Multifarious Evolutionary Pathways of a Nuclear RNA Editing Factor: Disjunctions in Coevolution of DOT4 and Its Chloroplast Target rpoC1eU488SL

Identifying DOT4 Orthologs and Phylogenetic Analyses

Arabidopsis thaliana editing factor DOT4 (AT4G18750, Hayes et al. 2013) was used as protein query in BlastP and TBlastN searches (Altschul et al. 1990) against the angiosperm (magnoliophyte) data in the NCBI protein database and in the Transcribed Shotgun Assemblies (TSA) and Whole Genome Shotgun sequences data, respectively (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The previous sampling of angiosperms with high-quality genomic and/or transcriptomic data (Hein and Knoop 2018) was extended by Anthurium amnicola, Persea americana, Helianthus annuus, and Kalanchoe fedtschenkoi to now comprise 121 flowering plant species. Determination of genome and/or transcriptome data quality was done as described previously (Hein and Knoop 2018). The MEGA alignment explorer (Tamura et al. 2013) was used for sequence alignment and processing. Gaps and missing or inaccurate C- and N-terminal sequences in evidently erroneous protein models could be corrected in most cases by TBlastN searches against respective nucleotide databases and rechecking nucleotide sequences for possible sequence errors. Camptotheca acuminata included in our previous taxon sampling (Hein and Knoop 2018) was here retained although its DOT4 homolog remains incomplete at present owing to a small assembly gap of about 60 amino acids in the TSA data.

Care was taken to exclude paralogous PPR proteins. All sequences were used as queries to check whether DOT4 was consistently identified as the closest Arabidopsis homolog and the growing sequence collection was sequentially rechecked with phylogenetic tree constructions to avoid inclusion of paralogs outside the DOT4 ingroup including the Amborella DOT4 ortholog. The ultimate DOT4 alignment is available from the authors upon request. “WebLogos” were created at http://weblogo.threeplusone.com (Crooks et al. 2004).

Phylogenetic Tree Construction

Final alignments were used for calculation of maximum likelihood phylogenetic trees using the IQ-tree webserver (Trifinopoulos et al. 2016) at http://iqtree.cibiv.univie.ac.at. The JTT+F+R5 model of sequence evolution was used as the best-fitting model identified with the implemented ModelFinder (Kalyaanamoorthy et al. 2017). Node reliability was determined from 1,000 bootstrap replicates with ultrafast bootstrap approximation “UFBoot” (Hoang et al. 2018).

Collection of Chloroplast Sequences and RNA Editing Predictions

Wherever available, chloroplast data were collected from completely determined chloroplast genome sequences matching the species of the nuclear genome taxon sampling. In a few cases, cp data were alternatively obtained from Whole Genome Shotgun sequences data, from closely related sister species (indicated by asterisks in fig. 2) or newly determined during this study (e.g., for Metrosideros, Aquilegia, Anthurium, Eichhornia, and Rauvolfia).

Plant Material and Molecular Work

Plant material for Am.trichopoda, Illicium oligandrum, Spirodela polyrhiza, An.amnicola, Agave americana, Phoenix dactylifera, Eichhornia paniculata, Ananas comosus, Aquilegia chrysantha, Banksia serrata, Camptotheca acuminata, Vaccinium macrocarpon, Coffea canephora, Rauvolfia serpentina, Glycine max, Phaseolus vulgaris, Vigna angularis, Cicer arietinum, Medicago truncatula, Trifolium pratense, Vicia faba, Metrosideros carminea, Citrus sinensis, Eutrema salsugineum, Capsella rubella, Symplocarpus foetidus, Lemna minor, Wolffia columbiana, Spathiphyllum wallisi, Dieffenbachia seguine, Amorphophallus titanum, Alocasia odora, and Colocasia esculenta as well as seeds for Lupinus angustifolius, V.angularis, Vic.faba, and C.arietinum were obtained from the Bonn University Botanic Garden. Seeds were kept on humid filter paper until germination and then transferred to soil for several days before nucleic acid preparation. Cocos nucifera, Epipremnum aureum, and Persea americana were obtained from local stores. Total plant nucleic acids were isolated using CTAB-based (Cetyltrimethylammonium bromide) protocols (Doyle and Doyle 1990; Liao et al. 2004). RNA preparations were alternatively done with the TRI reagent protocol (Sigma Aldrich) and different kits of the Macherey-Nagel NucleoSpin series. Random hexamers or gene-specific primers were used for cDNA synthesis with the Revert Aid First Strand cDNA Synthesis Kit (Thermo Scientific/Fermentas). Gene-specific primers (supplementary table 1, Supplementary Material online) were used to amplify rpoC1 and dot4 gene regions. Polymerase chain reaction (PCR) products were isolated from agarose gels using the NucleoSpin Extract II Kit (Macherey-Nagel) and 2–3 replicates were sequenced directly. Commercial Dideoxy (Sanger) sequencing was done at Macrogen Europe (Amsterdam, the Netherlands). Where RNA editing could not be detected at the rpoC1eU488SL site, procedures were repeated.

An RNA-linker-adaptor strategy was used to detect specific antisense RNA following and adapting published procedures (Bensing et al. 1996; Georg et al. 2010). RNA of Hordeum vulgare was prepared using a modified CTAB protocol (Liao et al. 2004). Following DNase treatment with DNase I (Thermo Fisher Scientific), RNA-5′-polyphosphatase (Epicentre) was used instead of tobacco acid pyrophosphatase to remove the terminal triphosphates at transcription initiation sites. A ribo-oligonucleotide (5′-GAUAUGCGCGAAUUCCUGUAGAACGAACACUAGAAGAAA-3′, Integrated DNA Technologies) was ligated to RNA 5′-monophosphate ends using T4-RNA ligase (Thermo Fisher Scientific). Purification steps performed with NucleoSpin RNA Plant (Macherey-Nagel) were applied after the respective enzymatic treatments. Random hexamers were used for cDNA synthesis prior to Reverse transcription polymerase chain reaction (RT-PCR) with an adaptor primer matching the 5′-region of the ligated ribonucleotide (5′-ATATGCGCGAATTCCTGTAGAACGAACA-3′) and a gene-specific primer in the upstream rps3 coding region (5′-TYGGTTTCAGACTTGGTACAACCC-3′). PCR products were isolated from agarose gels using the NucleoSpin Extract II Kit (Macherey-Nagel) and cloned into the pGEMT-easy vector (Promega). Dideoxy sequencing was done at Macrogen Europe (Amsterdam, the Netherlands).

Conservation of Chloroplast Editing Site rpoC1eU488SL and Its Corresponding Editing Factor DOT4

DOT4 (“Defectively Organized Tributaries,” A.thaliana locus At4g18750) is a PPR protein with 18 PLS-type PPRs and a terminal DYW domain that has been characterized as the site-specific RNA editing factor for editing site rpoC1eU488SL converting codon 163 from serine into leucine in the chloroplast rpoC1 mRNA (Hayes et al. 2013). This is the only documented, and expected, RNA editing event in the Arabidopsis rpoC1 mRNA. It is shared with the early-branching flowering plant Am.trichopoda (Hein et al. 2016). Inspecting other early angiosperm lineages, we now document altogether 13 ancient C-to-U RNA editing sites in the rpoC1 mRNA occurring in different patterns (fig. 1). Other than rpoC1eU488SL only rpoC1eU41SL appears to be more widely conserved between eudicots and monocots. We inspected the now available, high-quality genomic and transcriptomic data of 121 flowering plants for the presence of DOT4 orthologs.

Unequivocal DOT4 orthologs could be clearly identified in (nearly) all angiosperms where RNA editing rpoC1eU488SL remains necessary to re-establish the evolutionarily conserved leucine codon in the chloroplast rpoC1 gene (fig. 2 and supplementary fig. S1, Supplementary Material online). When DOT4 orthologs were identified, they feature full suites of 18 PLS-type PPRs like in the case of the Arabidopsis DOT4 protein with conserved positions 5 and “Last” (L) known to be important for RNA sequence recognition (Barkan et al. 2012) and with carboxyterminal DYW domains including the conserved cytidine deaminase signatures in most cases. However, we also identified several noteworthy exceptions that document a highly dynamic evolution of DOT4 and its target, which we will discuss in the following.

Frequent and Independent Degeneration and Loss of DOT4 after Loss of rpoC1eU488SL

DOT4 orthologs are missing altogether in many taxa where RNA editing rpoC1eU488SL has become obsolete after a cytidine-to-thymidine conversion in the chloroplast DNA (cpDNA) (fig. 2). This is the case for Lamiales, Myrtales, Solanales, and Zingiberales suggesting early losses in these orders. Indeed, checking upon all available cpDNAs in these orders, we find that the rpoC1eU488SL editing site is absent in all Myrtales and Solanales for which sequence information is available. In contrast, the simultaneous absence of rpoC1eU488SL and DOT4 in Allium, Arachis, and Macadamia (fig. 2) suggests more recent losses within the Asparagales, Fabales, and Proteales, respectively. Moreover, the DOT4 case now reveals an interesting evolutionary spectrum for the stepwise degeneration of an RNA editing factor upon loss of its editing target. Despite becoming unnecessary after C-to-T conversion at the previous rpoC1eU488SL editing site, DOT4 shows no evident signs of degeneration in Camelina (Brassicales), in Aquilaria (Malvales), in Aquilegia (Ranunculales), and in Dioscorea and only minor sequence deviations in otherwise highly conserved positions in Agave and in Eichhornia, likely indicating only recent conversion of the editing site (fig. 2).

In other cases, however, the degeneration of DOT4 after loss of its target becomes more pronounced. This is reflected not only by further amino acid exchanges in otherwise highly conserved positions but also by partial or complete deletions of the terminal DYW domain in Daucus (Apiales), Diospyros (Ericales), Helianthus (Asterales), Kalanchoe (Saxifragales), and Nelumbo (Proteales). The Caryophyllales display a particularly wide spectrum of DOT4 sequence degeneration after an early, partial DYW domain deletion, ultimately resulting in complete loss of DOT4 in Amaranthus hypochondriacus, whereas a degenerated DOT4 ortholog is still present in the sister species A. tricolor (fig. 2).

Losses of DOT4 without Previous Loss of the rpoC1eU488SL Editing Target

The above cases show that DOT4 orthologs are mostly retained, while the RNA editing target rpoC1eU488SL is present in the chloroplast genome and that DOT4 becomes stepwise degenerated and ultimately lost upon C-to-T conversions at the previous editing site. However, we now also identified cases where no DOT4 ortholog could be identified in the nuclear genomes although RNA editing in position rpoC1eU488SL was still expected: in Carica papaya, in Vic.faba, and in S.polyrhiza (fig. 2). It could be imagined that another editing factor paralog would act as a substitute for DOT4 in these cases. However, our cDNA studies showed that the genomic C indeed remained unedited in the rpoC1 transcripts (table 1) and that accordingly a serine would be retained instead of a leucine in this otherwise highly conserved position. Notably, in extending our cDNA sampling, we observed “partial” editing at the rpoC1eU488SL site in many cases, hence resulting in a mix of unedited and edited mRNAs, as also observed in independent studies (table 1). It remains to be seen whether differences in reported editing frequencies, like in the case of A.thaliana (table 1), may be due to variability among biological isolates, tissues, or developmental stages.

Because the two Alismatales taxa in the angiosperm-wide sampling (fig. 2) appeared to represent the full range between a species with an apparently intact DOT4 ortholog (An.amnicola) and a species with no recognizable DOT4 ortholog (S.polyrhiza), we chose to investigate the Alismatales more closely. The results for a sampling of 15 Alismatales species is shown in figure 3. The absence of RNA editing, and hence retention of an unedited cytidine, as initially detected in Spirodela, is shared with the other two Lemnoideae (duckweed) species in our sampling, Lemna minor and Wolffia columbiana. In the other taxa of the Araceae editing efficiencies vary in the full range from 100% in Spathiphyllum wallisi to no detectable RNA editing in Epipremnum aureum (fig. 3). Outside the Araceae, a C-to-T conversion has occurred in Potamogeton perfoliatus (pondweed) and the two Hydrocharitaceae taxa Elodea and Najas.

A DOT4 Duplication among Fabales

The DOT4 orthologs exist as single copies in the respective plant nuclear genomes and their phylogeny largely coincides with the phylogeny of flowering plants as currently understood, indicating common ancestry of a single ortholog (supplementary fig. S1, Supplementary Material online). One noteworthy exception are the Fabales where two DOT4 paralogs coexist in Vigna, Phaseolus, Glycine, Cajanus, and Lupinus (fig. 2 and supplementary fig. S1, Supplementary Material online). The phylogeny suggests a deep gene duplication early in Fabales creating two DOT4 copies, a (likely ancient) “GFW-type” and a (likely derived) “DFW-type,” so named after the terminal tripeptides in their DYW domains (fig. 4 and supplementary fig. S1, Supplementary Material online). The DOT4 duplicates may trace back to whole-genome duplication about 58 Myr in the history of the Papilionoideae subfamily ago (Cannon et al. 2015), to which all the species of the current Fabales taxon sampling belong. The DOT4 paralogs have been subsequently lost independently during Fabales diversification. Both copies are lost in Arachis coinciding with a C-to-T conversion at the former editing site and in Vicia where RNA editing activity has been lost (fig. 4). Evidently, this scenario needs to be tested with additional Fabales genome information and denser taxon sampling in the future, especially also outside the Papilionoideae, which show accelerated cp genome evolution (Schwarz et al. 2017).

Intriguingly, the DFW-type of the duplicated DOT4 paralogs shows better sequence conservation in the DYW domains when present in parallel with the GFW-type copy (fig. 4A), especially within and directly upstream of the PG-Box where the GFW-type copies strongly deviated in most species (see supplementary fig. S3, Supplementary Material online). This is also reflected in the relevant PPR amino acid positions 5 and “Last” (L) important for RNA recognition as exemplarily shown for V.angularis (fig. 4B). The long PPR array of DOT4 generally shows a very good match to its corresponding rpoC1eU488SL target (see also fig. 5) and no evidence for any reassignment to an alternative target can be deduced for the second DOT4 copy, which rather shows a degeneration of the positions identified as relevant for RNA targeting (fig. 4B). The GFW-type DOT4 factors might have lost their ancestral RNA editing functionality early considering that RNA editing is absent in Cicer, Medicago, and Trifolium where (only) the DFW-type DOT4 is lost (figs. 2 and 4). The taxa lacking rpoC1eU488SL editing represent the galegoid or “IR-lacking” IRLC subclade of Papilionoideae, which separated from the milletoid sister subclade about 54 Ma (Schwarz et al. 2017).

Loss of the rpoC1eU488SL Editing Target under Retention of DOT4 Orthologs

Whereas the earlier sections document losses of DOT4 despite retention of its editing target rpoC1eU488SL, exactly the opposite is observed among the Poales. The editing target site rpoC1eU488SL is ancestrally lost through C-to-T conversion after a split of Ananas (Bromeliaceae) in the remaining Poales (Poaceae), but unequivocal DOT4 orthologs remain conserved in their nuclear genomes, although with evident signs of degeneration in the terminal DYW domain (fig. 2 and supplementary fig. 3, Supplementary Material online). All Poaceae share a mutation of the HSE motif to HSS (supplementary fig. S3, Supplementary Material online). Because the glutamate (E) residue is of unequivocal importance for RNA editing function (Hayes et al. 2015), the DYW domain is likely unfunctional in Poaceae. Surprisingly, however, the retained DOT4 homologs in the Poaceae not only feature a conserved PPR array, but even have this extended by six additional, amino-terminal PLS-type repeats. Moreover, these six additional PLS-type PPRs feature amino acids in positions 5 and L following canonical PPR-RNA recognition rules, indicative of RNA-binding. However, these would not fit the sequence further upstream of the original editing target (fig. 5A).

We used the new TargetScan module implemented with the recent update of PREPACT (Lenz et al. 2018) to explore potential binding sites for the extended DOT4 homologs in the Poaceae. To this end we used the arbitrary weightings for purine and pyrimidine recognition as introduced before and (optionally) with additional weights of 40% for positions −3 to +2 to match the (former) rpoC1eU488SL editing site (supplementary fig. S2, Supplementary Material online). The latter was intended to account for potential, yet uncharacterized, sequence preferences imposed by the E1, E2, and (degenerating) DYW domain but no differences in top-scoring hits were seen without these extra weights (not shown). Searching for best matches upstream of RNA editing sites in all 15 angiosperm cp editome references currently available in PREPACT, the rpoC1eU488SL edit is identified with top scores despite the fact the six additional PPRs do not at all contribute to binding further upstream (supplementary fig. S2A, Supplementary Material online).

Scanning for matches in complete chloroplast genomes, we find that the extended PPR stretch of DOT4 in Poaceae would best fit to a sequence in the 5′-region of the rps3 coding sequence as exemplarily shown for the H.vulgare cpDNA, accession NC_008590 (supplementary fig. S2B, Supplementary Material online, fig. 5B). However, this sequence fit is in inverse sequence orientation suggesting binding to a potential antisense RNA, which is not immediately to be expected given the location of rps3 in the ribosomal protein cluster rps19-rpl22-rps3-rpl16. On the other hand, several publications have meanwhile reported on the presence of antisense transcripts in chloroplasts (see discussion). Most relevant to our above findings is a comprehensive study on the chloroplast transcriptome of barley (H.vulgare, Poaceae) systematically identifying and differentiating transcription start sites (TSSs) of the plastid-encoded RNA polymerase PEP and the nucleus-encoded (“phage-type”) RNA polymerase NEP (Zhelyazkova et al. 2012). This work identified altogether three TSSs (one of the PEP-type and two of the NEP-type) in antisense orientation within the rpl16 group II intron (rpl16i9g2) immediately downstream of the rps3 gene (fig. 5B). The TSSs were labeled according to their distance upstream of trnH-GUG (Zhelyazkova et al. 2012), the next gene orientated in that direction of transcription. A unique extension of the IRs among Poales cpDNAs places a second copy of trnH-GUG in antisense orientation between rpl2 and rps19 of the ancestral ribosomal protein gene cluster (fig. 5C). To independently confirm rps3 antisense transcripts and check whether they would indeed cover the candidate DOT4 binding site we treated total barley RNA with 5′-RNA-polyphosphatase prior to ligation of a ribo-oligonucleotide to the 5′-monophosphate ends of transcripts as an adaptor target. Followed by specific RT-PCR, this strategy indeed identified rps3 antisense RNAs in the proposed DOT4 binding region extending toward the 5′-end of rps3 for at least 52 nucleotides toward the binding site of a corresponding PCR primer. Intriguingly, the ribooligonucleotide adaptor strategy identified variable 5′-termini of the antisense RNAs likely indicative of ribonucleolytic processing, which clustered in the 5′-region of the proposed DOT4 target (fig. 5B). Dominating among RT-PCR clones was position +9 with five clones, followed by positions +7 and +10 with two clones each and position +5 (one clone). Only one clone revealed a larger antisense RNA with a 5′-end 48 nt. upstream of the suggested DOT4 target.

Aligning the H.vulgare sequences upstream of the antisense TSSs in rpl16i9g2 with homologous sequences in other taxa to check for conservation of the suggested promoter motifs (Zhelyazkova et al. 2012), we observe significant conservation among the two large Poaceae clades (BOP and PACMAD) but much less in Ananas or more distant taxa, most notably for the likely −10 box of the eubacterial-type PEP promoter (fig. 5D).

A persistence of DOT4 orthologs for some evolutionary time after loss of the rpoC1eU488SL editing site, especially of those with apparently intact DYW domains as, for example, in Camelina, Aquilaria, Aquilegia, or Dioscorea (see fig. 2) could be explained by a yet unidentified second editing target of DOT4. Scanning the 15 available angiosperm chloroplast editomes identifies ndhBeU611SL as a second-best match for a (likewise ancestral) chloroplast editing site (supplementary fig. S2C, Supplementary Material online). A significantly lower matching score than for rpoC1eU488SL (950 vs. 1,260) makes this site an unlikely editing target per se. Yet more importantly, edit ndhBeU611SL is also confirmed as an editing site in Solanales taxa like Nicotiana tabacum (supplementary fig. S2C, Supplementary Material online), which lack DOT4 altogether (fig. 2).