DNA is easy to work with, stable, and has been the focus of delivery methods for transgenics for decades. The first studies to demonstrate CRISPR gene editing in plant cells delivered DNA (Figure 2A) (Nekrasov et al., 2013; Shan et al., 2013; Li et al., 2013; Feng et al., 2013; Xie and Yang, 2013; Miao et al., 2013). Historically, plant cell delivery mechanisms focused on delivery of DNA; thus a wide range of options is available [reviewed in Ghogare et al. (2021) and Kausch et al. (2021)]. Some are described below in the section on vehicle choice. Some of these methods can only deliver DNA in a particular form, such as a single-stranded DNA excised from a circular plasmid, while others can accommodate variations in stranded-ness, length, and chemical modifications.

Most often, CRISPR reagents are prepared on a circular dsDNA plasmid using common cloning methods in E. coli. The nuclease and gRNA components can be prepared on separate plasmids, an approach which offers ease of mixing-and-matching. Alternatively the nuclease and gRNA(s) can be on a single plasmid, ensuring that they are co-delivered. Each can be expressed from its own promoter, or a single transcript unit can be used to express the nuclease and gRNA(s). When multiple gRNAs are needed, these can be expressed from separate promoters (often pol-III promoters such as pU3 or pU6), or from a multiplex array with tRNAs, self-cleaving ribozymes, or Cas12a-proccesed crRNAs expressed from pol-II promoters (such as pUBI). For a review of plasmid design considerations, see (Hassan et al., 2021). The purified plasmid(s) can be delivered using several of the vehicles described below in the section on vehicle choice. Alternatively, the plasmid can be delivered to Agrobacterium or a virus, which then deliver a portion of the plasmid to the plant cell (Gelvin, 2003; Liu et al., 2023). Some of these delivery vehicles are size-limited or require particular sequence elements, so it is important to consider the delivery vehicle during the design of the plasmid.

DNA is also a very useful cargo to deliver DNA templates for homology-directed repair (Huang and Puchta, 2019). These templates can also be on a circular dsDNA plasmid, or they can be delivered as a linear dsDNA PCR product or a ssDNA. One advantage of non-plasmid DNA is to allow chemical modifications that can aid the probability of homology-directed versus random blunt-ended integration. If a DNA template needs to be delivered, this may influence the decision towards also delivering nuclease and gRNA as DNA so that the reagents can be combined and co-delivered. However, it is not a requirement, and DNA templates can be co-delivered with mRNA and/or ribonucleoproteins.

Additionally, consider whether the process will employ a transgene as a visual marker, selectable marker, or additional trait. These transgenes could be delivered as DNA while the CRISPR reagents are delivered as mRNA or ribonucleoproteins (Figure 2B). Alternatively, it may be practical to deliver CRISPR reagents in the same DNA format, most often on the same plasmid (Figures 2A, C).

When reagents are delivered as DNA, the reagents can be expressed from their extra-chromosomal form. This is often referred to as “transient” expression (Figure 2C). The DNA can remain in an extra-chromosomal form for minutes, hours, or days (Tyurin et al., 2020). At some point after reaching the nucleus, a subset of the DNA reagents integrate into the genome by a variety of mechanisms that are only partially understood. If these reagents integrate into a chromosome, the transgenes will begin to replicate as part of the plant’s own genome, and if the genomic position is suitable, the transgenes will be expressed. Typically, this is referred to as “stable” expression.

The considerations for DNA delivery methods for transgene-free gene editing are different, and in fact opposite compared to transgenics. The ideal DNA delivery platform for transgenics results in a small amount of extra-chromosomal DNA, of which a large proportion integrates at low-copy. The ideal DNA delivery platform for gene editing results in a large amount of highly expressing extra-chromosomal DNA, of which a very small proportion (preferably none) integrates. Note that in scientific literature the term “transient” is often used as a proxy for extra-chromosomal and the term “stable” is used as a proxy for chromosomally integrated transgenes (Kim et al., 2010). Integration of extra-chromosomal DNA is essentially a one-way process, therefore researchers who are hoping to avoid making a transgenic plant using DNA-based delivery usually seek to avoid integration in the first place. However, as described below in the section on full transgenic selection, a transgene can be removed through either excision or through segregation during meiosis, and either option could be followed with selection against the transgene.

The main advantages of DNA cargo are the flexible in-house design of the reagent, DNA’s stability, the ability to deliver greater amounts over time by virtue of transcription and translation, and compatibility of DNA with a wide range of delivery vehicles. The main limitations of DNA cargo are that it must be transcribed and translated by the host cell, and, for transgene-free applications, DNA reagents present a challenge due to their propensity to integrate into the plant’s genome.

The clearest way to avoid transgenes is to employ delivery methods without exogenous DNA (Tsanova et al., 2021; Poddar et al., 2023), often termed less precisely as “DNA-free” methods (Figure 2D). In practice, the available cargo is either RNA or protein. Since these reagents are only available for a short time, both theorizing and empirical evidence point to a lower likelihood of off-target editing or unwanted reagent integration into either the targeted locus or random locations in the genome (Stoddard et al., 2016; Liang et al., 2017).

Delivery of gRNA and mRNA encoding a nuclease to plant cells has been reported by only a few groups, suggesting this method remains challenging and non-trivial to reproduce (Stoddard et al., 2016; Zhang et al., 2016; Imai et al., 2020). However, mRNA-based delivery is mainstream in mammalian tissue culture, as demonstrated by suppliers offering off-the-shelf mRNA versions of a range of gene editing reagents. Since RNA is a nucleic acid similar to DNA, many of the same delivery methods can be used, including protoplast transfection, gene gun particle bombardment, and appropriately selected nanocarriers, as described below in the section on vehicle choice. As illustrated in Figures 1A, B, delivery of mRNA means the reagents do not require cellular machinery for transcription, however, the delivery dose may need to be adjusted to account for the loss of amplification by transcription. For example, the Voytas lab found that expression of a fluorophore from mRNA was barely visible and editing by TALENS delivered as mRNA was ∼12-fold lower than with DNA delivery (Stoddard et al., 2016). Researchers must also consider that RNA is typically less stable than DNA, and that the gRNA must “wait” for the mRNA encoding nuclease to be translated into protein (Zhang et al., 2016). If synthetic gRNA is used, it can be chemically stabilized in a variety of ways (Kelley et al., 2016; Allen et al., 2021; Rozners, 2022). mRNA can be stabilized through the design of stabilizing 5′ and 3′ UTRs, 5′ caps, and appropriate poly-A tails (Stoddard et al., 2016). One reported advantage of RNA-based delivery is the observation of significantly fewer random insertions compared to a DNA-based plasmid (Stoddard et al., 2016).

The main advantages of RNA cargo are avoiding DNA, but still employing a nucleic acid. Since RNA reagents are typically produced from a plasmid, this method offers the same benefits of customizability, although the in-vitro transcription must be performed separately for each new reagent. RNA-based delivery has an extremely low risk of integrating reagents into the genome. Another advantage is the ability to deliver modified bases, for example, to target polymorphic sequences (Krysler et al., 2022). Due to its chemical similarity to DNA, several delivery vehicles that were designed for DNA can be adapted to RNA with only minor modifications. The main limitations of RNA cargo are that RNA is less stable than DNA and that RNA is typically synthesized or produced by in-vitro transcription, which is an additional expense compared to DNA delivery. In addition, the mRNA lacks the amplification step of transcription, so it is likely that a higher dose must be delivered for a comparable effect to DNA delivery.

Functional gene editors are complexes of gRNA and protein (ribonucleoproteins, RNPs, Figure 2), and many groups have demonstrated success in delivering these “ready-to-edit” reagents [Svitashev et al., 2016; Liang et al., 2017; Poddar et al., 2023 and many others, reviewed in Zhang et al. (2021)]. RNPs require neither transcription nor translation machinery from the host cell, although this does mean that a sufficient quantity must be delivered. The complexes are easily assembled in a simple aqueous buffer at ambient temperature in a matter of minutes, requiring nothing more than pipetting. The protein can be produced and purified in-house or obtained commercially. While purification of nuclease proteins is relatively routine, it is not yet trivial to obtain nuclease fusions (e.g., base editors, prime editors, etc.—see Figure 1C) thereby presenting a current limitation to the use of this technology for applications beyond the creation of random indels. gRNAs can be produced by in-vitro transcription in-house or commercially synthesized. While RNA is generally less stable than DNA, gRNAs are less prone to premature degradation when they are protected within the nuclease. Nevertheless, RNA modifications can be employed to stabilize the gRNAs (Poddar et al., 2023). Ribonucleoproteins can be useful in rapid screening to identify the best gRNA in a transient protoplast assay, for example, because new gRNAs can be purchased and tested without time-consuming steps to clone new gRNAs into plasmids.

The delivery vehicles for proteins are somewhat limited compared to those for nucleic acids. Particle bombardment into intact plant cells and PEG/Ca²⁺-mediated transfection into protoplasts are most common, although other methods are described below in the section below on vehicle choice. The protein component is more sensitive to denaturation, which means that the prepared RNPs must be maintained within a narrower range of temperature, pH, salinity, etcetera compared to DNA and RNA.

The main advantages of ribonucleoprotein cargo are that these reagents are ready-to-edit and that this is an approach entirely free of exogenous DNA, thus without risk of transgenes. The main limitations of ribonucleoprotein cargo are the difficulty in obtaining the purified protein of a custom gene editors depicted in Figures 1F, G and delivering a non-denatured RNP in sufficient quantity using a limited number of delivery vehicles.

The reagents must be matched to the desired outcome of gene editing. For simple single-locus knockout editing by double-strand break and non-homologous end joining, it is sufficient to deliver the appropriate nuclease with the appropriate gRNA, which can be accomplished in a DNA-free way using ribonucleoproteins (Woo et al., 2015; Svitashev et al., 2016). For larger deletions or when navigating genomes with polymorphisms between copies of a gene, a nuclease and multiple gRNAs may need to be delivered (Eid et al., 2021). Similarly, for base editing, the appropriate base editor and gRNA must be delivered. In all designs employing Cas9 and similar nucleases, one must consider whether a crRNA + tracrRNA complex or a sgRNA is more appropriate and cost-effective to synthesize. Avoiding double strand breaks offers an advantage in product purity—i.e., resulting alleles are more likely to be identical to each other rather than a mixture of small deletions/insertions/substitutions (Xie et al., 2022).

For more precise sequence replacement, an appropriate template must be co-delivered. If the template is in the form of DNA, this may influence the decision whether the remaining reagents are also delivered as DNA, or whether advantages remain to DNA-free delivery methods. Several characteristics of the DNA template can influence the choice of delivery method. Is the DNA template double-stranded or single-stranded? How long is the DNA template? Is the DNA template part of a larger molecule, such as a plasmid, or is the template its own fragment of DNA? Does the DNA template get excised from a plasmid using the CRISPR nuclease or another mechanism?

Templated editing can also be achieved without delivery of DNA template. Base editing enables a narrow range of sequence replacements without the need for a nucleotide template (Komor et al., 2016; Kantor et al., 2020). Prime editing allows the template to be delivered as RNA, typically in the form of an extension of the gRNA (Anzalone et al., 2019). Prime editing enables any substitution, precise deletions, and insertions up to 40 or even 80 nt using single or paired prime editors. Prime editors combined with an integrase enable much larger insertions, although these require the desired insert to be delivered as DNA which could possibly integrate in off-target locations (Anzalone et al., 2022; Yarnall et al., 2022). For a review of prime editors and base editors in plants, see (Xie et al., 2022; Azameti and Dauda, 2021).

To date, a wide range of gene editing reagents have been demonstrated using plasmid-based methods, but only a subset has been purified as ribonucleoproteins and demonstrated to function in plants. If considering ribonucleoprotein-based delivery, it is important to first establish whether the desired engineered nuclease, base editor, or prime editor is commercially available as protein or can be produced in-house. The scarcity of commercially available purified fusion proteins implies that the generation of transgene-free plants resulting from these approaches may depend on improvements of DNA-based delivery.

Several of the delivery vehicles described below have limited cargo capacity and may be unable to deliver an entire SpCas9 protein, mRNA, or DNA. This limitation is very familiar to researchers in the medical and mammalian fields, who are most concerned about size limitations to AAV-based delivery, thus much progress has been made on this topic recently. For context, the coding sequence of LbCas12a is >3.5 kb, SpCas9 is >4 kb, and the first published prime editors and base editors are typically >6 kb, in each case excluding any promoter or other expression element. Several small nucleases have been published with coding sequences as small as ∼1.5 kb. Full elaboration on the recent advances in small nucleases is beyond the scope of this review, however, the following reviews and publications may be helpful (Wu et al., 2021; Wang et al., 2022a; Wang et al., 2023; Al-Shayeb et al., 2022; Goltsman et al., 2022; Li et al., 2023; Liu et al., 2019a; Pausch et al., 2020).

Today, Agrobacterium-mediated delivery of T-DNA is widely used in many species to achieve consistent, low-copy transgene insertions and to deliver GE reagents encoded in transgenes. These methods take advantage of a pathogen’s ability to bypass the cell wall. Since this method’s beginnings in the 1970s, various protocol improvements have expanded the host range of Agrobacterium: tissue pre-treatments, culture with acetosyringone (reviewed in (Gelvin, 2003; Kausch et al., 2021)), use of supervirulent plasmids (Anand, 2018), assembly of very large T-DNAs directly in Agrobacterium (McCue et al., 2019), use of species other than A. tumefaciens (Cao et al., 2023), blocking host defense factors using the Pseudomonas Type III secretion system to deliver effectors (Raman et al., 2022), and using a variety of morphogens to improve regeneration of transformed tissue (Lowe et al., 2016; Gordon-Kamm et al., 2019; Hoerster et al., 2020; Debernardi et al., 2020; Che et al., 2022).

Agrobacterium delivers a single stranded DNA, called a T-DNA, excised from a plasmid and defined by Left Border and Right Border sequences. The mechanism uses several virulence proteins, in particular VirD2 and VirE to excise the T-DNA from the Ti plasmid, protect its ends and length, and pull it into the plant cell’s nucleus—VirD2 has a strong nuclear localization sequence. The multiple components of gene editing, such as the nuclease, guide RNAs, and repair template(s) are typically cloned into a single binary plasmid, along with any visual or selectable marker(s). The resulting plasmids can be quite large. If the relative amounts of any components should be optimized, this must usually be done through choice of expression elements, as the T-DNA is transferred as a single unit. T-DNAs ranging from 5 to 30 kb are used routinely, but larger T-DNAs, especially those in excess of 100 kb are transferred only at very low efficiency, reviewed in (Xi et al., 2018). DNA with synthetic chemical modifications, which might be desirable for a homology-directed repair template, cannot be introduced and maintained in the T-DNA region of the plasmid from which Agrobacterium transfers DNA into a plant cell.

Recently, modified strains of Agrobacterium with a Pseudomonas type III secretion system (PT3SS) have been demonstrated to deliver the protein AvrPto and fragments of GFP (Raman et al., 2022), reminiscent of how TALEs are delivered by the type III secretion system from Xanthomonas (Bogdanove, Schornack, and Lahaye, 2010). While particle bombardment (described below) can deliver ribonucleotide complexes, it is not clear whether the modified Agrobacterium would also deliver RNP complexes, or would need to deliver the protein and RNA components separately.

The main advantages of Agrobacterium as a vehicle are its ability to transfer an easily customized piece of DNA (and possibly a customized small protein) into a plant cell and that Agrobacterium protocols have been used and improved upon for decades. Undeniably, Agrobacterium is a powerful tool for low-copy integration of transgenes. The main limitations of Agrobacterium as a vehicle are the limited host range (many cultivars and species remain recalcitrant to Agro-transformation), requirement to clone relatively large binary plasmids, and that protocols are generally optimized for low-copy integration of transgenes, not for transgene-free gene editing outcomes.

Particle bombardment uses force to bypass the cell wall, which, while crude, means this method works for almost any species and cultivar. In addition, particle bombardment is a flexible technique where the desired protein, RNA, or DNA (circular or linear, double- or single-stranded) is randomly and non-covalently attached to millions of microscopic gold (or tungsten) microcarriers (0.6 μm and 1.0 μm diameter particles are common). Reagents are typically applied to microcarriers in an aqueous solution and precipitated using spermidine/CaCl₂/PEG, glycogen, or a cationic lipid reagent (e.g., TransIT-2020) (Ratnayaka and Oard, 1995; Svitashev et al., 2016; Imai et al., 2020; Miller et al., 2021; Dong et al., 2021). The particles provide mass and are accelerated with pressurized helium in a gene gun, thus providing sufficient force to overcome the cell wall. A very small fraction of these particles results in successful delivery of reagents to a plant cell which survives. The reagents must detach from the particle in the target cell compartment, often the nucleus, or be transported into it.

A frequently cited downside to particle bombardment is that the process damages DNA, causes occasional chromosome rearrangements (Yue et al., 2022), and creates DNA breaks into which the reagent DNA (or carrier DNA) may integrate randomly (Liu et al., 2019b). In addition, the loading of DNA onto particles may be uneven (i.e., many particles contain no DNA, while others contain many), and high copy tandem insertions of transgenes are not uncommon (Poddar et al., 2023). These disadvantages, combined with improvements in Agro-transformation techniques, led to a decrease in the popularity of particle bombardment for purposes of making transgenic plants.

A key reason for the revival of particle bombardment is that gold microcarriers can deliver protein as well as nucleic acids (Martin-Ortigosa and Wang, 2014; Miller et al., 2021). The particle bombardment protocol must be modified to prevent denaturing the gene editor RNPs, which can tolerate desiccation, but not suspension in pure ethanol (Svitashev et al., 2016). Several protocols for delivery of CRISPR RNPs have been published by both academic and industry groups (Svitashev et al., 2016; Liang et al., 2017; Poddar et al., 2023). While these protocols are reproducible, delivery of proteins (and ribonucleoproteins) by particle bombardment has not been optimized to the same extent as delivery of DNA (Ratnayaka and Oard, 1995; Martin-Ortigosa and Wang, 2014; Miller et al., 2021). For example, “no studies to our knowledge have been carried out to quantify the amount of Cas9-RNP that is truly adsorbed to gold particles by these methods. Such data would be valuable to avoid underloading or overloading gold particle preparations with Cas9-RNPs.” (Poddar et al., 2023). While studies from mammalian tissue culture indicate that transfected RNPs are degraded within <48 h of delivery, one recent report suggests that traces of RNPs delivered to plant cells by bombardment remain detectable, and possibly even active, for as long as 14 days (Poddar et al., 2023).

Another major advantage of particle bombardment is that this method works for a much wider range of species, cultivars, and explants, because it relies on force rather than a “clever” pathogen. Bombardment is typically used with immature embryos, callus, or explants which are regenerated through callus. In addition, in planta particle bombardment, targeting the embryonic shoot apical meristem from dissected imbibed wheat seeds, was published as an alternative, and used to deliver transgenes encoding CRISPR reagents (Hamada et al., 2017; Imai et al., 2020). Presumably a similar approach is possible for delivering RNPs in a DNA-free way, since the published approach does not rely on selection using transgenes.

The main advantages of particle bombardment as a vehicle are the ability to deliver anything that can be coated onto microcarriers and the ability to reach an exceptionally wide range of explants, cultivars, and species. The main limitations of particle bombardment as a vehicle are due to the imprecision of loading of reagents onto microcarriers and potential for damaging the genome through force.

Instead of bypassing the cell wall by force or using Agrobacterium to deliver DNA, several teams have borrowed another plant pathogen—viruses—to bypass the cell wall in a method called virus-induced gene editing (VIGE). This approach is analogous to AAV-mediated delivery to mammalian cells. The requirements for viral delivery of gene editing reagents to plant cells are host range and being able to fit the gene editing cargo into the genome of the virus. In some cases, the virus can fit a gRNA, but the nuclease must be expressed in the plant as a transgene, meaning that those methods are not exactly relevant to this review. For a recent succinct and illustrated review of viral delivery methods, see (Laforest and Nadakuduti, 2022). The benefits of viral vector delivery of editing reagents are two-fold. First is the possibility to produce heritable edits without the need for tissue culture, such as in (Li et al., 2021), and second is the ability to produce edited plants which do not contain the DNA encoding the editing components in their genomes.

Briefly, the approach involves delivery of the viral genome on a T-DNA vector via Agrobacterium infiltration of leaves on an intact plant. The T-DNA can integrate locally at the site of infiltration, but when expressed from the T-DNA, the viral genome containing CRISPR reagents can form viral particles and move systemically through the plant. Editing can thus theoretically occur from the expressed viral genome in many plant cells, which do not contain the DNA encoding the editing reagents. If meristematic cells are also “infected” with the mobile viral genome, editing in these cells can be propagated to embryos, thus resulting in a transgene-free edited plant obtained without the need for tissue culture. Alternatively virus-encoding Agrobacterium can be infiltrated into an intermediate host, such as Nicotiana benthamiana, to produce viral particles, which can be purified and applied to the target plants.

Though promising, this approach has five main challenges. The first challenge is the limitation on the size of a modification which can be tolerated within a typical plant viral genome. In most cases reported to date, the cargo capacity of compatible plant viral genomes is ∼1 kb, while the size of a typical Cas protein is ∼4 kb, and that of a Cas9 plant expression cassette is >6 kb. Thus, most studies have focused on viral delivery of gRNAs to a transgenic plant expressing a Cas9 protein. This approach is promising for species for which the elimination of tissue culture is the major benefit of this system, since Cas9 can be crossed out in subsequent generations. However, for species with vegetative propagation, this approach presents a limitation. At least one study reports successful delivery of an entire CRISPR-Cas cassette to tetraploid tobacco using a plant virus belonging to the Rhabdovirus family of negative strand RNA viruses with a larger cargo capacity (Ma et al., 2020). Additionally, in this case, the virus could be delivered directly to the plant without the need for an Agrobacterium vehicle. Despite this success, the challenge of engineering novel plant viral vectors such as this can limit the adoption of this approach. Additionally, due to host-range limitations of most viruses, additional viral genome engineering may be required to expand the range of species which could benefit from this approach, further deepening the technical challenge of its use. Another approach to overcoming the cargo size limitation is the use of distinct or satellite viruses (Maher et al., 2020). A promising approach to overcome the cargo size limitation of more commonly used viral genomes is to use tiny Cas proteins so that the entire CRISPR-Cas cassette would “fit” within 1 kb. Multiple Cas proteins less than one quarter the size of typical Cas9/Cas12a proteins have now been described (see previous section on cargo size). The approach of combining tiny Cas proteins with the viral delivery approach is promising for delivery both tissue culture-free and transgene-free benefits of the system to vegetatively propagated species.

A second challenge to the viral delivery of editing components is the extent of systemic movement, which directly translates into the efficiency of obtaining the desired edited plants. One successful approach to increase the cell to cell mobility of sgRNAs has been the addition of short sequences in the 3′ of Cas9 gRNAs such as tRNA motifs and FLOWERING LOCUS T (FT)-like motifs (Ellison et al., 2020; Li et al., 2021).

The third limitation is the fact that often the choice of an efficient viral vector is specific to a particular group of plants (often monocots vs. dicots), requiring the testing and optimization of viral vectors directly in the crop species or its relatives. For example, although some of the foundational research for viral delivery was done in Nicotiana species using either positive strand DNA Geminiviral or positive strand RNA viruses like tobacco rattle virus, for grass crops such as maize and wheat, foxtail mosaic virus (+RNA) and barley stripe mosaic virus (-RNA) have been used. For example, foxtail mosaic virus has been used to deliver CRISPR reagents to maize and Setaria (Mei, Beernink, et al., 2019) and sugarcane mosaic virus, which can tolerate insertions of at least 1809 nucleotides has been used to deliver GFP, Gus, and Bar into maize (Liu, et al., 2019). Recently TSWV, a virus with broad host range in dicots and even a few monocots (Cha and Mau, 1987) was reported to deliver full-sized Cas9 reagents (Liu et al., 2023).

Fourth, because of the non-targeted nature of systemic viral movement, obtaining a non-chimeric plant necessitates passage through a single cell bottle neck such as in seed formation, making this a challenging approach for vegetatively propagated crops, complex hybrids or crops with long generation times.

Finally, fifth, the absence of virus or viral genome from the resulting plants is crucial from a crop product perspective. Many viruses are excluded from the germline (single-cell-bottleneck), largely alleviating this concern. In some cases, the virus can be modified to be non-insect-transmissible and curable (Liu et al., 2023). In other cases, antivirals can be used to cure the viral infection after editing, although this may not be especially challenging in vegetatively propagated crops (Voytas, 2023).

The main advantages of viral vectors as a vehicle are the potential for transgene-free, and perhaps also tissue-culture free delivery of gene editing reagents. The main limitations of viral vectors as a vehicle are an especially narrow host range for each particular virus and a cargo size capacity typically too small to contain the entire expression cassette for gRNA(s) and a nuclease.

Many methods besides the gene gun and pathogen-based vectors have been tested, but to date none are widely adopted for use with intact plant cells. When the cell wall is removed to form a plant protoplast, the delivery vehicle only needs to bypass the cell membrane. In plant protoplasts and mammalian cell culture, lipofection, lipid nanoparticles (LNPs), electroporation, and PEG/Ca²⁺-mediated transfection can be used to deliver proteins, RNA, or DNA (Woo et al., 2015; Murovec et al., 2018; Andersson et al., 2018; Furuhata et al., 2019; Wu et al., 2020; Wang et al., 2020; Badhan et al., 2021). In the species where protoplast regeneration is practical, such methods may offer a convenient and extremely efficient method for transgene-free gene editing. Unfortunately, in most species, protoplast regeneration is not straightforward. An alternative explant that may be able to bypass the plant cell wall is a zygote shortly after gamete fusion, which can be transfected using similar methods as protoplasts (Toda et al., 2019). However, researchers are often limited to working on explants with mature cell walls.

Various nanomaterial-mediated delivery options have been published to overcome the plant cell wall. Nanoparticles offer entry into plant cells without the use of crude force (unlike particle bombardment) and thus hold the theoretical ability to provide tissue culture free delivery of reagents. However, reagent delivery is still limiting due to the custom chemistries required to reversibly bind different reagents to the nanoparticles. Single wall carbon nanotubes can deliver transient expression of plasmids up to ∼10 kb, including gene editing reagents, and siRNA (Zhang et al., 2022; Demirer et al., 2020, 2019). For a review of nanotube and nanoparticle methods, see (Demirer et al., 2021). Clay nanoparticles can deliver RNA into intact plant cells (Yong et al., 2022, 2021). While not peer-reviewed, carbon nanodots have been reported to deliver plasmid DNA to intact plant cells (wheat leaves) which expressed fluorophores and demonstrated editing. (Doyle et al., 2019). DNA nanostructures have been reported to deliver siRNAs and may be adaptable to delivering gene editing reagents (Zhang et al., 2019a). Another recent report describes transformation of maize pollen using DNA-coated magnetic nanoparticles (Wang et al., 2022b). Nanodelivery of RNPs has not yet been demonstrated in intact plant cells, but has been shown in mammalian cells (Wang et al., 2016; Mout et al., 2017; Mout et al., 2017; Lee et al., 2017; Wei et al., 2020). For a review on protein delivery to intact plant cells, see (Wang et al., 2021). For reviews of nanocarrier delivery methods including carbon nanotubes, carbon dots, and mesosporous silicon nanoparticles, see (Kumari and Pratap Singh, 2021; Mujtaba et al., 2021), and for a review focused on gene editing, see (Ahmar et al., 2021).

Besides nanomaterials, cell penetrating peptides, which employ endosomal entry and escape mechanisms, have been demonstrated to deliver DNA and proteins into plant cells after infiltration into leaves in dicot plants (Ng et al., 2016; Chuah and Numata, 2018; Terada et al., 2020). Combinations of cell penetrating peptides have recently been reported to enable delivery of reagents to intact plant cells, possibly through a mechanism similar to micropinocytosis (Miyamoto et al., 2022). For a review of peptide-based and nano-based delivery methods for gene editing into intact plant cells, see the review (Arya et al., 2021).

The advantages and limitations of these delivery methods are difficult to generalize and their adoption is not yet widespread.

The two delivery methods in this section take inspiration from the ways in which DNA, RNA, and proteins move into intact plant cells using mechanisms found in plant biology. Both of the following methods also have the advantage of being tissue-culture-free.

The natural process of fertilization requires genetic material from two cells to fuse. The delivery method called HI-EDIT (haploid-induction editing) takes advantage of pollen as a vehicle for gene editing reagents (Kelliher et al., 2019). First, a transgenic line is made which contains the gene editing reagents and a haploid inducer genotype. This line is used as the pollen donor. The target variety is pollinated by this pollen donor, receives and expresses the gene editing reagents, and editing can take place in the maternal genome. However, fusion of genomes from the pollen and from the egg does not occur. Instead, the genome from the pollen is rejected very early, along with the gene editing reagent transgenes, and the egg develops into a haploid zygote. This creates a short window when the editing reagents can be expressed and produce an edit. The haploid genome can be doubled to produce a homozygous edited, transgene-free plant without tissue culture. Note that this method is entirely different from delivering DNA to pollen, as in (Wang et al., 2022b).

Plant tissues communicate using a variety of small molecules, but also mobile RNAs. Recently, a method was published to work in several dicot species which takes advantage of this inherent biology to move gRNA and mRNA encoding gene editing reagents into the target plants. The method relies on grafting wild-type target scions onto transgenic rootstocks which deliver mobile mRNA and gRNA (Yang et al., 2023). This method produces mosaic plants, which must pass through a single cell bottleneck such as meiosis to produce the next-generation where the edits become fixed.

The main advantage of these methods is the potential for simultaneously tissue-culture-free and transgene-free gene editing and in enabling editing of extremely transformation-recalcitrant cultivars since the donor plant can be made in an easily transformable cultivar. The main limitations of both methods are that transgenic reagent donor plants must be generated, the haploid induction and/or grafting techniques are not possible in all species or cultivars, and the process is not very efficient even in the model where it was developed.