Gene Editing in Prunus Spp.: The Challenge of Adapting Regular Gene Transfer Procedures for Precision Breeding

2.1 Regeneration in Prunus salicina

Since the first work on in vitro propagation described by Rosati et al. [11], Japanese plum has been demonstrated to be suitable for in vitro regeneration [12] and genetic transformation [13]. Based on the use of hypocotyls isolated from mature embryos, several varieties, such as ‘Early Golden’, ‘Shiro’, ‘Angeleno’ and ‘Larry Ann’, were responsive to culture treatment in the presence of thidiazuron (TDZ) and indole-3-butyric acid (IBA) as supplements of regular Murashige and Skoog (MS) [14] media. These procedures were derived from a previous description for European plum (P. domestica) reported by Mante et al. [15] and later improved by Padilla et al. [16].

In our hands, the use of hypocotyls has allowed for a reproducible gene transfer and regeneration procedure for several Japanese plum varieties (Figure 1A–D). Slight variations from the primary protocol described by Urtubia et al. [13] have allowed expanding this procedure to the use of epicotyl sections as starting explants, increasing the number of useful material when mature fruits are collected. Agrobacterium-mediated transformation of Japanese plum explants requires co-culturing (2–3 days in darkness, usually using strain GV3101) and regeneration on LP-derived medium [17] supplemented with picloram (LP1 medium). For shooting, explants are transferred to MS medium supplemented with variable ratios between TDZ and IBA (MSR2 medium), depending on the variety. Shoot elongation is achieved by cultivation in MS03 medium, i.e., MS-based medium supplemented with benzylaminopurine (BAP). Finally, rooting is achieved by culturing these shoots in a woody plant medium (WPM) [18] supplemented with 1-naphthalene-acetic acid (NAA) and kinetin (WPM rooting medium). Regarding the efficiencies, variable and acceptable values have been obtained depending on the starting explant and the variety assayed, for instance, whereas ‘Larry Ann’ hypocotyl sections reach about 3%, epicotyl segments of the same genotype have shown 17%.

2.2 Regeneration in Prunus avium

Very few reports have successfully achieved regeneration, and first procedures used expanded leaves of commercial varieties, including ‘Lapins’, ‘Burlat’, ‘Napoleon’, and ‘Sweetheart’ [19, 20]. These works used WPM supplemented with the growth regulators NAA, BAP, and TDZ. Later, improved regeneration efficiencies were described for ‘Schneiders’, ‘Sweetheart’, ‘Starking Hardy Giant’, ‘Kordia’, and ‘Regina’ using leaf and nodal segments, which were achieved by the combined use of WPM plus Driver and Kuniyuki medium (DKW) [21] or applying Quoirin and Lepoivre medium (QL) [22]. As best growth regulators, these experiments included TDZ and IBA. Using a different approach, mature cotyledons cultured in solid QL medium bound to a 10-day darkness incubation regime led to the genotypes ‘Vista’, ‘Sunburst’, ‘Tehranivee’, ‘Vouge’, and ‘Heidelfingen’ to be regenerated [23]. That shooting capability was lost by some varieties when darkness was eliminated.

Better results have been reported in other commercially important genotypes, including sour cherry (P. cerasus L.) [7, 24], black cherry (P. serotina Ehrh.) [25], and the cherry rootstocks ‘Rosa’ (P. subhirtella autumno) [26], ‘Gisela 6’ (P. cerasus x P. canescens) [7], ‘Colt’ (P. avium x P. pseudocerasus) [27]. In general, these protocols have cultured leaf explants in media, such as QL plus BAP, and WPM plus BAP and IBA, to produce whole plants in a 4–6 month period.

Based on those previous descriptions and considering our results in P. salicina regeneration, we developed a consistent procedure for sweet cherry varieties, such as ‘Bing’, ‘Van’, and ‘Lapins’. The system is based on the use of epicotyl segments, which are first co-cultured with Agrobacterium (mostly strain GV3101) in MP2 medium (LP-derived medium supplemented with picloram) in darkness for 2–3 days. Subsequently, explants are transferred to B15-medium for budding induction; B15 medium is also based on LP micro- and macro-nutrients, supplemented with the growth regulators TDZ and 6-γ,γ-dimethylallylaminopurine (2iP). Shooting elongation is then achieved in D03 medium (DKW-based medium supplemented with BAP and IBA), first culturing the explants in Petri dishes and then using jars. Finally, elongated shoots are rooted in WPM medium (same as in plum regeneration) (Figure 2A–D). Observed efficiencies in this system reach about 4% (4.33% and 3.75% for ‘Lapins’ and ‘Bing’, respectively). Hypocotyls have also proven to be responsive to these procedures, although efficiencies are considerably lower.

2.3 Regeneration in Prunus persica

Peach highlights as one of the more recalcitrant species regarding the in vitro regeneration process. Although the first descriptions for in vitro propagation were described as early as 1982 [28], all the considerable research published to date pursuing the development of a reproducible regeneration system based on adult tissue as starting explants has shown this to be more complex than originally thought [2].

Considering those experiences and our own experimentation, we have been focused on improving efficiencies for the use of immature seeds as a source for explants [4], which in our hands represents the only reproducible procedure to achieve peach regeneration with a reasonable time frame. Since then, improvements on this pipeline have contributed to the generation of new individuals by an adequate protocol for the application of genetic engineering.

The protocol for peach regeneration using seed explants was formerly derived from descriptions found in the work by Gentile et al. [17], and it was initially applied on ‘O’Henry’ and ‘Rich Lady’ explants. One key point resulted in the use of immature cotyledons 70–90 days after bloom, which are cultured in LP modified medium. In general terms, after transformation and co-culture with A. tumefaciens strain GV3101, peach cotyledons are transferred to solid LP medium supplemented with BAP and NAA for callusing. Induced calli, usually observed after 3 to 6 months, are transferred to 03BAP-LP medium (LP medium supplemented with BAP and IBA) for budding induction. Shoot elongation is then achieved by transferring the shoots to CP2 medium (LP-derived medium supplemented with growth increased ratios of BAP/IBA). Finally, peach shoots are transferred to 03BAP-MP medium (LP medium using MS micronutrients plus growth regulators BAP and IBA) for elongation and rooting (Figure 3A–D). We have observed that embryo-derived explants from these immature fruits are also responsive to these treatments, and after infection, co-culture, and callusing in LP medium, explants must be cultured in 03BAP-MP medium.

For rooting, plantlets regenerated from either cotyledon or embryo explants may need to be placed for 1 month in M1 medium (MS-derived medium supplemented with NAA); after this period, plantlets can be cultured back to 03BAP-MP medium. The procedure to achieve complete in vitro rooted plants can be as short as 8 months. However, longer times can be needed. In terms of regeneration efficiency, around 2% of embryo explants from ‘Elegant Lady’ can regenerate into plants. The regeneration rate for cotyledons in ‘Elegant Lady’ is in the region of 0.5–1%.

3.1 Available genomes in Prunus and a CRISPR/Cas9 analysis tool for them

Prunus spp. belong to the Rosaceae family are among the most relevant groups in terms of genome drafts already available. The Genome Database for Rosaceae (GDR, https://www.rosaceae.org) provides public access to genomics, genetics, and breeding data for the family [29].

Peach was one of the first sequenced species in this group [30], and a reviewed version of the database (Peach v2.0) was recently made available [31]. The generated information was obtained from the doubled haploid cultivar ‘Lovell’, showing a relatively small genome size [265 Mb; diploid 2n = 16]. In addition to its short juvenile period (2–3 years) and self-pollination capacity, this condition makes this species a useful model for the Rosaceae family [32].

Genome assemblies for sweet cherry have been generated since 2017 for ‘Satonishiki’ [33], ‘Karina’ [34], and ‘Big Star*’ [35]. These works showed a slightly bigger genome size [353 Mb; diploid 2n = 16] compared to peach, although with a greater synteny between both genomes.

Just recently, a chromosome-level assembly for Japanese plum (P. salicina) has been described [36], revealing an intermediate genome size [284 Mb; diploid 2n = 16] relative to the other two species. Phylogenetic analysis showed P. salicina having a close relationship with the other two already sequenced species, P. mume [37] and P. armeniaca [38].

The availability of these genome drafts in Prunus enables advancement toward dedicated bioinformatics tools to carry out faster and safer analyses regarding the GE technology.

Targeted mutagenesis by CRISPR/Cas9 involves making guide RNAs (gRNAs) that target customized sequences in the genome of a cell to direct the Cas9 nuclease activity to generate double-strand DNA breaks close to the gRNA-joining location. Experimentally, this gRNA is a short synthetic RNA composed of a) a scaffold sequence that gives a secondary structure that is necessary for Cas9-recognition and assembly and b) a user-defined 20-nucleotide sequence (20-nt) spacer that determines the genomic target to be modified. Consequently, the target sequence recognized by the spacer will be a protospacer sequence contained in the target genome, located right next to a motif recognized by Cas9 for DNA cleavage, which is an NGG nucleotide arrangement called protospacer adjacent motif (PAM). Given its impact, the advent of CRISPR/Cas9 technology has led to the discovery and characterization of new nucleases recognizing different PAM motifs and even to Cas9 genetic engineering to develop modified enzyme activities [39].

Whereas the RNA scaffold sequence in the gRNA is operatively a regular, non-changing component in the system, the 20-nt spacer will be the key molecule to deduce from any available genome draft. As shown in Figure 4, from the genome information, it is possible to acquire the number of NGG motifs for each species, and from these data to consider the 20-nt adjacent to each one of these motifs, creating a database for all the putative gRNA + Cas9 recognition site in each genome. This database will represent all the recognition sites for the CRISPR/Cas9 system in each genome and will allow us to predict both the on-target as the possible off-target sequences. As known for most of the genome information, several drafts still need to be confirmed due to sequencing errors or uncertainness and that information is filled with unknown nucleotides (“N”), information that will be removed from its corresponding protospacer databases.

Due to the probability of having nucleotide mismatches at the time of recognition between the gRNA and the protospacer, we can expand the 20-nt + PAM databases to consider the occurrence of mismatches, starting from one mismatch to as many pairing errors as we want to analyze. These expanded databases will increase the number of candidate off-targets in a particular genome, improving the predictive power of our ex-ante analysis for the right gRNA selection.

3.2 Vectors for gene editing

Despite its novelty, the delivery to the cell of CRISPR/Cas9 relies on regular and efficient gene transfer technology. The status of gene transfer procedures in tree fruit crops has been recently reviewed, including GE [5]. That work stressed the additional problems in using GE technologies in these species, considering their clonally propagated nature. In fact, the use of stable transformation in these species for gRNA + Cas9 delivery into the cell (explants) is not suitable because, unlike annual crops for which the transgenes can eventually be bred out after the mission of GE is achieved, breeding out these transgene sequences is a difficult, labor-intensive, and frequently time extensive process. Also, the use of crosses could lead to eliminating the identity and valuable traits of the variety.

For these reasons, efforts to improve the delivery of the editing components need to be considered in the case of fruit tree crops. One of the more convenient procedures to overcome these inconveniences has been the use of non-transgene-involved GE by using a gRNA-Cas9 ribonucleoprotein complex coupled to protoplasts gene transfer and regeneration. This has been established as proof of concept in apple and grape cells with no final plant generation [40]. In this regard, massive use of this approach will depend on developing plant regeneration procedures from protoplast cells, which in tree fruit crops is quite limited [5].

Improvements in the delivery of GE tools have already been achieved by autonomously replicating viral vectors [41]. The single-stranded DNA (ssDNA) replicons from geminiviruses having been particularly useful. This group represents a large family of plant viruses with small (2.5–5.5 kb) genomes that replicate by rolling circle replication (RCR) in the plant cell nucleus. This process occurs through a double-stranded replication intermediate that also serves as a template for transcribing the viral open reading frames [42]. Bean yellow dwarf virus (BeYDV), the smallest member in the family, has been extensively explored as a molecular tool to improve gene expression by generating disarmed versions referred to as “LSL” (LIR-SIR-LIR) [43] or “deconstructed virus” vectors [44]. The three main elements from the virus replication machinery are retained in the LSL vectors, allowing virus replication by RCR to be emulated, thus enabling the transcriptional activation of the included expression cassettes. Two of these elements act in cis, the long and short intergenic regions (LIR and SIR, respectively) [45]. The SIR is the origin of replication for minus-strand synthesis and contains transcription termination signals. The LIR contains bidirectional promoter elements and provides a stem-loop structure, which is essential for initiating the RCR of the plus-strand [46]. The third element corresponds to the virus replication initiator protein (Rep/RepA), which acts in trans [47] and, therefore, must either be expressed by the viral replicon itself or be externally provided [43, 46].

In gene transfer experiments, the LSL vectors are activated by Rep’s nicking function acting on the LIR components arranged in tandem in the delivered plasmid, which results in the replicative release of the recombinant viral DNA cloned between them [43]. The released DNA then replicates episomally in the nucleus, leading to the efficient expression of the encoded genes [48]. The shuttle capability of LSL vectors was demonstrated when T-DNAs released BeYDV-derived replicons in Nicotiana benthamiana cells, allowing for the expression of exogenes, including the green fluorescent protein (GFP) and human vaccine antigens for human papillomavirus and human immunodeficiency virus [49]. More recently, BeYDV-derived replicons have been used in T-DNAs in tobacco [41], tomato [47], and potato [50] to efficiently deliver plant genome editing machineries, such as zinc-finger nucleases, transcription activator-like effector nucleases, and CRISPR/Cas9.

3.3 Building a universal LSL-based vector for gene editing in plants

Considering our weakness in establishing protoplast systems in tree fruit crops, and particularly in Prunus, we decided to expand the geminivirus technology by designing and building general vectors that allow for efficient gene transfer experiments in these species that could lead to eventually edited individuals.

Formerly, we built a universal version of a BeYDV-derived LSL vector by assembling all the fragments containing the important components of a DNA replicon, as shown in (Figure 5, “Agrobacterium T-DNA + LSL components”). In this vector, named pGMV-Universal (pGMV-U), we incorporated all the elements required for both viral RCR replication and CRISPR/Cas9 gene editing. These include pGMV-U components arranged into multiple gRNA expression cassettes, allowing the individual expression of up to four gRNAs, the sequence comprising a SIR element, the coding sequence for Rep/RepA, and the LIR sequences adjacent to the right and left T-DNA borders, respectively. Finally, the vector included a Cas9 expression cassette (Figure 5, “pGMV-Universal vector”). Overall, a 15,657 bp vector for plant gene-transfer experiments was made (Addgene #112797).

In a second instance, considering that selection in the gene transfer process is a very limiting decision step in tree fruit regeneration processes, we improved pGMV-U into a traceable version of it by adding a GFP expression cassette (Figure 5, “pGEF-Universal LSL vector”). This vector has been named pGEF-U and retains all the previous functions for a geminivirus-based GE vector in plants.

4.1 An example: editing of the TERMINAL FLOWER 1 gene in peach

As perennial trees, Prunus spp. are grown in temperate climates. In the early fall season, the decrease in temperature and daylight hours promotes entry to the dormancy state, an evolutionary phenomenon relevant for survival during the adverse conditions in winter. In dormancy, certain requirements must be completed to resume growth. The first requirement corresponds to cold accumulation during winter (endo-dormancy) and then heat accumulation at the beginning of spring (eco-dormancy). After an adequate cold accumulation, flowering takes place.

Plants continuously sense the environment (i.e., day and night conditions; seasonal cycles) to adapt their metabolism, growth, and development to these diverse conditions. These responses involve changes in gene expression programs to overcome a particular scenario until new conditions demand new responses. Diverse gene programs have been associated with dormancy and flowering signals and include cell cycle regulation, light perception, hormonal signaling, and stress response [52, 53, 54]. Two important genes participating at some level in these events are FLOWERING LOCUS T (FT) and TERMINAL FLOWER 1 (TFL1). Despite FT and TFL1 share a high (∼60%) amino acid sequence identity, they function in an opposite manner [55]. Whereas FT promotes the transition to reproductive development and flowering, TFL1 represses this transition [56]. As already mentioned, transgenic European plum over-expressing the poplar FT1 gene has led to altered dormancy requirement and continuous flowering, enabling what has been named “FasTrack” breeding technology [6]. On the other hand, the flowering repressive function of TFL1 has been reported in many species of Rosaceae, including Prunus spp. [57, 58, 59, 60]. In pear, silencing of PcTFL1 and PcTFL2 through RNAi resulted in individuals that started flowering as early as 4–6 months, with no adverse phenotypic effects [61]. Consequently, TFL1 is an appealing candidate for producing fast-breeding trees through loss-of-function experiments in stone fruits using CRISPR/Cas9 edition.

4.1.1 Design and strategy

The TFL1locus in Prunus persica comprises four exons and three introns and spans through 1.3 kb. We chose to simultaneously cleave two different located in exons 2 and 3, removing approximately half of the gene from the genome as a result (Figure 10). Guide gRNAs were selected using the CRISPR-Search tool (freely available at https://www.fruit-tree-genomics.com/). This tool, generated by our group, searches for the best pair combinations of gRNAs to induce the deletion of a defined target area within a genomic sequence. As mentioned, one of the main characteristics of our pGMV-U and pGEF-U vectors is their cargo capability for up to four different gRNAs. Thus, both chosen gRNAs targeting TFL1 were cloned into a single vector, named pGEF-TFL1. The “loss of a large gene fragment” approach has two distinct advantages: first, it allows us to maximize the chances of successfully inactivating our target genes by avoiding silent mutations, and second, it enables the screening of deletions through a straightforward PCR assay (Figure 10A and C).

We show results from the molecular analysis of embryo explants from Prunus persica transformed with the geminivirus-based GFP-Cas9 double gRNA vector pGEF-TFL1 (Figure 10B). Twenty-eight days post-infection, infected and uninfected embryo explants were processed for genomic DNA purification. DNA samples were analyzed through PCR, using primers located upstream from gRNA1 and downstream from gRNA2 (Figure 10A). As a result, we found an array of unedited and edited amplicons: the bigger PCR products, with a size around 950 bp, corresponding to the unedited form of the PpTFL1 gene (Figure 10C). The second group, comprised of shorter amplicons with varying sizes close to 380 bp, originated from the deletion of the DNA segment between the two gRNAs. The cloning and sequencing of these amplicons confirmed the successful edition of PpTFL1 (Figure 10D). These explants have already regenerated into several new individuals (Figure 10E) whose genotypic and phenotypic characteristics remain to be assessed.