Three-year-old sweet orange (Citrus sinensis, cultivar Valencia) plants were cultivated in a quarantine greenhouse facility at the Citrus Research and Education Center, Lake Alfred, FL, U.S.A, at temperatures ranging from 25 to 30°C. A specific permit was issued by the Division of Plant Industry, Florida Department of Agriculture and Consumer Service, for the location and activities. Before agroinfiltration, the plants were pruned to produce uniform shoots.

DNA sequences for all primers in this study are provided in Table S1 in File S1. The 35S promoter of CaMV, derived from the binary vector pBI121 (accession number AF485783), was cloned into pCambia1380 (accession number AF234301) using HindIII and BamHI to produce 1380-35S. pBI121 was obtained from The Arabidopsis Information Resource (Stock#: CD3-388). Cas9 containing a Flag tag at its N-terminus and a nuclear localization signal at its C-terminus was amplified from Addgene plasmid 42230 using a pair of primers, Cas9-5-BamHI and Cas9-3-EcoRI [20]. After BamHI-EcoRI digestion, the PCR product was inserted downstream of the CaMV 35S promoter to produce the construct 1380-35S-Cas9. A Nos terminator was amplified using the primers NosT-5-EcoRI and NosT-3-XhoI-AscI. The PCR products were digested with EcoRI and AscI and placed downstream of the Cas9 coding region to obtain 1380-Cas9. The sgRNA scaffold portion was amplified from Addgene plasmid 41819, using the primers sgRNA-5-BamHI and sgRNA-3-SacI [22], and inserted into pBI121 following digestion with BamHI and SacI to form 121-sgRNA.

The CaMV 35S promoter fragment was amplified using the primers CaMV35-5-XhoI and sgRNA-CsPDS2, and the sgRNA-NosT fragment was amplified using the primers sgRNA-CsPDS1 and NosT-3-AscI. Using three-way ligation, XhoI-digested CaMV35S and AscI-cut sgRNA-NosT were cloned into XhoI-AscI-treated 1380-Cas9 to construct 1380-Cas9:sgRNA targeting the CsPDS gene.

Sweet orange leaves were inoculated with either tap water or a culture of actively growing Xanthomonas citri subsp. citri (Xcc) re-suspended in sterile tap water (5×10⁸ CFU/ml). Eight hours later, the same inoculated leaf areas were subjected to agroinfiltration as described previously [31], with modifications (Fig. 1a). Recombinant Agrobacterium tumefaciens cells were cultured in 3 ml Luria broth (LB) medium with appropriate antibiotics at 28°C. One hundred microliters of overnight culture was grown in 100 ml fresh LB medium with 10 mM 2-(N-morpholino) ethanesulfonic acid (MES), pH5.6, and 40 μM acetosyringone (AS), as well as the appropriate antibiotics. Upon reaching OD₆₀₀ = 0.8, the inoculum was harvested and resuspended in MMA solution (10 mM MgCl₂, 10 mM MES, pH 5.6 and 200 μM AS) to a final OD₆₀₀ of 1.0. The suspension was left at room temperature for 2 h and infiltrated into the same area previously inoculated by Xcc. As a control, citrus leaves were subjected to agroinfiltration in the absence of prior Xcc inoculation.

Xcc-facilitated agroinfiltration was carried out with sweet orange leaves using A. tumefaciens cells transformed with Cas9/sgRNA binary plasmids (Fig. 2a) or with pCambia1301 (accession number AF234297), which carries a GUS construct [31]. Four days after agroinfiltration with pCambia1301-transformed Agrobacterium, GUS staining was performed as described previously (Fig. 1a) [32]. The β-Glucuronidase Reporter Gene Staining Kit (Sigma-Aldrich) was used for GUS staining following the manufacturer's protocols. After overnight staining, the citrus leaf samples were destained and subsequently photographed.

The GUS activity assay was performed as described previously [33], with modifications. Thirty milligrams of leaf tissue was ground in 300 μl of GUS extraction buffer (50 mM NaPO₄ (pH 7.0), 1 mM Na₂EDTA, 0.1% SDS, 0.1% Triton X-100, 10 mM DTT) using a mortar and pestle. The ground material was centrifuged at 4°C for 15 min at top speed (12,000× g) in a tabletop centrifuge. After centrifugation, 200 μl of supernatant was collected, and the amount of protein was measured using Bradford reagent (Bio-Rad, Hercules, CA). Twenty-five microliters of extract was mixed with 225 μl of GUS extraction buffer containing 1 mM 4-methyl-umbelliferyl glucuronide (MUG), and then the mixture was incubated at 37°C. After 30 min and 90 min, 100 μl of the reaction was added to 900 μl of 0.2 M Na₂CO₃, and the fluorometric values were determined using a VersaFluor Fluorometer (BioRad). The GUS activity is expressed as pmol of 4-methyl-umbelliferone (MU) per μg protein per min.

Four days after Xcc-facilitated agroinfiltration with 1380-Cas9:sgRNA-transformed Agrobacterium, genomic DNA was extracted from the sweet orange leaves using a Wizard Genomic DNA Purification Kit (Fig. 1a) (Promega) according to the manufacturer's protocol. The genomic DNA was dissolved in 100 μl of distilled deionized water, and the concentration was determined using a spectrophotometer.

To confirm the targeted modification of the CsPDS gene (accession number XM_006492918), 400 ng of genomic DNA was digested with MfeI for 3 hours at 37°C. Using the digested genomic DNA as the template, PCR was performed with the Phusion DNA polymerase (New England Biolabs) and a pair of primers, CsPDS-5-P1 and CsPDS-3-P2, flanking the target site in CsPDS. The PCR products were ligated into the PCR-BluntII-TOPO vector (Life Technologies), and 28 random colonies were selected for DNA sequencing.

The method used to measure the mutation rate was previously reported by Shan et al. [28]. After PCR amplification using non-digested genomic DNA as a template, the PCR products were treated with MfeI and analyzed on an agarose gel. By using AlphaImager EP (AlphaInnotech) software, the intensity of the ethidium bromide-stained DNA bands was quantified after background subtraction, and the band intensities were summed to obtain the total intensities. To measure the mutation frequency, the intensity of the uncut band was divided by the total intensity.

Potential off-target sequences were identified by searching the sweet orange genome database via the BLASTN tool [29] against the sgRNA target site within the CsPDS gene (TTGTGCACAAGCAATTGTAC). A total of 46 off-target sequences were found, of which 23 contained an MfeI site within the aligned sequence (Table S2 in File S1). Fourteen regions out of the 23 that lacked an MfeI site close to the target site were chosen for analysis. Attempts were made to amplify these regions by PCR using the primers listed in Table S1 in File S1. The eight regions amplified by PCR were analyzed by agarose gel electrophoresis, either with or without prior digestion by MfeI.

Previously, it has been difficult to implement agroinfiltration-mediated transient expression in citrus leaves (Fig. 1) [34]. Importantly, our results indicated that GUS expression was dramatically enhanced in sweet orange leaves using Xcc-facilitated agroinfiltration (Fig. 1b, Fig. 1c) (i.e., initial inoculation with Xcc followed eight hours later by agroinfiltration with recombinant A. tumefaciens cells containing a GUS gene (Fig. 1a)). In contrast, there was very low GUS activity in sweet orange leaves when agroinfiltration was performed alone, or when leaves were pretreated with tap water (Fig. 1b, Fig. 1c). Therefore, Xcc pre-treatment is very important for enhanced transient protein expression in citrus leaves.

One possible mechanism to explain the fact that Xcc pre-treatment could significantly increase the transient expression of GUS in agroinfiltrated sweet orange leaves is that in Xcc-inoculated citrus leaves, the PthA4 effector is known to be translocated from Xcc into plant cell nuclei, where it activates downstream target genes. This leads to excessive cell division (hyperplasia) and cell enlargement (hypertrophy) [35]–[37]. Interestingly, citrus epicotyl segments are being widely used as explants for Agrobacterium-mediated citrus transformation, owing to the higher proportion of actively dividing cells, which are presumed to be more susceptible to T-DNA, in epicotyl tissue [34]. It seems possible that in the case of Xcc-facilitated agroinfiltration, Xcc is eliciting excessive cell division, which might mimic the fast dividing epicotyl segment. Such rapidly dividing cells may also be more susceptible to A. tumefaciens transformation, thereby resulting in the enhanced GUS expression observed in treated sweet orange leaves (Fig. 1b, Fig. 1c).

To test the potential of the Cas9/sgRNA system to induce genetic modification in citrus, Xcc-facilitated agroinfiltration was employed to deliver the Cas9 endonuclease and an sgRNA. We first constructed the binary vector p1380-Cas9:sgRNA (Fig. 2a, Fig. S1 in File S1), which contains both Cas9 and an sgRNA targeting the CsPDS gene. CsPDS encodes a phytoene desaturase with homologs in rice, Arabidopsis and Nicotiana benthamiana, which have been successfully genetically modified using Cas9/sgRNA [25], [28], [38]. The 35S promoter of CaMV was used to drive the expression of Cas9 and the sgRNA targeting CsPDS. Cas9 was fused with a nuclear localization signal at the C-terminus (Fig. 2a) [20].

We expressed Cas9 and the CsPDS-targeting sgRNA in sweet orange leaf tissue using Xcc- facilitated agroinfiltration. Four days after agroinfiltration, sweet orange genomic DNA was extracted from the treated leaf areas (Fig. 1a) and digested using the MfeI restriction enzyme. The target sequence within the CsPDS gene contains an MfeI restriction site, which will be mutated if Cas9/sgRNA works as predicted. Thus, selective amplification of mutagenized CsPDS genes was used to verify targeted genome engineering in sweet orange. As expected, the CsPDS PCR product (480 bp) was observed in control leaves not treated with MfeI (Fig. 2b, lane 4). However, after MfeI digestion, the 480-bp band was only detected in the MfeI-cleaved genomic DNA of sweet orange leaves co-expressing Cas9 and the CsPDS-targeting sgRNA (Fig. 2b, lane 1) and not in sweet orange leaves expressing Cas9 alone or control leaves (Fig. 2b, lane 2 and lane 3). Two additional replicates showed similar results (Fig. S2 in File S1). This result indicated that Cas9/sgRNA successfully induced mutations in the targeted CsPDS gene.

The PCR products were purified from lane1 (co-expression of Cas9 and CsPDS-targeting sgRNA) and cloned for sequencing (Fig. 2c, Fig. S3 in File S1). A total of twenty-eight colonies were selected for sequencing. The sequencing results showed twenty-four clones with indels in the targeted sequence, which abolished the MfeI restriction site within the target region (Fig. 2c, Fig. S3 in File S1). Further detailed analysis revealed that the indels could be grouped into eleven different types, including deletions (from 1 bp to 12 bp) and nucleotide substitutions (Fig. 2c, Fig. S3 in File S1). Although the CsPDS gene was clearly targeted by Cas9/sgRNA (Fig. 2c, Fig. S3 in File S1), there were no visible albino spots on the treated sweet orange leaves. This result is consistent with results from Arabidopsis and Nicotiana benthamiana, whose leaves also did not show visible albino spots after Cas9/sgRNA-mediated PDS targeting via agroinfiltration [38].

To calculate the efficiency of CsPDS mutation caused by Cas9/sgRNA in citrus, PCR products were amplified from the genomic DNA of Cas9/sgRNA-treated and control leaf tissues, digested with MfeI, and subjected to gel electrophoresis. Based on the ratio of the intensities of uncut DNA bands to the total intensity of both cut and uncut bands from three independent experiments [28], the mutation rate was estimated to be 3.2%, 3.4% and 3.9%(Fig. 2d, Fig. S4 in File S1). In a previous report [38], the mutation rate caused by Cas9/sgRNA was 5.6% in Arabidopsis protoplasts and 37.7% in N. benthamiana protoplasts, whereas the mutation frequency was 2.7% in agroinfiltrated Arabidopsis leaves and 4.8% in N. benthamiana leaves. Thus, the mutation rate in sweet orange leaves is comparable to that induced by Cas9/sgRNA after leaf agroinfiltration in Arabidopsis and N. benthamiana.

It should be noted that the 35S promoter of CaMV was employed to drive sgRNA expression (Fig. 2a, Fig. S1 in File S1). It has been previously reported that the 35S promoter of CaMV could be utilized to transcribe the sequence of interest with no extraneous sequence at the 5′ end if the sequence was inserted at the transcription start site of the 35S promoter of CaMV [39]. Using the same scheme, the sgRNA was inserted at the transcription start site of the CaMV 35S promoter (Fig. S1 in File S1). In comparison to the plant U6 promoter, which constrains the first nucleotide at the transcription start site to be ‘G’ [28], the 35S promoter of CaMV seems to allow more flexibility in the sgRNA design, owing to a lack of constraint with regard to the first nucleotide. This was confirmed by the results.

We also analyzed potential off-target mutagenesis caused by Cas9/sgRNA. We identified a total of 46 potential off-target sequences by searching the C. sinensis genome database using the 20-bp target sequence of the CsPDS gene in BLASTN (Table S2 in File S1). Those off-target sequences contain 13 to 17 nucleotides identical to the CsPDS-targeting sequence (Table S2 in File S1). Fourteen regions out of the 46 identified sequences contained MfeI sites and did not have additional MfeI sites in close proximity to the off-target sequences. Thus, we selected those 14 sequences for MfeI-suppressed PCR analysis. Only 8 sequences were amplified successfully using PCR (Fig. S5 in File S1). None of the 8 amplicons showed evidence of Cas9/sgRNA-induced MfeI restriction site loss similar to that observed with the CsPDS target sequence (Fig. 2, Fig. S5 in File S1). Thus, off-target mutagenesis was not detected for CsPDS in our study, even though we could not rule out potential off-target mutagenesis in other sequences. This is most likely a case-by-case situation that depends on whether high-identity off-target sequences are present in non-targeted genes and whether we could select unique target sequences during our bioinformatic analysis, considering the recent findings of off-target mutagenesis induced by Cas9/sgRNA [40].

Our data complements recent studies in which the Cas9/sgRNA system was successfully developed to modify the genomes of plants such as Arabidopsis, tobacco, rice, wheat, and sorghum [41]. For the first time, the sweet orange genome has proven to be readily targeted via the Cas9/sgRNA system, with the aid of Xcc-facilitated agroinfiltration. It must be kept in mind that this study did not produce stable Cas9/sgRNA-transformed sweet orange. Thus, it is worth confirming that Cas9/sgRNA can also be employed to modify the sweet orange genome in transgenic plants. Given that Cas9/sgRNA works in transgenic citrus, Cas9/sgRNA-mediated citrus genome editing holds significant promise for studies of citrus gene function and for targeted genetic modification. In conclusion, Cas9/sgRNA appears to be a potentially valuable tool for creating new citrus cultivars with beneficial traits for both growers and consumers.