As it was demonstrated by Wen and co-authors, the dsODN insertion looks like a NHEJ-depended process. The dsODN’s role as a donor template in therapeutic knock-in strategies is currently under discussion, mainly because it carries a risk of random insertion to the DNA breaks, induced not only by nucleases, but by replication, or other stresses, which are constantly occurring throughout the cell cycle. The general use of dsODNs is predominantly associated with tagging methods detecting a nuclease off-target cleavage, such as Genome-wide Unbiased Identification of DSBs Enabled by Sequencing (GUIDE-Seq) (Tsai et al., 2015) and its analogues (Tang et al., 2018; Miller et al., 2019; Fan et al., 2022; Quan et al., 2022). In addition, asymmetric semi-dsODNs found an application in introducing the epitope tags to study protein functions (Fang and Shinkai, 2021). Available literature suggests that the precise dsODN insertion efficiency ranges from 4.4% to 52% for human cells, depending on targeted locus, cell type, and dsODN dosage (Table 4).

As for the dsODN structure optimization, it is considered that dsDNA cytotoxity is a DNA length-dependent effect due to the activation of cytosolic DNA-sensing pathways (Yu and Liu, 2021). However, reducing the dsODN’s length does not seem to increase insertion frequency (Wen et al., 2021) and does not enhance efficiency of genome editing. Moreover, addition of phosphorothioate linkages at the ends of the DNA strands helps to stabilize ODNs in cells, ensure nuclease resistance and facilitate ligation reactions, and both 5′ and 3′ end-phosphorylated dsODNs show higher integration frequencies in comparison with the dsODNs with 5′ end protection alone (Tsai et al., 2015; Papasavva et al., 2022).

When using dsODNs as a donor template, one should take into account the low specificity of their insertion and look for the ways to improve this characteristic. For this purpose, a “Ligation-Assisted Homologous Recombination” (LAHR) GE method has been proposed as an option (Zhao et al., 2022). It is based on CRISPR/Cas12a nuclease that forms DSBs with DNA overhangs and the dsDNA donor template containing homologous arms with overhangs complementary to the ends generated by Cas12a at the desired locus. The insertion efficiency of LAHR template with 80 bp homologous arms was 3 times higher (6% knock-in) than that of ssODN template.

SsODNs represent an attractive option as an HDR donor template for small controlled mutations or insertions. Main advantages of these templates includes affordable and scalable production, relatively low toxicity and reduced risk of integration by ligation mechanisms. Usually, 200 bp ssODNs containing at least 20 bp homology arms at both ends allow to insert 160 bp into the DSB locus by SSTR mechanism (Richardson et al., 2018).

Today ssODNs are widely used in biotechnology (Saha et al., 2019), cellular and animal models generation (Remy et al., 2017), and in therapeutic applications using Fok1-based nucleases and CRISPR/Cas9 (Chen et al., 2011, 2015; Di Stazio et al., 2021). While ssODN can be co-transfected with nuclease in the form of plasmid DNA (Johnston et al., 2019) and mRNA (Guo et al., 2018), the most promising results were achieved with ssODN paired with Cas9 RNP (Okamoto et al., 2019; Schubert et al., 2021). In a comprehensive study by Schubert et al., knock-in activity mediated by Cas9 RNP with HDR efficiency was tested at 254 genomic loci in Jurkat cells and 239 genomic loci in HAP1 cells, achieving up to 50% HDR rate for selected experimental conditions and genomic loci (Schubert et al., 2021).

The selected reports on the use of ssODN for knock-in in different cell types is presented in Table 5. HSPCs represent an important cell population, clinically relevant for ssODN-mediated repair. The majority of studies performed on HSPCs are focused on optimizing the correction of the HBB locus, which offers the hope of developing treatments for TDT and SCD associated with certain point mutations at this locus. DeWitt and co-authors have screened a set of ssODN template designs with asymmetric homology arms in a complex with Cas9 RNP with achievement of up to 33% HDR at the first exon of HBB (DeWitt et al., 2016). In a subsequent work by Magis and co-authors, a knock-in level of more than 30% was achieved at HBB in LT-HSCs. The corrected cells were then transplanted to NBSGW mice model. The erythroid differentiation enrichment indicating the survival advantage of corrected alleles in vivo was demonstrated (Magis et al., 2022). In a study by Pattabhi and co-authors, the codon-optimized ssODN donor for sickle correction was compared to the respective rAAV6 donor. The efficient repair rates in vitro were achieved for both templates (ssODN, 29.6%; rAAV6, 37.5%).

Importantly, in contrast to higher in vitro efficiency of rAAV6 vector, at 12–14 weeks post-transplant into recipient NBSGW mice, a ∼6-fold higher proportion of ssODN-modified cells persisted in vivo compared to recipients of rAAV6-modified HPSCs. It can be partially explained by the toxicity of viral vectors impairing HSC engraftment (Romero et al., 2019; Lattanzi et al., 2021). Hoban et al. guessed that longer ssODN strands with prolonged homology arms were optimal for achieving a high gene correction level, inferring the homology arm reduction inefficient. Using longer reverse-strand templates with introduction of silent mutation sites to block a nuclease recleavage allowed them to increase the knock-in from 15% up to 40% (Table 5) (Hoban et al., 2015).

It should be noted that as the ssODN represents a robust platform for screening of different structural modifications and experimental conditions, the results of the cited works cannot be easily presented in concise manner and one should address the original articles for detailed analysis of the study results. The majority of the basic parameters, such as the choice of targeting (T, ODN complementary to the CRISPR–Cas9 gRNA) or non-targeting (NT) strand, length, homology arms size and symmetry, orientation, addition of mutations preventing target re-cut, sequence optimizations, nuclear localization

Sequences as well as chemical modifications was tested in reported studies with only few general considerations that could be drawed from them. The preference for utilizing a donor oligo with sequences either complementary or non-complementary to gRNA may influence the outcomes of genome editing, as T strand binding to the Cas9 may reduce overall editing efficiency. However, the results of the studies addressing this question regarding the HDR rate varied with no universal strand preference concluded to date. Schubert and co-authors in the systematical manner assessed the knock-in activity mediated by Cas9 RNP/ssODN templates at 254 and 239 genomic loci in Jurkat cells and HAP1 cells respectively (Schubert et al., 2021). Phosphorothioate-modified templates contained 40 bp HA and an insert of a six base EcoRI restriction digest recognition site. The HDR rate defined as the precise insertion of the EcoRI sequence at the canonical cut site was quantified with comparisons between templates consisting of either the T strand or the NT strand. Interestingly, no statistical difference was observed in total editing when either the T or NT strand was used in Jurkat cells, while a significant difference in editing efficiency was observed in HAP1 cells (80.2% mean editing for NT strand vs. 67.8% for T strand was used). Consequently, for HAP1 a significantly higher mean HDR rate was observed across all sites when the NT strand was used (20.6%) than the T strand (15.2%). In contrast, a significantly higher mean HDR rate in Jurkat cells was observed for T strand compared to the NT strand (11.3% vs. 7.5%, respectively). The reasons for these cell-type specific differences are currently undeciphered. The same report also included the assessments of the strand preference for the inserts shifted from the Cas9 cut loci and therefore targeted outside optimal distance from DSB (see below). In that case, for PAM-distal insertions, the NT strand donor templates gave ∼6-fold higher mean HDR rate compared to the T strand, while, for PAM-proximal insertions, the T strand gave ∼10-fold higher HDR than the NT strand. Considering these cell type and gRNA specific differences, the testing of particular templates seems to be reasonable in each case.

The analysis of the significance of overall length of ssODN template across the reported studies did not lead to conclusive results among different experimental adaptations (Yang et al., 2013; Richardson et al., 2016). Studies of Cas9 DSB formation kinetics showed that upon binding to its target, Cas9 releases the PAM-distal non-target strand, creating the rationale for designing asymmetric homology arms for sgRNAs that target the sense strand with a long 5′ homology arm and a shorter 3′ annealing arm (DeWitt et al., 2016; Richardson et al., 2016). In a study by DeWitt et al., in a set of tested ssODN donors with Cas9 RNP, the best results of HBB gene correction were achieved using a template with a 111 bp 5′ arm and a 57 bp 3′ arm, proponing longer and asymmetric homology arms to be advantageous for some experimental installments. Such designs were shown to be optimal by other groups (Di Stazio et al., 2021). Other model proposed by Liang et al. implies that after Cas9 nuclease cleaves, both sides of the double-stranded break are recognized by the DNA repair machinery equally, based on observations that both the non-target and target asymmetric HA ssODNs enhanced HDR regardless of the orientation of the Cas9 nuclease compared to a standard symmetrical donor design. The optimal ssODN donor defined having 30–35 base arms 3′- to the insertion/repair cassette and greater than 40 bases on the 5′-end (Liang et al., 2017). However, in a comprehensive study by Okamoto et al. comparing different structures of ssODN templates for single-base substitution in a reporter gene model showed that the optimal ssODN had a 30–35 bp perfectly matched homology arms on both sides (Okamoto et al., 2019). Considering these discrepancies, one standard starting point that could be employed is to design ssODN donor templates with 40-bp homology arms (Schubert et al., 2021).

CRISPR/Cas9 can re-cut the repair site after the successful insert if the protospacer and PAM sequence remained unaltered, lowering HDR efficiency (Schubert et al., 2021). This outcome can be prevented by incorporation of blocking mutations into the donor template, which was shown by several independent groups (Paquet et al., 2016; Okamoto et al., 2019; Schubert et al., 2021). While there are reports showing lack of impact of these modifications in case of disruption of the seed recognition sequence by longer inserts (Liang et al., 2017), the introduction of PAM mutations can be considered an important general aspect for design of ssODN template ² .

Shubert et al. aimed to define a ruleset for the placement and a number of blocking mutations required to maximize HDR efficiency. HDR ssODN with Cas9 RNP complexes targeting two different genomic loci into HEK293 and K562 cells was tested with assessment of HDR rate. The HDR rate was low (<2%) for all conditions tested in the absence of any blocking mutation, and was greatly increased (8.0%–17.8%) by addition of a blocking mutation in the second or third base of the “NGG” PAM. Blocking mutations placed around the Cas9 cleavage site and near the 3′-end of the guide were also effective, with the impact reduced as the position of the blocking mutation moved PAM-distal. None of four DNA bases were preferred over others. The study showed that two blocking mutations may lead to more robust improvement in HDR efficiency than donor templates containing a single blocking mutation, while incorporating three or four blocking mutations did not further enhance HDR efficiency over the best combination of two (2 PAM or 1 PAM +1 seed). The additional PAM mutations negatively affected the HDR efficiency in some experimental settings, suggesting that there is a limit to the number of additional mutations that should be added to prevent Cas9 re-cleavage. Based on the results of the study the authors ³ suggested the HDR template tool² that facilitates the placement and number of blocking mutations required (Schubert et al., 2021).

Sequence optimization in regard of bp distance of the insert from the DBS location was also shown to be a factor affecting the ssODN knock-in efficiency. Most reports assessing this parameter show that the insert site should be placed in close proximity to DSB for optimal HDR efficiency (Inui et al., 2015; Liang et al., 2017; Schubert et al., 2021). Unfortunately, the compliance for this requirement is limited in CRISPR based systems as the location of a DSB site is defined by the presence of PAM, which is not the case for Fok1 based nucleases. As an example, Liang and co-authors using ssODNs and Cas9 RNP showed that gRNAs in close proximity (−7, −3, +3, and +5) to the insertion site produced the highest HDR rates (up to 10-fold higher comparing to + -10 bps from DSB site) (Liang et al., 2017). Importantly, as different gRNAs enabling different rate of overall DSB formation, this factor also influences the outcome in regard of HDR rate and should be considered when designing ssODNs\Cas9 editing system. Cas9 nickase RNP complexes targeting both strands in the PAM-out orientation was shown to be more efficient in inserting of templates between flanking nick sites at a location that would be otherwise considered sub-optimal for wild-type Cas9 designs. Alt-CRISPR HDR Design Tool proposed by Schubert et al. also include gRNA selection algorithms for Cas9 to balance the distance from the cut to mutation taking into account the on- and off-target scores of available gRNAs, as well as for Cas9 D10 A nickases, where the gRNA orientation and distance between nick sites is considered. Another predictive model, assisting the design of ssODNs for introduction of point mutations called CUNE (Computational Universal Nucleotide Editor)³ considering sequence and homology arms structure was created using machine learning algorithms based on the dataset containing 30 samples (unique HDR targets), from 126 experiments targeting a total of 744 mice (O’Brien et al., 2019).

The use of two sgRNA was shown to increase the HDR rate mediated by ssODNs donors. In a study by Di Stazio et al., the optimized HDR procedure with the use of double sgRNAs, asymmetric ssODN and triple transfection events enabled the increase the TNFα gene HDR rate from an undetectable level to 39% in HEK293 cell line model (Di Stazio et al., 2021). In addition, the cell cycle blocking at G2/M phase with nocodazole treatment was reported to increase HDR efficiency (Han et al., 2020). The transcription activity of targeted sequence was also shown to be affecting the knock-in rate, with transcriptionally active sites increasing DSB repair activity (Davis and Maizels, 2014).

As ssODN cannot readily transport trough the nuclear membrane, the addition of NLS-tagged ssODNs represent the strategy for optimization of HDR efficiency. In study by Han et al., 2020 the NLS-tagging of ssODNs enhanced SSTR and small genetic aberrations efficiency by 4-fold compared to the control (Han et al., 2020). Alternatively, the ssODN may be covalently bound to crRNA or even Cas9 enzyme with formation of chimeric RNPs, assisting the nuclear import of the template (Aird et al., 2018; Savić et al., 2019; Ghosh et al., 2022). It has been reported that incorporating chemical modifications such as phosphorothioate linkages may improve HDR when using ssODN donors (Renaud et al., 2016; Liang et al., 2017), while other studies does not confirm the benefit of phosphorothioate modifications of ssODNs regarding the HDR rates (DeWitt et al., 2016).

Despite this controversy, accounting for the relative simplicity and safety of this modification can be one of the basic considerations in ssODN template design. Another route to improve HDR frequency is by using chemical compounds that inhibit key DSB repair enzymes that play a role in the competing NHEJ pathway. Several chemical compounds have been reported to increase HDR mediated by ssODN templates, however this issue is out of the scope of the current review (Ma et al., 2018; Pavel-Dinu et al., 2019; Li et al., 2020).

To sum it up, general recommendations that can be considered when designing the ssODN template are as follows: introduction of 40 bp symmetric homology arms, addition of a blocking mutation in the PAM sequence, localization of insertion in maximal proximity to DSB site. The exact structural modifications and parameters such as targeting T or NT strand may be cell type and genetic locus specific and should be tuned for each experimental setting.

Linear long double-stranded DNA is the one of the most widely developed forms of donor template for clinical translation (Quadros et al., 2017; Mayo-Muñoz et al., 2018; Roth et al., 2018; Yang et al., 2022). Such templates can repair DSB using HR, MMEJ and SSA mechanisms, depending on the homology arm’s length. They are usually used to insert long fragments (>1 kb) and make it possible to encode several required sequences and obtain biallelic gene insertions as well as multiplex 2x/3x gene modifications (Agudelo et al., 2017; Barylski et al., 2020; Branton et al., 2020; Zhang et al., 2022).

The most studies that used linear long dsDNA as a donor template were carried out using the CRISPR/Cas9 (Ménoret et al., 2015), but applications of other nucleases were also reported (Mino et al., 2009; Jinek et al., 2013; Zhang et al., 2019; Yang et al., 2022). The usage of Cas9 nickase modified to reduce off-target double-strand breaks was also efficient in increasing of the template homologous insertion into the desired locus in some studies with reduced genotoxicity profile (Mali et al., 2013; Ran et al., 2013). A small number of works have been devoted to the application of the linear type dsDNA template on primary human cells. Model cell lines and organisms are the prevailing objects in this field (Gutierrez-Triana et al., 2018). The knock-in efficiency when using dsDNA linear templates reached 5%–70% (Table 6), but still remains relatively inefficient for many cell types (Chen et al., 2011). Efficiency was minimal in nerve cells and melanocytes, but was above 50% in human hematopoietic cell lines, human pronuclear zygotes, and some other model cells.

According to the literature data, successful HR requires the creation of dsDNA templates with length of homology arms within the range of 0.2–1 kb (Canaj et al., 2019; Zhang et al., 2022). Some studies showed that homology arms with a length of 100–200 bp demonstrate less HDR efficiency than 400–800 bp arms when targeting IL2RG locus in K562 cells (Hendel et al., 2014). The difference in knock-in efficiency between templates with homology arms of 400 and 800 bp was insignificant. This fact allows assuming that for the most efficient nuclease-induced HR in human cells, the homology arms at both ends must be approximately 400–800 bp in length (Hendel et al., 2014; Zhang et al., 2022). Jian-Ping Zhang’s team confirmed this hypothesis in HEK293T cells, where the dsDNA template with homologous arms of 200–300 bp showed 0.22% of HDR efficiency, while the template with homology arms of 600–800 bp increased HDR up to 10% (Zhang et al., 2017). Same group reported that the combined use of double cut donors with homology arms of 600 bp with cell cycle regulators (Nocodazole and CCND1) increased knock-in by up to 20%–30% (Zhang et al., 2017).

An important factor that affects the HDR efficiency is the length of dsDNA donor template. It has been found that ultra-long (2 kb) and ultra-short (<250 bp) DNA templates lead to a decrease in HDR efficiency compared to 951 bp dsDNA (DiNapoli et al., 2020). Moreover, the toxicity of linear dsDNA templates is length-dependent (Paludan et al., 2019), especially for primary human cells (Zhao et al., 2006; Hornung and Latz, 2010; Luecke et al., 2017; Roth et al., 2018). The delivery of CRISPR-Cas9 as RNP can partially overcome this problem (Roth et al., 2018; Nguyen et al., 2020). The activation of intracellular dsDNA sensors in response to non-viral donor template delivery should be taken into account (Piras and Kajaste-Rudnitski, 2021) as an important toxicity mechanism. According to some literature data, the temporary inhibition of cytosolic DNA sensors increased the viability and CAR insertion rates after DNA transfection of primary human T-cell (Kath et al., 2022).

Yao and co-authors tested a set of configurations of linearized dsDNA templates with long 800 bp homology arms (Yao et al., 2018). The dsDNA linearized in vitro by digestion with two restriction enzymes with 0–2000 bp non-homologous junk sequence adjacent to Has, was compared with the two types of plasmid donors either or not containing sgRNA binding sites next to HAs. The dsDNA donor that did not contain a non-homologous sequence adjacent to the homology arms (“Tild” donor) demonstrated better knock-in efficiency (33%) than either the plasmid donor with sgRNA binding sites (21%) or the plasmid lacking sgRNA binding sites (0%) at the Actb gene in mouse zygotes. OCT4-sgRNA, Cas9 mRNA and the Tild OCT4 donor were co-introduced into human pronuclear zygotes. The Tild template has demonstrated higher knock-in efficiency than the plasmid HR template (21% vs. 2% for a single blastomere and 71% vs. 11% for a whole embryo.

David N. Nguyen et al., (2020) attempted to integrate the Cas9 nickase with RNP and CTS in the HDR template to enhance nuclear import and knock-in rate. The HDR template was modified to encode 20 bp Cas9 target sequences at the ends of the homology arms. The use of CTS-modified HDR templates in conjunction with dCas9-NLS RNP increased the HDR up to 1.5% compared to 0.6% in control samples. Next, HDR template was synthesized with short CTS (sCTS) 16 bp long and catalytically active Cas9-NLS RNP was used instead of dCas9-NLS RNP. The sCTS provide binding to Cas9 but exclude cutting (Jiang and Doudna, 2017). This approach was reported to increase the insertion efficiency of 1.5 kb dsDNA into the TRAC locus of human T-cell (5% vs. 1.5%).

Another proposed approach to improve knock-in efficiency is the use of linear dsDNA along with an asymmetric sgRNA site located at the 3′ end (DiNapoli et al., 2020). The application of such a technology at mitfa gene (b692) using resulted in a 9% phenotypic recovery in the embryos. Christopher D Richardson et al. investigated the interaction of Cas9 with a donor template in order to increase the efficiency of HDR (Richardson et al., 2016). It was demonstrated that Cas9 asymmetrically releases the 3′ end of the cleaved DNA strand, which is not complementary to sgRNA. Based on these data, the Cas9-dsDNA-ssDNA complex was developed, which was reported to achieve the level of HDR up to 16% in human cell lines (HEK293, K562).

One of the limiting factors that reduce the knock-in efficiency of templates delivered as linear dsDNA is their multimerization and non-homologous end joining during reparation. According to the literature data, the chemical protection of the template ends by biotin or carbon spacers (SpC3) prevented multimerization and resulted in knock-in increase up to 9.5% among the survived medaka zygotes (Gutierrez-Triana et al., 2018). Ghanta and co-author’s group on human cells tested the similar approach (Ghanta et al., 2021). The dsDNA donors (566 bp) with 90 bp homology arms were modified at the 5′-terminus with 2′OMe-RNA fused to triethylene glycol of different length. The HDR frequency was increased up to 52% for triethylene glycol compared to 25% when using unmodified dsDNA in model HEK293T cells. Moreover, 2′OMe-RNA/triethylene glycol-modified donors participated in the same number of HDR events.

Lee et al. selected several modifications for dsDNA template ends to increase the gene knock-in (Lee et al., 2019). They included an amine group with a C6 linker, C12 linker to the donor template and conjugated the linker’s amine groups with their N-hydroxysuccinimide esters. Additionally, the phosphate linkages were replaced with stable phosphorothioate linkages at the donor template ends. The modified dsDNA template has been tested on GAPDH locus in HCT116 cells. The dsDNA templates modified with phosphorothioate was reported to be efficient with 1.8х increase of knock-in. At the same time, C6 and C12 linkers enhanced knock-in rate to 4x and 3.8x respectively. It is important to note that all 5′-modifications of the template resulted in an increase in the knock-in level, while the 3′-modifications showed no changes or a decreased the gene correction rate.

Summarizing what has been set forth above, the structure of long dsDNA templates is important for efficient knock-in. Variation of homology arms and transgene lengths greatly affects the knock-in efficiency. According to the works described above, templates with 400–800 bp homology arms and approximately 1000 bp in length are the most promising. Also, chemical DNA end protection and the application of cytosolic DNA sensors inhibitors can be proposed as potential strategies to increase the efficiency of dsDNA incorporation into the required loci and preserve the proportion of viable transfected cells.

While historically dsDNA was firstly used as synthetic donor templates with subsequent wider adoption, its shortcomings may hamper therapeutic gene editing applications. Such applications include cases of duplication of homology arms or partial incorporation of the dsDNA template as it is more readily inserted by the dominant NHEJ mechanism (Wen et al., 2021) at off-target DSBs or endogenous DSBs, as well as high toxicity and activation of inflammatory response by dsDNA templates which impair cell viability and transfection efficiency (Yu and Liu, 2021). To avoid these undesired consequences, researchers have turned their attention to long single-stranded DNA templates as homology recombination donors. While the most common forms of homology recombination in somatic cells are completely dependent on the action of the Rad51 recombinase (Chen et al., 2008), lssDNA donors are utilized in Rad51-independent SSTR pathway (Gallagher et al., 2020; Gallagher and Haber, 2021). The key advantage of the lssDNA donors in terms of safety is the absence of integration by direct ligation of donors with DSB end by NHEJ and minimal risk of non-specific insertions in loci independent of target homology, translating in higher knock-in specificity (Roth et al., 2018). Moreover, end-joining ligation reactions assemble linear dsDNA molecules into concatemers in eukaryotic cells, limiting the number of individual donor molecules and their ability to diffuse to their DSB target sites, which is not the case for ssDNA.

The ssDNA was also demonstrated to exhibit lower toxicity (Roth et al., 2018), which is especially important for applications involving scarce populations of primary cells, such as HSPCs. Currently long ssDNA donors are mostly need to be generated manually by one of the reported methods, all of which are comparably laborious. Due to this aspect, the data regarding the application of lssDNA donors is limited, especially in primary human cell types (Table 7).

The successful use of lssDNA donors with Cas9 RNP’s for generation of gene modified animal strains was demonstrated in a series of reports (Bennett et al., 2021; Zhang et al., 2022). However, the quantitative assessment of the knock-in efficiency in these cases is challenging. Using Cas9 DNA plasmid and lssDNA donors the knock-in of the methylated SP3 promoter was demonstrated by Katayama and Andou with estimated efficiency of about 16% (Katayama and Andou, 2021). Roth et al. reported the successful correction of IL2RA c.800delA mutation in T-cell using D10 A Cas9 nickase and long ssDNA donor, demonstrating 24.9% IL2RA expression rate. However, this result should be considered with caution, as addressed frameshift mutation could be corrected both by HDR as well as NHEJ, presumably due to some of the small indels restoring the open reading frame (Roth et al., 2018). In the most comprehensive study to date by the same group, the performance of lssDNA donors harboring different structural modifications was evaluated across a variety of clinically relevant primary cell types including CD4⁺ T-cell, CD8⁺ T-cell, regulatory T-cell, NK cells, B cells, HSPC’s and gamma-delta T-cell with ultra-high HDR efficiencies (>80–90%) in selected optimal conditioned experiments. Importantly, in all evaluated cell types lssDNA templates demonstrated significantly lower toxicity, increased knock-in efficiency, and generated higher absolute knock-in cell yields comparing to dsDNA donors containing the corresponding structural modifications (Shy et al., 2022). Based on the developed protocol, the group created a GMP-compatible method for fully non-viral CAR-T cell manufacturing, demonstrating knock-in efficiencies of 46%–62% and generating yields of >1.5 × 10⁹ CAR + T-cell.

While the number of systematical studies regarding gene correction implying lssDNA is limited, the current data suggests that length, concentration, and addition of CTS hairpins may be important factors influencing the results of editing.

The experiments assessing the role of the template length included comparison of tNGFR (1500 bp), IL2RA-GFP (2267 bp) bearing lssDNA templates of similar structure targeted to a IL2RA locus and BCMA-CAR (2923 bp) targeted to TRAC locus with 78.5%, 38%, 39% maximum transgene expression rate, correspondingly, confirming the observations that shorter templates facilitate higher HDR efficiency (Shy et al., 2022).

In a comparative study by Li et al., a GFP (∼700 bp) insert flanked by various sizes HA (36–700 bp) was targeted to RAB11A in HEK293T cells with a near-exponential relationship between homology length and knock-in efficiency: longer homology arms were shown to drive higher efficiency, with 95% of maximal efficiency reached using ∼400 bp arms (Li et al., 2019).

As the nuclear transport of DNA template represent one of the key barriers for efficient gene correction, the modification of the template structure by introduction of Cas9 Target Sequences (CTS), allowing the co-electroporated RNPs to bind the templates and facilitate their delivery was proposed (Nguyen et al., 2020). Shy et al., 2022 screened 10 different template CTS designs using short 113–195 bp ssDNA templates (Shy et al., 2022). While the majority of applied ssCTS configurations increased knock-in efficiency, the group considered the design that incorporated CTS sites in annealed complementary oligonucleotides to be optimal in terms of efficiency, toxicity and preparation procedures. Addition of CTS also increased the knock-in rates of lssDNA donors. CTS sites with scrambled gRNA sequence had not increased knock-in efficiency, suggesting specific recognition of the gRNA sequence. Importantly, only the 5′ CTS was shown to be functional in the studied designs which could reflect requirements for RNP binding and orientation, intracellular trafficking, or interference with repair machinery during 3′ annealing of long ssDNA (Shy et al., 2022).

In report by Lin-Shiao and co-authors the similar approach involved ∼3.5 kb multigene lssDNA cassette which contained truncated Cas9 target sequences to increase shuttling into the nucleus, complexed to DNA nanostructures. This approach enabled up to 20% knock-in to IL2RA in primary human T-cell (Lin-Shiao et al., 2022).

5′-Modifications may improve potency and efficacy of DNA donors for precision genome editing. Ghanta et al., (2021) reported that addition of RNA: TEG (triethylene glycol) at the 5′-end of a long (800 bp) ssDNA donor significantly boosted HDR in HEK293T model. The frequency of HDR increased with the dose of ssDNA donor, reaching maximal HDR (22.5%) at 6–8 pmol donor amounts.

A panel of small molecule inhibitors reported to enhance knock-in efficiency of lssDNA donors in primary human T-cell including NU7441 and M3814 (DNA-PK inhibitors), the Trichostatin A (HDAC class I/II Inhibitor), XL413 (CDC7 inhibitor), Alt-R HDR Enhancer (an IDT’s proprietary NHEJ inhibitor). At optimal concentrations, M3814 showed the largest effect size (∼49% increase), followed by XL413 (∼46% increase), NU7441 (∼43% increase), Alt-R HDR Enhancer (∼29% increase), and TSA (∼16% increase). By introduction of MT and MTX inhibitor combinations Shy et al. were able to achieve a knock-in rate of >90% in some experiments involving 1.5–2.7 kb lssDNA donors with CTS structural modifications (Shy et al., 2022).

Results of HDR experiments invoving lssDNA show that this type of templates is highly promising for therapeutic GE due to the improved safety profile. The starting points when designing lssDNA templates are symmetrical 400 bp homology arms, introduction of CTS sequences and sequence optimization for overall repair template length reduction.