastreviruses interacts with Geminivirus RepA binding protein 1 (GRAB1) and GRAB2 (Xie et al., 1999) and the retinoblastoma-related protein (RBR; Horváth et al., 1998). It may be possible that expression of mastrevirus Rep and RepA in plant cells affects levels of homologous recombination. To test this hypothesis, we compared gene targeting levels in tissue-delivered p35SZ.D T-DNA or p35SZ.D and pREP T-DNA. Here, we observed a significant increase in blue sectors when pREP was delivered (Figure 4D). To confirm that delivery of pREP T-DNA to plant cells results in the expression of both Rep and RepA, we used RT-PCR to detect Rep and RepA transcripts. To avoid detecting RepA transcripts produced by Agrobacterium, we transfected protoplasts with the pREP T-DNA plasmid. Total RNA was extracted 24 h after transfection, and we detected unspliced (RepA) and spliced (Rep) transcripts (Supplemental Figure 12). These results demonstrate that pleiotropic activity of Rep and/or RepA promotes gene targeting (Figure 4D). Taken together with the results in the previous two sections, high-frequency gene targeting using GVRs is due to synergy between targeted double-strand breaks, replication of repair template sequence, and pleiotropic activity of Rep and RepA.

Recovery of Calli and Plantlets with a Precise DNA Change

To demonstrate the feasibility of GVRs for generating fully developed plants with a desired sequence change, we regenerated cells harboring the corrected GUS:NPTII transgene. We chose to regenerate leaf cells because our previous data demonstrate the usefulness of GVRs in whole leaves. Leaves were syringe infiltrated with two strains of Agrobacterium containing pLSLZ.D and pREP. Five days later, infiltrated leaves were removed from the plant, surface sterilized, and placed on regeneration media with kanamycin (Figure 5A). To track the progression of regenerating cells with the GUS:NPTII transgene, subsets of leaf discs were stained in X-Gluc solution 0, 7, 14, 21, 42, and 49 d after plating. We observed GUS expression from leaf discs stained between 0 and 21 d (Figures 5A to ​to5D).5D). After day 42, we observed shoots with GUS expression (Figures 5E and ​and5F).5F). To confirm that tissue staining blue harbors the corrected GUS:NPTII transgene, we extracted genomic DNA from blue calli and performed PCR with primers designed to amplify sequence from the repaired GUS:NPTII transgene. Because our negative control samples (noninfiltrated tissue) did not produce calli or shoots, we chose to analyze genomic DNA from callus tissue that did not stain blue in the pLSLZ.D and pREP samples. We observed amplification of sequence from all four blue-stained tissues and not from tissue that did not stain blue (Figure 5G). Sequencing of all four PCR products confirmed the presence of an NPTII repair template-transgene junction, consistent with repair of the reporter by gene targeting (Figure 5H). These findings demonstrate that GVRs can be used to introduce exact DNA changes in plant cells and that these cells maintain the ability to regenerate.

Plant Material

Tobacco (Nicotiana tabacum var Xanthi) plants, both wild type and transgenic (Wright et al., 2005), were grown at 21°C with 60% humidity under a 16-h-light and 8-h-dark cycle. Leaves (fully expanded upper leaves) from 4- to 6-week-old tobacco plants were used for Agrobacterium transformation. When comparing gene targeting or targeted mutagenesis frequencies between different T-DNA molecules, one leaf per plant was infiltrated.

Agrobacterium Transformation

T-DNA vectors were introduced into Agrobacterium (GV3101/pMP90) using the freeze-thaw method. Single transformed colonies were grown overnight at 28°C in 3 mL of Luria-Bertani starter culture containing kanamycin (50 μg/mL) and gentamycin (50 μg/mL). The next day, 1 mL of starter culture was used to inoculate 50 mL of Luria-Bertani culture containing kanamycin (50 μg/mL), gentamicin (50 μg/mL), 10 mM MES, and 20 μM acetosyringone. After reaching an OD₆₀₀ of ∼1 (around 16 h), cells were pelleted and resuspended to an OD₆₀₀ of 0.2 using infiltration buffer (10 mM MES, 150 μM acetosyringone, and 10 mM MgCl₂, pH 5.6). Resuspended cultures were incubated at room temperature for 2 to 4 h. For experiments requiring the coinfiltration of two different Agrobacterium strains, cultures were mixed together in a 1:1 ratio, or, for three different strains, in a 1:1:1 ratio. When comparing two different Agrobacterium samples, the ODs of the solutions were adjusted accordingly to ensure equal concentrations of individual Agrobacterium strains. Leaves from tobacco plants were infiltrated with Agrobacterium using a 1-mL syringe. Immediately following infiltration, plants were watered and covered with a plastic dome to maintain high humidity. Plastic domes were removed ∼24 h after infiltration.

GUS Assay

To visualize cells expressing GUS, leaf tissue was vacuum infiltrated with X-Gluc solution (10 mM phosphate buffer, 10 mM EDTA, 1 mM ferricyanide, 1 mM ferrocyanide, 0.1% Triton X, and 1 mM X-Gluc). Leaf tissue was then incubated in X-Gluc solution at 37°C for ∼24 h. Chlorophyll was removed by incubating stained leaf tissue in 80% ethanol for ∼1 week.

Tobacco Regeneration

Leaves from 4- to 5-week-old tobacco plants were infiltrated with pLSLZ.D and pREP. As a control for selection and regeneration of kanamycin resistant cells, additional leaves were infiltrated with pCAMBIA2301 (T-DNA that contains a kanamycin resistance and also the GUS gene). Two separate experiments were performed; for each experiment, three to five leaves from different plants were infiltrated with each T-DNA or T-DNA pair. Five days after infiltration, leafs were removed from the plant and surface sterilized by submersion in 10% bleach for ∼20 min. Leaves were then washed three times in sterile water. Sterilized leaf tissue was cut into ∼1 × 1-cm discs using a surgical blade. Leaf discs were plated onto regeneration media (4.4 g Murashige and Skoog medium with vitamins, 1 mg/L 6-benzylaminopurine, 0.1 mg/L 1-naphthaleneacetic acid, 30 g Suc, and 8 g agar, pH 5.7) containing kanamycin (100 μg/mL) and Timentin (200 μg/mL). As a negative control for tobacco selection, nontransformed leaf tissue was sterilized and plated. Discs were transferred to new plates weekly.

PCR-Based Detection of Circularized GVRs and the GUS:NPTII Transgene

To detect the presence of circularized GVRs, total DNA was extracted from plant tissue and used as a template for PCR. Oligonucleotides were designed to be homologous to DNA sequence located upstream and downstream of LIR sequence in the circular replicon (5′-GTTTCACTTCACACATTATTACTG-3′ and 5′-TGTTGAGAACTCTCGACGTCCTGC-3′). To minimize strand transfers during amplification, PCR was performed using the Expand Long Template PCR system (Roche).

To detect the presence of a repaired GUS:NPTII transgene, total DNA was extracted from leaf or callus tissue and used as a template for PCR. Oligonucleotides were designed to be homologous to DNA sequence located within the corrective sequence on the repair template and sequence ∼1 kb downstream of the region of homology harbored on the repair template (5′-GTCGGTGAACAGGTATGGAAT-3′ and 5′-CTACATACCTCGCTCTGCTAATC-3′).

Detection of ZFN-, TALEN-, and Cas9-Induced Mutations

Experiments that used GVRs to mediated expression of ZFNs, TALENs, and CRISPR/Cas elements for targeted mutagenesis were performed similarly: One side of a leaf was delivered pLSL, and the other side was infiltrated with pLSL and pREP. Five to seven days after infiltration, total DNA was extracted and assessed for mutations at the appropriate sequence-specific nuclease target site. To detect Zif268:FokI-induced mutations, the gus:nptII target locus was amplified using primers that are homologous to sequences 270 bp upstream and 214 bp downstream of the MseI site (present within the spacer region of the Zif268 binding sequences; 5′-AAGGTGCACGGGAATATTTCGCGC-3′ and 5′-GCCATGATGGATACTTTCTCG-3′). The resulting amplicons were digested for 16 h at 37°C with MseI and resolved by gel electrophoresis. Cleavage-resistant amplicons were gel purified, cloned into pJET1.2, and sequenced. To detect TALEN-induced mutations at the SurA and SurB loci, total DNA was first predigested with AluI for 16 h at 37°C. Digested DNA was used as a template for a PCR with primers designed to amplify a 573-bp product encompassing the T30 binding site (5′-GACGGTGCAGAAAGTGAAGTA-3′ and 5′- TATGGCCCAGGAGTGTCTAA-3′). The resulting amplicons were digested for 16 h at 37°C with AluI and resolved by gel electrophoresis. Cleavage-resistant amplicons were gel purified, cloned into pJET1.2, and sequenced. To detect Cas9-induced mutations, the target DNA sequence present within SurA and SurB was amplified by PCR (forward, 5′-TGACATCTGGTGGATTAGGAGC-3′; reverse, 5′-CAATCACATCCAACAAGTATGGC-3′). The resulting 408-bp amplicons were digested with AlwI for 16 h at 37°C. Cleavage-resistant products were gel purified and cloned into pJET1.2 and sequenced.

Data Collection and Statistics

Following staining in X-Gluc, the density of blue sectors was quantified in leaf tissue. Three nonoverlapping images were taken for each infiltrated region (leaf halves) using a camera attached to a stereoscope (Nikon). Each image captured 1.8 cm² of leaf tissue, totaling 5.4 cm² for each leaf half. Image locations were directed to capture the entire number of blue sectors for each leaf half. For samples where the total number of blue sectors exceeded 5.4 cm² of area, regions with the highest density of blue sectors were imaged. Leaves without detectable GUS activity were excluded from the data analysis. ImageJ software was used to calculate the number of blue sectors per image. P values in Figures 4 to ​to66 were generated using a paired, two-tailed Student’s t test. P values in Figures 2 and ​and33 were generated using a two-tailed Student’s t test.

Accession Numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: ADH1, AT1G77120.1; mild strain of BeYDV, {"type":"entrez-nucleotide","attrs":{"text":"DQ458791","term_id":"91980531"}}DQ458791; SurA, {"type":"entrez-nucleotide","attrs":{"text":"X07644.1","term_id":"19776"}}X07644.1, SurB, {"type":"entrez-nucleotide","attrs":{"text":"X07645.1","term_id":"19778"}}X07645.1; pCPCbLCVA.007, {"type":"entrez-nucleotide","attrs":{"text":"AY279345","term_id":"31747149"}}AY279345; and pCPCbLCVB.002, {"type":"entrez-nucleotide","attrs":{"text":"AY279344","term_id":"31747148"}}AY279344.

We thank members of the Voytas lab for the helpful discussion and insights with this work. We thank Dominique Robertson for the kind donation of the CaLCuV plasmids. This research was supported by a grant to D.F.V. from the National Science Foundation (DBI 0923827) and by a grant to J.G.-H. from Fundación Alfonso Martín Escudero.

N.J.B. and D.F.V. designed the research. N.J.B., J.G.-H., and T.C. performed the experiments. N.J.B. analyzed the data. N.J.B., T.C., and P.A.A. contributed reagents/materials/analysis tools. N.J.B. and D.F.V. wrote the article.