Conservation of the moss RNA editing factor PPR78 despite the loss of its known cytidine-to-uridine editing sites is explained by a hidden extra target

Abstract

Cytidine (C)-to-uridine (U) RNA editing in plant organelles relies on specific RNA-binding pentatricopeptide repeat (PPR) proteins. In the moss Physcomitrium patens, all such RNA editing factors feature a C-terminal DYW domain that acts as the cytidine deaminase for C-to-U conversion. PPR78 of Physcomitrium targets 2 mitochondrial editing sites, cox1eU755SL and rps14eU137SL. Remarkably, the latter is edited to highly variable degrees in different mosses. Here, we aimed to unravel the coevolution of PPR78 and its 2 target sites in mosses. Heterologous complementation in a Physcomitrium knockout line revealed that the variable editing of rps14eU137SL depends on the PPR arrays of different PPR78 orthologues but not their C-terminal domains. Intriguingly, PPR78 has remained conserved despite the simultaneous loss of editing at both known targets among Hypnales (feather mosses), suggesting it serves an additional function. Using a recently established RNA editing assay in Escherichia coli, we confirmed site-specific RNA editing by PPR78 in the bacterium and identified 4 additional off-targets in the bacterial transcriptome. Based on conservation profiles, we predicted ccmFNeU1465RC as a candidate editing target of PPR78 in moss mitochondrial transcriptomes. We confirmed editing at this site in several mosses and verified that PPR78 targets ccmFNeU1465RC in the bacterial editing system, explaining the conservation and functional adaptation of PPR78 during moss evolution.

IN A NUTSHELL

Background: Plant-type RNA editing refers to the site-specific deamination of cytidines (C) to uridines (U) in organellar transcripts. Pentatricopeptide repeat (PPR) proteins of the PLS subfamily are key players in this process. In mosses, PPR editing factors consist of an RNA-binding PLS-type PPR array and a C-terminal DYW-type cytidine deaminase domain. PPR78 is 1 of 9 organellar RNA editing factors in the model moss Physcomitrium patens. It edits 2 mitochondrial targets to different extents: the cox1eU755SL site is edited efficiently, whereas the rps14eU137SL site is edited partially. Establishing the cox1 site is also crucial in other mosses, whereas editing of rps14eU137SL is highly variable, ranging from nonexistent to very efficient.

Question: We aimed to unravel the coevolution of PPR78 and its 2 target sites in mosses. What causes the variability in rps14 editing efficiency? What happens to PPR78 if its 2 target sites are lost due to genomic C–T conversions creating a preedited state?

Findings: An early species sampling was widely extended, focusing on the Hypnales (feather mosses). In this moss order, the cox1 site is preedited, making editing obsolete, and the rps14 site is edited inefficiently or not at all, while PPR78 orthologues remain conserved in most species. A complementation study of a Physcomitrium knockout (KO) mutant with PPR78 orthologues from different mosses revealed that variable rps14 editing is defined by the proteins themselves, specifically their PPR arrays. We successfully expressed functional PPR78 in Escherichia coli. Off-target profiling helped predict a third editing target site of PPR78 in Hypnales: ccmFNeU1465RC. We functionally assigned PPR78 to the ccmFN site in bacterio, providing evidence that PPR78 is evolutionarily retained in Hypnales to serve this additional editing target.

Next steps: In the future, assignments of further editing sites to PPR78 may be explored. The region responsible for variable rps14eU137SL editing capacities of PPR78 orthologues may be narrowed down to certain PPR repeats. In addition, ccmFNeU1465RC editing functionalities of PPR78 orthologues may be tested by KO studies in another moss or heterologous expression.

Introduction

RNA editing by site-specific cytidine (C)-to-uridine (U) deamination significantly affects the chloroplast and mitochondrial transcriptomes in land plants (embryophytes) and mainly acts as a repair mechanism to reestablish conserved amino acid codons (Takenaka et al. 2013b; Yan et al. 2018; Small et al. 2020; Ichinose and Sugita 2021; Knoop 2023). The abundance of plant organelle RNA editing sites varies widely. During plant evolution, this process was uniquely lost altogether in the complex thalloid marchantiid liverworts (Steinhauser et al. 1999; Groth-Malonek et al. 2007; Rüdinger et al. 2008; Rüdinger et al. 2012; Dong et al. 2019), but more than a thousand edits affect the organellar RNAs in certain hornworts (Gerke et al. 2020), lycophytes (Hecht et al. 2011; Oldenkott et al. 2014), or ferns (Zumkeller et al. 2023). Flowering plants (angiosperms) generally feature dozens of C-to-U edits in their chloroplasts and hundreds in their mitochondria (Giegé and Brennicke 1999; Bentolila et al. 2013; Edera et al. 2018; Small et al. 2020).

Early functional studies of plant organelle RNA editing used in vitro analysis (e.g. Miyamoto et al. 2002; van der Merwe et al. 2006; Hayes and Hanson 2007), in organello analysis (e.g. Farré et al. 2001; Staudinger and Kempken 2003), experimentation with transplastomic plants (e.g. Chaudhuri and Maliga 1996; Reed et al. 2001; Chateigner-Boutin and Hanson 2002; Karcher et al. 2008), or e.g. analysis making use of the allotetraploid ancestry of tobacco (Nicotiana tabacum) (Tillich et al. 2006). The latter approaches in particular provided the first insights linking chloroplast RNA editing events with their specific cofactors and found that RNA editing activities may be retained when editing targets are lost by C-to-T conversion in an organelle genome. The first clear assignment of specific RNA editing factors to chloroplast and mitochondrial target sites, however, ultimately relied on reverse genetic approaches using the flowering plant model Arabidopsis (thale cress, Arabidopsis thaliana) (Kotera et al. 2005; Zehrmann et al. 2009).

Many RNA editing protein factors have meanwhile been characterized in angiosperms, ultimately revealing complex multiprotein editosomes that act on most editing targets (Bentolila et al. 2012; Takenaka et al. 2012; Takenaka 2014; Sun et al. 2016; Andrés-Colás et al. 2017; Guillaumot et al. 2017; Gutmann et al. 2017; Sandoval et al. 2019; Yang et al. 2022; Wang et al. 2023). In contrast, the organelle RNA editing machinery appears to be much simpler in the spreading earthmoss (Physcomitrium, [previously Physcomitrella] patens) as a model system (Rensing et al. 2020). Single proteins containing an upstream array of RNA-binding pentatricopeptide repeats (PPRs) responsible for the specific recognition and binding of RNA targets and a terminal DYW-type cytidine deaminase domain serve as specific C-to-U RNA editing factors in Physcomitrium. Proteins of this kind are also at the core of the RNA editing processes in flowering plants, but here, they seem to rely on additional helper proteins at most C-to-U editing sites (Bentolila et al. 2012; Takenaka et al. 2012). In particular, the functions of RNA target recognition and cytidine deamination have been separated in many cases in angiosperms, and the respective truncated protein components need to be reassembled in trans with the help of additional protein factors (Boussardon et al. 2012; Andrés-Colás et al. 2017; Guillaumot et al. 2017).

The DYW domain was long suspected to carry the cytidine deaminase function given its phylogenetic distribution and its highly conserved catalytic and zinc-binding signature motifs (Salone et al. 2007; Iyer et al. 2011; Boussardon et al. 2014; Hayes et al. 2015, 2013; Wagoner et al. 2015). It was recently confirmed (beyond a reasonable doubt) that the DYW domain performs specific C-to-U conversions (Oldenkott et al. 2019; Hayes and Santibanez 2020; Takenaka et al. 2021). It must be noted that the early concept of a carboxyterminal DYW domain linked via “E” and “E+” motifs to upstream PPR arrays (Lurin et al. 2004) has been replaced by the definition of “E1” and “E2” motifs with distant similarity to tetratricopeptide repeats (TPRs) and an amino-terminally longer DYW domain beginning with a characteristic PG box motif (Hayes et al. 2013; Cheng et al. 2016).

The considerably simpler (and likely evolutionary ancestral) makeup of RNA editing in P. patens as a representative for an early branching land plant clade has ultimately led to the complete mutual assignment of its 13 organelle RNA editing sites to 9 DYW-type RNA editing factors in the moss (Ichinose et al. 2013, 2014; Schallenberg-Rüdinger et al. 2013; Schallenberg-Rüdinger and Knoop 2016). Among these, PPR78 appeared to be particularly attractive for further studies for several reasons. Firstly, PPR78 acts on 2 mitochondrial target sites in parallel, cox1eU755SL (cytochrome c oxidase subunit 1 gene) and rps14eU137SL (protein 14 of the small ribosomal subunit gene) (Rüdinger et al. 2011; Uchida et al. 2011; Schallenberg-Rüdinger et al. 2017). Secondly, and despite converting serine into highly conserved leucine codons in the 2 mitochondrial proteins, the PPR78 knockout (KO) mutant shows no significant phenotype, allowing straightforward downstream experimentation via functional complementation studies. Thirdly, in contrast to the highly efficient editing of its cox1eU755SL target (with >99% C-to-U conversion), editing at the rps14eU137SL site is both overall less efficient and also highly variable among mosses. Finally, the engineering of PPR78 protein chimeras showed that its DYW domain can be functionally replaced by that of the paralogous editing factor PPR79 (Schallenberg-Rüdinger et al. 2017).

Here, we show that the variable efficiencies to edit the rps14eU137SL site are dictated by the inherent properties of the PPR78 orthologues in different moss species. Moreover, PPR78 has been retained even when RNA editing at its 2 known targets was lost during moss evolution, pointing to an additional moonlighting function for the editing factor. We successfully expressed PPR78 in the recently established heterologous RNA editing assay system in Escherichia coli (Oldenkott et al. 2019). Aside from operating at its native targets, PPR78 revealed a remarkably small number of only 4 off-targets in the E. coli background transcriptome, likely reflecting high target specificity for the evolutionarily conserved PPR78 protein. By optimizing the present PPR-RNA recognition code concept (Barkan et al. 2012; Yagi et al. 2013; Takenaka et al. 2013a; Yan et al. 2019) for PPR78 using its off-target profile, we identified ccmFNeU1465RC in the mitochondrial ccmFN transcript as a further candidate editing target in moss mitogenomes. We confirmed ccmFNeU1465RC to be edited in several moss species, including taxa lacking the 2 previously known editing sites but retaining PPR78. Finally, we were able to confirm ccmFNeU1465RC as an additional editing target for PPR78 in the bacterial assay system, explaining the evolutionary retention of PPR78 in taxa where the 2 initially identified targets were lost. Fittingly, however, PPR78 orthologues appear to be absent in Hypnum (feather moss) species once RNA editing at all 3 target sites has become obsolete.

Results

Variable RNA editing of the 2 known targets of PPR78 among mosses

PPR78 is a mitochondrial RNA editing factor previously characterized to address 2 RNA editing sites in P. patens: cox1eU755SL and rps14eU137SL (Rüdinger et al. 2011; Uchida et al. 2011; Schallenberg-Rüdinger et al. 2017). PPR78 features a typical makeup, with an array of 20 P-, L-, and S-type PPRs followed by the E1 and E2 motifs and a terminal DYW domain (Fig. 1A). An earlier phylogenetic investigation of the 2 targets of PPR78 in P. patens had already revealed that they are variably present and edited in other mosses. Notable in that regard is the likely recent loss of the rps14eU137SL target by C-to-T conversion in the closely related bonfire moss (Funaria hygrometrica), residing in the same family as Physcomitrium (Funariaceae), and a strongly reduced efficiency of rps14eU137SL editing among some mosses of the Hypnales (Schallenberg-Rüdinger et al. 2017).

Figure 1.

Structure and phylogenetic conservation of PPR78 and its targets. A) The P. patens mitochondrial RNA editing factor PPR78 has a characteristic makeup with a “PLS-type” array of PPRs featuring “long” (L) and “short” (S) variants along with the canonical P-type PPRs, followed by the E1 and E2 motifs with distant similarity to TPRs (tetratricopeptide repeats) linking to a carboxyterminal DYW cytidine deaminase domain. The 3 most C-terminal PPRs feature unique conservation profiles and are labeled as P₂, L₂, and S₂. PPR78 has 2 targets for C-to-U conversion in mitochondria of P. patens: cox1eU755SL and rps14eU137SL. The editing site nomenclature (Lenz et al. 2010) indicates affected gene, editing to U (eU), position in the coding sequence and resulting codon change, here serine (S) to leucine (L). The PPR-RNA matching code (Barkan et al. 2012) for amino acid identities in positions “5” and “last” (5 + L) and corresponding nucleotides for P- and S-type PPRs (T/S + N, a; T/S + D, g; N + S, c; N + D, u; N + N, y) reveals mostly matches (green shading) but also transitions (yellow) and mismatches (magenta). B) A cladogram for the phylogeny of selected mosses following NCBI taxonomy (Schoch et al. 2020) and recent phylogenetic insights (Schlesak et al. 2018) with a focus on species for which genome, transcriptome, and mitogenome data are available is shown, along with the status of the 2 previously known targets of PPR78 and the presence of PPR78 orthologues identified in available sequence data (Johnson et al. 2016; Carpenter et al. 2019; Gutmann et al. 2020) or via PCR amplification on available moss samples. Editing site ccmFNeU1465RC is a third target identified in the course of the present study. A current lack of data is indicated with “n.d.” The absence of detectable C-to-U RNA editing or a T on mtDNA level making editing obsolete in a preedited state is indicated with lighter and darker background shading in orange. Weaker (<50%) or stronger (≥50%) C-to-U conversions at a given site are indicated with lighter or darker green background shading. RNA editing efficiencies are listed in detail in Supplemental Table S1. Green asterisks indicate species with PPR78 orthologues selected for further investigation. Alignment and phylogeny of PPR78 orthologues are given in Supplemental Fig. S1 and File S2, respectively. *Means nucleotide information was obtained from another species of the same genus.

We extended the data sampling, adding several freshly collected Hypnales samples and including currently available moss genome and transcriptome data (Fig. 1B). We found a wide conservation of PPR78 orthologues in mosses by scanning available databases and sequencing our own PCR amplification products (Fig. 1B, Supplemental Figs. S1 and S2). More variability was seen, however, for the 2 known mitochondrial RNA editing targets of PPR78 (Fig. 1B, Supplemental Table S1). As expected based on a previous study (Schallenberg-Rüdinger et al. 2017), the cox1eU755SL target was confirmed to be lost altogether in the superorder Hypnanae, with a C-to-T conversion in their mitogenomes, while it is edited efficiently in all other species carrying a C at the DNA level (Fig. 1B). In contrast, maximum variability is seen for RNA editing at the rps14eU137SL target among mosses, ranging from highly inefficient or even no detectable editing at all in the majority of Hypnales species to highly efficient editing in more distant mosses, such as redshank (Ceratodon purpureus) (Fig. 1B). Notably, the total absence of detectable RNA editing at the rps14eU137SL site within the family Hypnaceae appears to be accompanied by an as yet unique lack of PPR78 orthologues in the genus Hypnum (Fig. 1B).

Variable RNA editing efficiencies at the rps14eU137SL target are a function of the different moss PPR78 orthologues

Heterologous complementation of the P. patens PPR78 KO line with its orthologue from F. hygrometrica previously showed that the latter can fully complement editing at both identified targets despite the loss of the rps14 target in the Funaria mitogenome by C-to-T conversion (Schallenberg-Rüdinger et al. 2017). We then explored cases of variable RNA editing efficiencies at the rps14eU137SL target in more distant taxa, ranging from highly efficient editing in C. purpureus to inefficient editing in tree-skirt moss (Pseudanomodon attenuatus) to no detectable editing in brocade moss (Callicladium imponens) (Fig. 1B). We cloned the respective PPR78 orthologues, retaining the N-terminal targeting sequence of the Physcomitrium orthologue into the usual complementation setup (Fig. 2A) between the strong ACTIN promoter and the NOS terminator and flanked by homologous regions of the P. patens intergenic (PIG) region (Okano et al. 2009; Schallenberg-Rüdinger et al. 2017). At least 3 independent lines obtained after transformation using recombinant constructs were investigated for the complementation of RNA editing at the 2 targets of PPR78 (Supplemental Table S2). None of the transgenic Physcomitrium complementation lines showed a recognizable phenotype under laboratory growth conditions (Fig. 2B).

Figure 2.

Expressing distant PPR78 orthologues in the P. patens PPR78 KO line. A) Coding sequences of PPR78 orthologues from C. imponens, P. attenuatus, and C. purpureus were combined with the PpPPR78 N-terminus and cloned into the PIG2.0 vector, between the ACTIN1 promoter of rice (Oryza sativa) (Horstmann et al. 2004) and the NOS terminator, followed by a hygromycin resistance cassette. The construct is flanked by sequences, which are homologous to “PIG,” a Physcomitrium intergenic region (Okano et al. 2009). B) PPR78 editing sites cox1eU755SL and rps14eU137SL are edited to different extents in the investigated species. The Physcomitrium KO line lacks editing in both sites (Rüdinger et al. 2011). Native rps14eU137SL editing efficiencies were mirrored upon overexpression of respective moss PPR78 orthologues in the Physcomitrium KO (green shading), while cox1eU755SL editing was efficiently complemented by the orthologues of P. attenuatus and C. purpureus but not at all by that of C. imponens. Data of the plant line exhibiting highest rps14eU137SL editing are shown (all replicates are listed in Supplemental Table S2). No phenotypes could be observed after plants were grown on Knop plates, at 21 °C with a 16 h light (40% intensity) and 8 h dark period for 3 wk (scale bar: 3 mm).

An examination of the RNA editing status at the 2 PPR78 targets in the respective complementation lines suggested that RNA editing is dependent on the orthologues of the different mosses, thus making the influence of other factors in the respective species of origin less likely. The C. purpureus (Cp) orthologue led to highly efficient C-to-U RNA editing at both sites, very much like in the native environment. Interestingly, the PPR78 orthologue of P. attenuatus (Pa) edited the cox1eU755SL target in Physcomitrium very efficiently, although editing at this site is obsolete in Pseudanomodon itself. In contrast, and despite strong overexpression (Fig. 2B, Supplemental Table S2 and Fig. S3), RNA editing at the rps14eU137SL target remained weak using PPR78 of Pseudanomodon, with a maximum of 18% of C-to-U conversion, which is similar to the low degree (10% to 20%) of editing in the native situation (Fig. 2B). The PPR78 orthologue of C. imponens, however, was unable to complement RNA editing activity at either of the 2 target sites, which is in line with the absence of editing at rps14eU137SL and the C-to-T conversion at the cox1eU755SL site in its native environment, even though overexpression was confirmed (Fig. 2B, Supplemental Fig. S3).

Variable editing of PPR78 orthologues does not result from defects in their C-terminal domains

The lack of functional complementation in the Physcomitrium PPR78 KO line by the C. imponens orthologue suggested that this orthologue might harbor a dysfunctional DYW domain despite its overall conservation, including the typical cytidine deaminase signature motifs (Supplemental Fig. S1). To test whether the different RNA editing efficiencies of the orthologues rely on the downstream domains or the upstream PPR arrays, we created reciprocal chimeras between PPR78 of P. patens (Pp) and the 3 different orthologues under study (Fig. 3). To exclude effects from the choice of fusion points, we tested chimeric PPR78 variants with either the DYW domain alone or the entire E1-E2-DYW stretch swapped between the respective orthologues for complementation of the PPR78 KO line (Fig. 3, Supplemental Table S2).

Figure 3.

Chimeric constructs swapping protein domains of PPR78 orthologues. Sequences encoding PPR78 protein chimeras were cloned, swapping the respective domains of 3 different moss orthologues with fusion points either between the upstream PPR array and the E1 motif (top) or between the E2 motif and the DYW domain (bottom). Investigating RNA editing in the PPR78 KO background revealed efficient RNA editing at both known target sites for the terminal domains of all PPR78 orthologues, including those of C. imponens, after linking them to the upstream regions of PPR78 from P. patens. Only partial RNA editing (P. attenuatus) or none at all (C. imponens) was found for the inverse exchanges. For each chimeric construct, editing of the plant line with the highest observed rps14eU137SL editing is shown. Results of additional replicates are listed in Supplemental Table S2.

The experiments showed that variability of RNA editing at the PPR78 targets did not result from differences in the C-terminal domains of the respective orthologues (Fig. 3). Most notably, the E1-E2-DYW domains of the C. imponens (Ci) orthologue when combined with the PLS stretch of Physcomitrium PPR78 efficiently edited both targets (Fig. 3) despite the absence of editing at the rps14eU137SL target in its native environment and the lack of editing function upon expression of the complete orthologue in the Physcomitrium KO line (Fig. 2B). Similarly, chimeras with C-terminal domains of PpPPR78 replaced with those of CpPPR78 and PaPPR78 complemented editing at both target sites efficiently.

Analysis of the respective reciprocal chimeras in which the terminal domains of Physcomitrium PPR78 were fused to the upstream PPR arrays of its lower performing orthologues from Pseudanomodon and Callicladium demonstrated that the different performances in RNA editing relied on the PPR arrays rather than the terminal structures (including the DYW domain). While moderate partial RNA editing of both targets was observed for the Pseudanomodon/Physcomitrium chimera, no editing was detected for the Callicladium/Physcomitrium chimera (Fig. 3).

Variable editing of PPR78 orthologues does not result from characteristic differences in their PPRs

Comparing the amino acid residues in key positions 5 and last (5 and L) within their PPRs, which are crucial for nucleotide discrimination, revealed 4 cases that set the PPR78 orthologues with efficient RNA editing at the rps14eU137SL site (in Physcomitrium and Ceratodon) apart from their counterparts with low or even absent RNA editing (in Pseudanomodon and Callicladium), respectively (Fig. 4A). Such characteristic differences were discernible for L-type PPRs L-17 (MD vs. VD), L-11 (VD vs. LD), and L-5 (M/ID vs. ME). Additionally, a characteristic change was observed for position L of PPR S-4 in CiPPR78, showing an ND amino acid combination where all other orthologues feature NN (Fig. 4A). Accordingly, we examined the functional consequence of these amino acid exchanges and introduced the corresponding mutations into the PPR78 of Physcomitrium for testing in the KO background (Fig. 4B). None of the 5 single amino acid mutations introduced into the L-type PPRs led to any reduction of the highly efficient RNA editing in the complementation lines. The same held true for a double mutant with modified L-5 and L-17 PPRs (Fig. 4B).

Figure 4.

Characteristic differences in the PPRs of PPR78 orthologues. A) Comparing PPR78 orthologues with efficient RNA editing at the rps14eU137SL target (P. patens and C. purpureus) with those showing inefficient (P. attenuatus) or absent editing (C. imponens) reveals differences in L-type PPRs L-17MD vs. L-17VD, L-11VD vs. L-11LD, and L-5MD or L-5ID vs. L-5ME and in S-type PPR S-4NN vs. S-4ND. B) Corresponding point mutants L-17MD > VD, L-11VD > LD, L-5ID > MD, L-5ID > MD, and L-5ID > IE and a double mutation L-17MD > VD + L-5ID > ME were introduced into PPR78 of P. patens, which was tested in the KO background and revealed very efficient RNA editing at both targets (rps14eU137SL, cox1eU755SL). Another mutant PPR (S-4NN > ND) revealed slightly less efficient editing at the rps14 editing site. The reverse exchange S-4ND > NN in the C. imponens orthologue could not rescue RNA editing at either site in the P. patens KO background. Maximal observed editing efficiencies for each construct in 3 independent plant lines are given. Individual results of all replicates are listed in Supplemental Table S2.

Similarly, only a mild reduction of RNA editing at the rps14eU137SL target was observed for the S-4NN > ND mutant, although this amino acid combination should conceptually allow for even better recognition of the U in position −7 of the rps14 vs. the C in the cox1 target (Fig. 4B). Finally, we introduced the reverse mutation S-4ND > NN into the C. imponens orthologue. Like for its native counterpart, however, no RNA editing at the known targets was identified for this mutant either, although overexpression of the construct was successful (Supplemental Table S2 and Fig. S3).

Exploring PPR78 in the heterologous E. coli RNA editing assay

Given the recent success to functionally express 2 P. patens editing factors, PPR56 and PPR65, in E. coli (Oldenkott et al. 2019; Yang et al. 2023a, 2023b), we aimed to utilize a similar approach for PPR78. However, equivalently designed constructs did not yield detectable RNA editing with the PPR78 coding sequence fused to an upstream His₆-MBP (maltose-binding protein) tag (Fig. 5, Supplemental Data Set 1). The successful functional expression of PPR78 turned out to require changes to the experimental setup, including a larger N-terminal extension of the native PPR78 coding sequence beyond its PPR array and an extended target sequence environment (Fig. 5, B and C). Adding 34 native amino acids upstream of the most amino-terminally recognized PPR (L-20) ultimately delivered successful editing of the targets in the heterologous setup. An extended target spanning 100 nucleotides up- and downstream of the editing site, respectively, enhanced the cox1 editing rates significantly compared with the regular target design (with 40 nucleotides of native sequence up- and 5 downstream of the editing position). Surprisingly, editing at the rps14eU137SL site was nearly complete even using the short target variant in the heterologous bacterial system (>99%), whereas the editing of cox1eU755SL reached only ca. 70% even with the extended target variant. As such, the bacterial setup in this case rather inverts the in planta situation, with near-complete editing at cox1eU755SL and partial editing at rps14eU137SL in P. patens.

Figure 5.

Functional analysis of PPR78 in the E. coli assay. A) The heterologous RNA editing assay setup in E. coli is based on GATEWAY cloning of an RNA editing factor flanked by attB attachment sequences downstream of the coding sequence of the maltose-binding protein (MBP) and a His₆ tag. Target sequences comprising 40 bp upstream and 5 bp downstream of the C to be edited were cloned downstream of the coding sequence of the editing factor upstream of a T7 terminator in pETG41Kmod (Oldenkott et al. 2019). Expression of the cotranscript was driven by the T7 promoter and controlled by an IPTG-inducible lac operator (His₆ = 6×histidine tag, RBS = ribosome binding site). B) Unlike than previously observed for PPR56 and PPR65, 14 additional amino acids of the native CDS upstream of the first PPR L-20 (as identified at https://ppr.plantenergy.uwa.edu.au/ppr) and a short target of 46 bp alone were not sufficient to detect RNA editing in the case of PPR78 (Supplemental Data Set 1). Here, RNA editing could only be detected with longer N-terminal extensions of 34, 54, 69, or 89 amino acids, and higher RNA editing efficiency of ca. 70% was only found for a spliced 200 bp variant of the cox1 target site. C) In contrast to the native situation, the rps14 target was edited more efficiently than the cox1 target in E. coli, with up to 100% editing detected even in a short 46 bp target variant. Data for individual replicates are given in Supplemental Data Set 1. D) Four weakly edited off-targets of PPR78 were revealed by RNA-Seq of the E. coli transcriptome upon expression of PPR78:n89L20-DYW. Like the native targets, the off-targets revealed a lack of expected nucleotide identities in positions −21, −19, −15, and −10 but showed unexplained selectivity in positions −11, −8, −4, and +1. RNA editing is indicated for the best-performing replicate; for details, see Supplemental Table S4. The conservation profile was created using WebLogo (Crooks et al. 2004).

To gain more insights into the functionality of PPR78, we used the most efficiently working PPR78 construct with the 89 amino acid extension upstream of PPR L-20 for RNA-Seq analysis to identify off-targets in the E. coli transcriptome. Notably, the data revealed only 4 weakly edited off-targets (Fig. 5D, Supplemental Tables S3 and S4). Like the 2 natural targets, the 4 off-targets matched expectations very well for PPRs according to the PPR-RNA code: P-15TN:a-18, S-13TD:g-16, S-10ND:u-13, P-9TN:a-12, P-6ND:u-9, and S-4NN:y-7. In contrast, expected matches were not seen for upstream PPRs P-18ND (expected u) and S-16NS (expected c) in the off-targets, which is in accordance with the lack of expected fits in those positions in the native targets. An exception in the asdeU113SF target is the matching c-19 opposite of PPR S-16NS, which intriguingly directly flanks the exceptionally mismatching c-18 opposite of P-15TN (Fig. 5D). Equally notable is the unexpected presence of a-15 opposite of PPR P-12NN in 2 of the off-targets, exactly like in the case of the rps14 target and a lack of selectivity opposite of PPRs S-7ND and P-3ND. Then again, a strong and unexpected preference was observed for a-4, g+1, and pyrimidines in position −11.

Searching for additional candidate targets of PPR78

Surprisingly, PPR78 orthologues turned out to be highly conserved among mosses, including several species of the diverse genera Plagiothecium, Entodon, Isothecium, and Callicladium, where editing at the cox1 site was obsolete due to a C-to-T conversion in the mtDNAs and where we could not detect C-to-U editing at the rps14 site (Fig. 1B). Consequently, we considered the possibility that another essential target may be edited by PPR78 in these and other mosses to explain its phylogenetic conservation. To approach this issue, we used the TargetScan function of PREPACT (Lenz et al. 2018) to identify other potential targets in moss mitogenomes, where a thymidine (T) is present in the mitogenome of Physcomitrium but a potential C editing target at the corresponding site might be present in other moss mtDNAs. We use a weighting matrix following just the PPR-RNA code rules for P- and S-type PPRs (Barkan et al. 2012) and added a known bias for pyrimidines and against guanosine in position −1 directly upstream of an editing position (Supplemental Fig. S4A). This placed the 2 native targets of PPR78 top among all 11 documented mitochondrial RNA editing sites in the Physcomitrium mitogenome (Supplemental Fig. S4B). However, these targets only ranked number 8 and 11, respectively, among possible matches when scanning across all mitochondrial protein-coding sequences with either a C or T (for a preedited state with T at the mtDNA level) in the editing position (Supplemental Fig. S4C).

We subsequently used the additional insights from the conservation profile by joining the 4 identified off-targets in E. coli to the sequences of the 2 native targets (Fig. 5D) to create an adjusted weight matrix (Fig. 6A). To this end, weights were reduced for PPRs with unexpectedly low selectivity (P-18ND, S-16NS, P-12NN, and P-7ND) opposite of positions −21, −19, −15, and −10 and increased to take into account the dominance of guanosine in position +1, adenosine in position −4, and the preference for pyrimidines in position −11 (Fig. 6A). The adjusted matrix identified the 2 known mitochondrial editing sites of PPR78 in the cox1 and rps14 genes, respectively, together with a sequence match in the ccmFN gene featuring a T in the editing position as the best-scoring PPR78 targets in the mitochondrial coding sequences of Physcomitrium (Fig. 6B). The latter could point to an RNA editing site in other mosses, named ccmFNeU1465RC, that causes a change of an arginine into a cysteine codon, which is present in a preedited state in Physcomitrium, making this editing obsolete in the model moss. Scanning available mitogenomes, we indeed found a C to be present at this position in several mosses in support of this hypothesis. Checking for potential C-to-U conversion, we indeed verified the RNA editing event ccmFNeU1465RC in several mosses (Figs. 1B and 6C), including most Hypnales species.

Figure 6.

A third candidate editing target for PPR78: ccmFNeU1465RC. A) An adjusted TargetScan weight matrix for PPR78, which accounts for the results of off-target profiling (Fig. 5D), with reduced weighting for positions −21, −19, and −10, altered weighting for positions −15 and −5, and added weights for positions −17, −14, −11, −8, −4, and +1, was used as input for the TargetScan module of PREPACT (Lenz et al. 2018). Position weighting was deleted opposite of 4 PPRs that failed to reveal the expected selectivity (blue boxes), and conversely, additional weights were introduced at positions that revealed unexpected selectivity (magenta boxes). B) The weight matrix under A was used to scan all P. patens mitochondrial protein-coding regions for candidate targets (max. score value 1,350, top right). Nonmatching nucleotides in candidate targets are labeled in magenta, nucleotides with some (<30%) degree of matching are labeled in dark red, and matching ones (30% to 100%) are labeled in green. Native targets cox1eU755SL and rps14eU137SL (edited C in blue font) appear on top of matching sequences in list positions #1 and #3 with scores of 1,317 together, with a sequence in the ccmFN gene featuring a T at the target site (list position #2), which could reflect the preedited state of an RNA editing site (ccmFNeU1465RC) in other moss species. The next-best match is a position 3 nucleotide downstream of the stop codon of the nad2 gene. C) Efficient RNA editing at the ccmFNeU1465RC site was revealed in several investigated mosses (compare Fig. 1B) including C. imponens, P. attenuatus, and C. purpureus. In contrast, and as seen for the cox1eU755SL site, a T is already encoded at the mtDNA level in the ccmFN gene of the 2 shown Hypnum species, as well as in P. patens.

Confirmation that PPR78 acts on ccmFNeU1465RC as a third target

Since we demonstrated that PPR78 is amenable to functional analysis in E. coli (Fig. 5), we tested its performance on the newly identified ccmFNeU1465RC editing site in this assay system. To that end, we created a mutant version of the candidate Physcomitrium ccmFN sequence in which we replaced the T with C and cloned it into the usual setup. Indeed, we found that Physcomitrium PPR78 was able to edit the ccmFNeU1465RC site very efficiently (Fig. 7), although it exists in a preedited state in nature, with a T in the mitogenome of P. patens.

Figure 7.

Verifying ccmFNeU1465RC as a target of PPR78. The E. coli RNA editing assay was used to verify whether RNA editing site ccmFNeU1465RC could indeed be targeted by PPR78. Given the results from experimentation with the previously known targets (Fig. 5), ccmFN sequences of 200 bp encompassing the candidate target were cloned downstream of the CDS of Physcomitrium PPR78 with the extension of 34 amino acids upstream of PPR L-20. Editing efficiency of >99% was observed with the Physcomitrium ccmFN target, including a T-to-C replacement at the C to be targeted, and moderate editing (up to 41%) was observed in the Pseudanomodon target. Data for individual replicates are given in Supplemental Data Set 1.

Finally, we replaced the ccmFN target of Physcomitrium with that of P. attenuatus, whose ccmFNeU1465RC site is partially edited (46%; Fig. 6). The RNA editing of the Pseudanomodon ccmFNeU1465RC target was only moderate, with up to 41% of C-to-U conversion in the E. coli setup (Fig. 7), which is in congruence with a transition mismatch of S-10ND with target position c-13 and an additional mismatch of PpPPR78 P-12NN with target position g-15. The latter position, however, did not seem to contribute to target selectivity, as judged from the off-target analysis (Fig. 5D).

Discussion

The lack of a recognizable phenotype upon knocking out the RNA editing factor gene PPR78 in P. patens is surprising (Rüdinger et al. 2011; Uchida et al. 2011). The resulting deficit of RNA editing in the cox1 and rps14 mRNAs leads to the retention of serine codons where leucine codons are evolutionarily conserved. This is particularly striking for the lack of editing at the cox1eU755SL site. The leucine residue and its peptide sequence environment (GHPEVYILILPG, edited codon underlined) is highly conserved not only in plants (with the corresponding RNA editing event also existing e.g. in lycophytes and ferns) but also even in metazoa, protists, and bacteria. Along the same lines and in contrast to rps14eU137SL editing, we could not identify a single case of inefficient cox1eU755SL editing among mosses (Fig. 1B). The leucine codon introduced by editing rps14eU137SL and its sequence environment are somewhat less conserved. This may indicate less functional pressure, and it could also explain the lower editing efficiencies or even the absence of editing at this site altogether in some mosses, especially within the order of Hypnales (Fig. 1B).

The absence of an evident phenotype for the Physcomitrium PPR78 KO line was, however, advantageous for the functional complementation studies reported here. Hence, we could create multiple recombinant constructs including the PPR78 orthologues of 3 other mosses, which showed different editing efficiencies at the PPR78 targets in their native environments, ranging from highly efficient editing at the rps14 site in C. purpureus to much less efficient editing in P. attenuatus and even totally absent editing of rps14eU137SL in C. imponens (Fig. 1B). The resulting RNA editing efficiencies at the rps14 site in the respective complementation lines reflected the native editing status, although the expression levels of the inserted orthologues were up to >20,000 times higher than wild-type PPR78 expression in P. patens (Supplemental Table S2 and Fig. S3), which is naturally very low, as generally observed for most site-specific PPR editing factors (Lurin et al. 2004; Fuchs et al. 2020; Oldenkott et al. 2020). Notably, the respective orthologues efficiently edited cox1eU755SL in the PPR78 KO complementation lines (Fig. 2, Supplemental Table S2). Hence, the results reveal that the RNA editing efficiencies at the rps14 site directly depend on the interplay of the different orthologues with the target RNA but do likely not result from other species-specific factors (Fig. 2B).

Neither the different enzymatic capacities of their terminal DYW cytidine deaminase domains (Fig. 3) nor the intriguing differences in the 5th and last amino acid positions of 3 L-type PPRs and 1 S-type PPR between the 4 PPR78 orthologues under study (Fig. 4) could account for the observed functional differences with respect to rps14eU137SL editing. Together with previous findings, this suggests that the roles of L-type PPRs may in fact be variable or depend on the context within a given array of PPRs. When matching the PPR binding code, an L-type PPR may be functionally relevant, as observed for PPR L-5TD in PPR65 with a matching guanidine (Oldenkott et al. 2019). In contrast, a change of the 5th amino acid of the L₂ motif of moss PPR56 toward a match (VD to ND opposite of C) reduced editing at 1 target site and even abolished editing at a second target site (Yang et al. 2023b). A recently investigated large collection of off-targets for PPR56 in human cells indicated certain preferences for nucleotides opposite of L-type PPRs (Lesch et al. 2022); the modification of these nucleotides indeed led to a reduction of editing by PPR56 in the bacterial editing system (Yang et al. 2023b). An earlier study suggested that L-type PPRs are relevant for editing but not necessarily for RNA binding (Matsuda et al. 2020).

Most notably, even the DYW domain of the C. imponens PPR78 orthologue, which is not required for editing at the cox1eU755SL site due to a C–T conversion in the mitogenomes of Hypnales (Fig. 1B), operated faithfully when it replaced that in its Physcomitrium counterpart (Fig. 3), although the complete Callicladium orthologue failed to do so. The retention of PPR78 orthologues in this and similar cases likely points to an additional role for the PPR78 editing factor.

Despite its highly attractive features as a model organism, the creation and maintenance of numerous transgenic Physcomitrium lines is labor-intensive. More importantly, the plant-engineering approach has its limits given that there is no straightforward way to easily manipulate the plant mitochondrial genome for introducing modified or alternative targets for an RNA editing factor, although the first successful experiments using TALENs to modify the Arabidopsis mitochondrial genome were recently described (Kazama et al. 2019; Forner et al. 2023). Hence, a heterologous setup like the bacterial assay system using E. coli is a welcome alternative.

Compared with PPR56 and PPR65, the 2 previously analyzed Physcomitrium RNA editing factors (Oldenkott et al. 2019; Yang et al. 2023b, 2023a), PPR78 required an extended native amino-terminal sequence fused to the upstream protein tags (His₆ and MBP) for functional expression (Fig. 5B). Regarding this point, it is noteworthy that predictions of some PPRs receive very low scores with available bioinformatic tools (https://ppr.plantenergy.uwa.edu.au/). Notably, this is often the case for the most N-terminal PPRs, which frequently also reveal more variability between orthologues. We noted that PPR L-20 of PPR78 is only weakly supported or not identified at all in some taxa (Supplemental Fig. S5). Moreover, the further upstream sequence ends with asparagine (N), which is among the signature amino acids for the last positions of PPRs according to the PPR-RNA code, pointing to the previous existence of a highly degenerate additional upstream P-type PPR (see Fig. 5B, Supplemental Fig. S1). Hence, it may come as no surprise that the retention of additional native sequence upstream of PPR L-20 proved necessary for the functional expression of PPR78 in E. coli.

Furthermore, the previously used small target sequence environment (with 40 bp upstream and 5 bp downstream around an editing site) required an upstream extension for efficient editing at the cox1 target sequence by PPR78 (Fig. 5C). This finding supports the recent observation that sequences upstream of the region ultimately targeted by the PPR stretch of an RNA editing factor can contribute significantly to the efficient editing of certain targets (Yang et al. 2023b). Notably, early studies using other experimental approaches also found that an extended RNA target sequence can be beneficial for enhanced RNA editing (Reed et al. 2001; Hayes and Hanson 2007).

It must be kept in mind, however, that complementation studies or expression in heterologous systems always carry the risk of artifacts e.g. through variably strong overexpression, potentially different secondary structures of RNA targets, or the influence of other RNA binding factors. For example, we found that the transcript levels of an overexpressed RNA editing factor and the resulting degree of RNA editing observed do not always correlate in all cases (see Supplemental Table S2 and Fig. S3) and that downstream processes at the protein level could play a role in this process.

The successful functional establishment of PPR78 in the E. coli assay system allowed us to study (i) its performance on off-targets and (ii) selected alternative candidate targets. Considering the first point, we detected very few (4) off-targets of PPR78 (Fig. 5D), which is even smaller than the number of off-targets (6) of PPR65 and much smaller than the more than 130 off-targets identified for PPR56 (Oldenkott et al. 2019; Yang et al. 2023a). The findings for PPR78 and PPR65 are actually more in line with those for 2 synthetic PPR-RNA editing factors recently tested, with only 1 or no off-targets detected, respectively (Royan et al. 2021; Bernath-Levin et al. 2022). Hence, PPR56, with its high number of additional targets, might turn out to be the exception rather than the rule. It is presently unclear which properties cause this wide discrepancy. Perhaps, an overall more flexible PPR array might help PPR56 bind to RNA targets with an adaptive induced fit mechanism, whereas other editing factors (including PPR65 and PPR78) are structurally more rigid and inflexible, or perhaps downstream E repeats and the terminal DYW domain exert strong selectivity (Ruwe et al. 2019; Maeda et al. 2022; Yang et al. 2023a). PPR78 and its targets, which are widespread among mosses (Fig. 1B, Supplemental Fig. S2), share a long history of coevolution and therefore likely underwent mutual adaptions toward increased target specificity. This may also be the case for PPR65 and its conserved editing target ccmFCeU103PS along the moss phylogeny (Schallenberg-Rüdinger et al. 2013) compared with a likely more recently evolved editing factor such as PPR56, whose targets are not shared with distant moss taxa.

In any case, even the small number of PPR78 off-targets in E. coli (only 4) has helped us understand which positions of a target matter more or less than expected based on the predictions according to the PPR-RNA code (Fig. 5C). Adapting these insights for a weight matrix used to scan for alternative targets revealed ccmFNeU1465RC as an additional candidate target (Fig. 6). We confirmed the ccmFNeU1465RC candidate position as a true editing site in many mosses (Fig. 1B). We also found highly efficient C-to-U conversion by PPR78 at this newly identified target site in E. coli (Fig. 7). From a phylogenetic viewpoint, this additional editing site targeted by PPR78 perfectly explains the retention of the gene for this RNA editing factor in taxa where editing at the cox1 and rps14 targets is obsolete or nonexistent (Fig. 1B). Finally, and quite intriguingly, no PPR78 orthologues could be identified in the 2 species of the genus Hypnum lacking rps14 editing and with C-to-T conversions at both the cox1 and ccmFN targets, strongly supporting this hypothesis. Fitting this viewpoint, examples of the quick degeneration and loss of RNA editing factors following the loss of their targets have been reported among angiosperms (Hayes et al. 2012; Hein et al. 2016; Hein and Knoop 2018; Hein et al. 2019, 2020).

Currently less well explored, but even more interesting in terms of functional evolution, are extensions or shifts in functionality of plant RNA editing factors. Interesting examples include the chloroplast editing factor DOT4 in flowering plants (Hayes et al. 2013), which has lost its known editing target but acquired 6 additional PPRs among Poaceae and could possibly now bind to a chloroplast antisense RNA (Hein et al. 2019, 2020). Along similar lines, the evolutionarily ancient editing factor QED1 has evidently adapted toward the new target ndhBeU872SL in chloroplasts of Brassicaceae (Loiacono et al. 2022). While these 2 cases of chloroplast editing factors are experimentally amenable to using transplastomic approaches to engineer chloroplast DNAs, studying plant mitochondrial editing factors currently relies on heterologous systems, as demonstrated here with the E. coli assay system.

Among the most valuable additional benefits of heterologous expression is the ability to identify and evaluate off-target data sets, as performed here for PPR78. This approach delivers unbiased information on target recognition features that would be hard to obtain otherwise. In the case of PPR78, the lack of selectivity of nucleotides opposite to the amino-terminal PPRs P-18ND and S-16NS is less of a surprise than the similar case in which positions −15 and −10 juxtaposed with the downstream PPRs P-12NN and S-7ND also feature amino acids fitting the PPR-RNA code rules. Perhaps even more relevant are the unexpectedly conserved target positions, such as the pyrimidines opposite of PPR L-8TD or the adenosine in position −4 and the preference for guanosine in position +1. Notably, however, position +1 was found to be less relevant for target recognition by other PPR editing factors such as PPR56 and PPR65 (Oldenkott et al. 2019; Lesch et al. 2022; Yang et al. 2023a). This likely supports the concept that the target recognition and specificity of different RNA editing factors can be quite variable depending on the interplay of motifs and domains within each protein. Importantly, the off-target data allowed us to put more or less weight on positions for target predictions, which ultimately helped us identify ccmFNeU1465RC as a third target of PPR78, explaining its conservation and adaptation during the evolution of mosses. With the identification PPR78 orthologues in the early branching genera Polytrichum, Tetraphis, and Andreaea, we assume a very early origin of PPR78 in the evolution of mosses, which is explained by the emerging necessity of editing cox1eU755SL (Fig. 1B, Supplemental Fig. S2). Conversely, however, no PPR78 orthologues were discernible in the available genomic data of 3 yet earlier diverging peat moss species, Sphagnum palustre, Sphagnum fallax, and Sphagnum magellanicum, which is in line with presence of T at the important editing sites cox1eU755SL and ccmFNeU1465RC.

Materials and methods

Cloning of complementation constructs

Nucleic acids were prepared using the cetyl trimethylammonium bromide (CTAB) protocol (Doyle and Doyle 1990). PPR78 protein-coding sequences were amplified from moss DNAs using Phusion or Q5 proofreading polymerase (Thermo Fisher Scientific, New England Biolabs) and oligonucleotides containing SgsI endonuclease recognition sites (Integrated DNA Technologies). After FastDigest (FD)-SgsI treatment (Thermo Fisher Scientific), coding sequences were introduced into the FD-SgsI-treated, gel-purified (Macherey–Nagel kit), and dephosphorylated (FastAP, Thermo Fisher Scientific) PIG2.0 expression vector. The PIG_AN vector (Schallenberg-Rüdinger et al. 2017), which contains regions homologous (HRs) to the PIG region (Okano et al. 2009), was modified by inserting a multiple cloning site containing AseI, NotI, LguI, and EcoO109I recognition sites. The resulting vector was named PIG2.0 (accession number OR802131). Plasmids were isolated and purified using a NucleoBond Xtra Midi kit (Macherey–Nagel), and constructs were verified by Sanger sequencing (Macrogen Europe). To generate chimeric constructs or mutations, overlap extension PCRs (Horton et al. 1989) were performed. Sequences of oligonucleotide used in this study are given in Supplemental Data Set 2.

Prior to transformation, the PIG2.0 backbone was removed via restriction digestion (SmiI, LguI, NotI; Thermo Fisher Scientific FastDigest or New England Biolabs CutSmart), and the linearized construct was gel-purified (Macherey–Nagel kit).

Transforming complementation constructs

Protonema of the PPR78 KO line of P. patens ecotype Gransden (Rüdinger et al. 2011) was cultured in liquid Knop medium (pH 5.8) supplemented with a microelement solution (Reski and Abel 1985; Oldenkott et al. 2020). Seven to 10 d before transformation, 15 mg of protonema material was harvested and transferred into 150 mL of Knop liquid medium. The pH was reduced to 4.5 3 d prior to transformation. Protoplasts were isolated using a driselase enzyme mix (kindly provided by Dr. M. Sugita, Nagoya, Japan). The protoplasts were transformed with linearized DNA constructs (ca. 18 µg) using PEG, and the transformed protoplasts were incubated in regeneration medium for 10 d and plated on solid Knop medium on a cellophane layer as described before (Hohe et al. 2004). To select stably transformed plantlets, the plants were alternately grown on Knop medium with or without 15 µg/mL hygromycin B (Roth). Plants were grown for a minimum of 2 mo under standard growth conditions (21 °C with 16 h light [∼70 µmol/m²/s¹, OSRAM LUMILUX Cool Daylight HO 39W/865 and Radium Bonalux Super NL 39W/840 bulbs] and 8 h dark period).

Genotyping of mutant lines

DNA was isolated from transgenic P. patens lines following a modified fast extraction protocol (Edwards et al. 1991). The presence of the transgene construct was verified by Taq-based PCR (Promega, Blirt) using construct-specific oligonucleotide primers (Supplemental Data Set 2). PCR products were gel-purified (Macherey–Nagel kit, Blirt kit, or Nippon Genetics kit) and Sanger sequenced (Macrogen Europe). The integration of constructs into the PIG region (Okano et al 2009) was verified using Phusion or Q5 polymerase (Thermo Fisher Scientific, New England Biolabs) and a forward primer that binds to the upstream genome region of the construct insertion site and a reverse primer that binds to the ACTIN1 promoter region, which drives construct expression (Supplemental Data Set 2).

RNA editing analyses of mosses and transgenic P. patens lines

Nucleic acids were prepared using the CTAB protocol (Doyle and Doyle 1990). DNA was digested by DNaseI treatment (Thermo Fisher Scientific) before cDNA was synthesized using random hexamer primers and RevertAid Reverse Transcriptase (Thermo Fisher Scientific). The Taq-based (Promega, Blirt) PCR assays contained primers framing the mitochondrial editing sites of PPR78 (Supplemental Data Set 2). PCR products were gel-purified (Macherey–Nagel kit, Blirt kit, or Nippon Genetics kit) and Sanger sequenced (Macrogen Europe). Sequencing chromatograms were analyzed using BioEdit 7.0.5.3 (Hall 1999), and RNA editing was determined by dividing the T peak by the sum of C and T peaks in the editing site.

Quantification of construct expression

To quantify construct expression in transgenic Physcomitrium lines, gametophytes of the investigated plant lines were grown on 3 separate Knop plates (biological replicates), wrapped with Leukosilk, and grown at 21 °C under a 16 h light (40% intensity) and 8 h dark period (Percival Scientific growth chamber E-41L2) for 4 to 6 wk. RNA was isolated from the samples using a kit (Macherey–Nagel), and DNA was depleted by DnaseI treatment prior to reverse transcription (Thermo Fisher Scientific) using oligo-d(T)₂₃VN primers (IDT). Taq-based (Blirt) RT-PCR with qPCR primer combinations (Supplemental Data Set 2) was performed to examine cDNA quality and rule out DNA contamination. RT-qPCR was performed using a BioRad CFX96 Touch real-time system and EvaGreen master mix (BIOTIUM) with construct-specific primers and an amount of cDNA corresponding to 12 ng of RNA. The adenine phosphoribosyltransferase (Ade-PRT) gene (Pp3c8_16590) was used as a reference (Le Bail et al. 2013). Three technical replicates (individual qPCR reactions with the same cDNA as template and with the same PCR reagents) were processed for each biological replicate, and a wild-type sample was included in each run. Melting curves were controlled using BioRad CFX Maestro 2.3 software, and expression levels relative to mean wild-type expression were calculated using the delta-delta-Ct method (Livak and Schmittgen 2001) in Excel.

RNA editing assays in E. coli

Protein and target coding sequences were amplified from PIG2.0-based plasmids or plant DNA (see above). Short targets were generated by hybridization of oligonucleotides (IDT). E. coli RNA editing experiments using the investigated PPR78 constructs, which were cotranscribed from petG41K_MCS with their 46 or 200 bp targets, were performed and analyzed as outlined before (Oldenkott et al. 2019; Yang et al. 2023b). RNA editing values given in the manuscript are the respective mean of all biological replicates (independent primary E. coli clones) with individual RNA editing assays listed in Supplemental Data Set 1.

Total RNA sequencing and off-target detection

To investigate off-targets of PPR78 in the E. coli transcriptome, editing assays were performed as described before (Oldenkott et al. 2019; Yang et al. 2023b). Total RNA was isolated from different replicates using an RNA purification kit (Macherey–Nagel). Remaining traces of DNA were digested with DNaseI (Thermo Fisher Scientific). RNA-Seq library preparation (rRNA depletion) and subsequent Illumina sequencing (150 bp paired-end) were performed by Novogene. A complete list of RNA-Seq samples is given in Supplemental Table S3. RNA-Seq raw data were quality-checked using FastQC (www.bioinformatics.babraham.ac.uk/projects/fastqc). Adaptor and reads mapping to the respective petG41K-based expression plasmid and the Rosetta2DE3-specific pRARE2 plasmid were removed using BBDuk of the BBTools suite (bbduk.sh, Bushnell B., sourceforge.net/projects/bbmap/) with default settings and the options out/outm, k = 31, and hdist = 2. The remaining trimmed reads were mapped to the BL21DE3 genome (accession CP001509.3) using BBMap of the BBTools suite (bbmap.sh, Bushnell B., sourceforge.net/projects/bbmap/) with default settings and the options outu/outm, mdtag = true, sam = 1.4, subfilter = 3, and pairedonly = true. The output SAM files were converted into BAM format using “samtools view” of SAMtools 1.10 (Li et al. 2009), deduplicated (“samtools fixmate” and “samtools markdup”), coordinate-sorted (“samtools sort”), and indexed (“samtools index”). SNPs compared with a control RNA-Seq data set derived from Rosetta2DE3 cells, possessing an empty petG41K, were called by JACUSA v1.3.0 using call-2 command options -a H:1, M, B, Y; -T 1.56; -f V; -p 1; -c 10; -m 30 (Piechotta et al. 2017). Only SNPs that were called in at least 2 replicates, but not in samples expressing PPR56 (Yang et al. 2023b), were selected by a custom-made R script (kindly provided by Simon Zumkeller). Potentially edited (G > A/C > T) and total read counts in the individual SNP positions were reviewed using IGV, the Integrated Genomics Viewer (Robinson et al. 2011), and sites with edited (G > A/C > T) reads also appearing in the PPR56 control data set were excluded. Mapping qualities surrounding the investigated positions were evaluated using Tablet (Milne et al. 2013), and candidate SNPs that only occurred in RNA-Seq read ends were excluded. RNA editing rates were defined as the ratio of edited to total RNA reads in the respective position.

Phylogenetic analysis

Alignment of protein sequences (see Supplemental File 1) was done using the MUSCLE algorithm (Edgar 2004) as implemented in MEGA 7 (Kumar et al. 2016) using default settings. Maximum likelihood (ML) phylogenetic analyses were performed with IQ-Tree version 1.6.12 (Minh et al. 2020) in the IQ-TREE webserver (http://iqtree.cibiv.univie.ac.at). The best-fitting substitution model was determined automatically with ModelFinder (Kalyaanamoorthy et al. 2017). Branch support was determined using the Ultrafast Bootstrap option (Hoang et al. 2018) with 1,000 replicates. The bootstrap consensus tree was chosen for display (see Supplemental File 2 and Fig. S2).

Accession Numbers

Sequences of the major proteins and genes investigated in this study can be found under the following accession numbers in the GenBank or Phytozome database: PPR78 (XP_024361076.1), cox1 (ARI44036.1), rps14 (ARI44072.1), ccmFN (ARI44043.1), and Ade-PRT (Pp3c8_16590). Sequence data from this article and plasmid sequence of PIG2.0 can be found in the GenBank data libraries under accession numbers OR474045 to OR474058 and OR802131. RNA-Seq raw data generated in this study were deposited in the NCBI Sequence Read Archive (SRA) database under BioProject PRJNA1004579. All custom scripts used for off-target detection are available on GitHub (https://github.com/SMZuckerle/off-target-RNA-editing).

Acknowledgments

The authors gratefully acknowledge funding of this project by the Deutsche Forschungsgemeinschaft (SCHA 1952 2-2) and computing time and support provided by the Paderborn Center for Parallel Computing (PC2). We are very grateful to Dr. Bernard Goffinet, Peter Tautz, and the Berlin and Bonn University Botanic Gardens for their help in obtaining moss materials. We also wish to thank students in lab courses and internships (notably Shyam Ramanathan and Mirjam Thielen) for early experimental contributions and Monika Polsakiewicz and Sarah Schwirblat for technical assistance. We thank Yingying Yang for providing a total RNA sample isolated from E. coli Rosetta2DE3 cells expressing His₆-MBP for 20 h as a control and Philipp Gerke and Simon Zumkeller for their kind help with bioinformatic questions.

Author contributions

E.L. and M.S.-R. designed the study and experiments with input by V.K. E.L. analyzed editing sites and PPR78 orthologue presence in mosses and did most of the molecular cloning and subsequent experimental work in P. patens and E. coli, including the analyses of RNA-Seq data. B.O. started cloning of the first constructs and introduced E.L. to work with Physcomitrium and high-performance computing. M.S. and V.D. contributed to the generation and analyses of transgenic plant lines. The study program was supervised by M.S.-R. and V.K. Figures were created by E.L. and V.K. V.K. wrote the manuscript, which was read, commented on, and approved by all coauthors.

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. Alignment of PPR78 orthologues.

Supplemental Figure S2. Phylogram of PPR78 orthologues.

Supplemental Figure S3. Quantification of the expression of selected complementation constructs by RT-qPCR.

Supplemental Figure S4. Scanning for additional targets of PPR78 with TargetScan.

Supplemental Figure S5. Comparative predictions of PPRs for the PPR78 orthologues.

Supplemental Table S1. Editing efficiencies of PPR78 targets in mosses.

Supplemental Table S2. Details on P. patens PPR78 KO complementation lines.

Supplemental Table S3. Samples for RNA-Seq studies in E. coli.

Supplemental Table S4. Detailed results for off-target identification.

Supplemental File 1. PPR78_orthologues_alignment.fas.

Supplemental File 2. PPR78_orthologues_phylogeny.nwk.

Supplemental Data Set 1. Studies with recombinant PPR78 constructs in E. coli.

Supplemental Data Set 2. Oligonucleotides used in this study.

Funding

This research was supported by a grant (SCHA 1952 2-2) from the Deutsche Forschungsgemeinschaft (DFG) awarded to M.S.-R.

Data availability

The data underlying this article are available in the article and in its online supplementary material.

Author notes

The authors responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plcell/pages/General-Instructions) are Elena Lesch ([email protected]) and Mareike Schallenberg-Rüdinger ([email protected]).

Conflict of interest statement. None declared.