Potato (Solanum tuberosum L.) is among the world’s most important food crops. Beyond serving as a vital energy source, potato tubers provide essential nutrients, including vitamins B and C, phenolic compounds, minerals, and high-quality protein, reinforcing their significance for global food security (Burgos et al., 2020; Devaux et al., 2020). Additionally, potatoes are a key source of starch, a renewable bulk product widely used in food and industrial applications (Dupuis and Liu, 2019). However, potato cultivation requires high inputs to manage abiotic and biotic stresses, challenges exacerbated by climate change. Traditional breeding methods, while effective, are time-consuming and complicated by potato complex genetics, characterized by a tetraploid inheritance, high heterozygosity, and inbreeding depression (Bonierbale et al., 2020).

Recent advances in genome editing, particularly the clustered regularly interspaced short palindromic repeat (CRISPR) system, offer precise and efficient tools for targeted genetic modifications. The CRISPR/Cas9 system (Jinek et al., 2012) is established as the most widely used gene-editing platform due to its simplicity, efficiency, and cost-effectiveness (Zhu et al., 2020; Gao, 2021; Cardi et al., 2023). In its simplest application, CRISPR/Cas9 facilitates targeted mutagenesis via a single guide RNA (sgRNA) that directs the Cas9 nuclease to a specific genomic site, where it induces a double-stranded break (DSB). This type of break is primarily repaired through the error-prone non-homologous end joining (NHEJ) pathway (Puchta, 2005), leading to small insertions or deletions (indels). Alternatively, the presence of short regions with homology (microhomologies) flanking the Cas9-induced DSB can trigger the microhomology-mediated end joining (MMEJ) pathway, resulting in larger deletions whose size is determined by the distance between the microhomologies (Sfeir et al., 2024). Targeted mutagenesis has a prominent role for studying gene functions and engineering commercially valuable traits in several crops, including potato (Hofvander et al., 2022; Tuncel and Qi, 2022).

However, many agronomically important traits require precise modifications in coding or regulatory regions rather than simple gene knockouts (Gilbertson et al., 2025). When a donor repair template (DRT) is available, the homology-directed repair (HDR) pathway can be activated, enabling precise insertions or substitutions. The DRT molecule is designed with a desired insert and flanking homology arms (HAs) that facilitate the incorporation of edits into the genome. HDR-mediated gene editing has been demonstrated in different plant species, including potato, though at a low frequency of precise recombination events (Butler et al., 2016; Hegde et al., 2021). The reason is that HDR is typically infrequent in somatic plant cells, as it is limited to the S and G2 phases of the cell cycle, unlike the most frequent NHEJ, which operates throughout the entire cycle (Schmidt et al., 2019). Additionally, inefficient DSB induction and poor DRT availability near the DSB further constrains HDR activation (Cermak, 2021).

Given its potential in crop improvement, increasing HDR efficiency in plant species remains a critical research focus. Strategies to enhance HDR include promoting conditions that favour HDR over NHEJ and increasing local DRT availability (Chen et al., 2022; Singh et al., 2023). While various methods to boost HDR frequencies have been tested in animal systems (Singh et al., 2023), studies in plants remain limited (Cermak, 2021).

Research in animal models suggest that DRT structure significantly influences HDR activation. Key factors include HA length, the ratio between HA and insert fragments, the strandedness of the DRT molecule (single-stranded [ss] vs. double-stranded [ds] DNA), and sequence orientation in ssDNA molecules (Baker et al., 2017; Quadros et al., 2017; Miura et al., 2018; Bai et al., 2020; Ranawakage et al., 2020). For example, a systematic evaluation in mice demonstrated that for dsDNA donors, HDR efficiency increases sharply as HAs extend from 200 bp to 2,000 bp, with more moderate gains observed for HAs longer than 2,000 bp and up to 10,000 bp (Baker et al., 2017). In human cells, Zhang et al. (2017) reported a similar trend, with HDR efficiency gradually increasing as HAs extended from 50 bp to 900 bp, although sequences as short as 50 bp still enabled 6%–10% HDR efficiency (Zhang et al., 2017). Regarding the use of ssDNA as donors, high HDR efficiency appears achievable even with short HAs. For instance, in mice, combining ssDNA DRTs with ribonucleoprotein (RNP) delivery of editing components resulted in HDR efficiencies ranging from 8.5% to 100% for HAs of 50–100 nucleotides, even for large inserts (>800 bases) (Quadros et al., 2017; Miura et al., 2018). Similar findings were reported in zebrafish, where ssDNA outperformed dsDNA for HDR-mediated editing, even with 40 nucleotides-HAs (Bai et al., 2020; Ranawakage et al., 2020).

DRTs as ssDNA molecules can be used in one of two possible orientations relative to the sgRNA recognition sequence. The “target” orientation coincides with the strand that is recognised by the sgRNA, whereas the “non-target” orientation corresponds to the opposite strand containing the PAM sequence. While some studies in animals have indicated that there may be a preference for using either orientation (Paix et al., 2017; Skarnes et al., 2019), other studies demonstrate that the optimal orientation may be dependent on the target locus and its sequence (Ranawakage et al., 2020). Thus far, no empirical analysis of this parameter for specific target sites and donor molecules have been conducted in plants.

Despite extensive research conducted in animal models, research on DRT structure and its impact on HDR efficiency in plants remains limited. Jiang et al. (2021) studied how HA length influences HDR efficiency in Nicotiana benthamiana protoplasts transfected with CRISPR/Cas9 RNPs, finding that HAs longer than ∼35 and up to ∼64 nucleotides achieved highest HDR efficiency (∼45%) in the pool of protoplasts (Jiang et al., 2021). However, all other parameters related to DRT structure were kept invariable. Given the lack of established guidelines for optimal DRT design, this study investigates how DRT structure influences HDR efficiency in potato. To this end, we employed ribonucleoprotein (RNP)/DRT transfections in potato protoplasts, combined with Next-generation sequencing (NGS), to precisely quantify editing outcomes. Our results demonstrate that DRT structure significantly impacts HDR efficiency, with ssDNA donors in the target orientation outperforming other structures at three of the four tested genomic loci. In contrast, the length of HAs appeared to have a comparatively minor effect on HDR efficiency, within the tested range of 30–97 nucleotides. Efforts to enhance HDR by modulating DNA repair pathways or increasing DRT availability, using strategies commonly successful in animal systems, did not improve outcomes. Furthermore, sequencing revealed a high-frequency of targeted insertions, likely driven by alternative repair mechanisms, highlighting the potential for efficient knock-in integration at non-coding regions in the potato genome.

Gene editing using the CRISPR/Cas9 system is a powerful tool for genetic engineering and is significantly accelerating the improvement of agronomically important traits in a wide range of crop species. HDR can mediate precise changes, but its efficiency in somatic plant cells remains low (Cermak, 2021). In our study, we investigated conditions with the potential to favour HDR-mediated gene editing in potato using a protoplast transfection system coupled with NGS analysis.

Since the induction of a DSB is an essential initiating step for HDR (Puchta et al., 1996), and HDR efficiency in plants is positively correlated with DSB frequency (Zhang Z. et al., 2022; Li et al., 2024), we began by evaluating four different sgRNAs targeting the SS1 gene. Editing efficiency varied markedly, ranging from 3.82% for guide T13 (statistically indistinguishable from the mock control) to 74.19% for guide T7 (Figure 1B). These results underscore the well-established impact of sgRNA selection on editing efficiency. Although various bioinformatic tools can predict sgRNA efficacy (Konstantakos et al., 2022), predictions often diverge from experimental outcomes (Concordet and Haeussler, 2018). This observation is even more noticeable in gene editing applications in plants, as most of the efficiency predictors have been trained using empirical data derived from animal models (Naim et al., 2020). Our findings reaffirm that sgRNA performance is best validated empirically, and the protoplast transfection and NGS analysis performed in our study, represent a rapid and reliable approach to this endeavour.

Due to tetraploid nature of the potato variety used in our study and high allelic variability in the SS1 targeted region, only the sgRNA designed on T13 matched all four alleles (Figure 1A). In contrast, sgRNAs targeting T7 and T12 were not predicted to bind any of the alternate alleles, which was confirmed experimentally (data not shown). The sgRNA designed on T17, however, targeted a region containing a single-nucleotide polymorphism (SNP) in one allele, located distal to the PAM (Figure 1A). Interestingly, sequencing data showed T17 edited all four alleles with comparable efficiency, including the SNP-containing allele (Figure 1C). This supports earlier findings that Cas9 tolerates single mismatches distal to the PAM (Jiang and Doudna, 2017; Feng et al., 2018; Es et al., 2019; Lee et al., 2019; Modrzejewski et al., 2020; Wang et al., 2021), highlighting implications for sgRNA selection in regions with allelic variation and also the prediction of potential off-target effects in highly similar loci.

In addition to efficient DSB induction, HDR also requires the presence of a DRT carrying the desired insert. Although extensively studied in animal models, the impact of DRT structure on HDR efficiency in plants is poorly understood. In a previous report using N. benthamiana protoplasts, Jiang et al. (2021) examined how HAs length affects HDR. Other DRT parameters were constant in that study, as ssDNA molecules with target orientation were used in all cases. Furthermore, the efficiency of HDR was determined with a reporter system based on GFP activity restoration upon HDR-mediated editing of its coding sequence (Jiang et al., 2021). While highly valuable for facile and inexpensive analysis of the efficiency of different HDR components, a more thorough analysis through high-throughput sequencing may characterise outcomes more accurately, allowing precise comparisons between different strategies. Our data show that HDR efficiency in potato is influenced by DRT structure, with ssDNA donors matching the target strand orientation (ss-T donors) outperforming other structures in three out of four tested loci (Figures 2B, 5A,C). Notably, for the target site on SS1, this was consistent across varying HA lengths, as indicated by our two-way ANOVA results (Table 1). This result indicates that for ss-T donors, HAs as short as 30 nucleotides are sufficient to mediate HDR, consistent with the findings reported in N. benthamiana (Jiang et al., 2021). However, a limitation of our study is the fixed insert length, leaving open the question of how the ratio between 30 nucleotides-HAs and insert fragments may impact HDR efficiency.

ssDNA donors are generally more effective than dsDNA in various animal systems, such as zebrafish (Bai et al., 2020) and mammalian cells (Yeh et al., 2019; Zhang X. et al., 2022; Jin et al., 2025). The increased efficiency is often attributed to synthesis-dependent strand annealing (SDSA) mechanism, which requires only short homologous sequences (30–40 nucleotides) to trigger precise insertions (Paix et al., 2017; Jin et al., 2025). In contrast, HDR using dsDNA donors typically requires much longer HAs (0.5–1 kb) (Baker et al., 2017; Zhang et al., 2017). It is also generally accepted that ssDNA represents a less cytotoxic cargo than dsDNA for donors in animal cells (Zhang X. et al., 2022). While further investigation is needed to determine if these factors influence ssDNA performance in potato, our data indicate a consistent reduction in targeted mutagenesis when dsDNA donors were included in the transfection (Figure 2C; Supplementary Figure S4). This observation could be explained by two non-exclusive scenarios. First, the presence of dsDNA donors might reduce overall CRISPR/Cas9 activity. A lower rate of DSBs would directly decrease HDR efficiency in these conditions. One possible explanation is an in vitro interaction between the RNP complex and dsDNA donors before transfection, which may interfere with target DNA interrogation in the cell. To test this hypothesis, delivering the CRISPR/Cas9 components via vector-based expression could help to prevent potential pre-transfection interactions. Alternatively, reduced targeted mutagenesis could result from dsDNA-associated cytotoxicity, as widely reported in animal systems (Nguyen et al., 2020; Zhang X. et al., 2022; Shy et al., 2023). This hypothesis could be evaluated by assessing protoplast viability prior to PCR amplification of the target locus. These scenarios assume that variations in transfection efficiencies related to specific donors can be disregarded in our system.

Regarding ssDNA strand orientation, ss-T donors were generally more effective in our study, consistent with what has been previously proposed based on Cas9 cutting dynamics (Jiang and Doudna, 2017). Multiple studies using biochemical, structural, and single-molecule approaches support that the non-target strand is typically cut and released first, which would make it more accessible for interaction with a complementary ssDNA donor for HDR (Richardson et al., 2016; Paix et al., 2017; Wang et al., 2018; Wang et al., 2023). Our data support this, with ss-T donors outperforming in three out of four loci (Figures 2B, 5A,C). Importantly, these differences were not merely due to overall CRISPR/Cas9 activity, as for SS1 similar frequencies of targeted mutagenesis were observed for ss-T and ss-NT donors (Figure 2C), while for EID1 and SES genes, targeted mutagenesis was even higher in experiments using ss-NT donors (Supplementary Figure S4). Nevertheless, inconsistent strand preference at the LNK2 target site suggests this may not be a generalised rule, aligning with some observations in animal systems (Ranawakage et al., 2020; Schubert et al., 2021).

When a DRT is available, insertions can also occur via non-HDR pathways such as NHEJ or MMEJ (Hsu et al., 2021; Van Vu et al., 2021; Kumar et al., 2023; Vu et al., 2024). Our analysis of reads containing the desired insert, regardless of perfect HA recombination (“targeted insertions”), indicates these events are relatively frequent outcomes in the potato genome. Using ss-T donors with 30-nt HAs, we observed average targeted insertion frequencies of 24.89%, 9.62%, 2.83%, and 12.50% for SS1, EID1, LNK2, and SES, respectively (Figures 2D, 5A–C), while corresponding HDR frequencies were much lower (0.80%, 3.15%, 1.25% and 0.22%, respectively). Similar trends were seen with ss-NT and dsDNA donors. Furthermore, our analysis targeting SS1 suggest that targeted insertions frequency increases as the length of the donor molecules decreases (Figure 2D). Although our method for quantifying targeted insertions includes HDR-mediated events (see Section 5.7), the difference between values of the two parameters determined for each target site, provides a clear indication that most of the insertions obtained were mediated by imprecise or alternative repair mechanisms, rather than by perfect recombination repairs. High-frequency targeted insertions using ssDNA donors have also been reported in N. benthamiana protoplasts (Hsu et al., 2021). By using ss-NT donors, the authors reported targeted insertions reaching frequencies of 10.5%–13.6%, based on single-cell genotype analysis (Hsu et al., 2021). Furthermore, plant regeneration conducted on the transfected protoplasts resulted in 29.3%–31.8% of plants carrying targeted insertions, while only one regenerated plant (8.3% of the total analysed), displayed a precise insertion mediated by HDR (Hsu et al., 2021).

Determining the exact repair mechanism underlying the high-frequency targeted insertion of ss-T donors, is beyond the scope of this study. However, our analysis on the SS1 target gene suggests compatibility with the MMEJ pathway (Sfeir et al., 2024). Features of this pathway include the presence of microhomologies at deletion junctions following DSB induction, which result in deletion patterns “guided” by these microhomologies, as well as insertions generated through DNA synthesis using short homologous sequences as templates, albeit with less fidelity than HDR (Schmidt et al., 2019; Van Vu et al., 2021; Sfeir et al., 2024). Furthermore, high-frequency insertions mediated by MMEJ have been reported in plants (Schmidt et al., 2019). Two lines of evidence from our study support the involvement of MMEJ in mediating targeted insertions of ss-T donor with 30-nt HAs in SS1. First, MMEJ-compatible mutations (i.e., deletions flanked by microhomologies) were observed in a significant fraction of all mutated reads across experiments targeting the T17 site in SS1 (Supplementary Table S1; Supplementary Figure S2). This suggests MMEJ could be actively repairing DSBs in potato, as observed in other plant species (Tan et al., 2020; Weiss et al., 2020). Second, transfections including the ss-T donor predominantly yielded insertions shorter (10–30 nt) than the full donor length (66 nt) due to partial duplications of the flanking homology arms (Supplementary Figure S1). This outcome is compatible with imprecise base pairing between short HA regions and the target site, followed by DNA synthesis, leading to incomplete donor incorporation at the target locus (Sfeir et al., 2024). In contrast, analysis of the equivalent dsDNA donor revealed higher prevalence of insertion lengths closer to the full donor sequence, suggesting that the majority of these events could be mediated by the NHEJ pathway (Supplementary Figure S1). Likewise, in Setaria viridis, dsDNA short donors led to up to 51.1% targeted insertions in protoplasts, whose sizes matched a complete HAs duplication, pointing out at a vast majority being mediated by the NHEJ mechanism (Kumar et al., 2023). However, inferring exact mechanisms based solely on editing outcomes remains challenging and further research efforts are needed to determine precise molecular mechanisms behind targeted insertions observed in our study, possibly using plants defective in specific pathways (Schmidt et al., 2019).

To address the discrepancy between HDR and targeted insertions, we tested chemical inhibition of NHEJ in potato protoplasts, as a strategy to boost HDR in the SS1 locus. Contrary to genetic suppression of NHEJ (Cermak, 2021; Chen et al., 2022), chemical modulation of repair mechanisms has not been extensively studied in plants. One of tested molecules, NU7441, has been widely used in animal systems to inhibit DNA-dependent protein kinases (DNA-PKcs) that act in the canonical NHEJ mechanism, leading to increased HDR (Robert et al., 2015; Schimmel et al., 2023; Shy et al., 2023). Even though no plant homologs of DNA-PKs have been described (Schmidt et al., 2019), NU7441 was previously applied at a concentration of 1 µM during tomato callus regeneration to enhance HDR efficiency in that species, albeit with moderate effectiveness (Vu et al., 2021). This prior evidence prompted us to test different concentration of NU7441 in our protoplast system. Consistent to the absence of its targeted element in plants, adding NU7441 to the protoplasts culture medium at concentrations of 0.5, 1, or 5 µM did not significantly impact HDR efficiency in the SS1 locus. Moreover, in agreement with its previous application in tomato (Vu et al., 2021), neither of the tested concentrations resulted in negative impacts on the NHEJ mechanism, represented in our study by the quantification of targeted mutagenesis (Figure 3) and the mutation profile analysis (Supplementary Figure S2). These results suggest that incorporation of NU7441 to the culture medium is not a viable strategy for modulating DNA repair in potato, at least for the evaluated concentrations.

We also tested Alt-R HDR Enhancer V2, generally described as a NHEJ inhibitor (Kath et al., 2022; Shy et al., 2023), and TSA, a histone deacetylase inhibitor (Robert et al., 2016). Alt-R HDR Enhancer V2, effective in animal systems (Kath et al., 2022), had no detectable effect on targeted mutagenesis or HDR efficiency in our protoplast system (Figure 3). Further testing at varying concentrations, other than as employed here following manufacturer’s recommendation, may clarify whether dosage or plant-specific factors are limiting its efficacy. Regarding TSA, although generally described as a NHEJ inhibitor, it has shown mixed effects on DNA repair modulation (Singh et al., 2023). For instance, in human cells TSA interfered with NHEJ mechanism, by inhibiting the deacetylation of key factors, such as Ku70 and Ku80, limiting their access to the DSB (Robert et al., 2016). Additionally, TSA increased the duration of the G2 phase in the cell cycle in animal cells, contributing to higher HDR efficiency (Li et al., 2020; Shy et al., 2023). Conversely, in the absence of donor molecules, TSA increased the targeted mutagenesis in animal cells, possibly due to a higher accessibility of the CRISPR/Cas9 components to due to an open chromatin state at target sites, and activation of alternative end-joining repair mechanisms (Li et al., 2020). Previous to our study, concentrations of 0.1 µM–10 μM TSA, increased targeted mutagenesis rates in lettuce and tobacco protoplasts (Choi et al., 2021), potentially by enhancing RNP access through chromatin relaxation. Here, we tested TSA at various concentrations in potato and in presence of a donor molecule. However, tested concentrations did not significantly influence either targeted mutagenesis or HDR (Figure 3). Since TSA’s impact on genome editing may depend on the basal chromatin state at specific target sites, the higher targeted mutagenesis previously reported in lettuce and tobacco may not be generalizable to all loci. Given its promising use in other systems and observation of higher CRISPR/Cas9 activity in other plant systems, additional investigation of TSA would be relevant in potato. Altogether, our results using chemical molecules with potential modulation effects on DNA repair mechanisms provide evidence that directly translating strategies validated in animal studies remains challenging, possibly due to differences in the components and molecular mechanisms operating in plant systems (Schmidt et al., 2019).

In an attempt to further enhance HDR efficiency, we investigated a strategy to increase local availability of DRT for SS1 gene editing. To this end, we incorporated truncated Cas9 target sites (CTS) at the ends of the HAs in the donor molecules. This approach has previously been shown to increase HDR efficiency by up to threefold in human cells transfected with dsDNA donors (Nguyen et al., 2020), and has also been applied successfully to ssDNA donors (Shy et al., 2023). In human cells, CTSs comprising 16 bp of the target sequence enable Cas9 to bind, but not cleave, the DRT, thereby enhancing its nuclear localization through co-translocation with the RNP complex. In contrast, the inclusion of CTS at the ends of the 30-nucleotide HAs in the ss-T donor designed for the SS1 locus significantly reduced HDR efficiency in our potato system (Figure 4A). A similar reduction was observed with the corresponding dsDNA donor, although in this case, the difference compared to the control was not statistically significant (Figure 4A). Based on targeted mutagenesis analysis, we concluded that the inclusion of CTSs reduced overall CRISPR/Cas9 activity (Figure 4D). In previous applications of this strategy in human cells, pre-incubation of the RNP complex with the CTS-containing donor was essential to achieve high HDR efficiency. This may be due to the necessity of forming a stable interaction between the RNP and the CTS-containing DRT, enabling their joint translocation into the nucleus (Nguyen et al., 2020). In our study, such interactions may have interfered with Cas9 activity at the genomic target site in potato, possibly due to competitive binding or steric hindrance. This observation supports our earlier hypothesis regarding potentially deleterious interactions between dsDNA donors and Cas9 prior to protoplast transfection. Additionally, a potential decrease in DNA cleavage efficiency in strategies that tether the donor molecule to the RNP complex, has been suggested by other authors (Jin et al., 2024; Jin et al., 2025). Alternative successful strategies to enhance local DRT availability have been explored in rice, such as the use of RNA donor molecules as extensions of the sgRNA (Butt et al., 2017), or fusion of the Agrobacterium-derived VirD2 protein to Cas9, which enables tethering of a ssDNA donor to the editing complex (Ali et al., 2020; Tang et al., 2023). Despite the reduced efficacy observed in our system, the demonstrated success of CTS-based donors in human cells and their compatibility with RNP-based, transgene-free editing strategies suggest that further optimisation of this approach could still hold promise for improving gene editing efficiency in potato. To this end, the employment of an inactive dCas9 variant to fuse to the CTS-containing donor (Nguyen et al., 2020), while maintaining the RNP complex targeting the desired gene free of any interactions previous to the transfection, could be a strategy to test.

In this study, we explored conditions that have potential to favour HDR-mediated gene editing in potato. The employed combination of protoplasts with NGS offers an effective platform for rapidly assessing gene editing components and conditions for downstream applications. One main bottleneck of this platform is the isolation of high-quality protoplasts, typically considered labour-intensive. However, simplifying and optimising this step in potato and other plant species would promote broader use of this approach for validating gene editing tools. Our study on how HDR efficiency is influenced by the DRT structure, indicate that ssDNA donors matching the CRISPR/Cas9 target strand may be generally more efficient than other configurations, providing insights for future genome editing strategies in this species. Additionally, short ssDNA donors are prone to high-frequency insertions in the potato genome, mediated by alternative mechanisms other than HDR. While this may be a limitation of their application to precise modifications of coding sequences, this approach holds potential for efficient targeted insertions of cis-regulatory elements to modulate target gene expression in potato. Recent identification of short cis-regulatory elements in this species (Zeng et al., 2019; Wan et al., 2024; Zhu et al., 2024) coupled with efficient CRISPR/Cas9-mediated targeted insertion approaches, would contribute to fine-tune gene expression as a key driver of phenotypic novelty.