One C-to-U RNA Editing Site and Two Independently Evolved Editing Factors: Testing Reciprocal Complementation with DYW-Type PPR Proteins from the Moss Physcomitrium (Physcomitrella) patens and the Flowering Plants Macadamia integrifolia and Arabidopsis

DYW-type PPR proteins independently evolved in mosses and flowering plants restore RNA editing defects across wide phylogenetic distances.

Abstract

Cytidine-to-uridine RNA editing is a posttranscriptional process in plant organelles, mediated by specific pentatricopeptide repeat (PPR) proteins. In angiosperms, hundreds of sites undergo RNA editing. By contrast, only 13 sites are edited in the moss Physcomitrium (Physcomitrella) patens. Some are conserved between the two species, like the mitochondrial editing site nad5eU598RC. The PPR proteins assigned to this editing site are known in both species: the DYW-type PPR protein PPR79 in P. patens and the E+-type PPR protein CWM1 in Arabidopsis (Arabidopsis thaliana). CWM1 also edits sites ccmCeU463RC, ccmBeU428SL, and nad5eU609VV. Here, we reciprocally expressed the P. patens and Arabidopsis editing factors in the respective other genetic environment. Surprisingly, the P. patens editing factor edited all target sites when expressed in the Arabidopsis cwm1 mutant background, even when carboxy-terminally truncated. Conversely, neither Arabidopsis CWM1 nor CWM1-PPR79 chimeras restored editing in P. patens ppr79 knockout plants. A CWM1-like PPR protein from the early diverging angiosperm macadamia (Macadamia integrifolia) features a complete DYW domain and fully rescued editing of nad5eU598RC when expressed in P. patens. We conclude that (1) the independently evolved P. patens editing factor PPR79 faithfully operates in the more complex Arabidopsis editing system, (2) truncated PPR79 recruits catalytic DYW domains in trans when expressed in Arabidopsis, and (3) the macadamia CWM1-like protein retains the capacity to work in the less complex P. patens editing environment.

INTRODUCTION

Cytidine-to-uridine (C-to-U) RNA editing takes place in the organelles of all land plants except the complex-thalloid liverworts from the Marchantiopsida (Groth-Malonek et al., 2007; Rüdinger et al., 2012; Dong et al., 2019). The C-to-U conversion most often changes codon identities in transcripts and therefore ensures translation of correct proteins involved in such fundamental processes as photosynthesis in chloroplasts and oxidative phosphorylation (OXPHOS) in mitochondria (reviewed by Fujii and Small, 2011; Small et al., 2020). In most flowering plants, some 30 to 50 Cs (in plastids) and ∼400 to 500 Cs (in mitochondria) are specifically converted to U (Edera et al., 2018; Sloan et al., 2018; Small et al., 2020).

Multi-protein complexes called editosomes perform C-to-U RNA editing and have been characterized in angiosperms (Sun et al., 2016). Individual pentatricopeptide repeat (PPR) proteins are core components of the editosome. PPR proteins were identified as specific factors that recognize RNA sequences upstream of the editing target site via their PPR repeats (Barkan et al., 2012), thus defining the C to be edited (Kotera et al., 2005; Zehrmann et al., 2009). The ∼10 to 25 nucleotide-long RNA target upstream of the editing site is bound in a 1 nucleotide/1 PPR manner by the PPR protein (Choury et al., 2004; Takenaka et al., 2004; Hammani et al., 2009; Barkan et al., 2012). All specific PPR editing factors belong to the PLS-type subfamily of PPR proteins. PLS-type PPR arrays not only consist of the canonic 35–amino acid P-type repeats but also contain longer L-type (35 to 36 amino acids) and shorter S-type (31 to 32 amino acids) variants, typically arranged as tandem PLS triplet repeats (Lurin et al., 2004). P- and S-type PPRs bind to individual nucleotides via their fifth and last amino acid in accordance with a RNA-PPR binding code that was initially proposed by Barkan et al. (2012; Figure 1) and refined in several subsequent studies (Takenaka et al., 2013; Yagi et al., 2013; Yin et al., 2013; Yan et al., 2019). The function of L-type PPRs remains elusive (Barkan et al., 2012; Kindgren et al., 2015; Oldenkott et al., 2019).

Figure 1.

Characterization of Arabidopsis CWM1 and P. patens PPR79.

(A) and (B) Models of the DYW-type and E+-type PPR proteins P. patens PPR79 (A) and Arabidopsis CWM1 (B) aligned to the assigned targets of both species. The nomenclature of RNA editing sites is based on Rüdinger et al., (2009) starting with the gene name (nad5), RNA editing from C-to-U (eU), the position of the C in the transcript counted from the first position of the start codon, and the amino acid before and after editing, here Arg (R) to Cys (C). Both PPR proteins feature an array of PLS-repeats downstream of the N-terminal mitochondrial target peptide (red). The last PLS triplet, differing in amino acid conservation, is labeled P2L2S2 according to Cheng et al., (2016). The PPR array is numbered backward starting from the terminal S2 repeat (Hein and Knoop, 2018). The respective fifth and last amino acid positions, critical for PPR-RNA interaction, are displayed below each PPR. DYW, DYW domain; E1 and E2, E motifs. Shown below the PPR protein model are aligned pre-mRNA targets of nad5eU598RC from P. patens and Arabidopsis and ccmBeU428SL, ccmCeU463RC, and nad5eU609VV from Arabidopsis. Ribonucleotides are shaded according to the PPR-RNA binding code proposed by Barkan et al. (2012) based on amino acid identities at the fifth and last amino acid positions: T/S+N: A, T/S+D: G, N+D: U, N+S: C, N+N: Y. Colored backgrounds indicate a match in green, a transition among pyrimidines (Y) or purines (R) in light blue, and a mismatch in magenta shading, respectively. The L-motifs and juxtaposed ribonucleotides are indicated by light gray shading. The ribonucleotides are numbered backward, starting from the editing site (underlined and capitalized C). RNA editing efficiencies in the wild type (WT) and mutant lines for Arabidopsis and P. patens are displayed next to the RNA stretch. The slash indicates editing efficiencies of different plant lines (Arabidopsis: cwm1-1 [left], cwm1-2 [right]; P. patens: ecotype Gransden [left], ecotype Reute [right]). Mean values are displayed for silent site nad5eU609VV, as variability in RNA editing has been observed in Col-0 (100%/80%/78%) and in the mutant lines (cwm1-1: 0%/42%/40%; cwm1-2: 0%/36%/39%). Additional Cs in the RNA targets undergoing C-to-U RNA editing are underlined.

Many editing factors carry carboxy (C)-terminal extensions downstream of their PPR arrays: the E1 motif, the E2 motif, and the DYW domain. The DYW domain is named after the three conserved amino acids Asp (D), Tyr (Y) and Trp (W) at its end (Lurin et al., 2004). The C-terminal DYW domain is crucial for the cytidine deaminase function during C-to-U conversion (Boussardon et al., 2014; Wagoner et al., 2015; Oldenkott et al., 2019; Hayes and Santibanez, 2020) and was shown to have structural and sequence similarity to cytidine deaminases (Salone et al., 2007; Iyer et al., 2011). The DYW domain also includes a conserved peptide motif at its N terminus: the PG-Box (PGxSWIEV), which participates in RNA editing through an unknown mechanism and is retained in many C-terminally truncated editing factors from flowering plants (Okuda et al., 2007; Hayes et al., 2013; Cheng et al., 2016). The E1 and E2 motifs, with similarity to tetratricopeptide repeats, are located upstream of the DYW domain. They are necessary for efficient editing (Okuda et al., 2009, 2014; Cheng et al., 2016) but have not yet been assigned to a clear function. They might function as spacers, placing the DYW domain in the correct position to convert the C to be edited, or as interaction partners with other subsidiary editing factors (Okuda et al., 2014; Ramos-Vega et al., 2015; Bayer-Császar et al., 2017; Ruwe et al., 2019).

Many characterized angiosperm editing factors are C-terminally truncated, lacking at least parts of the DYW domain. Several such E+-type PPR proteins are able to recruit DYW domains in trans to gain full editing functionality (Boussardon et al., 2014; Andrés-Colás et al., 2017; Diaz et al., 2017; Guillaumot et al., 2017). Moreover, additional editing factors such as organellar RNA recognition motif-containing proteins (ORRMs; Sun et al., 2016) or multiple organellar RNA editing factors/RNA editing factor interacting proteins (MORF/RIPs; Bentolila et al., 2012; Takenaka et al., 2012) also influence RNA editing in flowering plants.

The moss Physcomitrium (Physcomitrella) patens is evolutionarily distant from flowering plants and constitutes an excellent model for comparative evolutionary studies. Although historically named Physcomitrella patens, thorough phylogenomics studies have shown that Physcomitrella patens emerged from within Physcomitrium (McDaniel et al., 2010; Beike et al., 2014; Medina et al., 2019). For this reason, we use P. patens to mean Physcomitrium patens throughout this article and encourage our colleagues in the moss community to do the same, as suggested in a recent article (Rensing et al., 2020). P. patens features a very simple organellar RNA editing scenario: only 2 editing sites in plastids (Miyata and Sugita, 2004) and 11 RNA editing events in mitochondria (Rüdinger et al., 2009). Nine complete DYW-type PPR proteins are each responsible to edit one or two editing target sites (Schallenberg-Rüdinger et al., 2013; Ichinose et al., 2014). Additional editing factors have not been identified in the moss yet (Rensing, 2008; Schallenberg-Rüdinger and Knoop, 2016; Small et al., 2020). Two P. patens RNA editing factors were even shown to successfully edit their respective targets in the bacterium Escherichia coli with similar efficiencies as in planta and thus even in the absence of additional components from plant organelles (Oldenkott et al., 2019).

The conservation of specific RNA editing factors and coevolution with their assigned editing sites were described for several editing factors in angiosperms (Hayes et al., 2012; Hein and Knoop, 2018; Sun et al., 2018; Hein et al., 2019, 2020; Loiacono et al., 2019) and in mosses (Rüdinger et al., 2011). The conservation of an editing factor normally depends on the retention of its assigned editing sites in the organelles. Exceptional cases where the editing factor survives the loss of its corresponding editing targets may be explained by neo-functionalization, as for example the P. patens PPR43 protein with a diverged DYW domain, which acts as a splicing factor (Ichinose et al., 2012), or the ortholog of DEFECTIVELY ORGANIZED TRIBUTARIES4 (DOT4) among Poales, which may be involved in binding of an antisense transcript for the ribosomal protein S3–encoding gene rps3 in chloroplasts (Hein et al., 2019).

Here, we aimed to test whether editing factors of different and evolutionarily distant editing systems adjusted to the same editing site would be able to operate in the reciprocal genetic system. The genetically amenable model systems P. patens among the mosses and Arabidopsis (Arabidopsis thaliana) among angiosperms are ideal test cases. They are well studied, their RNA editing systems evolved in different directions (reduced versus complex), and these species are separated by more than 400 million years of evolution (Rensing et al., 2008).

RESULTS

The RNA Editing Site nad5eU598RC Is Shared between P. patens and Arabidopsis

The growing collection of plant organellar editomes, coupled with the availability of new software tools, allows for the straightforward identification of RNA editing positions shared between organelles of different taxa (Lenz et al., 2018). We identified three RNA editing sites that are conserved between P. patens (Rüdinger et al., 2009) and the angiosperm Arabidopsis (Giegé and Brennicke, 1999; Bentolila et al., 2013): two in the ccmFC gene (encoding the cytochrome C biogenesis factor F), ccmFCeU103PS and ccmFC122SF (for editing site nomenclature; Figure 1), and one in nad5 (encoding subunit 5 of the NADH-ubiquinone oxidoreductase), nad5eU598RC. Editing sites ccmFCeU103PS and ccmFC122SF are assigned to editing factors PPR65 and PPR71, respectively, in P. patens (Tasaki et al., 2010; Schallenberg-Rüdinger et al., 2013), but no factors have been identified for the corresponding editing events in Arabidopsis. However, nad5eU598RC has been assigned to editing factors in both species (Uchida et al., 2011; Schallenberg-Rüdinger et al., 2013; Hu et al., 2016). Indeed, the E+-type PPR protein CELL WALL MAINTAINER1 (CWM1; Hu et al., 2016) and the DYW-type PPR protein PPR79 (Uchida et al., 2011; Schallenberg-Rüdinger et al., 2013) are responsible for the editing event nad5eU598RC in Arabidopsis and P. patens, respectively. Moreover, the target sequences of CWM1 and PPR79 upstream of nad5eU598RC are nearly identical in the two species (Figure 1), suggesting that reciprocal complementation with the two evolutionarily distant editing factors might succeed.

A Knockout of the P. patens Editing Factor PPR79 Exhibits a Sporophyte Generation Phenotype in the Reute Strain

The P. patens editing factor PPR79 is a complete, nontruncated DYW-type PPR protein, with the full suite of E1, E2, and DYW extensions behind 17 PLS-type PPR repeats, that targets the nad5eU598RC editing site (https://ppr.plantenergy.uwa.edu.au/ppr/; Figure 1A; Cheng et al., 2016). An earlier study had shown that ppr79 knockout (ko) plants in the P. patens background Gransden exhibited a reduced growth rate of the protonema, the filamentous haploid tissue of the moss, relative to the wild type (Uchida et al., 2011). In the course of this study, we noticed that the lack of PPR79 (in Gransden) also influenced the growth of the gametophores, the dominant haploid growth stage of mosses. Gametophores from ppr79 Gransden ko lines (Schallenberg-Rüdinger et al., 2013) grew more slowly compared to the wild type and showed a dwarf phenotype with small, spikier leafy structures (Figure 2A). Overexpression of PPR79 in the ppr79 ko line recovered both the RNA editing defect at the nad5eU598RC site and the slow-growth phenotype, confirming that PPR79 is essential for RNA editing of the nad5 site in P. patens.

Figure 2.

Phenotypic Analysis of P. patens ppr79 ko.

(A) Comparison of 4-week old gametophytes for the wild type (WT) and ppr79 ko of the two P. patens ecotypes Gransden and Reute grown on Knop agar. Bar = 1 mm. Both the wild-type Reute and ko Reute ppr79 can produce viable sporophytes. Bar = 1 mm.

(B) Boxplots displaying the average number of mature (postmeiotic, white) and immature (premeiotic, dark gray) sporophytes developed per individual gametophyte (n = 35 plantlets). Center lines, medians; box limits, 25th and 75th percentiles; whiskers extend 1.5 times the interquartile range; individual data points are plotted as open gray circles. ppr79 Reute ko plants show a significant reduction of mature (P < 3.7 × 10⁻¹⁶) and immature (P < 1.1 × 10⁻⁰³) sporophytes per plant.

Mosses are land plants with a dominant haploid gametophyte growth phase. Their sexual reproduction cycle eventually leads to diploid sporophytes, which grow on the gametophytic tissue and ultimately produce haploid spores. To investigate whether sexual reproduction was impaired by the loss of PPR79 in P. patens, we generated an independent ppr79 ko line in the P. patens ecotype Reute. The ecotype Reute was recently established as an alternative P. patens laboratory strain (Hiss et al., 2017) to the widely used strain Gransden, which is largely impaired in sporophyte production (Meyberg et al., 2020). ppr79 Reute ko plants showed the same growth phenotype as ppr79 Gransden ko plants (Figure 2A). We scored sporophyte formation and observed that ppr79 Reute ko plants produced far fewer sporophytes relative to the Reute wild-type background when grown under same growth conditions in a pilot experiment (only 65 sporophytes for ppr79 Reute ko versus 250 sporophytes for Reute on a total of 40 plantlets each; Supplemental Data Set 1). To gain a clearer picture of the consequences of PPR79 loss, we repeated the experiment but scored mature and immature sporophytes separately. Again, we measured drastically reduced numbers of sporophytes, both immature and mature (Figure 2B; Supplemental Data Set 1). The few mature ppr79 Reute ko sporophytes that developed were nonetheless able to produce fertile spores, which germinated and yielded new gametophores when spread on solid growth medium (Supplemental Figure 1). Hence, although less efficiently than the wild type, ppr79 Reute ko plants can complete their entire sexual life cycle.

Defects in Mitochondrial Respiratory Chain Complexes in the P. patens ppr79 ko

To investigate the consequences of unedited nad5 transcripts on mitochondrial function in P. patens caused by the loss of PPR79, we isolated crude plant mitochondrial extracts from ppr79 Reute ko and from the wild-type strains Reute and Gransden to test the composition of their mitochondrial protein complexes by two-dimensional (2D) blue native (BN)/SDS-PAGE (Figure 3). All five mitochondrial complexes of the OXPHOS system were clearly visible in the mitochondrial extracts of the wild-type plants. NAD5 is one of the core subunits of complex I, which forms a supercomplex together with complex III₂. We identified peptides for various complex I and complex III subunits, including NAD5, by mass spectrometry of the same wild-type samples (Supplemental Figure 2). By contrast, supercomplex I+III₂ was missing in the ppr79 Reute ko samples following 2D BN/SDS gel electrophoresis (Figure 3), although the abundance of the other OXPHOS complexes was unchanged. We conclude that the phenotype of ppr79 ko plants may result from diminished complex I functionality, which is reminiscent of the phenotype associated with the loss of the corresponding angiosperm editing factor CWM1 (Hu et al., 2016).

Figure 3.

Comparative Analysis of the Mitochondrial Proteomes of the Wild-Type P. patens Ecotype Reute and ppr79 Reute ko.

Proteins were separated by 2D BN/SDS polyacrylamide gel electrophoresis and stained with silver. The molecular masses of standard proteins are given in between the two 2D gels. The identities of the OXPHOS complexes are indicated above the gels: V, complex V, ∼650 kD; III₂, dimeric complex III, ∼500 kD; I+III₂, supercomplex formed of complex I and dimeric complex III, ∼1500 kD; F1, F1-part of complex V, ∼300 kD; IV, complex IV, ∼200 kD; and II, complex II, ∼160 kD. Complexes F1, III₂, V, and supercomplex I+III₂ were identified based on visual comparison to Arabidopsis 2D BN/SDS gels (Klodmann et al., 2011) and by mass spectrometric analysis (Supplemental Figure 2). Chloroplast contaminants, that are subunits of PSI, PSII, and the light-harvesting complex II (LHCII), are indicated by magenta circles. Supercomplex I+III₂ position is indicated by a blue frame in the wild-type (WT) Reute and the corresponding position in ppr79 Reute ko by a gray frame.

The E+-Type PPR Protein CWM1 Edits Four Sites in Arabidopsis Mitochondria

Arabidopsis cwm1 mutant plants show a mild root growth phenotype (Hu et al., 2016). In addition, we observed reduced seed germination rates on soil (Supplemental Figure 3). The Arabidopsis editing factor CWM1 features 16 PPRs and C-terminal E1 and E2 motifs but only a partial DYW domain that is truncated downstream of the PG-Box (Figure 1B; Hayes et al., 2013). Accordingly, it is classified as an E+-subclass PPR protein (Cheng et al., 2016). The canonical tandem PLS structure is also imperfect, as the PPR array contains two P-type tandem repeats at positions P–6 and P–7 and P–10 and P–11 (Figure 1B; note that PPR repeats are counted backward from C-terminus to N-terminus). CWM1 does not exclusively address the RNA editing site nad5eU598RC but also the editing sites ccmCeU463RC and ccmBeU428SL in the ccmC and ccmB transcripts, respectively (encoding cytochrome c biogenesis subunits B and C; Hu et al., 2016). In addition, we observed that editing at the silent site nad5eU609VV, 11 nucleotides downstream of nad5eU598RC, was affected by lack of CWM1 (Figure 1B). RNA editing at nad5eU598RC and ccmCeU463RC was abolished in both Arabidopsis cwm1 alleles, while editing at the ccmBeU428SL site was strongly reduced in both cwm1 mutants (Hu et al., 2016). In our analysis, we also detected no editing of ccmBeU428SL in cwm1-1 (Figure 1B). Overexpression of CWM1 in the cwm1-1 and cwm1-2 mutant backgrounds restored RNA editing at all sites in both mutant lines, confirming that CWM1 is a key specificity factor for the four RNA editing sites in Arabidopsis mitochondria.

CWM1 and PPR79 Have Distinct Evolutionary Origins

Phylogenetic analyses should reveal whether PPR79 and CWM1 are true orthologs derived from the same ancestor or whether the two factors emerged independently in mosses and flowering plants. A search for CWM1-like sequences in high-quality genomic and transcriptomic data presently available from 123 angiosperms identified 103 putative CWM1 orthologs, including the Arabidopsis CWM1 (Figure 4). The CWM1 editing factor appears to have been lost at least six times independently during flowering plant evolution. No CWM1 duplication events were recognized. Interestingly, we identified putative CWM1 orthologs with C-terminal DYW domains in the three early-emerging angiosperm orders Proteales, Ranunculales, and Caryophyllales (Figure 4). In avocado (Persea americana), which belongs to the order Laurales, we identified a DYW fragment downstream of the predicted stop codon, suggesting an evolutionarily recent truncation of the protein.

Figure 4.

Evolution of RNA Editing Factor CWM1 in Angiosperms.

Cladogram of 123 angiosperms. Asterids, Proteales, Ranunculales, Caryophyllales, and Liliopsida have been collapsed in the left panel and Rosids in the right panel. The cladogram is based on the current understanding of angiosperm phylogeny (Open Tree of Life; The Angiosperm Phylogeny Group, 2009; Hinchliff et al., 2015). Putative CWM1 orthologs were identified for all species except the ones marked by red branches. Most identified putative orthologs belong to the E+-subclass of PPR proteins (Cheng et al., 2016). Only species in the early-branching Caryophyllales, Proteales, and Ranunculales feature C-terminal DYW domains (marked in blue). A remaining DYW coding sequence was also identified downstream of the CWM1 stop codon in P. americana (underlined).

We next checked the RNA editing status for known CWM1 targets in selected angiosperm species (Figure 5). The nad5eU598RC site already carried a thymidine (equivalent to a pre-edited site) in the mitochondrial genome of several angiosperms. Evolutionarily derived species whose nad5eU598RC and ccmCeU463RC sites were pre-edited had also lost CWM1 (Figure 5). However, red wild einkorn wheat (Triticum urartu) and sugar beet (Beta vulgaris) also lost CWM1, despite retaining ccmCeU463RC editing. Conversely, ccmBeU428SL remained editable in some species of the Rosids and in Catharanthus, although they lacked a putative CWM1 ortholog. We hypothesize that ccmCeU463RC editing may have been adopted as an extra target by CWM1, while ccmBeU428SL editing was likely taken over by another factor during angiosperm evolution. Consistent with our hypothesis, the ccmBeU428SL site is still edited to low levels in the Arabidopsis cwm1 mutants (Figure 1; Hu et al., 2016).

Figure 5.

Overview of CWM1-Associated RNA Editing Targets in Selected Angiosperms.

Overview of CWM1-associated RNA editing targets (nomenclature based on Arabidopsis CDS) in selected angiosperms with (+) or without a putative CWM1 ortholog (–). Related proteins with an apparent complete DYW domain are indicated by a red +. Black dots: RNA editing verified by RT-PCR at the respective position. Gray dots, potential RNA editing sites (cytidines at DNA level); open circles, unedited cytidines; X, pre-edited state with a thymidine in the mitochondrial genome.

We identified putative PPR79 orthologs in 35 moss species from a publicly available moss deep sequencing of the transcriptome (RNA sequencing) data set (Supplemental Table 1; Johnson et al., 2016; Gutmann et al., 2020). The requirement of nad5eU598RC editing was highly conserved among mosses, and the position was only rarely pre-edited in individual species (Supplemental Table 1; Supplemental Data Set 2; Liu et al., 2019). The ccmC and ccmB editing sites edited by CWM1 in Arabidopsis were present in some moss species as well. The respective upstream target sequences matched only moderately with the PPR array of PPR79 (Supplemental Figure 4; Supplemental Data Set 2; Liu et al., 2019).

We then constructed a phylogenetic tree based on the C-terminal domain (PLS2-E1-E2-DYW) of all nine P. patens RNA editing factors, a set of 27 Arabidopsis PPR-type RNA editing factors featuring complete DYW domains, all 35 identified putative PPR79 orthologs in mosses, and the 103 putative CWM1 orthologs in angiosperms as described above. All CWM1 and PPR79 orthologs clustered together within the moss and the angiosperm clades, respectively (Figure 6). The CWM1 and PPR79 phylogenies roughly followed the known phylogenies of angiosperms and mosses (Supplemental Figure 5; The Angiosperm Phylogeny Group 2009; Hinchliff et al., 2015; Johnson et al., 2016; Liu et al., 2019), as previously observed for phylogenies based on other PPR protein orthologs (Hein et al., 2019, 2020). However, the CWM1 and PPR79 clades were not sister clades but were rather separated by closer branching editing factors, indicating their independent origins and evolution. Moreover, we failed to identify putative orthologs for CWM1 or PPR79 in the available genome or transcriptome data of other plant clades (gymnosperms, ferns, hornworts, liverworts, or lycophytes; Gutmann et al., 2020).

Figure 6.

Phylogenetic Analysis of the PPR Proteins CWM1 from Arabidopsis and PPR79 from P. patens.

Phylogram of 103 angiosperm putative CWM1 orthologs, including Arabidopsis CWM1 (collapsed; see Supplemental Figure 5B for the full angiosperm branch), 35 putative orthologs of P. patens PPR79 (collapsed for 21 Hypnales; see Supplemental Figure 5A for the uncondensed moss branch), 27 Arabidopsis DYW-type PPR proteins identified as editing factors (collapsed; listed in Methods), and all 9 DYW-type PPR P. patens paralogs (Schallenberg-Rüdinger et al., 2013). Accession numbers for the moss proteins are displayed behind the species name (e.g., VBMM-2051258 for OneKP accessions or c24535 g1 i1 m22706 for accessions from Johnson et al., 2016). The Maximum Likelihood phylogenetic tree was calculated with IQ-tree (http://iqtree.cibiv.univie.ac.at/; Trifinopoulos et al., 2016) using the JTT+F+R6 model of sequence evolution. The alignment of the terminal domains (P2L2S2-E1-E2-DYW) used for tree construction contained 283 parsimony informative sites (Supplemental Data Set 5). Node confidence was determined from 1000 bootstrap replicates and is shown at nodes where exceeding 70%.

CWM1 and PPR79 in Reciprocal Complementation Experiments

Although the CWM1 and PPR79 proteins have distinct evolutionary origins and show differences in their PPR array structures, their targeting sequences upstream of nad5eU598RC were nearly identical. The angiosperm nad5 target showed a U-to-C exchange at position –10 opposite to PPR P–7NN and a guanosine (G) to adenosine (A) exchange at position –13 opposite to PPR P–10TS (Figure 1), the designation of individual PPRs is according to the nomenclature proposed by Hein and Knoop (2018): first, the PPR repeat type (P, S, or L), followed by the position of the repeat counting backward from C- terminus to N-terminus, and the amino acids at the fifth and last position within the PPR). Considering the current RNA-PPR binding code (Figure 1), the U-to-C exchange is not expected to influence the binding of Arabidopsis CWM1 to the P. patens target, as the PPR repeat responsible for the recognition of this position (PPR P–7NN) does not distinguish between C and U. The combination of Thr (T) and Ser (S) residues within PPR P–10TS has yet to be assigned to a nucleotide preference (Barkan et al., 2012, Barkan and Small 2014). In the case of the P. patens PPR79 protein, the two nucleotide changes in the Arabidopsis nad5 target would disfavor binding of PPR79 to the target RNA, as PPR S–7ND predominately recognizes U and PPR P–10TD preferably binds to G. Considering the overall number of matches, reciprocal complementation (including recognition of nad5eU598RC in the respective other genetic background) should be possible for CWM1 and PPR79.

P. patens PPR79 Can Complement the Arabidopsis cwm1 Mutant Even without Its DYW Domain

To investigate the functionality of the P. patens editing factor PPR79 in the Arabidopsis background, we expressed both the original PPR79 coding sequence and a truncated version lacking the C-terminal DYW domain in the cwm1 background (Figure 7). As two independent cwm1 mutants (cwm1-1, cwm1-2) were available (Hu et al., 2016), we transformed both mutants with the same constructs.

Figure 7.

Functional Complementation Test of Arabidopsis cwm1 Mutant Lines.

(A) RNA editing in the Arabidopsis cwm1 mutants can be restored by transformation with the Arabidopsis CWM1 gene, different versions of P. patens PPR79, and chimeras between the two editing factor genes. Red and blue backgrounds indicate species origins for creation of chimeric proteins. All constructs are listed in Supplemental Table 4.

(B) Chromatopherograms of all known CWM1 targets in the respective top transgenic lines (with highest editing efficiencies for nad5eU598RC) in the Arabidopsis cwm1 mutant background. Percentages of C-to-U conversion are given below. The editing positions under consideration are highlighted by underlining and red shading. Partially edited sites are labeled as pyrimidine ambiguities (Y). Full editing data for all transgenic lines are listed in Supplemental Table 2.

The full-length P. patens PPR79 protein reconstituted RNA editing in the two cwm1 mutant backgrounds not only for the nad5 editing site shared between P. patens and Arabidopsis but also for the two other sites in ccmC and ccmB. Depending on the transgenic line, we observed RNA editing efficiencies ranging from 7 to 51% for nad5eU598RC and from 16 to 34% for ccmCeU463RC. RNA editing of ccmBeU428SL increased to 45 to 73% relative to the cwm1 mutants (Figure 7; Supplemental Table 2). The extent of editing of the silent site nad5eU609VV was variable in the cwm1 mutant backgrounds (0 to 40%), while editing of the same site in the transgenic lines reached 20 to 48%, preventing us from unequivocally assigning it to PPR79 activity alone. We confirmed that all transgenic lines expressed the inserted transgenes by RT-PCR. Transgenic lines with low (<10%) nad5 editing efficiencies showed lower expression of the transgene than most other lines (Supplemental Figure 6). Another RNA editing site, ccmBeU485SL, which is not affected by CWM1, was efficiently edited in the various transgenic lines, as expected (range between 96 and 100%; Supplemental Table 2).

To test the role of the DYW domain, present in PPR79 but absent in CWM1, we also generated transgenic lines expressing a truncated version of PPR79, lacking the DYW domain, in the cwm1 mutants. Interestingly, even such truncated PPR79∆DYW version rescued the RNA editing defects at the affected sites (Figure 7), some lines even reaching higher editing efficiencies than lines expressing full-length PPR79, suggesting that P. patens PPR79, like CWM1, can recruit a DYW domain in trans when expressed in Arabidopsis (Figure 7; Supplemental Table 2).

As control, cwm1 mutant lines transformed with Arabidopsis CWM1 displayed the wild-type levels of editing at the four affected editing sites. These lines had higher efficiencies than mutant lines transformed with PPR79 (Figure 7; Supplemental Table 2). The lower editing activity exhibited by heterologous P. patens PPR79 constructs relative to the homologous Arabidopsis CWM1 constructs may be due to a reduced fit between the PPR stretch and the editing targets. To test this possibility, we also generated two chimeric fusion constructs between CWM1 and PPR79. In one construct, we added the DYW domain from PPR79 after the PG-Box of CWM1 (CWM1-DYWPPR79). In the second construct, we replaced the complete set of C-terminal domains from CWM1 with those from PPR79 (CWM1PLS-E1E2DYWPPR79). Both chimeric proteins were able to edit the nad5 RNA editing sites, as well as the ccmB and ccmC sites (Figure 7; Supplemental Table 2) with the same high efficiencies as Arabidopsis CWM1 transformed in the Arabidopsis cwm1 mutant background, demonstrating that the P. patens DYW cytidine deaminase can be added for functional complementation of the mutant, with or without the upstream E motifs.

Arabidopsis CWM1 Does Not Reconstitute nad5eU598RC RNA Editing in P. patens

In a reciprocal set of experiments, we tested whether constructs for Arabidopsis CWM1 might rescue the editing defects of the P. patens ppr79 ko mutant (Figure 8). In an initial control experiment, we observed full restoration of nad5eU598RC editing when we expressed P. patens PPR79 in the ppr79 Gransden ko mutant background. Similarly, the putative PPR79 ortholog from the closely related moss Funaria hygrometrica fully restored RNA editing at the target site (Figure 8; Supplemental Figures 7 and 8).

Figure 8.

Functional Complementation Test of the P. patens ppr79 ko.

(A) Physcomitrium patens PPR79, the Funaria hygrometrica putative PPR79 ortholog, Arabidopsis CWM1, and chimeric constructs between P. patens PPR79 and Arabidopsis CWM1 (background coloring as in Figure 7) were introduced into the ppr79 ko line. All constructs are listed in Supplemental Table 4.

(B) Only full-length PPR79 and its putative ortholog from F. hygrometrica complement the ppr79 ko. RNA editing efficiencies of nad5eU598RC of the ppr79 transgenic lines are given. Results of one independent transgenic line each are displayed as chromatopherograms (others are listed in Supplemental Table 3), which are labeled according to Figure 7. Single gametophores of ppr79 ko transgenic lines were grown on non-selective Knop media for 8 weeks in long-day conditions. Bars = 1 mm. Transgenic lines with restored RNA editing at nad5eU598RC display the wild-type phenotype, while transgenic lines expressing CWM1 or CWM1+PPR79 fusion constructs display the mutant phenotype (Supplemental Figure 8).

Next, we introduced a construct where the CWM1 mitochondrial target sequence was added upstream of the PPR79 PPR array, to assess mitochondrial targeting efficiency of the CWM1 mitochondrial target sequence in P. patens. This construct fully restored RNA editing in P. patens, demonstrating that the CWM1 N-terminal signal peptide was functional for targeting and import into P. patens mitochondria (Supplemental Table 3). However, full-length Arabidopsis CWM1 protein failed to reconstitute editing at the nad5eU598RC site in the P. patens ppr79 ko background, as did the chimeric CWM1-PPR79 variants with the added PPR79 C termini (Figure 8). The transgenes for each construct were expressed in their corresponding backgrounds (at least three independent lines; Supplemental Table 3) with similar expression levels for the PPR79 and CWM1 versions (Supplemental Figure 9), excluding low transgene expression as the cause for the inability to restore RNA editing.

Macadamia CWM1 Can Complement the ppr79 P. patens Mutant

Our phylogenetic survey revealed CWM1 orthologs with a complete DYW domain in early-branching angiosperms, macadamia (Macadamia integrifolia) being one of them (Figure 6). There are differences at the fifth and last positions in the PPR repeats between the Arabidopsis and the macadamia proteins. However, considering the PPR-RNA binding code only, none of these differences should reduce the potential of the macadamia homolog to bind to the four editing targets normally recognized by its Arabidopsis counterpart (Figure 9). The macadamia ortholog also shows one additional match due to a P–8NG to P–8NN change compared to Arabidopsis CWM1.

Figure 9.

CWM1 Ortholog from Macadamia Can Reconstitute RNA Editing in Arabidopsis and P. patens.

(A) Macadamia CWM1 PPR array compared to Arabidopsis CWM1. The fifth and last positions differing from Arabidopsis CWM1 are indicated by blue letters. The N-terminal sequence of Arabidopsis CWM1 used for transformation is highlighted with red shading. PPR binding code matches of macadamia CWM1 and candidate RNA targets of macadamia, Arabidopsis (nad5eU598RC, ccmBeU428SL, ccmCeU463RC, and nad5eU609VV), and P. patens (nad5eU598RC) are shown with shading and labeling of PPR-RNA interactions as in Figure 1.

(B) RNA editing efficiencies of the respective CWM1 targets in macadamia and transgenic lines in the Arabidopsis cwm1 mutant or P. patens ppr79 ko expressing macadamia CWM1, displayed as chromatopherograms and percentage of C-to-U conversion as in Figures 7 and 8.

The nad5 target sequence in macadamia was identical to the one in Arabidopsis. Likewise, RNA editing was functionally conserved at sites nad5eU598RC, ccmBeU428SL, and ccmCeU463RC (Figure 9). Interestingly, we failed to detect any C-to-U conversion for silent site nad5eU609VV in macadamia (Figure 9). We chose the macadamia CWM1 ortholog to test its ability to edit nad5 in P. patens and nad5, ccmC, and ccmB in Arabidopsis.

To ensure mitochondrial import into Arabidopsis and P. patens mitochondria, we used a chimera that added the Arabidopsis CWM1 transport signal at the N terminus of the macadamia CWM1-like protein (Supplemental Table 4). The introduction of the macadamia CWM1-like gene into the Arabidopsis cwm1 mutant background restored RNA editing at all four affected sites, surprisingly even the silent nad5eU609VV editing event not found to be edited in macadamia itself (Figure 9). The other affected sites (in the ccmC and nad5 transcripts) were also edited efficiently, with 100% editing in one transgenic line. These results confirmed that the macadamia CWM1-like protein is a functional ortholog of Arabidopsis CWM1. Only the ccmB site was edited at a lower efficiency, with one transgenic line showing no detectable C-to-U conversion at site ccmBeU428SL. The low editing efficiency at the ccmB site and the editing of the silent nad5eU609VV site indicate that the functional spectrum of an editing factor is defined both by the protein itself and by its respective genetic environment, most likely owing to different protein–protein interactions.

We next tested the macadamia CWM1-like protein in reciprocal experiments in P. patens ppr79 ko background. Macadamia CWM1, expressed in the ppr79 ko, efficiently restored editing at the nad5 editing site nad5eU598RC, in contrast to its Arabidopsis ortholog, which lacks a DYW domain. Editing efficiencies reached 40 and 100%, respectively, in two independent transgenic lines (Figure 9B), correlating well with transgene expression (Supplemental Figure 9). The successful rescue of P. patens nad5 editing by macadamia CWM1, but not by its Arabidopsis counterpart, suggests that the original DYW domain is necessary in cis to perform efficient editing in the heterologous P. patens system.

The P. patens PPR79 Editing Factor Tolerates Single Amino Acid Modifications in Planta

Since the macadamia CWM1 ortholog, but not Arabidopsis CWM1 itself, efficiently edited nad5eU598RC in P. patens, we wondered how tolerant the P. patens editing system would be toward less perfectly fitting editing factors. We therefore changed several amino acids in individual PPR79 PPR repeats and tested these variants for rescue of RNA editing in the P. patens ppr79 ko background. First, we modified the last position in PPR repeats P2–3 and S–4 from Asp (D) to Thr (T), while maintaining the same Asn (N) residue at the fifth position. The combination of N and T at the fifth last positions, respectively, was recently assigned to pyrimidines, with a preference toward C in P-type PPR proteins (Yan et al., 2019). Both individual changes (P2–3ND > P2–3NT and S–4ND > S–4NT) still fully restored editing (Figure 10). We then changed a key residue in the binding stretch of PPR P–6TN, opposite a matching A, to P–6TD, which is predicted to predominately bind to G. The resulting PPR79 variant edited the nad5 site completely in transgenic lines in the ppr79 ko background. We also introduced another modification at a position thought to be important: P–6TN to P–6NS, the former recognizing purines with preference for A, the latter expected to bind to C. This new variant still preserved full PPR79 functionality, as nad5eU598RC was edited to 100% in the resulting transgenic lines. Finally, we changed a conceptually mismatching PPR repeat toward a theoretically matching one: P–9NN (opposite A) was changed to the fitting P–9TN, or the P–12TS (opposite U) was changed to the fitting P–12NN. Just as we observed for the previous set of mismatches above, single changes of PPR repeats toward a better match did not change the editing efficiency of nad5eU598RC (Figure 10; Supplemental Table 3).

Figure 10.

PPR79 Variants with Mutations Expected to be Relevant for RNA Binding Can Fully Rescue RNA Editing in the ppr79 ko.

Individually modified PPRs are shown in color in the PPR79 model. Protein modifications and resulting change in PPR-RNA interaction are displayed below the respective native PPR79 repeats. The color code follows the PPR-RNA binding code proposed by Barkan et al. (2012) as in Figure 1. The resulting RNA editing efficiencies remain 100% at the nad5eU598RC site in at least two independent lines for each of the six mutants.

DISCUSSION

Single Complete Editing Factors from Two Evolutionarily Distant Editing Systems Are Compatible

PLS-type PPR proteins are the key factors for the recognition of RNA editing sites by binding upstream of their targets in land plant organelle transcripts. Previous studies have shown that related editing factors within a plant clade can in principle rescue the loss-of-function mutant of another species (Okuda et al., 2008; Schallenberg-Rüdinger et al., 2017).

Here, we show that editing factors from plant clades separated by more than 400 million years of evolution can successfully perform editing in distant species. These editing factors were phylogenetically unrelated, but functionally analogous rather than homologous. As objects of study, we chose the P. patens DYW-type PPR protein PPR79 and the Arabidopsis E+-type PPR protein CWM1, both targeting the nad5eU598RC editing site in the mitochondrial nad5 transcript.

CWM1 acts naturally within the more complex Arabidopsis editing system with single editing factors involved in editing at multiple sites (Kim et al., 2009; Brehme et al., 2015) and complex combinations of editing helper factors required for editing (Bentolila et al., 2012; Takenaka et al., 2012; Andrés-Colás et al., 2017; Diaz et al., 2017; Guillaumot et al., 2017; Shi et al., 2016; Sun et al., 2016; Malbert et al., 2020). By contrast, PPR79 participates in the much simpler P. patens RNA editing system, with only nine complete DYW-type PPR proteins that each edit one or two editing sites. Critically, no additional factors are needed for efficient RNA editing as the P. patens DYW-type PPR proteins PPR56 and PPR65 also efficiently edit their targets when expressed in bacteria (Oldenkott et al., 2019). PPR65 can even edit its target in vitro (Hayes and Santibanez, 2020).

Here, we demonstrate that the P. patens DYW-type PPR79 is fully functional in Arabidopsis mitochondria when it is introduced in the Arabidopsis cwm1 mutant background. In this heterologous system, PPR79 edited the nad5 editing site nad5eU598RC, which is common to both P. patens and Arabidopsis as well as the ccmCeU463RC and ccmBeU428SL sites, although those latter editing sites do not exist in the P. patens mitochondrial genome. Hence, P. patens PPR79, an evolutionarily distant editing factor with a terminal DYW domain, can replace its derived analog CWM1, which normally relies on recruitment of a DYW domain in trans. We wondered whether a similarly shortened version of PPR79 would be functional in the heterologous environment.

Recruitment of the DYW Domain by Truncated P. patens PPR79 in Arabidopsis

A truncated version of PPR79 lacking the DYW domain also restored RNA editing at the three nonsilent affected sites when expressed in the Arabidopsis cwm1 mutant. We conclude that the truncated P. patens PPR79 protein already possesses an inherent capacity to recruit a DYW domain in trans, congruent with earlier reports on protein–protein interaction between PPR79 and different members of the seed plant specific family of MORF/RIP editing factors (Schallenberg-Rüdinger et al., 2013; Gutmann et al., 2020).

Since CWM1 lacks a complete DYW domain, it must recruit a cytidine deaminase for its function. DYW2, a short PPR protein with a complete DYW domain, was recently identified as the top candidate to be recruited by different E+-type editing factors (Andrés-Colás et al., 2017; Guillaumot et al., 2017; Malbert et al., 2020). The higher protein abundance of DYW2 in Arabidopsis mitochondria compared to the low abundance of specific editing factors supports its widespread and general function in RNA editing (Fuchs et al., 2020). Fittingly, DYW2 mutant plants display reduced editing efficiencies at the sites assigned to CWM1, supporting the combined action of DYW2 and CWM1 (Andrés-Colás et al., 2017; Guillaumot et al., 2017). Furthermore, both proteins were identified to act in one protein complex in mitochondria (Rugen et al., 2019). We failed to identify a DYW domain in Arabidopsis with high similarity to the macadamia CWM1 DYW domain. The DYW domain of the DYW-type PPR protein encoded by At4g02750 was the closest candidate but shared only 52% identity with the macadamia CWM1 DYW domain. It is therefore unlikely that CWM1 was split into two proteins, one responsible for recognition (CWM1) and one with the cytidine deaminase activity (DYW domain), only to associate for proper function, as it was observed for Arabidopsis PPR protein CHLORORESPIRATORY REDUCTION4 (CRR4) and DYW1 (Boussardon et al., 2012). The degeneration of the DYW domain of CWM1, rather than a split into two proteins, is further supported by the CWM1-like locus in P. americana. A degenerated DYW domain fragment was identified downstream of an early stop codon, suggesting truncation and degeneration of the domain.

A truncated PPR79 lacking the DYW domain (equivalent to an E+-type PPR protein) was also able to recruit a DYW domain when expressed in Arabidopsis, to form an editing complex that reached editing efficiencies up to 85% (for the ccmBeU428SL site; Supplemental Table 2). Whether PPR79 recruits DYW2, as it is assumed for CWM1, or other DYW domains present in Arabidopsis mitochondria, remains an open question and needs to be investigated further.

Notably, a truncated version of another P. patens DYW-type PPR protein, PPR78, was able to edit only 2 to 4% of its cox1 targets when highly expressed in its natural P. patens mitochondrial environment (Schallenberg-Rüdinger et al., 2017). Most likely, the competence for recruiting DYW domains in trans is much more established in Arabidopsis, where the P. patens editing factor PPR79 can interact with MORFs/RIPs and thereby extend its RNA editing capacity. Supporting this idea, the RNA binding activity of designer PLS-type PPR proteins was shown recently to be increased by L-repeat–bound MORF proteins (Yan et al., 2017).

Reduced Editing Capacity of PPR79 in Arabidopsis May Rely on Structural Effects and Differences in RNA Binding Positions

Structural differences between the PPR arrays of PPR79 and CWM1 may directly influence their respective editing efficiencies of the nad5eU598RC site in Arabidopsis. Focusing on the relevant RNA binding positions following the PPR-RNA recognition code, PPR79 features the S-type PPR S–10TD opposite A instead of a better matching G and the S-type PPR S–7ND opposite C instead of a conceptually better fitting U. This is in contrast to CWM1, whose PPR motif P-type PPR P–7NN binds U and C equally well. The amino acid combination TS found in P-type PPR P–10 of CWM1 is not assigned yet to a nucleotide. Both CWM1 and the CWM1-PPR79 chimera led to a complete functional complementation of the Arabidopsis cwm1 mutants, further supporting the notion that the PPR array is crucial for editing efficiency differences between CWM1 and PPR79 rather than the C-terminal domain. Similarly, RNA editing efficiency decreased when modified versions of the chloroplast editing factors CHLOROPLAST BIOGENESIS19 (CLB19) and ORGANELLE TRANSCRIPT PROCESSING82 (OTP82) carrying changes in binding positions were transformed into the respective mutant background (Kindgren et al., 2015). Single additional mismatches in those two Arabidopsis editing factors were however relatively well tolerated in planta (Kindgren et al., 2015). Likewise, we documented here robustness to single amino acid changes at fifth and last positions of RNA binding PPR repeats for PPR79 in the P. patens background.

It was astonishing to us that the Arabidopsis ccmC and ccmB target sites were edited by PPR79, since the N-terminal half of its PPR stretch does not fit well to the targets (Figure 1). This finding suggests that N-terminal PPR motifs in PLS-type PPR proteins are less important for recognition. Congruently, the numerous off-targets of the P. patens editing factors PPR56 and PPR65 in E. coli transcripts are less conserved in their upstream regions opposite of the N-terminal PPR repeats (Oldenkott et al., 2019), and many PLS-type proteins assigned to their editing sites show mismatches or even PPR repeats with unusual amino acid combinations in their N-terminal PPR array region (EdiFACTs; Takenaka et al., 2013; Lenz et al., 2018). This offers a contrast with previous findings for P-type PPR proteins with PPR motifs becoming misaligned toward the C-terminus (Miranda et al., 2018). The number of mismatches alone clearly does not explain the different editing efficiencies of the four editing sites assigned to CWM1. For the three nonsilent editing sites, we showed here that they were also edited by PPR79. The core nad5 target, however, matches both editing factors best and is edited less efficiently than the other sites in most transgenic lines. The combination of helper proteins, RNA structural effects, and yet uncharacterized amino acid motifs interacting with the RNA target might cause different editing efficiencies.

Neither Arabidopsis CWM1 nor the CWM1-PPR79 chimeras tested here resulted in editing of the nad5 site when expressed in P. patens, although the CWM1 PPR array conceptually matches the P. patens nad5 target even better than the PPR79 binding stretch matches the Arabidopsis nad5 target. We hypothesize that CWM1-PPR79 chimeras we tested may be nonfunctional due to incompatibilities between upstream PPR stretches and C-terminal domains, as previously shown for other P. patens DYW-type PPR protein chimeras (Ichinose and Sugita, 2018). In such cases, recruitment of other DYW domains would allow the chimeras to work in Arabidopsis, but not in P. patens. We cannot rule out the possibility that the CWM1 chimeras may be misfolded or rapidly degraded when expressed in P. patens. P. patens lacks the helper proteins typical for angiosperm editosomes, which may act as stabilizing factors (Shi et al., 2016).

An Angiosperm DYW-Type PPR Protein Is Fully Functional in a Moss Editing System

The macadamia CWM1-like protein edited the nad5eU598RC site in P. patens up to 100%. Ancestral versions of editing factors in an otherwise derived angiosperm RNA editing system can therefore function in the basic P. patens editing system. Here, additional editing factors are clearly dispensable.

This is particularly intriguing, because the RNA editing apparatus of macadamia consists very likely of the same type of proteins as the Arabidopsis editosome. We identified proteins with homology to Arabidopsis E-type, E+-type, and DYW-type PPR, MORF, and ORRM proteins in publicly available macadamia transcriptome data sets (National Center for Biotechnology Information [NCBI] BioProject: PRJNA533518). We even identified a sequence with high similarity to DYW2 on macadamia contig GHOE01034924.1. Since the macadamia genome sequence has not been completed, we do not know how many editing factors are encoded by the genome. Such information is, however, available for the close relative Indian lotus (Nelumbo nucifera) among the Proteales (https://ppr.plantenergy.uwa.edu.au/ppr/). Indian lotus also possesses a CWM1-like protein with a complete DYW domain (Figure 4). The Indian lotus genome encodes 106 E-type, 133 E+-type, and 149 DYW-type PPR proteins in total, even more than Arabidopsis with a total of only 197 E-, E+-, and DYW-type PPR proteins (Cheng et al., 2016). These numbers fully align with previous analyses finding that the core set up of angiosperm editing factors has evolved early in the first angiosperms (Gutmann et al., 2020) and that angiosperms mostly lost editing sites and editing factors over time (Fujii and Small, 2011). The macadamia CWM1-like protein also largely restored RNA editing when expressed in the Arabidopsis cwm1 mutant background, although with reduced efficiency at the ccmB site, pointing to detailed differences of target recognition between the macadamia and Arabidopsis CWM1 proteins. Although the critical amino acids at the fifth and last positions in PPRs globally fit the targets, amino acid substitutions likely contribute to different sequence recognition specificities, which might, besides the C-terminal domain functionality, be another reason behind the different results seen when we expressed the two angiosperm CWM1 orthologs in P. patens. The efficient recruitment of further factors in trans might, likewise or additionally, play a role in the lower editing efficiency of ccmB by the macadamia CWM1-like protein. Notably, MORF3, MORF8, and ORRM4 have been shown to influence editing of the ccmBeU428SL site (Takenaka et al., 2012; Shi et al., 2016), but not the other CWM1-affected sites in Arabidopsis.

Compatible DYW-Type PPR Proteins as Basis for Designer Editing Factors

This study makes the case for complete DYW-type PPR proteins as the best candidates to modify organellar transcripts in plants. That a P. patens editing factor operates properly in Arabidopsis mitochondria while a macadamia homolog can rescue editing of the P. patens ppr ko mutant demonstrates the general robustness of their function in organelles across plants. The core editing apparatus appears to be conserved throughout the evolution of land plants, even though RNA editing factor families expanded independently in each land plant clade (Gutmann et al., 2020). Surprisingly, we also discovered that P. patens editing factors can be turned into functional E+-type PPRs when expressed in Arabidopsis by deleting their C-terminal DYW domain, although it should be noted that the P. patens genome does not encode for a single E+-type editing factor. Because these truncated PPR proteins remain functional, they must be inherently capable to recruit a cytidine deaminase function in trans. Conversely, the macadamia CWM1-like protein can edit the nad5 target in P. patens without the need of any additional seed plant-specific editing factor.

Plant mitochondrial genome transformation is not yet established (Larosa and Remacle 2013). The first successful attempts of mitochondrial genome modification were recently published for rice (Oryza sativa) and rapeseed (Brassica napus), using transcription activator-like effector nucleases (Kazama et al., 2019). Engineering of nuclear-encoded DYW-type PPR proteins that would work in organelles without additional factors may allow mitochondrial gene expression to be influenced in diverse plant systems. P-type PPR proteins have already been established as inducible switches to activate gene expression in chloroplasts (Rojas et al., 2019). Precise C-to-U modification via DYW-type PPR proteins to introduce start or stop codons in organellar transcripts would likewise modulate translation potential of transcripts. Alternatively, single amino acid modifications may be introduced in some of the more poorly understood proteins encoded by the mitochondrial genome, opening a new avenue to inhibit or disrupt their function.

Influence of Mutated nad5 on P. patens Development

Several existing RNA editing mutants already provide valuable resources to investigate the influence of single mutations in mitochondrial genes (Takenaka et al., 2019). To date, 35 editing factors responsible for RNA editing sites in the mitochondrial nad genes have been characterized for rice, maize (Zea mays), and Arabidopsis (Takenaka et al., 2019). Most of the corresponding mutant plants show growth retardation phenotypes, in addition to impaired seed development and germination (Murayama et al., 2012; Xie et al., 2016; Li et al., 2018; Xiao et al., 2018; Takenaka et al., 2019). Correlating these phenotypes with individual complex I subunits remains difficult, however, as most editing factors are assigned to more than one editing site in several mitochondrial genes (Sosso et al., 2012; Zhu et al., 2012; Arenas et al., 2014; Yap et al., 2015). For example, the Arabidopsis cwm1 mutant affects nad5, ccmB, and ccmC.

By contrast, the P. patens ppr79 ko mutant results in a single U-to-C mutation in nad5 at the nad5eU598RC editing site, causing retention of a basic Arg instead of Cys at amino acid position 199 in the NAD5 protein. Cys-199 is highly conserved and located closely to the exit of one of the four proton channels within the membrane arm of complex I (it is located at the intermembrane space–exposed side of the membrane arm; see Supplemental Figure 10; Fiedorczuk et al., 2016; Senkler et al., 2017). The ppr79 ko and the resulting editing defect had a moderate impact on the growth of gametophores and protonema in the haploid generation but led to a significantly reduced number of developed sporophytes in the diploid generation compared with the wild-type P. patens plants.

This study included a biochemical investigation of the OXPHOS system in P. patens. We identified all five OXPHOS complexes in the wild-type ecotypes Reute and Gransden, although complex I was only detected as a supercomplex together with complex III₂. In the ppr79 ko, by contrast, the supercomplex remained below our detection limit. The apparent loss of complex I may have resulted from the mutated NAD5 protein translated from the nonedited nad5 transcript, which might affect the assembly of the membrane component, similar to previous observations with the Arabidopsis tang2 mutant with a defect in nad5 splicing (Colas des Francs-Small et al., 2014). Other NAD editing mutants also showed impaired incorporation of subunits and subsequent reduction of complex I levels (Ligas et al., 2019). Alternatively, the supply of NAD5 subunits for complex I assembly may be limited. Indeed, the translational apparatus prefers edited transcripts for yet unknown reasons (Planchard et al., 2018; Small et al., 2020). A reorganization of mitochondrial translation and a negative feedback loop activated by reduced respiratory chain activity has been proposed for different Arabidopsis complex I mutants (Planchard et al., 2018).

Regardless of the detailed molecular defects, it is worth noting that upon switching our analysis to the newly established P. patens Reute ecotype, we discovered that the ppr79 Reute ko phenotype was more pronounced during sexual propagation relative to the vegetative growth phase. Respiration mutants of flowering plants share the same difference in phenotypic evocation, likely caused by a mitochondrial bottleneck dysfunction during a developmental phase when plants rely on full mitochondrial performance (Touzet and Meyer, 2014). With a biotechnological outlook in mind, the demonstrated function of RNA editing factors in distant genetic systems is intriguing. Although the results presented here for RNA editing factors from three distant plant species clearly document their widely analogous function, it is also apparent, that we have much left to learn about compatibility of domain fusions in DYW-type PPR proteins and RNA target recognition beyond the currently understood PPR-RNA code.

METHODS

Plant Materials and Growth Conditions

We received macadamia (Macadamia integrifolia) plant materials from the Bonn Botanical Garden (catalog no. xx-0-B-0633374). Seeds for the homozygous Arabidopsis (Arabidopsis thaliana) T-DNA insertion lines cwm1-1 (SALK_017325C) and cwm1-2 (SALK_124160) were kindly provided by Lieven De Veylder, Ghent University, Belgium (Hu et al., 2016). We grew Arabidopsis plants in growth chambers at 21°C and 65% humidity, with a 16-h-light (photosynthetic photon flux density of 82 μmol/m²/s, neon tubes, Osram L36W/840 Lumilux Cool White)/8-h-dark cycle. We harvested green leaves from the rosettes of 2- to 3-week-old plants grown on potting soil (Ökohum Anzuchterde) mixed with 20% (w/w) vermiculite and 0.5 mg L^-1 fertilizer (Osmocote Exact Mini). We confirmed the presence of the T-DNA insertions in the mutants by PCR with primers At CWM1_-82_F, At CWM1_742_R, and SALK_LB1.3 (Supplemental Table 5). For phenotypic characterization of mutants and transgenic lines, we grew all plants alongside the wild-type accession Columbia-0 (Col-0) under the same conditions. Physcomitrium patens strain Gransden (Rensing et al., 2008) and P. patens ecotype Reute (Hiss et al., 2017) gametophores were cultivated on modified Knop medium plates (250 mg L⁻¹ KH₂PO₄, 250 mg L⁻¹ KCl, 250 mg L⁻¹ MgSO₄ × 7H₂O, 1000 mg L⁻¹ Ca(NO₃)₂ × 4H₂O, 12.5 mg L⁻¹ FeSO₄ × 7H₂O, 0.22 µM CuSO₄, 0.19 μM ZnSO₄, 10 μM H₃BO₃, 0.1 μM Na₂MoO₄, 2 μM MnCl₂, 0.23 μM CoCl₂, and 0.17 μM KI, pH 5.8, 1% [w/v] agar; Rüdinger et al., 2011) at 21°C, with a 16-h-light (photosynthetic photon flux density of 65 μmol/m²/s, neon tubes, Osram HO 39W/865 Lumilux Cool Daylight)/8-h-dark cycle.

Generation of P. patens ppr79 Knockout Plant Lines

We generated knockout lines in P. patens PPR79 by homologous recombination after polyethylene glycol–mediated transformation (Hohe et al., 2004) of P. patens strain Gransden (Rensing et al., 2008) and P. patens ecotype Reute protoplasts (Hiss et al., 2017) as described in Rüdinger et al. (2011). The two homologous regions, HR1 and HR2, were generated by PCR to target the resistance cassette (Nos-promoter:neomycin phosphotransferase II – Nos-terminator or the in-frame β-GLUCURONIDASE gene and neomycin phosphotransferase II resistance cassette, respectively) to the PPR79 locus (Pp3c5_7610V3.1, P. patens nuclear genome v3.3; Lang et al., 2018) at PLS-repeat P-6 (behind nucleotide position 1314 counting from start AUG, PLS repeat nomenclature; see Figure 1). Knockout plants were identified by genotyping PCR with primers Pp PPR79 UTR_F/R/rev and Pp PPR79 KO_F/R (Supplemental Table 5). P. patens lines were cultivated as described above.

Nucleic Acid Preparation

Fresh green leaf material from 2- to 3-week-old Arabidopsis Col-0 plants was used for preparation of DNA and RNA. DNA was prepared as described in Edwards et al. (1991). Total RNA was isolated from Arabidopsis leaf tissue using the NucleoSpin RNA Plant kit (Macherey-Nagel) with a subsequent DNase I treatment (Thermo Fisher Scientific) to remove residual genomic DNA. Total nucleic acids of P. patens and macadamia were isolated following the Cetyltrimethylammoniumbromid protocol (Doyle and Doyle 1990). The resulting nucleic acid preparations were then treated either with RNase A (Thermo Fisher Scientific) for genotyping PCR or amplification of mitochondrial fragments in macadamia or with DNase I (Thermo Fisher Scientific) for further cDNA synthesis.

cDNA Synthesis and RT-PCR

Arabidopsis

We performed first-strand cDNA synthesis with Moloney Murine Leukemia Virus Reverse Transcriptase, RNase H Minus Point Mutant (Promega) according to the manufacturer’s instructions with up to 2 µg of RNA and a 5 µM mixture of oligo(dT) (Thermo Fisher Scientific) or random hexamer primers (Roth). After annealing of the oligonucleotides, the reaction mixture was incubated for 30 min at 42°C and subsequently for 30 min at 55°C. The mixture was diluted with double-distilled water (1:2), after which cDNA fragments were amplified with FIRE Polymerase (Solis Biodyne) with specific forward and reverse primers (Supplemental Table 5).

P. patens and Macadamia

First-strand cDNA synthesis was performed using the RevertAid recombinant Moloney Murine Leukemia Virus Reverse Transcriptase kit (Thermo Fisher Scientific) following the manufacturer’s instructions. For RNA editing analysis of organellar transcripts, we set up assays with 6.25 µM hexanucleotide random primers mix (Roth), whereas we used 5 µM oligo(dT) primers (Thermo Fisher Scientific) for verification of transcription of nuclear genes. DNase I–treated RNA (250 ng) was used per 10 μL of cDNA synthesis assay. The assays were incubated at 42°C for 4 h. For each RT-PCR assay, 2 μL of the cDNA solution was used as template. cDNA fragments were amplified with Go Taq Polymerase M300 (Promega) using specific primers (Supplemental Table 5).

Analysis of RNA Editing Sites

The amplified cDNAs generated above were sequenced by Macrogen Europe (https://dna.macrogen-europe.com). Specific primers were used for different fragment amplification (the PCR program was adapted with 35 cycles of amplification; see Supplemental Table 5). We screened the resulting sequences for Cytidine-to-thymidine differences. We compared the relative height of the respective nucleotide peaks in the sequence chromatograms to estimate the RNA editing levels at each editing site. We performed peak height calculations with DNADynamo (Blue Tractor Software) or BioEdit version 7.1 (https://bioedit.software.informer.com/7.2/). Chromatopherograms were visualized using SnapGene software (GSL Biotech; available at snapgene.com).

Generation of Transgenic Lines in the Arabidopsis cwm1 Mutants

cwm1-1 and cwm1-2 mutant plants were transformed by the floral dip method (Clough and Bent, 1998) with various constructs where the cauliflower mosaic virus 35S promoter drives the expression of the CWM1 Col-0 coding sequence (CDS; At1g17630.1) or different full-length or truncated transgene constructs (Supplemental Table 4). The CDS of full length (N-terminally elongated according to Schallenberg-Rüdinger et al., 2013) and DYW-truncated versions of PPR79 (Pp3c5_7610V3.1, P. patens nuclear genome v3.3; Lang et al., 2018) were cloned into an expression vector with a hygromycin phosphotransferase resistance cassette (pMDC83 or pMpGWB102; Ishizaki et al., 2015) via Gateway LR cloning. We followed the same steps for the open reading frames of CWM1 and CWM1/ PPR79 chimeras corresponding to the PLS stretch of CWM1 (up to amino acid 604) linked to either the E1-E2-DYW extension continuity of PPR79 (starting at amino acid 639) or the PLS stretch plus E1-E2 of CWM1 (up to amino acid 679) and the DYW domain of PPR79 (starting at amino acid 714) alone. Owing to missing information on the 5′ end of macadamia CWM1, it was fused downstream of the Arabidopsis CWM1 mitochondrial targeting sequence (up to amino acid 58), which we confirmed as functional both in Arabidopsis and in P. patens. All constructs are listed in Supplemental Table 4.

Selection of primary Arabidopsis transformants was conducted on half-strength Murashige and Skoog (MS) medium (with MS-salts and 1% [w/v] agar; Duchefa Biochemie) containing 10 µg/mL hygromycin after seed stratification in the dark for 2 d at 4°C. Resistant seedlings were transferred to soil to set seeds; the presence of the transgenes was verified by PCR. Expression of the transgenes was also verified by reverse transcription PCR (30 cycles of amplification; Supplemental Table 2) and compared to the expression of native ACTIN2 (At3g18780 for primers; see Supplemental Table 5, 30 cycles of amplification). Relative expression levels are additionally calculated in relation to the CWM1 expression in the wild-type ecotype Col-0 (Supplemental Figure 6).

Transformation of P. patens ppr79 Gransden ko

The P. patens knockout ppr79 was transformed with the same open reading frames that were used for the generation of transgenic lines in the Arabidopsis cwm1-1 and cwm1-2 mutant backgrounds. The expression of the open reading frames was placed under control of the rice actin1 promoter and the nos terminator. The generation of constructs and insertion in the P. patens intergenic (PIG) region (Okano et al., 2009) were performed as described by Schallenberg-Rüdinger et al. (2017). In addition, the ppr79 knockout was transformed with a PPR79 ortholog (Supplemental Figure 7) amplified from Funaria hygrometrica (GenBank: JF501603.1; Rüdinger et al., 2011).

To verify that Arabidopsis CWM1 was properly targeted to P. patens mitochondria, we generated a fusion construct between the N-terminal sequence of CWM1 (first 75 amino acids, including the first L1-17 PPR; see Figure 1) and P. patens PPR79 (starting at amino acid 64) by overlap-extension PCR (Supplemental Table 4; Higuchi et al., 1988). The complementation lines were selected on solidified Knop plates with 1% (w/v) agar containing 25 µg/mL hygromycin B.

We genotyped the transgenic lines by PCR according to Schallenberg-Rüdinger et al. (2017) with primers ActinP_F and NosT_R to confirm the presence of the transgene, Pp_PIG_F, and Pp_PIG_R for the insertion of the construct into the PIG region (35 cycles of amplification each). Finally, we tested the correct orientation of the construct in the PIG region with primer combination Pp_PIG_F and ActinP_R. If a transgenic line showed all correct bands for PCRs with both primer combinations ActinP_F/NosT_R and Pp_PIG_F/ActinP_R, it was considered a stable transgenic line. Transformed lines giving PCR products for PCR reactions with primer combinations ActinP_F/NosT_R and Pp_PIG_F/R were assumed to carry the transgene, but not integrated into the PIG region by homologous recombination. These transgenic lines were considered to be ectopic (Supplemental Tables 2 and 5; Schallenberg-Rüdinger et al., 2017).

We tested the expression of the transgene via RT-PCR with specific primers (Supplemental Table 5, 30 cycles of amplification). Expression of the transgene was compared to the expression of the endogenous gene ELONGATION FACTOR 1-ALPHA (Ef1α; PHYPADRAFT_97715) with primers Pp EF1α_F/R (30 cycles of amplification). Relative expression levels were calculated in relation to PPR79 expression in the wild-type ecotype Gransden (Supplemental Figure 9; Supplemental Table 3).

Sporophyte Induction in P. patens

Sporophyte induction of P. patens ecotype Reute and ppr79 ko in P. patens ecotype Reute was conducted according to a protocol modified from Hiss et al. (2017). In a pilot experiment, we transferred five P. patens gametophytes onto one Petri dish (55 × 14 mm; Sarstedt) containing Knop medium solidified with 1% (w/v) agar. For the wild type and P. patens ko lines, we prepared eight plates each, for a total of 40 plants per line. The plants were incubated in long-day conditions (21°C, 16-h light/8-h dark cycle) until each plantlet carried at least six gametophores. The plants were then transferred to short-day conditions (8-h-light/16-h-dark cycle) at 18°C and incubated according to Hiss et al. (2017). In contrast to the established protocol, the sporophytes were counted 8 weeks after watering instead of 4 weeks to account for the retarded growth of the ppr79 ko lines. For sporophyte evaluation, we scored mature sporophytes (postmeiotic) and immature, premeiotic sporophytes separately, as described previously (Hiss et al., 2017). We then repeated this experiment by scoring sporophytes from the wild-type and ppr79 ko plantlets individually for each of 35 individual plantlets per line. Statistical significance of differences in sporophyte production per plantlet was tested with the Welch two sample t test (implemented in Excel [Microsoft]; Supplemental Data Set 1). Boxplots were generated with RStudio (https://rstudio.com) and BoxPlotR (http://shiny.chemgrid.org/boxplotr).

Seed Germination Experiments in Arabidopsis Col-0 and cwm1 Mutant

We sowed 16 seeds per pot for the wild-type Col-0 and the cwm1 mutants cwm1-1 and cwm1-2 on soil and in triplicates. We scored germinated seeds every day between 10 and 28 d after sowing. We calculated means and sds for each line.

Phylogenetic Analyses of PPR Proteins

A collection of putative CWM1 orthologs (Hu et al., 2016) was identified by BLAST using the CWM1 protein sequence as query against a nonredundant protein database (BLASTP) and against a translated nucleotide database (TBLASTN; Altschul et al., 1990) at NCBI (www.ncbi.nlm.nih.gov), making sure to include only true CWM1 ortholog for each angiosperm taxon sampled according to Hein et al. (2019), with additional data for plume poppy (Macleaya cordata), cape sundew (Drosera capensis; Droseraceae, Caryophyllales), and common buckwheat (Fagopyrum esculentum; Polygonaceae, Caryophyllales). We identified putative orthologs of PPR79 by TBLASTN search against the moss RNA-sequencing data sets published by Johnson and colleagues and the OneKP PPR data set (https://ppr.plantenergy.uwa.edu.au/onekp/; Johnson et al., 2016; Gutmann et al., 2020). PPR sequences were aligned in the MEGA alignment explorer with the MUSCLE tool (Tamura et al., 2013; Kumar et al., 2016, 2018) followed by manual adjustment.

We used putative angiosperm CWM1 orthologs, moss PPR79 orthologs, and a set of 27 characterized Arabidopsis DYW-type editing factors for phylogenetic analyses. The Arabidopsis proteins were as follows: CREF3, CREF7, CRR22, CRR28, DOT4, ECD1, ELI1, LPA66, MEF1, MEF7, MEF10, MEF11, MEF14, MEF22, MEF26, MEF29, MEF35, OTP81, OTP82, OTP84, OTP85, OTP86, RARE1, REME1, REME2, VAC1/ECB2, and YS1. Because of major differences in PPR array makeup of PPR proteins, we restricted phylogenetic tree calculations to the C-terminal domain sequences consisting of the terminal P2, L2, and S2 PPRs and the E1, E2, and DYW extensions (each ∼300 amino acids; Supplemental Data Set 3). We constructed a Maximum Likelihood phylogram using IQ-tree webserver (http://iqtree.cibiv.univie.ac.at; Supplemental Data Set 4; Trifinopoulos et al., 2016). We applied the JTT+F+R6 model for sequence evolution as the best fitting model identified by ModelFinder (Kalyaanamoorthy et al., 2017). We determined the confidence of each node from 1000 bootstrap replicates with ultrafast bootstrap approximation UFBoot (Hoang et al., 2018).

Isolation of Mitochondria from P. patens

Starting from single gametophores, P. patens wild-type (Gransden and Reute) and ppr79 ko (Reute) plant material was grown in 150 mL of liquid Knop medium and homogenized (Polytron PT 2500 E) weekly until plants reached the filamentous protonema stage. The plant material was then transferred to aerated (∼0.2 volumes of air per volume of medium each minute). Five-liter flasks with 4 liters of liquid Knop medium and grown in long-day conditions on a shaker at 125 rpm for 1 week. The plant material was then harvested and homogenized again, before being split into two aerated 5-liter flasks with 4 liters of liquid Knop medium and grown in same conditions for an additional week.

We isolated mitochondria from the P. patens plant material following a protocol developed for Arabidopsis leaves (Keech et al., 2005), with the following modifications. We harvested the protonema cell culture using a miracloth sieve, followed by grinding in disruption buffer (0.3 M Suc, 60 mM TES, 25 mM sodium pyrophosphate, 10 mM KH₂PO₄, 10 mM EDTA, 1 mM Gly, 1% [w/v] Polyvinylpyrrolidone 40, pH 8.0, and 1% [w/v] BSA, 50 mM sodium ascorbate, 20 mM Cys added directly before use) with sea sand for 10 min on ice. Disrupted cells were centrifuged twice at 2500g for 5 min at 4°C to remove sand and broken cell walls. Mitochondria (and comigrating organelles, e.g., chloroplasts) were pelleted by centrifugation at 15,100g for 15 min at 4°C and resuspended in washing buffer (0.3 M Suc, 10 mM TES, and 10 mM KH₂PO₄, pH 7.4). The suspension was homogenized with two gentle strokes in a Teflon homogenizer and loaded onto Percoll gradients (1:1 mixture of Percoll and Percoll-buffer [0.6 M Suc, 20 mM TES, 2 mM EDTA, 20 mM KH₂PO₄, and 2 mM Gly, pH 7.5]). Percoll gradients were prepared by centrifugation at 39,000g for 40 min at 4°C. After sample loading, centrifugation was performed at 15,000g for 20 min at 4°C. Mitochondria were retrieved from the lower third of the gradient and washed by repeated sedimentation (centrifugation for 20 min at 17,200g and 4°C) and resuspension in washing buffer. Finally, the mitochondrial pellet was resuspended in washing buffer at a concentration of 0.2 g/mL, flash frozen in liquid nitrogen, and stored at –80°C until further use. With the newly established isolation protocol, the purity of the mitochondrial fraction was ∼75%.

Protein Gel Electrophoresis Procedures

We performed 2D BN/SDS-PAGE as described previously (Klodmann et al., 2011, Wittig et al., 2006). For protein solubilization, we used a buffer containing 5% (v/v) digitonin (30 mM HEPES, 150 mM potassium acetate, and 10% [v/v] glycerol, pH 7.4) to retain native protein conformations. 2D BN/SDS gels were stained first with colloidal Coomassie Brilliant Blue R 250 as described previously (Neuhoff et al., 1988). For increased sensitivity, we subjected gels to silver staining following a protocol by Heukeshoven and Dernick (1988).

Mitochondrial OXPHOS complexes were identified on 2D gels by visual comparison with corresponding 2D separations of Arabidopsis mitochondrial protein complexes (Klodmann et al., 2011) and by mass spectrometry.

Mass Spectrometry

For direct protein identification, we cut out selected spots of the 2D BN/SDS gels and prepared them for mass spectrometric analysis by carbamidomethylation, tryptic digestion, and peptide extraction (Klodmann et al., 2010). Tandem mass spectrometry analysis of the extracted peptides was performed as described previously (Klodmann et al., 2011). Results were queried against a P. patens database from NCBI (downloaded 21.8.2018) and a modified version of The Arabidopsis Information Resource (TAIR10; www.arabidopsis.org) using an in-house Mascot Server (www.matrixscience.com).

Accession Numbers

Sequence data from this investigation can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: Arabidopsis: CWM1 (At1g17630), ACTIN2 (At3g18780), CREF3 (At3g14330), CREF7 (At5g66520), CRR22 (At1g11290), CRR28 (At1g59720), DOT4 (At4g22050), ECD1 (At3g49170), ELI1 (At4g37380), LPA66 (At5g48910), MEF1 (At5g52630), MEF7 (At5g09950), MEF10 (At3g11460), MEF11 (At4g14850), MEF14 (At3g26780), MEF22 (At3g12770), MEF26 (At3g03580), MEF29 (At4g30700), MEF35 (At4g14050), OTP81 (At2g29760), OTP82 (At1g08070), OTP84 (At3g57430), OTP85 (At2g02980), OTP86 (At3g63370), RARE1 (At5g13270), REME1 (At2g03880), REME2 (At4g15720), VAC1/ECB2 (At1g15510), YS1 (At3g22690). P. patens: PPR45 (XP_024389794.1), PPR56 (XP_024403439.1), PPR65 (XP_024374050.1), PPR71 (XP_024395351.1), PPR77 (XP_024375995.1), PPR78 (XP_024361076.1), PPR79 (XP_024376960.1), PPR91 (XP_024400211.1), PPR98 (XP_024367418.1), Ef1α (PHYPADRAFT_97715).

Supplemental Data

• Supplemental Figure 1. Spore germination assay of P. patens ppr79 Reute ko and wild type Reute.

• Supplemental Figure 2. Mass spectrometry analyses of subunits of OXPHOS complexes from P. patens separated by 2D BN / SDS PAGE.

• Supplemental Figure 3. Germination assays of Arabidopsis Col-0 compared with the cwm1-1 and cwm1-2 mutants.

• Supplemental Figure 4. P. patens PPR79 and Arabidopsis consensus sequence of moss RNA targets of ccmBeU428SL, ccmCeU463RC and nad5eU609VV.

• Supplemental Figure 5. Uncondensed clades of PPR79 and CWM1 orthologs of the phylogram in Figure 6.

• Supplemental Figure 6. RNA editing efficiency and relative expression of Arabidopsis cwm1 transgenic lines shown in Figure 7.

• Supplemental Figure 7. Funaria hygrometrica PPR79 ortholog and P. patens /F. hygrometrica RNA target nad5eU598RC.

• Supplemental Figure 8. Phenotype of P. patens transgenic lines shown in Figures 1 and 8

• Supplemental Figure 9. RNA editing efficiency and relative expression of P. patens ppr79 ko transgenic lines shown in Figure 8.

• Supplemental Figure 10. Plant protein model of complex I based on atomic structures of complex I from the domestic cow (Bos taurus;Fiedorczuk et al., 2016) with subunits missing in plants removed (Senkler et al., 2017).

• Supplemental Table 1. RNA editing of site nad5eU598RC in moss species with a putative ortholog of P. patens PPR79.

• Supplemental Table 2. Arabidopsis cwm1 ko transgenic lines.

• Supplemental Table 3. P. patens ppr79 ko transgenic lines.

• Supplemental Table 4. Constructs used for transformation of P. patens ppr79 ko and Arabidopsis cwm1 ko.

• Supplemental Table 5. List of oligonucleotides.

• Supplemental Data Set 1. Evaluation of sporophyte induction in wild type Reute and ppr79 Reute ko.

• Supplemental Data Set 2. Analysis of putative RNA editing sites nad5eU598RC, ccmBeU428SL, ccmCeU463RC and nad5eU609VV in moss mitochondrial coding sequences.

• Supplemental Data Set 3. Alignment of putative PPR79 orthologs from mosses and putative CWM1 orthologs from angiosperms for phylogenetic tree calculation.

• Supplemental Data Set 4. Newick format of the cladogram shown in Figure 4 and the phylogenetic tree shown in Figure 6.

• Supplemental Data Set 5. Accession numbers for all putative PPR79 and CWM1 orthologs used for phylogenetic analysis.

DIVE Curated Terms

The following phenotypic, genotypic, and functional terms are of significance to the work described in this paper: