Ecient Targeted Gene Insertion in Maize using Agrobacterium-Mediated Delivery

CRISPR-Cas is a powerful DNA double strand break technology with wide-ranging applications in plant genome modication. However, the eciency of genome editing depends on various factors including plant genetic transformation processes and types of modications desired. Agrobacterium infection is the preferred method of transformation and delivery of editing components into the plant cell. While this method has been successfully used to generate gene knock-outs in multiple crops, precise nucleotide replacement and especially gene insertion into a pre-dened genomic location remain highly challenging. Here we report an ecient, heritable, selectable marker-free site-specic gene insertion in maize using Agrobacterium-mediated delivery. Advancements in maize transformation and new vector design enabled targeted insertion with frequencies as high as 8–10%. Importantly, these advancements allowed not only an improvement of the frequency but also of the quality of generated events. These results further enable the application of genome editing for trait product development in a wide variety of crop species amenable to Agrobacterium-mediated transformation.


Introduction
Demonstration that the presence of a DNA double strand break (DSB) increases frequency of homologous recombination at the target site by more than 1,000-fold 1,2 has led to development of several classes of site-directed nucleases (SDNs) -homing endonucleases, zinc-nger nucleases (ZFNs), transcription activator-like effector nucleases (TALENs) and CRISPR (clustered randomly interspaced short polydromic repeats)-Cas (CRISPR associated) nucleases -capable of cutting genomic DNA and generating DSBs at predetermined locations 3,4 . Although all SDNs have been demonstrated to be useful DSB reagents, CRISPR-Cas quickly became the technology of choice for most laboratories due to its high activity, versatility, and low cost 5,6,7 .
In eukaryotic cells, DSBs can be repaired via two highly-conserved pathways -non-homologous end joining (NHEJ) and homology directed repair (HDR) 8,9 . NHEJ is prone to imperfect repair and may lead to variable size insertions and/or deletions (INDELs) resulting in transcription frame-shifts and consequent gene knockouts. If two guide RNAs (gRNAs), with homology to sites anking a gene or a DNA fragment of interest are used, it can result in a deletion of the entire sequence between the two sites 10,11 . HDR enables a precise insertion of a donor sequence to a speci c target site and relies on the presence of homology regions to sequences upstream and downstream of the DSB. Thus, by using a repair template that contains nucleotide alterations, speci c changes to the coding sequence of an endogenous gene can be introduced. Alternatively, if the repair template includes a new sequence (e.g. gene of interest), the DSB repair can result in the sequence insertion into a speci c chromosomal site avoiding endogenous gene disruption often observed during random integration. Moreover, this approach allows for consecutive insertions of several genes into the same chromosomal region (gene stacking), which can signi cantly simplify breeding programs 12,13,14,15 .
In planta 36,37,38,39 and intra-genomic 14,40 homologous recombination approaches have been used to address plant transformation ine ciency and quality of targeted insertion events. These approaches rely on stable random integration of the repair template, anked with SDN target sites, into the genome. Introduction of SDN (usually by crossing) releases repair template and simultaneously generates DSB at the intended target site. Long generation time and low e ciency have been two major limitations of this method in plants. However, a promising new approach with high e ciency intra-genomic targeted gene insertion has recently been reported 41 .
It has been suggested that linear DNA can be a better substrate for HDR-mediated DSB repair and increase frequency of targeted gene insertions 42,43 . Flanking repair template with target sites and release of linear repair template has been shown to provide a 2 to 5-fold improvement of targeted gene insertion in human culture cells in comparison to circular plasmid DNA 44 . However, in several experiments conducted in plant species using particle bombardment delivery of editing components, no such effect was demonstrated 25,45,46,47 .
Here we report on HDR-mediated targeted gene insertion in maize with frequencies as high as 3% using advanced Corteva Agriscience Agrobacterium-mediated transformation. We also demonstrate that anking repair template with Cas9-gRNA target sites to release the repair template from the T-DNA further increases frequency of targeted insertions by approximately 3-fold. Advances in plant transformation combined with improvement in vector design allowed us to signi cantly increase not only frequencies (up to 8-10%) but also improve the quality of HDR-mediated gene insertion events. Moreover, our results demonstrate that targeted insertion of a repair template with and without selectable marker gene occur with comparable frequencies. These results further enable the application of genome editing for trait product development in a wide variety of crop species amenable to Agrobacterium-mediated transformation.

Vector design
Three different T-DNA vectors were used in this study (Fig. 1). Two nearly identical vectors ( Fig. 1A 50 anked with regions of homology (HR1 and HR2). The only difference between the two vectors is the presence of TS45 sequences with protospacer adjacent motif (PAM) anking HR1 and HR2 fragments in the second vector leading to release of the repair template from the T-DNA upon target sites cleavage by Cas9 nuclease (Fig. 1B). This vector design allowed for evaluation of the repair template release effect on the frequency of the HDR-mediated targeted gene insertion. In addition, the presence of two selectable marker genes, one inside the repair template (NptII) and one outside (Hra), allowed to compare the effect of a selectable marker position on the frequency and quality of insertion events.
The third vector (Fig. 1C) contained two morphogenic genes, Wus2 and Bbm, regulated by Axig1 and PLTP promoters, maize Ubiquitin promoter regulated Streptococcus pyogenes Cas9, Polymerase III U6 promoter regulated gRNA for genomic target site TS45, the repair template comprised of the trait gene, Arabidopsis carboxylesterase 20 (CXE-20) 51 under rice Actin promoter 52 anked with homology regions (HR1 and HR2) and TS45 target sequences with PAM, and selectable marker gene NptII under maize Ubiquitin promoter. In this vector, a selectable marker gene is placed downstream from the repair template. With this con guration, cleavage of the target sites anking repair template at transient stage results not only in the release of repair template, but also in separation of selectable marker gene (NptII) from the rest of T-DNA (Fig. 1C). This design allowed us to compare selectable marker gene position effect on both transformation and targeted gene insertion frequencies.
Described vectors were transformed into Agrobacterium LBA4404 cells containing pVIR9 accessory plasmid, which provides increased virulence and T-DNA transmission to maize embryo cells 53 .

Targeted gene insertion
Results of the four experiments conducted using constructs described above are summarized in Table 1.
In Experiment 1, 889 and 865 maize immature embryos (PH184C genotype) were transformed using Agrobacteria carrying T-DNA constructs shown in Figs. 1A and 1B, respectively. Embryos were then subjected to tissue culture and regeneration procedures described in Methods section. A total of 343 and 425 T0 plants, respectively, were regenerated on G418 (Geneticin) containing media using selectable marker gene (NptII) within the repair template and analyzed as illustrated in Fig. 2. HDR-mediated insertion events were rst detected by diagnostic HR1 and HR2 junction qPCR (Fig. 2, A and B). Putative insertion events were further validated by long PCR spanning the entire insertion using genomic primers outside of HR1 and HR2 homology regions (Fig. 2, A and C). This approach replaced Southern blot hybridization analysis commonly used to identify T0 plants with putative perfect insertion among HR1/HR2 junction PCR positive events 19 . Long PCR products were analyzed by agarose gel  (Table 1, Experiment 1). Long PCR analysis demonstrated that for the two treatments, 6 (1.7%) and 26 (6.1%) T0 plants, respectively, had putative perfect gene insertions. In addition, different proportions of putative perfect events (50% and 67%) were observed among HR1/HR2 junction positive events for constructs without and with anking target sites, respectively (Table 1, Experiment 1).
The exact same constructs and experimental design were used in Experiment 2 to validate high frequency and quality of Agrobacterium-mediated gene insertion observed in Experiment 1. A total of 1155 and 1185 embryos were transformed and produced 591 and 342 T0 plants using constructs without and with target sites anking the repair template, respectively. For T-DNA vector without target sites, 14 T0 plants (2.4%) were HR1/HR2 junction qPCR positive, while for the construct with anking target sites, 33 T0 plants (9.6%) were HR1/HR2 positive when analyzed by junction qPCR (Table 1, Experiment 2). Notably, long PCR analysis performed on HR1/HR2 junction positive T0 plants again demonstrated higher proportion of putative perfect HDR-mediated gene insertion events (64%) for the construct with repair template anked with target sites. For the construct without anking sites, this number was 50% (Table 1, Experiment 2). Thus, Experiments 1 and 2 showed very close results con rming e cient targeted gene insertion using Agrobacterium-mediated delivery. In addition, anking repair template with Cas9-gRNA target sites not only improved HDR frequency by approximately 3-fold, but also the quality of the insertion events as indicated by the long PCR data.
To compare insertion frequencies of repair templates with and without selectable marker gene, we used T-DNA construct shown in Fig. 1B. In this experiment (Experiment 3), after Agrobacterium infection, a total of 2884 transformed embryos were divided in two groups. For the group of 909 embryos, tissue culture steps and T0 plant regeneration were performed on selective media containing G418 using NptII as selectable marker gene (inside the repair template), while for 1975 embryos, these steps were conducted using imazapyr as selective agent using Hra as selectable marker gene (outside of the repair template). The number of embryos used in the treatment with imazapyr selection was intentionally doubled to compensate for potentially lower frequency of the HDR-mediated insertion events. A total of 320 and 766 plants were regenerated using G418 and imazapyr selection, respectively. With selectable marker gene (NptII) inside the repair template, 33 (10.3%) regenerants were HR1/HR2 positive by junction qPCR, while 64 (8.3%) plants were HR1/HR2 positive when selectable marker gene (Hra) outside the repair template was used (Table 1, Experiment 3). Long PCR analysis demonstrated lower proportion of putative perfect insertion events in the group of T0 plants with selectable marker gene outside the repair template (53% vs. 60%). Experiment 4 was designed to test frequency of targeted insertion using a trait gene, CXE-20, and effect of selectable marker gene position (downstream of repair template, Fig. 1C) on transformation and trait gene insertion frequencies. In addition, to test potential differences in gene insertion frequency between genotypes, this experiment was conducted using a different Corteva Agriscience inbred line, PH1V5T, which has higher regenerability. One thousand immature embryos were transformed and 1403 T0 plants regenerated in this experiment. A total of 116 (8.3%) T0 plants were positive for insertion based on HR1/HR2 junction qPCR analysis, and 55 (3.9%) plants were identi ed to have putative perfect integration events by long genomic PCR ( It is important to emphasize, that in the four Experiments conducted, the proportion of long PCR positive events varied and probably re ects differences in vector design and position of selectable marker genes (Table 1). For the construct without anking target sites, this number was 50% in both Experiments 1 and 2. In Experiments 1, 2, and 3 when the repair template was anked with TS45 target sites and selectable marker was inside the repair template, 60 to 67% of all HR1/HR2 positive events were also long PCR positive. When selectable marker gene (Hra) upstream of the repair template was used, this number was 53% (Experiment 3). In Experiment 4 when selectable marker gene (NptII) was placed downstream from the repair template, the proportion of long PCR positive events dropped to 47%.

Transmission to next generation and segregation analysis
To evaluate transmission of targeted gene insertions to next generation, segregation pattern, and attrition rates, a group of 67, long PCR-positive T0 plants from Experiments 1 and 3, were sent to the greenhouse. A total of 49 T0 plants grew to maturity, were fertile and used as pollen donors to cross with wild-type PH184C plants to generate T1 progeny ( Table 2).
Fifteen to forty T1 plants (based on seed availability) for each of the 49 T0 events were then analyzed for the presence of targeted gene insertion and T-DNA components (Cas9, gRNA, Bbm, Wus2, Hra, and NptII). Out of 49 T0 events, 33 showed 1:1 segregation for the targeted gene insertion, consistent with stable Mendelian inheritance of a mono-allelic locus. From the remaining 16 events, 8 transmitted gene insertion allele to 20-25% of T1 progeny likely due to chimeric nature of the T0 plants. Other 5 T0 plants had insertions in both alleles -one allele containing a putative perfect insertion, while the second one was either HR1 or HR2 junction-qPCR positive. Progeny of one T0 plant was 100% positive for gene insertion likely indicating a perfect bi-allelic insertion event. The remaining 2 events did not show transmission of the insertion to T1 generation also likely due to chimeric nature of the original T0 plants. Positive for gene insertion and T-DNA-free progenies were obtained from 39 T0 plants. In the remaining 8 events, random T-DNA integration likely occurred either into the same chromosome as targeted gene insertion or into more than one genomic site. In these cases, additional round(s) of backcross are need to complete segregation of T-DNA components. Gene insertion veri cation by sequencing T1 progeny plants of 13 insertion-positive and properly segregated T0 events from Experiment 1 were further analyzed by sequencing to verify quality and integrity of the insertions. Long PCR products from 13 selected T1 plants (one plant representing each of the 13 T0 events) were sequenced by NGS and assembled into contigs. Comparison of all 13 contigs showed no sequence variation con rming that every plant had precise, HDR-mediated targeted gene insertion.

Discussion
Targeted gene insertion is the most challenging type of genome editing in plants. Its success depends on multiple factors, including transformation and regeneration processes, e ciency of genome editing components delivery, copy number of the repair template and its availability at the DSB site, size of the DNA sequence to be inserted, presence or absence of a selectable marker gene in the repair template, and HDR e ciency at a given target site. In addition, gene insertion experiments usually have high attrition rate of generated events due to the complexity of the insertion, plant chimerism, transmission to the next generation, and ability to segregate intended edits from helper genes (e.g. Cas9, gRNA, and selectable marker). Different approaches have been suggested and tested to overcome some of these issues and increase the overall success rate of targeted gene insertions: particle bombardment and Agrobacterium transformation, delivery of editing components as DNA, RNA, and ribonucleoprotein (RNP), suppression of NHEJ and activation of HDR pathways, viral replication to increase repair template copy number, linear and circular repair template molecules, tethering repair template to Cas9-gRNA RNP complexes, inducible, cell cycle-speci c and developmental promoters, intragenomic and in planta approaches, and exploiting different insertion mechanisms -HDR, NHEJ, homology mediated end joining (HMEJ), and microhomology mediated end joining (MMEJ) 54,55,56 . However, despite new genome editing approaches and better understanding of underlying mechanisms of DSB repair, the progress in developing a simple, robust, and reliable process for targeted gene insertion in plants at a higher e ciency has remained limited.
Agrobacterium-mediated transformation is the preferred method for delivery of editing components into the plant cell 57 . This system is reproducible, amenable for a broad variety of crop species (both monocots and dicots) and cost effective since it doesn't require special equipment and associated consumables. In addition, Agrobacterium-mediated delivery is less invasive, has lower attrition rate during regeneration stage 58,59 , and results in lower frequency of potential chromosomal rearrangements observed in particle bombardment experiments 60,61 . The major limitation of the Agrobacterium-mediated approach is a low number of T-DNA molecules transmitted into the plant cell during infection resulting in low copy number of repair template and, therefore, low frequency of targeted gene insertion. This disadvantage can be overcome by repair template copy number increase using geminivirus replication system 56 . However, this approach requires a more complicated vector design, results in uncontrolled DNA replication, may lead to unintended random integration of DNA into the genome, and can result in lower regeneration frequency 35 .
Due to low frequencies of genome editing via HDR, targeted insertion experiments require hundreds of regenerated events to recover plants with the desired outcome. Many important crop species, especially their elite genotypes, have low regenerability and are extremely challenging for complex HDR-based edits.
A rapid maize transformation system 48 that relies on expression of two morphogenic genes, Bbm and Wus2 62 , allows recovery of transgenic plants from transformed immature embryos at high frequencies.
The use of morphogenic genes has also signi cantly improved transformation e ciency of elite maize genotypes 62 , allowing a faster trait product development. Furthermore, it has been suggested that DSB repair via homologous recombination occurs predominantly during the late S, G2 and early M phases of the cell cycle, while being actively suppressed at G1 stage 63,64,65,66 . Consequently, e cient HDR-based genome editing is largely restricted to actively proliferating cells. Therefore, an additional bene t of using morphogenic genes in genome editing experiments is associated with an effect of BBM and WUS2 proteins on stimulation of cell division, providing an HDR-friendly cellular environment.
Another improvement of maize Agrobacterium-mediated transformation has been accomplished by developing a new ternary vector system that utilizes an optimized accessory plasmid pVIR9 53 . This system increased transient T-DNA delivery and recovery of stable callus events, resulting in six-to-sevenfold improvement over conventional random transformation using the plasmid pSB1 in an elite maize inbreds. The combination of morphogenic genes and new ternary vector design resulted in high transformation e ciency, increased number of T-DNA molecules transmitted, and high frequency of plant regeneration. Implementation of this enhanced Agrobacterium transformation protocol allowed us to generate HDR-based insertion events at target site TS45 with frequencies up to approximately 3% (Table 1). However, further optimization of Agrobacterium-mediated gene insertion was advantageous.
Target sites anking repair template was described in intragenomic gene targeting experiments 36,37,38 , where they were used to release the repair template from random integration loci. This design has an additional advantage as linear repair template potentially increases frequency of HDR-based gene insertion 42,43 . It might be even more advantageous considering recent results indicating that damaged DNA might be transported to speci c loci at the nuclear for further repair 67 . Target sites anking repair template have been used in recent particle bombardment experiments in Arabidopsis and rice 25,45,46,47 . However, a positive effect of this vector design on frequency of gene insertion could not be veri ed as no controls were provided in these experiments. Contrary, introduction of anking target sites has resulted in 2-5-fold increase of integration frequencies relative to circular plasmids in human culture cells 44 . In our experiments, direct comparison of T-DNA vectors with and without target sites anking repair templates, consistently demonstrated an approximately 3-fold increase of targeted gene insertion frequency, resulting in about 8-10% frequency of HDR-mediated gene insertion events based on the number of T0 plants analyzed (Table 1).
Target site TS45 used in this study was identi ed as one of the best sites with high HDR frequency in previous particle bombardment experiments 15 . Selection of this site was relevant for this study as it allowed to generate high number of HDR-based gene insertion events, compare results of different experiments and the corresponding controls, and draw reliable conclusions.
Another important factor that can signi cantly affect the outcome of gene insertion experiments, but rarely discussed, is the event attrition rate. For example, in this report, the frequency of HR1/HR2 junction qPCR positive events (approximately 8-10%) drops by approximately 50% after long PCR analysis. Most likely, this is the outcome of the concurrent action of HDR and NHEJ pathways often resulting in cointegration of various DNA molecules (vector DNA and/or genomic DNA sequences) into the target site 19,68 . In addition, our results indicate that attrition rate may depend on different parameters, including T-DNA vector con guration, such as position of selectable marker gene, regenerants survival rate, plant fertility, and chimerism. In this report, the combined attrition rate ranges from 64 to 80%, with the highest lost observed in the experiment with selectable marker gene outside the repair template (Table 2).
In summary, the advancements in Agrobacterium-mediated maize transformation 48,53,62 combined with optimized vector design, enabled e cient targeted gene insertions in maize. Our experiments showed reliable and reproducible results for two genotypes with different transformability, two different genes, and for constructs with a selectable marker both inside and outside repair templates. These results open new opportunities for accelerated precision breeding in a wide range of crop species amenable to Agrobacterium-mediated transformation.

Plant material
Maize (Zea mays L.) inbred lines PH184C and PH1V5T were obtained from internal Corteva Agriscience sources.

Maize transformation
Two Corteva Agriscience inbred lines used in this study, PH184C and PH1V5T, are proprietary. All plants used as a source of immature embryos were grown in greenhouse conditions. Ear harvest, immature embryo isolation, Agrobacterium-mediated transformation, and plant regeneration were performed as previously described 69 . T0 and T1 plant analysis Genomic DNA was extracted from leaf tissue as previously described 70 . Quantitative PCR (qPCR) was used to detect T-DNA components (Cas9, gRNA, Bbm, Wus2, Hra, and NptII) and mutation frequency at endogenous target site TS45 using Qiagen QuantiTect Multiplex PCR Master Mix (Qiagen, Germany) with the primers and probes listed in supplementary Table S1. Junction PCR assays were used to detect gene insertion at the target site TS45. In this assay, PCR ampli cation of the target region was coupled with a nested qPCR using primers and probes listed in supplementary Table S2. PCR was performed using 2x Extract-N-amp PCR Ready Mix (Sigma, USA) or 2x Phusion Flash High-delity PCR Master Mix (Thermo Fisher Scienti c, USA) according to the manufacturer's recommendations. For nested PCR used in screening of the insertion events, the rst PCR was carried out in 5 µL of reaction mixtures for 20 cycles.
Fifteen µL of the reaction mixture containing 2xTaqMan Master Mix (LGC, USA) and primers then were added, and the second PCR was performed using LightCycler 480 (Roche Life Science, Germany) for 30 cycles. Data were analyzed using the Endpoint Genotyping Software (Roche Life Science, Germany). Long PCR was performed using Extensor Master Mix (Thermo Fisher Scienti c, USA) or LongAmp® Hot Start Taq 2X Master Mix (NEB, USA) with primers and probes listed in supplementary Table S2. All qPCR copy number assays were calibrated using Corteva Agriscience proprietary housekeeping gene assay (VIC).
Integrity of the gene insertion was con rmed in T1 plants by NGS sequencing of long genomic PCR amplicons. Two sets of sequencing primers (22 forward and 22 reverse) were used in the analysis (Supplementary Table S3). Approximately 250 bp-long reads were then assembled into contigs and compared using Sequencher v. 4.8.

Data Availability
The authors declare that the data supporting the ndings are available within the paper or are available from the corresponding author upon reasonable request. Corteva Agriscience™ will provide plasmids to academic investigators for non-commercial research under an applicable material transfer agreement subject to proof of permission from any third-party owners of all or parts of the material and to governmental regulation considerations. Completion of a stewardship plan is also required. The Corteva Agriscience inbred lines PH184C and PH1V5T described in this research are proprietary.

COMPETING INTERESTS
The authors declare no competing interests.
conducted transformation and plant regeneration; B.L. and C.S. analyzed plant materials; D.P., P.B. and S.S. analyzed data; P.B. and S.S. wrote the manuscript.