De novo RNA base editing in plant organelles with engineered synthetic P-type PPR editing factors

Abstract

In plant mitochondria and chloroplasts, cytidine-to-uridine RNA editing is necessary for the production of functional proteins. While natural PLS-type PPR proteins are specialized in this process, synthetic PPR proteins offer significant potential for targeted RNA editing. In this study, we engineered chimeric editing factors by fusing synthetic P-type PPR guides with the DYW cytidine deaminase domain of a moss mitochondrial editing factor, PPR56. These designer PPR editors (dPPRe) elicited efficient and precise de novo RNA editing in Escherichia coli as well as in the chloroplasts and mitochondria of Nicotiana benthamiana. Chloroplast transcriptome-wide analysis of the most efficient dPPRe revealed minimal off-target effects, with only three nontarget C sites edited due to sequence similarity with the intended target. This study introduces a novel and precise method for RNA base editing in plant organelles, paving the way for new approaches in gene regulation applicable to plants and potentially other organisms.

Graphical Abstract

Open in new tabDownload slide

RNA editing is a post-transcriptional process that introduces specific nucleotide changes in RNA sequences, which is essential for the proper functioning of many eukaryotic cells [1, 2]. As a result, it can alter the genetic information encoded in the DNA, allowing for the synthesis of proteins with amino acid sequences that differ from those dictated by the genomic template [3]. The ability to reprogram RNA editing specificity offers significant potential for biotechnological applications in basic research, medicine, and agriculture. It enables precise correction of genetic mutations, modulation of gene expression, regulation of protein function, and adaptation of cellular responses to various conditions.

In plants, RNA editing primarily involves the conversion of cytidines to uridines (C-to-U), and less frequently, uridines to cytidines (U-to-C) in organellar transcripts [4–9]. In angiosperms like Arabidopsis thaliana, over 30 cytidines in chloroplast transcripts and >500 in mitochondrial transcripts are specifically edited [10, 11]. These modifications are crucial for the expression of functional proteins involved in essential metabolic and bioenergetic pathways [12–17]. The specificity and efficiency of RNA editing in plants is mediated by nuclear-encoded pentatricopeptide repeat (PPR) proteins. PPR proteins are a class of RNA-binding proteins composed of tandem helix-turn-helix repeats of ∼35 amino acids [18–20]. They are specific to eukaryotes and besides RNA editing, they regulate various post-transcriptional steps in organelles, from RNA stabilization to translation [21]. In plants, PPR proteins can be subdivided into two subfamilies based on the type and arrangement of their PPR arrays [22, 23]. P-type PPR proteins contain long stretches of canonical 35 amino acid P-PPR motifs and are widespread in eukaryotes, while PLS-type PPR proteins consist of arrays of P, L (long, 35–36 aa), and S (short, 31–32 aa) motifs and are specific to land plants performing organellar RNA editing [23, 24]. Genetic studies have implicated P-type PPR proteins in various post-transcriptional processes such as RNA stabilization, RNA cleavage, intron splicing, and mRNA translation [21]. In contrast, PLS-type PPR proteins are the key players in RNA editing [8]. They bind precisely to an RNA sequence located in the cis-element 6 nucleotides upstream of the target C to guide the deamination reaction (Fig. 1A). Each PPR motif (P, L, and S) aligns with a ribonucleotide in the cis-element in a parallel orientation and the RNA base recognition is mainly determined by a combinatorial code involving the identity of two amino acids in each P- and S-type PPR motifs: amino acid in position five distinguishes purines (A, G) from pyrimidines (C, U) and the last amino acid defines preferences for amino (A, C) or keto ribonucleobases (G, U) [25–28]. The final P–L–S motif triplet (called P2–L2–S2) in PLS–PPR editing factors is often followed by E1 (for extension 1) and E2 motifs. The C-to-U conversion is catalyzed by a DYW deaminase domain of ∼20 kDa, which is either naturally appended to the C-terminus of the PPR editing factor or recruited in trans [29–35]. Additionally, unusual SS-type PPR motifs were recently identified in several PLS-type PPR proteins of the lycophyte plant Selaginella and consists of tandem repeats of the S motif [22, 23] (Fig. 1A).

The modular architecture of PPR proteins and the elucidation of the PPR base recognition code have enabled the rational design of synthetic PPR proteins using consensus P, L, and S motifs with specific nucleotide binding capabilities [36–40]. In vivo applications of such synthetic P-PPR or PLS-PPR scaffolds for gene expression control in plants have been demonstrated in several contexts and include chloroplast RNA stabilization [41], translational activation [42], and RNA editing [39]. These studies have shown that synthetic PPR proteins can effectively mimic the functionality of their natural Arabidopsis counterparts in the respective mutant lines. Although these experiments provide proof-of-concept for synthetic PPR scaffolds being functional in chloroplasts, the potential to engineer these proteins for novel functions in gene expression by targeting new RNA sites in organelles has not yet been fully realized and requires further exploration. Additionally, the application of synthetic PPR scaffolds in mitochondria, which are associated with cytoplasmic male sterility in plants [43] and inherited myopathies in humans [44], remains largely unexplored.

In this study, we aimed to create de novo RNA editing in plant organelles by engineering chimeric editing factors. To this end, synthetic P-type PPR guides were fused to the DYW cytidine deaminase domain of a moss mitochondrial editing factor, PPR56 [45]. We demonstrated that these designer PPR editors function independently of cofactors and successfully edit their RNA targets in Escherichia coli, as well as in the chloroplasts and mitochondria of Nicotiana benthamiana plants. Transcriptome-wide analysis of off-target effects of the most efficient dPPRe in N. benthamiana chloroplasts, which achieved an editing efficiency of up to ∼70% at its designated target C site in rpl2 gene, revealed minimal off-target activity, affecting only three C sites with conserved cis-elements similar to the target site. This study is the first to demonstrate precise de novo RNA editing in both plant organelles—chloroplasts and mitochondria—using engineered synthetic P-type PPR repeats. It also marks the first successful use of synthetic PPR in mitochondria, offering a novel strategy for regulating gene expression by targeting RNA effectors, with potential applications in both plant systems and other biological contexts.

Materials and methods

Primers used in this study are listed in Supplementary Table S5 and DNA and protein sequences are provided in Supplementary Fig. S6.

dPPRe transgene cloning

The DNA sequences encoding the Arabidopsis chloroplast RecA or yeast mitochondrial Cox4 transit peptide (68 and 29 amino acids, respectively) fused to a 3×HA tag and dPPRe were synthesized and cloned into pDONR^TM/Zeo by Thermo Fisher Scientific. For plant expression, the genes were subsequently cloned into binary vectors pMDC32 (2×35S promoter) [46] or pGWB2 (35S promoter) [47] using LR clonase II (Thermo Fisher Scientific, Waltham, MA, USA) following the manufacturer’s protocol.

For expression in E. coli, the coding sequences of mature dPPRe without the organellar transit peptide and 3×HA tag were amplified from the pDONR^TM/Zeo templates. Target sequences comprising 100 bp up- and downstream of the cytidine to be edited, respectively, were amplified from N. benthamiana genomic DNA. dPPRe and RNA target encoding sequences were combined via Overlap Extension polymerase chain reaction (PCR) and inserted into petG41K_MCS using FastDigest restriction endonucleases KpnI and SgsI (Thermo Fisher Scientific).

Bacterial RNA editing assays

Escherichia coli RNA editing assays were performed as outlined in Oldenkott et al. (2019). In short, constructs were transformed into E. coli Rosetta 2(DE3) competent cells (Novagen), which were grown in liquid culture until reaching an optical density (OD₆₀₀) of 0.4–0.6 before inducing recombinant dPPRe (rdPPRe) expression by the addition of 0.4 mM isopropryl β-d-1-thiogalactopyranoside (IPTG). Cells were harvested after 20 h of incubation at 16°C under continuous shaking. RNA editing was determined by reverse transcriptase-polymerase chain reaction (RT-PCR) and Sanger sequencing (Macrogen) and rdPPRe expression was verified by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) and Coomassie blue staining. The mean RNA editing rates (ERs) were calculated from three biological replicates (independent primary E. coli clones that underwent individual RNA editing assays). RNA editing frequencies in the complementary DNA (cDNA) were quantified through the MacVector application by measuring the heights of overlapping T and C peaks in Sanger chromatograms at the position of the target cytidine.

Transient expression in N. benthamiana

For transient protein expression, 6-week N. benthamiana plants were infiltrated with Agrobacterium tumefaciens GV3101 carrying the dPPRe transgenes. Agrobacterium cultures were grown overnight, resuspended in infiltration buffer (10 mM MgCl₂, 10 mM 2-(N-morpholino)ethanesulfonic acid, pH 5.6, and 100 μM acetosyringone) to an OD₆₀₀ of 0.7, and infiltrated into the abaxial side of fully expanded N. benthamiana leaves using a needleless syringe. A Tomato bushy stunt virus-encoded 35S:P19-HA silencing suppressor gene construct [48] (P19) was co-expressed with dPPRe transgenes to increase protein expression. Mixtures of P19/dPPRe agrobacteria were prepared in a 1:1 (v/v) ratio as required. P19 alone was used as the control. Infiltrated plants were maintained under controlled growth conditions and leaf discs were harvested at 2- and 3-day post-infiltration (dpi) for protein analysis by western blot using monoclonal anti-HA antibody (H9658 clone) from Sigma–Aldrich. Entire leaves were also collected at the same time points for RNA extraction using the TRIzol reagent, following the manufacturer’s protocol, to assess RNA editing and gene expression levels by Sanger sequencing of RT-PCR products or RNA-seq. RNA editing frequencies were quantified as previously described or by GATK analysis as later described.

Plant RNA sequencing

Total RNA was extracted from N. benthamiana leaves infiltrated with dPPRe^rpl2 and P19 constructs or P19 alone at 2-dpi or 3-dpi by Trizol extraction followed by phenol/chloroform extraction and ethanol precipitation. RNA was treated by DNase Turbo (Thermo Fisher Scientific) to remove contaminant DNA. Library preparation (ribosomal RNA-depleted) and 2×150-bp pair-ended Illumina sequencing were done by Novogene (https://www.novogene.com/), generating ∼105–138 mio reads per library. Libraries were prepared and sequenced for three independent biological replicates of P19, dPPRe^rpl2 (2×35S)/P19 and dPPRe^rpl2 (35S)/P19 transient expression.

The quality of the reads was assessed with FastQC 0.12.1 and adapters were trimmed with Trim-galore 0.6.10. Sequences with a quality lower than 30, with >5 ‘N’s or shorter than 20 bp after trimming were removed. The remaining reads were mapped against the N. benthamiana chloroplast genome SRR7540368 [49] (https://github.com/WangSoybean/Nicotiana_chloroplast_genomes) combined with the dPPRe^rpl2 nucleotide sequence with HISAT2 2.2.1 using the -k 50 option and default parameters. Mapped reads were counted with featureCounts 2.0.1 with options -O -M –minOverlap 5, to consider multi-mapped reads. Differential expression analysis was done using the DESeq2 1.42.0 package.

Chloroplast RNA editing sites were identified from the HISAT2 aligned reads with GATK 4.5.0.0 according to the GATK best practice recommendations [50, 51] (https://gencore.bio.nyu.edu/variant-calling-pipeline-gatk4/). Initially, MarkDuplicatesSpark, HaplotypeCaller, and SelectVariants were run with default parameters, and VariantFiltration with options “QD_filter” -filter “QD < 2.0”, -filter-name “FS_filter” -filter “FS > 60.0”, -filter-name “MQ_filter” -filter “MQ < 40.0”, -filter-name “SOR_filter” -filter “SOR > 4.0″, -filter-name “MQRankSum_filter” -filter “MQRankSum < -12.5”, -filter-name “ReadPosRankSum_filter”, -filter “ReadPosRankSum < -8.0”. This initial run served as input for BaseRecalibrator, ApplyBQSR, and BaseRecalibrator steps. Thes e steps were then repeated with the same options to call single nucleotide polymorphisms (SNPs) between the chloroplast transcriptome and genome. RNA editing sites were identified by examining C-to-T or G-to-A conversions. De novo RNA editing sites induced by dPPRe^rpl2expression were identified by detecting base conversions specifically present in experimental libraries (dPPRe^rpl2/P19) but absent in control libraries (P19). The editing sites introduced by the expression of dPPRe^rpl2 were considered significant if they were detected in at least two independent libraries. ERs were calculated as the ratio between the number of SNP reads and read depth at the specific base position. The spearman correlation test (cor.test in R 4.2.0) was used to assess the correlation between the ER and expression level of dPPRe^rpl2.

The MA-plot and correlation test graphs were generated using the ggplot2 3.5.1 R package. Secondary structures were predicted with the mfold server (RNA folding form version 2.3) at http://www.unafold.org/ using default parameters and a folding temperature of 24°C.

Expression of rdPPRe

The E. coli recombinant dPPRe (rdPPRe) expressing vectors were cloned in one step by DNA assembly of overlapping DNA fragments using the NEBuilder HiFi DNA Assembly Master Mix following the manufacturer’s instructions (New England Biolabs). The nucleotide fragments coding for the predicted mature dPPRe (i.e. lacking the transit peptide) or the pMAL vector were amplified using Q5 High-Fidelity DNA polymerase (New England Biolabs) with primers k1504/k1505 and k1502/k1503, respectively. The final vectors express mature dPPRe^rpl2 and dPPRe^ndhB fused in frame to the N-terminal maltose binding protein tag (MBP) and were transformed into BL21 (DE3) pLysS cells. The expression of rdPPRe was induced overnight at 16°C by adding 0.5 mM IPTG, with constant agitation at 220 rpm. Bacterial cells were then lysed in a cold buffer containing 30 mM Tris–HCl, pH 7.5, 400 mM NaCl, 10% glycerol, 0.05% 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (CHAPS), 5 mM β-mercaptoethanol, and an ethylenediaminetetraacetic acid-free protease inhibitor cocktail (Roche). Soluble rdPPRe was first affinity-purified using an amylose column (New England Biolabs) and subsequently resolved on a Superdex S200 column (GE Healthcare Life Sciences) in lysis buffer lacking protease inhibitors. Peak fractions containing rdPPRe were pooled and dialyzed overnight in elution buffer containing 50% glycerol. The purity of the tandem affinity-purified protein was assessed by SDS–PAGE followed by Coomassie Brilliant Blue staining. Recombinant proteins were stored at −20°C and used within 10 days. Recombinant maltose-binding protein (rMBP) was purified following TEV digestion of rHCF107 as described [52].

Gel mobility shift assays

Synthetic RNAs (Integrated DNA Technologies) were purified on a denaturing polyacrylamide gel and 5′-end–labeled with [γ-32P]–ATP and T4 polynucleotide kinase. Unincorporated radio-labelled nucleotides were removed by filtration on illustra Microspin G-25 columns (GE Healthcare) followed by phenol-chloroform extraction and ethanol precipitation. Binding reactions contained 150 mM NaCl, 30 mM Tris–HCl, pH 7.5, 4 mM Dithiothreitol, 0.04 mg/ml bovine serum albumin, 0.5 mg/ml heparin, 10% glycerol, 0.01% CHAPS, 10 units RNaseOUT (Invitrogen), and 15 pM radiolabeled RNA. Reactions were incubated for 60 min at 25°C and resolved on 8% native polyacrylamide gels. Results were visualized on a Typhoon phosphorimager. Data quantification was performed with ImageQuant TL software (Amersham).

Statistics and reproducibility

All experiments were repeated at least three times. The number of samples for each experience is provided in the figures. The statistical significance of the RNA ER differences between experimental and control groups or between experiments in Fig. 3C was assessed by t-test using GraphPad Prism 8.4.0 (t-values and degrees of freedom can be found in Supplementary Table S2).

Results

Engineering synthetic dPPRe for precise RNA base editing

To date, two synthetic editing factors with PLS-type or SS-type PPR protein arrangements have been engineered to edit the chloroplast rpoA-78,691 site in vivo, a site naturally targeted by the endogenous CLB19 PPR editing factor in Arabidopsis [36, 39]. The design of these factors involved concatenating synthetic domains derived from consensus P, L, S, P2, L2, S2, E1, E2, and DYW motifs to create protein scaffolds that mimic natural plant editing factors (Fig. 1A). Here, rather than replicating the natural architecture of plant editing factors, we took a different approach to design PPR editors (dPPRe) for precise de novo RNA editing in plant organelles. We hypothesized that any sufficiently long PPR tract with RNA-binding specificity for the sequence located six nucleotides upstream of a target cytidine (C) could effectively position the DYW domain and enable cytidine deamination in vivo.

Unlike PLS-type PPR arrays, which contain L motifs whose contribution to RNA binding is unpredictable, we reasoned that P-type PPR arrays would be more suitable for reprogramming RNA editing specificity in vivo. This is due to their clear and predictable RNA-binding specificity, as all P motifs are involved in sequence recognition [25, 26]. To test this, we constructed guides composed of a tandem array of 10 synthetic P-type PPR motifs, termed designer PPR (dPPR) [37] (Fig. 1B), which we and others have previously demonstrated to possess programmable RNA-binding specificity in planta [41, 42, 53]. A synthetic consensus P2–S2–L2 triplet was added after the last P motif to replicate the structural arrangement found in natural PPR editing factors [39]. While the P2 motif is expected to confer RNA-binding specificity similar to P motifs, the roles of the L2 and S2 motifs in editing factors remain less clear [28, 54]. A guide composed of 11 P-type PPR repeats should be sufficient to specifically target an 11-nucleotide RNA sequence within plant chloroplast and mitochondrial genomes of few hundred kilobase pairs.

For the cytidine deamination reaction, we appended the C-terminal E1–E2–DYW domains from the Physcomitrium patens mitochondrial editing factor PPR56 to the synthetic [P]₁₀–P2–L2–S2 scaffold. This particular E-DYW domain was chosen due to PPR56’s demonstrated ability to catalyze cytidine deamination across various biological environments, including plant mitochondria [45] and cytoplasm [54], E. coli [28, 33], and human cell cytoplasm [55]. To stabilize the chimeric protein, an N-terminal domain from the maize chloroplast-localized PPR protein PPR10 [56] was added upstream of the first P motif, resulting in the engineered dPPRe. Finally, we applied the PPR code to the [P]₁₀-P2 motifs to target the RNA sequence located six nucleotides upstream of the selected cytidines, respectively.

For in planta editing assays, an N-terminal chloroplast transit peptide from Arabidopsis RecA (first 68 amino acids) or a mitochondrial presequence from yeast Cox4 (first 29 amino acids), both of which have been shown to efficiently direct fusion proteins to the respective organelles in plant cells [41, 57], were added to the dPPRe constructs. A 3×HA tag was also included for immunodetection.

Initially, we engineered two chloroplast dPPRe constructs, dPPRe^rpl2 and dPPRe^ndhB, to target specific cytidines in the rpl2 and ndhB mRNAs of N. benthamiana, whose chloroplast genome has been recently sequenced [49] (Fig. 1B). These chloroplast genes encode a subunit of the large subunit of the chloroplast ribosome and the NADH dehydrogenase-like complex, respectively. Natively, these specific cytidines are not edited in the chloroplast transcriptome of N. benthamiana [49]. Editing the cytidine in the duplicated rpl2 gene at positions 87,669 or 154,376 would affect the last base of the codon (CUC > CUU) without altering the amino acid sequence of the RPL2 protein (Leu > Leu) (Fig. 1B). In contrast, editing the cytidine at position 14,727 in ndhB mRNA is expected to change a CAA codon to UAA, introducing a premature stop codon and truncating the NdhB protein.

Before expressing the constructs in planta, we verified their functionality using a well-established heterologous E. coli RNA editing assay that allows co-expression of rdPPRe without their transit peptide as N-terminal fusion proteins with a six-histidine and MBP tags, along with their RNA targets [33] (Fig. 2). Upon IPTG induction of rdPPRe^rpl2 and rdPPRe^ndhB expression in E. coli, a specific ∼130 kDa protein band corresponding to the expected molecular weight of the recombinant proteins accumulated in the bacterial total protein extracts (Fig. 2A). Sanger sequencing of cDNA from these extracts confirmed that rdPPRe successfully edited their target cytidines in E. coli (Fig. 2B and Supplementary Table S1). Specifically, rdPPRe^rpl2 edited its target cytidine with an efficiency ranging from 49% to 62% (average of 57% ± 6.8%), while rdPPRe^ndhB exhibited a lower editing efficiency, averaging 33% ± 3.7%. These results demonstrate that our chimeric dPPRe are functional for RNA editing in vivo and that synthetic P-type PPR arrays can effectively guide RNA editing by the DYW domain, similar to PLS- and S-type scaffolds.

Expression of dPPRe in plants induces de novo RNA editing in chloroplasts

Next, we evaluated the RNA editing activities of dPPRe^rpl2 and dPPRe^ndhB in chloroplasts through agroinfiltration-mediated transient gene expression in N. benthamiana. To enhance transient expression levels, dPPRe constructs were co-expressed with the tombusvirus suppressor of silencing, P19-HA [48]. Initially, we optimized assay conditions by testing in vivo protein accumulation of the most efficient dPPRe from E. coli, dPPRe^rpl2, when driven by a single or double 35S promoter (2×35S), a strong and constitutive promoter, at 2 and 3 days post-infiltration (dpi). Immunodetection with anti-HA antibodies revealed that the 2×35S promoter facilitated higher accumulation of dPPRe^rpl2 protein (expected molecular weight of ∼97 kDa) in N. benthamiana leaves compared to the single 35S promoter, with protein levels peaking at 3 dpi for both promoters (Fig. 3A).

In parallel, we assessed RNA ERs at the rpl2-87,669/154,376 sites in chloroplasts upon dPPRe^rpl2 expression by cDNA sequencing. The results demonstrated that dPPRe^rpl2 induced editing of the target cytidine in rpl2 mRNA to an average of 38% at 2 dpi, increasing to 63% at 3 dpi while the expression of P19-HA alone did not cause editing (Fig. 3B and C, and Supplementary Table S2). A t-test confirmed the editing increase to be highly significant (P-value < 0.05). Notably, editing efficiency at the target site could reach up to 77% at 3 dpi (Fig. 3C and Supplementary Table S2). These findings suggest that dPPRe^rpl2 functions as an effective RNA editing factor in plant chloroplasts, with higher activity observed at 3 dpi, indicating a doseable effect of dPPRe^rpl2 on rpl2 editing efficiency in vivo.

Subsequently, we expressed dPPRe^ndhB in N. benthamiana and examined RNA editing at ndhB-14,727 site at 3 dpi. Like dPPRe^rpl2, dPPRe^ndhB showed higher protein accumulation with the 2×35S promoter (Fig. 3A). However, dPPRe^ndhB induced a lower ER in planta, achieving an average of 19% C-to-U conversion at its target site, reflecting the results obtained in E. coli (Fig. 3B and C, and Supplementary Table S2). This difference in editing efficiency between dPPRe^rpl2 and dPPRe^ndhB in E. coli and in planta was not due to differences in protein accumulation (Figs 2A and 3A). To determine whether this could be attributed to a difference in dPPRe affinity for the respective RNA molecules, electrophoretic mobility shift assays were performed using rdPPRe fused to the N-terminal MBP, and 19-nt RNAs spanning the −17/+1 cis-elements of the target cytidines, respectively, and with minimal expected secondary structures (Fig. 4A and B, and Supplementary Fig. S1). The results showed that rdPPRe^rpl2 and rdPPRe^ndhB bind their RNA targets with sequence specificity, whereas rMBP alone exhibited no RNA-binding activity (Fig. 4B). Quantification of the binding affinity revealed that both proteins bind their respective RNA targets with an approximate K_D in the nanomolar range, with rdPPRe^ndhB displaying a higher affinity for its RNA sequence than rdPPRe^rpl2 (K_D ∼2.6 nM versus 14.6 nM) (Fig. 4C and Supplementary Fig. S2). These results indicate that RNA-binding affinity is not responsible for the lower in vivo editing activity of dPPRe^ndhB compared to dPPRe^rpl2.

Since PPR protein binding to RNA can be hindered by the formation of RNA secondary structures [15, 58, 59], we hypothesize that the stronger hairpin structure predicted at the dPPRe^ndhB binding site in vivo could interfere with RNA recognition (Supplementary Fig. S3). This structural constraint may reduce the accessibility of the target sequence, thereby impairing dPPRe^ndhB binding and leading to the lower RNA editing efficiency observed in vivo.

Overall, these findings demonstrate that chimeric dPPRe proteins, composed of synthetic P-PPR guides and a mitochondrial E1–E2–DYW domain from moss, are functional in chloroplasts of angiosperms for de novo RNA editing. The results also suggest that the local folding of RNA target sequences plays an important role in determining dPPRe editing efficiency in vivo, consistent with previous reports for natural PPR editing factors [28, 60].

dPPRe^rpl2 exhibits minimal off-target activity in planta

To evaluate the off-target RNA editing activity of dPPRe, we focused on the most efficient variant, dPPRe^rpl2, and analyzed potential off-target sites in N. benthamiana chloroplasts. We performed RNA sequencing (RNA-seq) on samples from three independent agroinfiltration experiments. In these experiments, leaves were infiltrated either with the P19 protein alone (control) or with dPPRe^rpl2/P19 driven by either a 35S or 2×35S promoter, resulting in nine cDNA libraries. To assess the relationship between dPPRe^rpl2 expression levels and RNA ERs, RNA was extracted at two time points: 2 dpi for one cDNA library and 3 dpi for two cDNA libraries in each of the three experiments.

Differential expression analysis using DESeq2 revealed no significant changes in chloroplast gene expression across the samples, with the exception of the dPPRe^rpl2 nuclear transgene, which was significantly overexpressed in the experimental samples (Fig. 5A and Supplementary Table S3). This finding indicates that dPPRe^rpl2 expression does not disrupt the overall regulation of chloroplast gene expression.

Next, we used the GATK4 toolkit to identify de novo C-to-U RNA editing sites triggered by dPPRe^rpl2 expression (Supplementary Table S4). As expected, the target cytidine in the rpl2 gene was edited (Fig. 5B). However, we also detected three off-target editing sites in the matK, rpoC1, and psbB chloroplast genes. Editing at the rpl2 target site and the matK-2329 site was consistently observed across all dPPRe^rpl2-expressing libraries, while editing at the rpoC1-22,069 and psbB-75,097 sites was detected in only five of the six libraries (Fig. 5B).

We quantified the ER at these four sites in each experimental library in relation to the expression levels of dPPRe^rpl2 (Fig. 5B). Consistent with previous protein immunodetection results, the 2×35S promoter drove higher expression of dPPRe^rpl2 than the single 35S promoter, and dPPRe^rpl2 expression was higher at 3 dpi compared to 2 dpi for both promoters. The quantification revealed that the rpl2 target site had the highest ER across all libraries, reaching up to 67.8%, followed by matK-2329 (ER ≤ 46.3%), rpoC1-22,069 (ER ≤ 31.4%), and psbB-75,097 (ER ≤ 21.9%) (Fig. 5B). These results suggest that dPPRe^rpl2 preferentially guides RNA editing towards its intended cytidine target over off-target sites in vivo. Moreover, comparison of ERs across independent experiments indicated a positive correlation between dPPRe^rpl2 expression levels and RNA editing efficiency (Fig. 5B), particularly at the rpl2 target site, as confirmed by a Spearman correlation test (P-value < 0.05) (Fig. 5C and Supplementary Fig. S4).

To further investigate the off-target activity of dPPRe^rpl2, we aligned the cis-element sequences spanning from nucleotide −16 to +1 relative to each off-target site with that of the rpl2 target site. This analysis revealed a conserved nucleotide sequence among the off-target and rpl2 cis-elements (Fig. 5D). Notably, the first 11 nucleotides, predicted to be bound by the synthetic [P]₁₀-P2 guide, exhibited high conservation, with all off-target sites sharing 8 nucleotides with the rpl2 target site within this region. Additionally, a U was observed at position −3, and an A immediately downstream of the edited cytidine in all off-targets.

Further analysis of the nucleotide preferences of the 11 PPRs in dPPRe^rpl2 showed that some motifs tolerate mismatches along the 11-nucleotide target sequence (Fig. 5D). The number and position of these mismatches within the PPR–RNA duplex appeared to influence in vivo editing efficiency, likely by affecting protein-RNA interactions. For example, the most efficiently edited off-target, matK-2329, contained two mismatches positioned one nucleotide apart in the central region of the dPPR binding site. In contrast, the less efficiently edited sites, rpoC1-22,069 and psbB-75,097, exhibited two to three mismatches distributed differently: rpoC1-22,069 had three mismatches spread across the 5′ and 3′ ends of the RNA, while psbB-75,097 had two mismatches located between the 5′ end and the middle of the RNA. These findings suggest that the dPPR guide is more tolerant of mismatches in the central region of its binding site than at the 5′ or 3′ ends.

dPPR editors are active in plant mitochondria

Finally, we evaluated the potential of dPPRe for de novo RNA editing in plant mitochondria. For this purpose, we engineered dPPRe^nad7 to edit a specific cytidine in the nad7 mRNA within N. benthamiana mitochondria, which encodes a subunit of the respiratory complex I (Fig. 1B). Since the mitochondrial genome of N. benthamiana has not yet been sequenced, we selected this editing site based on the sequence of the mitochondrial nad7 gene from Nicotiana tabacum, assuming that the sequence is conserved between the two species. Similar to the editing event at ndhB-14,727, RNA editing at the selected cytidine in the mitochondrial nad7 gene is expected to create a premature stop codon in the mRNA (CAA > UAA) (Fig. 1B).

The editing activity of rdPPRe^nad7 was assessed in E. coli (Fig. 2). Upon induction, rdPPRe^nad7 accumulated to a much higher level in E. coli compared to rdPPRe^rpl2 and rdPPRe^ndhB, leading to RNA editing at the nad7 target site with an average efficiency of 74% (Fig. 2B). While the secondary structure of nad7 RNA expressed in E. coli is not predicted to be stronger than that of rpl2 or ndhB RNAs (Supplementary Fig. S3), the higher accumulation of rdPPRe^nad7 compared to rdPPRe^rpl2 and rdPPRe^ndhB may be responsible for its increased editing efficiency in bacteria (Fig. 2A). Finally, dPPRe^nad7was expressed under the control of a 2×35S promoter in N. benthamiana leaves and successfully accumulated by 3 dpi, as confirmed by anti-HA immunodetection. cDNA sequencing confirmed nad7 sequence and revealed that the expression of dPPRe^nad7 in plant cells induced RNA editing at the target cytidine within mitochondria, achieving an ER of up to 42% (Fig. 3B and C). This result demonstrates that synthetic dPPRe are functional ribocytidine editors in plant mitochondria.

Discussion

The results of this study highlight the significant potential of synthetic designer PPR RNA editors, as a powerful tool for precise RNA base editing in plant organelles, with successful applications demonstrated in both chloroplasts and mitochondria. Our engineering of dPPRe constructs, which incorporate arrays of synthetic P-type PPR motifs [37] fused with the DYW cytidine deaminase domain from the moss PPR56 editing factor, represents a significant advancement over previous designs that relied on PLS-type or SS-type PPR arrangements.

Previously, the first synthetic PLS-PPR editing factor (referred to as dsnPLS^CLB19) was capable to substitute the function of the natural CLB19 PLS-PPR editing factor at the rpoA-78,691 site in the respective Arabidopsis knockout line [39]. However, efficient editing was only possible in the presence of plant protein cofactors from the RIP/MORF family (Multiple Organellar RNA editing Factors). Structural and biochemical studies suggest that MORF interaction with PLS-PPR editing factors, particularly with the L motifs, enhances RNA-binding efficiency by inducing conformational changes [61]. Consequently, the application of synthetic PLS-PPR scaffolds may be limited to angiosperm plants that possess MORF proteins [62, 63].

To overcome the MORF-dependency, a subsequent SS-type PPR scaffold (dsnS^CLB19) was developed using a similar approach. dsnS^CLB19 edited its RNA target in E. coli without plant MORF cofactors, achieving an ER below 50%, similar to dsnPLS^CLB19 [36]. However, the effectiveness of the dsnS scaffold in planta and the broader applicability of dsnPLS- and dsnS-type editors for reprogramming RNA editing specificity in plant organelles have not been tested yet. In contrast, our study demonstrates that synthetic P-type dPPR editors can induce de novo, site-specific RNA editing in both chloroplasts and mitochondria achieving high efficiency above 60%. This significantly expands the RNA editing toolkit for plant systems.

Why plants have evolved different types of PPR arrays to fulfill distinct functions remains an open question [64]. A direct comparison of the performance of designer P-, S-, and PLS-type PPR arrays fused to DYW deaminase domains, in terms of efficiency and specificity, would be a valuable experiment to conduct in the future.

However, the preference for PLS-type PPR arrays over classical P-type arrays in land plants for RNA editing site recognition may be linked to the function and positioning of L-type motifs. Unlike P- and S-type PPRs, L-type motifs do not follow a clear recognition code and do not seem to be directly involved in the recognition of individual nucleotides. This characteristic could have facilitated the conserved recognition of specific editing sites in messenger RNAs throughout plant evolution. Given that RNA editing sites frequently occur at the second position of codons and that PLS-type PPRs bind approximately six nucleotides upstream of the editing site, L-motifs are typically positioned opposite the third codon position—an area generally less conserved across species due to the degeneracy of the genetic code. By omitting these variable positions, PLS-type PPRs may have ensured greater target recognition stability over evolutionary time [64]. However, this evolutionary advantage does not necessarily translate into a benefit for the design of synthetic editing factors.

The DYW deaminase of PPR56 has been shown to function in various biological systems including human cells, plants, and bacteria without perturbating the target cell [33, 45, 55]. In this study, we demonstrated that when fused to a programmable P-type PPR array, this deaminase can rewrite the genetic information within a target RNA by enzymatically converting a specific cytidine in both plant chloroplasts and mitochondria. This makes dPPRe a versatile tool for site-directed RNA base editing. However, like other site-directed RNA editing approaches, dPPRe face challenges related to editing efficiency and target scope. Although dPPRe can achieve a satisfactory RNA editing efficiency over 50% for biotechnological applications, certain plant PPR editing factors, such as the native PPR56, can nearly fully convert their cytidine target in vivo [33, 45]. The lower efficiency of the DYW domain of PPR56 in dPPRe may be due to the cooperative role of this domain with the PPR array for RNA recognition and editing [65, 66]. This cooperation might not be fully maintained in synthetic PPR scaffolds, potentially resulting in reduced in vivo activity.

In addition, similar to ADAR (Adenosine Deaminases Acting on RNA) and APOBEC (Apolipoprotein B Editing Catalytic polypeptide) base editors [67–69], the DYW deaminases of natural PPR editing factors and their associated E1–E2 domains have nucleotide context preferences, preferring cytidine that are embedded in specific RNA sequences [65, 66, 70]. For instance, the −3/+1 nucleotide sequences surrounding the dPPRe^rpl2 off-target cytidines in chloroplasts are rich in U/C and A (Fig. 5D), aligning with the described preferences of PPR56 [54, 65, 70]. This nucleotide context preference might also explain the lower editing efficiency of dPPRe^ndhB in vivo: sequence analysis of dPPRe^rpl2 off-targets revealed a strict preference of PPR56’s E1–E2–DYW domain for U at position −3, while the ndhB target C site has an A at this position (Fig. 1B). While this nucleotide preference is beneficial for the engineering of site-specific RNA editors by increasing editing precision, it limits the number of editable cytidines in a given transcriptome.

Furthermore, this study provides important insights into the precision and efficiency of RNA editing achieved by dPPRe in planta. While dPPRe^rpl2 exhibited robust ERs at its target site in the rpl2 gene in planta, we also detected a few off-targets. These off-target sites, however, shared significant sequence similarities with the intended target, particularly in the regions predicted to interact with the synthetic PPR motifs. Interestingly, our results showed that when synthetic P-type PPR motifs are fused with the C-terminal P2–L2–S2–E1–E2–DYW domains to create dPPRe, these motifs primarily tolerate mismatches in the central region of their RNA binding site (Fig. 5D). This indicates that, in the context of dPPRe, the P-type PPR motifs may exhibit somewhat different binding behavior compared to purely synthetic P-type dPPR scaffold (composed entirely of P-type PPR motifs), which typically shows more permissiveness to mismatches at the 3′ end of the RNA binding site [38]. In contrast, natural PLS-type PPR editing factors have been shown to tolerate mismatches more frequently at the 5′ end of the binding site, with less tolerance towards the 3′ end, close to the editing site [33, 71]. Given that dPPRe combines features of both P-type and PLS-type architectures, it is not surprising that mismatch tolerance is primarily constrained to the central region, while stricter complementarity is required at the 5′ and 3′ ends. This suggests that the hybrid nature of dPPRe imposes a unique pattern of mismatch tolerance, influenced by both the P-type and PLS-type elements, leading to a distinct RNA binding specificity. Although the synthetic S2 motif used in this study was previously reported to preferentially recognize A in vivo and contribute to the RNA site specificity in synthetic PLS editing factors [39], our findings suggest otherwise. In fact, all four RNA bases are represented at this position in the in vivo target sites of dPPRe^rpl2 (Fig. 5D).

These findings underscore the importance of carefully designing PPR motifs to minimize unintended interactions, as PPR–RNA interactions are not solely defined by the PPR RNA-binding code itself. For instance, when we applied the PPR code to predict binding sites for the [P]₁₀-P2 motifs of dPPRe^rpl2 and possible cytidine targets within the chloroplast genome of N. benthamiana using the Prepact Tool TargetScan [72], the designated cytidine target in rpl2 was identified as the top hit (position 154,376; Supplementary Fig. S5). However, none of the other predicted targets were edited in plants expressing dPPRe^rpl2, and the three off-targets observed in planta were not among the top 40 predicted matches. This discrepancy lies in the difficulty to evaluate the energetic costs of mismatches and their positional effects within a PPR–RNA duplex, competing RNA secondary structures or protein occupancy, and the influence of the C-terminal DYW domains on nucleotide specificity [54, 65, 70, 73]. While one possible strategy to enhance specificity could be to increase the length of synthetic PPR tracts, previous studies indicate that elongating the dPPR array beyond 11 motifs does not systematically improve target discrimination. In fact, increasing the number of PPR motifs can, in some cases, lead to greater tolerance of mismatches rather than enhanced specificity [38, 53]. A deeper understanding of these factors will significantly enhance our ability to predict PPR–RNA interactions and improve the specificity of dPPRe design without compromising editing efficiency in the future.

Recently, the use of the E. coli editing system and a plant cytosolic editing assay, demonstrated that the DYW domains from plant editing factors besides PPR56 show some variations in their flexibility towards the nucleotide context around the cytidine target site [54, 70]. These DYW domains offer great promise for broadening the editing site-specific capability of dPPRe. Although synthetic E1, E2, and DYW domains engineered from consensus amino acid sequences are currently available [39], their preferred nucleotide context remains unknown and complicates their use for de novo RNA editing in vivo.

In conclusion, the successful application of dPPRe in E. coli, chloroplasts and mitochondria emphasizes their versatility for targeted RNA editing and provides the first mitochondrial RNA base editor tool kit. Mitochondrial dysfunction is associated with a range of phenotypic abnormalities in plants, including cytoplasmic male sterility, which is of considerable interest in hybrid seed production [74]. The ability to precisely edit mitochondrial genes could therefore have profound implications for the development of new plant varieties with improved traits. Beyond plant systems, dPPRe hold great promise for mitochondrial research and associated human diseases [44]. Given that PPR56 is active for RNA editing in human cells [55], dPPRe technology could pave the way for innovative therapeutic strategies targeting mitochondrial genes. As we continue to refine and expand the capabilities of dPPRe, this technology holds the potential to become a versatile and indispensable tool in organellar genetic research.

Acknowledgements

The authors thank Volker Knoop for useful discussions and Sarah Schwirblat for technical assistance.

Author contributions: S.M., E.L., M.S.-R., and K.H. designed and conceived the experiments. E.L. performed the E. coli experiments. S.M., Sh.G., and K.H. performed the in planta work. K.H. performed EMSA assays. S.G. performed bioinformatic analysis. S.M., E.L., S.G., and K.H. analyzed the results. K.H. supervised the work, made the figures, and drafted the manuscript. S.M., E.L., and M.S.-R. edited the manuscript.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by Agence National de la Recherche [ANR-18-CE20-0013 to K.H.] and Centre National de la Recherche Scientifique. Work of M.S-R. and E.L. was funded by the Deutsche Forschungsgemeinschaft [SCHA 1952/2-2]. Funding to pay the Open Access publication charges for this article was provided by Centre National de la Recherche Scientifique.

Data availability

The read data were deposited to the NCBI Sequence Read Archive database with the Bioproject identifier GSE276199.

Notes

Present Address: Institut für Biologie und Biotechnologie der Pflanzen, Abteilung Grüne Biotechnologie, Universität Münster, 48143 Münster, Germany

References

Author notes