Homology-directed repair using next-generation CRISPR/Cpf1-geminiviral replicons in tomato

Genome editing via the homology-directed repair (HDR) pathway in somatic plant cells is very inefficient compared to error-prone repair by nonhomologous end joining (NHEJ). Here, we increased HDR-based genome editing efficiency approximately 3-fold compared to a Cas9-based single-replicon system via the use of de novo multiple replicon systems equipped with CRISPR/LbCpf1 in tomato and obtained replicon-free but stable HDR alleles. The efficiency of CRISPR/LbCpf1-based HDR was significantly modulated by physical culture conditions such as temperature and light. Ten days of incubation at 31°C under a light/dark cycle after Agrobacterium-mediated transformation resulted in the best performance among the tested conditions. Furthermore, we developed our single-replicon system into a next-generation multiple replicon system that effectively increased HDR efficiency. Although this approach is still challenging, we showed the feasibility of HDR-based genome editing of a salt-tolerant SlHKT1;2 allele without genomic integration of antibiotic markers or any phenotypic selection. Self-pollinated offspring plants carrying the HKT1;2 HDR allele showed stable inheritance and germination tolerance in the presence of 100 mM NaCl. Our work may pave the way for transgene-free editing of alleles of interest in asexually as well as sexually reproducing plants.

the best performance among the tested conditions. Furthermore, we developed our INTRODUCTION 35 In plant somatic cells, double-strand DNA breaks (DSBs) are efficiently repaired by a 36 nonhomologous end joining (NHEJ) mechanism, which dominates over the homology-directed 37 repair (HDR) pathway (Jiang et al., 2013;Puchta, 2005). NHEJ repair usually leads to various 38 types of mutations including DNA sequence insertions, deletions (Hsu et al., 2014;Zetsche et al., 39 2015), chromosome rearrangement, or chromosome relocation (Ferguson and Alt, 2001;40 Richardson et al., 1998;Varga and Aplan, 2005). Early in the 1990s, a transgenic approach using 41 yeast mitochondrial I-Sce I endonuclease as a DSB inducer was adopted in attempts to 42 investigate the mechanisms of DSB repair in plants, especially gene targeting via the HDR 43 pathway in plant somatic cells (Fauser et al., 2012;Puchta et al., 1993), which have been the 44 indel mutations are introduced during NHEJ-mediated repair (Baltes et al., 2014;Zetsche et al., 68 2015). CRISPR/Cpf1 complexes were recently successfully applied for gene targeting in 69 plants (Li et al., 2018), providing alternative options for T-rich target site selection. 70 Because of the highly efficient replication of geminivirus genomes and their single-stranded 71 DNA nature, these genomes have been used as perfect DNA template cargo for gene targeting in  al., 2003;Needham et al., 1998;Zhang and Mason, 2006) systems, due to their highly 75 autonomous replication inside host nuclei and the ability to reprogram cells (Gutierrez, 1999;76 Hanley- Bowdoin et al., 2013). Furthermore, Rep/RepA has been reported to promote a cell 77 environment that is permissive for homologous recombination to stimulate the replication of 78 viral DNA. Interestingly, it has been reported that somatic homologous recombination is 79 promoted by geminiviral infection (Richter et al., 2014). The above characteristics of geminiviral 80 replicons have been shown to make them perfect delivery tools for introducing large amounts of 81 homologous donor templates to plant nuclei. Likewise, the movement and coat proteins of a bean 82 yellow dwarf virus (BeYDV)-based replicon were removed and replaced with Cas9 or TALEN 83 to improve gene targeting in plants (Baltes et al., 2014;Butler et al., 2016;Cermak et al., 2015; 84 Dahan-Meir et al., 2018;Gil-Humanes et al., 2017;Hummel et al., 2018). The LbCpf1 complex, 85 which was subsequently discovered and adapted for plant genome editing in 2015, has not been 86 tested in combination with geminiviral replicon systems for plant gene targeting. 87 Despite higher success rates in gene targeting in plants using the geminiviral replicon system, 88 most of the reported cases have required markers associated with the edited alleles, while the 89 selection and regeneration of HDR events from edited cells are still challenging (Butler et al.,90 2016; Gil-Humanes et al., 2017;Hummel et al., 2018). In addition, the effective application of 91 replicon cargos in plant gene targeting has been shown to be limited by their size (Baltes et al.,92 2014; Suarez-Lopez and Gutierrez, 1997). Therefore, plant gene targeting, especially in cases 93 of marker-free alleles, still requires improvement. We hypothesized that the combination of 94 the repeatedly cutting nature of a CRISPR/Cpf1 complex and the highly autonomous 95 replication of de novo-engineered geminiviral replicon systems could overcome the efficacy 96 barrier of marker-free gene targeting via the HDR pathway in plants. Here, we report 97 significant improvement of homology-directed repair using next-generation CRISPR/LbCpf1-98 geminiviral replicons in tomato and the successful application of the system to target a marker-99 free salt-tolerant HKT1;2 allele. Through this work, we aimed to increase HDR efficiency for 100 practical application in a fast crop breeding scenario (Hickey et al., 2019).

102
The CRISPR/LbCpf1-based geminiviral replicon system is feasible for performing HDR in 103 tomato 104 To test the hypothesis above, we re-engineered a bean yellow dwarf virus (BeYDV) replicon 105 to supply a high dose of homologous donor templates and used the CRISPR/LbCpf1 system  Figure 2). The LbCpf1 system 124 using two guide RNAs for targeting the ANT1 gene, a key transcription factor controlling the 125 anthocyanin pathway, showed a much higher HDR efficiency, of 4.51±0.63% (normalized to 126 an overexpression construct (pANT1 ox , Figure 1B)), than the other control constructs, 127 including the "minus Rep" (pRep -) and "minus gRNA" (pgRNA -) constructs. LbCpf1 system-128 based HDR was visualized by the presence of purple calli and/or shoots ( Figure 1C and 1D), 129 and its efficiency was similar to that of a CRISPR/SpCas9-based construct (pTC217) (Cermak 130 et al., 2015) included in the same experiment ( Figure 1C) or used in hexaploid wheat with the 131 same scoring method (Gil-Humanes et al., 2017). It is worth noting that the normalized HDR 132 efficiencies reported from this study (see Materials and Methods section) using transformed 133 cell-based efficiency are calculated differently from those reported in the initial work by 134 Čermák and coworkers (2015); the previous authors used the transformed cotyledon-based 135 efficiency, which is approximately one order of magnitude higher than the cell-based 136 efficiency. The data obtained from this experiment revealed that functional geminiviral 137 replicons were crucial for increasing HDR efficiencies of the Cpf1 complex. This result shows 138 the feasibility of highly efficient HDR in plants using Cpf1 expressed from a geminiviral 139 replicon, thus expanding the choices of molecular scissors for gene targeting in plants.

140
Favorable physical conditions significantly increase the HDR efficiency of the 141 CRISPR/LbCpf1-based geminiviral replicon system 142 In seeking suitable physical conditions for Agrobacterium-mediated delivery and DSB repair 143 using our HDR tool in tomato somatic cells, we investigated various incubation regimes at 144 early stages posttransformation. Short-day conditions have been shown to have strong impacts 145 on intrachromosomal recombination repair (ICR) in Arabidopsis (Boyko et al., 2005). We 146 tested whether the same could be true for the gene targeting approach in tomato. Using 147 various lighting regimes, including complete darkness (DD), short (8 hours light/16 hours 148 dark; 8 L/16 D)-and long (16 L/8 D)-day conditions, we found that the HDR efficiencies 149 achieved under short-and long-day conditions were higher than those under DD conditions in 150 the case of LbCpf1 but not SpCas9 and reached 6.62±1.29% (p<0.05, Figure 1E). Considering 151 the similar repair activities observed after DSBs were generated by either of the CRISPR/Cas 152 systems, it was quite difficult to explain why the light conditions only affected LbCpf1-based 153 HDR in this experiment compared to the dark treatment. There must be unknown 154 mechanism(s) that facilitate LbCpf1-mediated HDR in a light-dependent manner.

155
Temperature is an important factor controlling ICR (Boyko et al., 2005), CRISPR/Cas9-based 156 targeted mutagenesis in plants (LeBlanc et al., 2018), and CRISPR/Cpf1-based HDR in zebrafish and Xenopus by controlling genome accessibility (Moreno-Mateos et al., 2017). 158 Pursuing the approach for the improvement of HDR, we compared the HDR efficiencies of 159 the pHR01 and pTC217 systems subjected to various temperature treatments under an 8 L/16 160 D photoperiod, since the two nucleases (SpCas9 and LbCpf1) may respond differently. Our 161 data revealed that within a temperature range of 19-31°C, the somatic HDR efficiency 162 increased with increasing temperature ( Figure 1F). Notably, at 31°C, LbCpf1 showed an HDR 163 efficiency (9.80±1.12%) that was more than 2-fold higher than that of SpCas9 (p<0.05) and temperature. Interestingly, the LbCpf1 complex was shown to be highly active only at high 172 temperatures (i.e., more than 29°C), which partially explains the higher HDR efficiencies 173 observed at high temperatures in this experiment. Briefly, a comparison of data on plant HDR 174 between Cas9-and Cpf1-based systems at different temperatures and under short-day 175 conditions is presented to reveal the best conditions for plant HDR improvement.

176
A multiple-replicon system outperformed the single-replicon system in HDR-based GE.

177
The size of viral replicons has been shown to be inversely correlated with their copy numbers 178 (Baltes et al., 2014;Suarez-Lopez and Gutierrez, 1997). In an approach to overcome the 179 replicon size limitation, we designed and tested the novel idea of using a T-DNA system that 180 potentially produces multiple replicons ( Figure 2A, and Supplemental Figure 3). Compared to 181 pHR01, a multiple-replicon system designed to release donor templates from replicon 2 182 (MR02) but not replicon 1 (MR01) showed a significant increase in the HDR efficiency by 30% 183 and reached up to 12.79±0.37% ( Figure 2B and Supplemental Table 1). Temporal evaluation 184 of donor template levels between the HDR tools showed significantly higher levels of MR02 185 at 3 days posttransformation (dpt) compared to those of pHR01 and MR01 ( Figure 2C).

188
Under the same conditions and calculation methods, the combination of our multiple replicons 189 with LbCpf1 significantly increased HDR efficiencies by 3-4-fold compared to those of the 190 Cas9-based replicon systems. We also confirmed the release of three circularized replicons 191 from the single vector used in this work ( Figure 2D) by PCR amplification using circularized 192 replicon-specific primers (Supplemental Table 2).

193
In another test of the multiple replicon system, we overexpressed two key proteins involved in    Figure 3B). The PCR products were sequenced to identify junction sequences.

224
A majority of the events (11/16) showed sequences corresponding to perfect right arm 225 integration through HDR repair, and 5/16 events showed a combination of HDR and NHEJ 226 repair with an NHEJ fingerprint at the 5' terminus of the pNOS sequence (Supplemental 227 Figure 5A, with event C1.8 highlighted in blue) or even RB integration at the left junction 228 boundary (Supplemental Figure 6). All of the sequences amplified from the left junctions 229 showed perfected DNA sequence exchange via the HDR pathway (Supplemental Figure 5B).

230
The results obtained in these analyses revealed the common features of products repaired via  The HDR allele was stably inherited in offspring by self-pollination as well as backcrossing 244 To confirm stable heritable edits, we grew genome-edited generation 1 (GE1) plants ( Figure   245 3C) obtained from the self-pollination of LbCpf1-based HDR GE0 events and identified a 246 segregating population with a purple phenotype (Supplemental Table 5  Practical successful editing by HDR using marker-free approaches 267 To show the applicability of our HDR system to practical plant genome editing, we sought to 268 use it to edit a potentially agronomic trait, and salinity tolerance was chosen as the target trait.

269
High-affinity K + Transporter 1;2 (HKT1;2) plays an important role in the maintenance of K+  Table 7). In comparison with the first report on the marker-free gene targeting showed allele segregation following Mendelian rules ( Figure 5C). The true HKT1;2 N217D 301 HDR alleles in the GE1 plants were ultimately confirmed by Sanger sequencing. Furthermore, 302 we successfully generated HDR-based SlEPSPS1 events with an ~1% efficiency using this 303 replicon system without using herbicide for selection (data not shown), thereby validating the 304 feasibility of our replicon systems for practical applications. It is worth noting that most of the 305 elite alleles in plants do not associate with any marker, and hence, a highly efficient marker-free 306 system is in high demand.

307
Thus, through the application of various approaches, our study showed a large improvement 308 of HDR efficiency in tomato somatic cells. The HDR allele was stably inherited in subsequent Our study of HDR improvement was conducted using tomato (Hongkwang cultivar, a local 334 variety) as a model plant. All the binary vectors were transformed into Agrobacterium 335 tumefaciens GV3101 (pMP90) using electroporation. Agrobacterium-mediated transformation 336 was used to deliver editing tools to tomato cotyledon fragments (Supplemental Figure 12).

395
HDR efficiencies were recorded in at least three replicates and were statistically analyzed and 396 plotted using PRISM 7.01 software. In Figure 1C, multiple comparisons of the HDR 397 efficiencies of the other constructs with that of pRepwere performed by one-way ANOVA 398 (uncorrected Fisher LSD test, n=3, df=2, t=4.4; 4.4 and 1.5 for pTC217; pHR01 and pgRNA -, 399 respectively). In Figure 1E, pairwise comparisons of the HDR efficiencies of pTC217 and        promoter were designed to be inserted at a position 142 bp upstream of the ANT1 start codon.

553
The cutting sites of the two guide RNAs used in this study are indicated by two black arrows.      The right junctions (amplified by ZY010F/TC140R) of the events were confirmed to be perfectly 683 adapted to HDR repair (Supplemental Figure 5A), but the left junction could not be amplified 684 (using the UPANT1-F1/NptII-R1 primer pair, Figure 1A).      . Sequence alignment shows the perfectly edited HKT1;2 N217 to D217 allele with the WT allele as a reference. The nucleotides highlighted in the discontinuous red boxes correspond to intended modifications for N217D, PAM and core sequences (to avoid recutting). (B) HDR construct layout for HKT1;2 editing. There is neither selection nor a visible marker integrated into the donor sequence. The NptII marker was used for the enrichment of transformed cells.
(C) Morphology of the HKT1;2 N217D edited event compared to its parental WT in greenhouse conditions. Scale bar = 1 cm. Supplemental Figure 2. The de novo-engineered geminiviral amplicon (named pLSL.R. Ly) and its replication in tomato.
(A) Map of pLSL.R.Ly. The DNA amplicon is defined by its boundary sequences (long intergenic region, LIR) and a terminated sequence (short intergenic region, SIR). The replication-associated protein (Rep/RepA) is expressed from the LIR promoter sequence. All of the expression cassettes of HDR tools were cloned into the vector by replacing the red marker (Lycopene) using a pair of type IIS restriction enzymes (BpiI, flanking ends are TGCC and GGGA). Left (LB) and right (RB) denote the borders of a T-DNA.
(B) Circularized DNA detection in tomato leaves infiltrated with pLSL.R. Ly compared to those infiltrated with pLSLR. Agrobacteria containing the plasmids were infiltrated into tomato leaves (Hongkwang cultivar), and infiltrated leaves were collected at 6, 8 and 11 dpi and used for the detection of circularized DNAs. N: water; P1: positive control for pLSL.R. Ly; positive control for P2: pLSLR; Cx: Control samples collected at x dpi; Ixy: infiltrated sample number y collected at x dpi; I11 V: sample collected from leaves infiltrated with pLSLR at 11 dpi. PCR products obtained using primers specific to GAPDH were used as loading controls.  N C6 I61 I62 I63 C8 I81 I82 I83 C11 I111 I112 I113 I11V  The right junctions (amplified by ZY010F/TC140R) of the events were confirmed to be perfectly adapted to HDR repair (Supplemental Figure 5A), but the left junction could not be amplified (using the UPANT1-F1/NptII-R1 primer pair, Figure 1A). Sequencing of the left junction region showed a ligation event between the RB of the T-DNA and the 3' break in the upstream ANT1 promoter sequence via NHEJ. Red dotted line: ligation boundary.