The P-type pentatricopeptide repeat protein DWEORG1 is a non-previously reported rPPR protein of Arabidopsis mitochondria

Gene expression in plant mitochondria is mainly regulated by nuclear-encoded proteins on a post-transcriptional level. Pentatricopeptide repeat (PPR) proteins play a major role by participating in mRNA stability, splicing, RNA editing, and translation initiation. PPR proteins were also shown to be part of the mitochondrial ribosome (rPPR proteins), which may act as regulators of gene expression in plants. In this study, we focus on a mitochondrial-located P-type PPR protein—DWEORG1—from Arabidopsis thaliana. Its abundance in mitochondria is high, and it has a similar expression pattern as rPPR proteins. Mutant dweorg1 plants exhibit a slow-growth phenotype. Using ribosome profiling, a decrease in translation efficiency for cox2, rps4, rpl5, and ccmFN2 was observed in dweorg1 mutants, correlating with a reduced accumulation of the Cox2 protein in these plants. In addition, the mitochondrial rRNA levels are significantly reduced in dweorg1 compared with the wild type. DWEORG1 co-migrates with the ribosomal proteins Rps4 and Rpl16 in sucrose gradients, suggesting an association of DWEORG1 with the mitoribosome. Collectively, this data suggests that DWEORG1 encodes a novel rPPR protein that is needed for the translation of cox2, rps4, rpl5, and ccmFN2 and provides a stabilizing function for mitochondrial ribosomes.


Results
DWEORG1 is a P-type PPR protein targeted to mitochondria with a similar expression pattern to rPPR proteins. The composition of the Arabidopsis mitoribosome revealed the presence of 11 ribosomal PPR proteins whose accumulation levels are elevated and identical to other mitoribosomal subunits 15,16,36 . Such high production levels are highly divergent from most mitochondria-targeted PPR proteins that correspond to poorly abundant mRNA-specific trans-factors 7 . However, several other PPR proteins that are predicted to target mitochondria accumulate similarly to rPPR proteins but were not reported to be part of the Arabidopsis mitoribosome 37 . To determine the role that such PPR proteins could fulfill, we characterized the function of the AT3G15590 PPR protein. A recent analysis of the mitochondrial proteome 37 showed that the AT3G15590 protein is an abundant P-type PPR protein with 926 copies/mitochondrion, similar to the abundance of rPPR proteins (442 to 2,189 copies/mitochondrion). The gene AT3G15590 of A. thaliana consists of three exons and two introns (Fig. 1a). Its open reading frame codes for a PPR protein of 610 amino acids in length and a molecular weight of 69.39 kDa. This PPR protein contains nine P-type PPR motifs and no additional C-terminal domains. The AlphaFold prediction 38,39 of DWEORG1 shows repeats consisting of pairs of antiparallel α-helices matching the typical structure of PPR proteins 7,8 (Supplementary Fig. S1). Using TargetP (http:// www. cbs. dtu. dk/ servi ces/ Targe tP), a putative mitochondrial localization signal was predicted at the N-terminal region of AT3G15590 (Fig. 1b).
Its mitochondrial localization was confirmed by transforming Arabidopsis leaf protoplasts with a DWEORG1-eGFP fusion protein. The mitochondria were stained with MitoTracker, and the transformed protoplasts were analyzed with a confocal laser scanning microscope. The red MitoTracker signal overlapped with the GFP fluorescence of the fusion protein (Fig. 1c), confirming the mitochondrial localization of AT3G15590.
Since the abundance of AT3G15590 resembles that of rPPR proteins, the transcription pattern of its gene was compared with that of rPPRs and other kinds of P-type PPRs using GENEVESTIGATOR 40  www.nature.com/scientificreports/ tissues, very similar to that of known rPPR proteins (Fig. 1d). The only difference to rPPRs is seen in the phase of stem elongation, where the expression of DWEORG1 exceeds that of the rPPR genes. Compared with other P-type PPR protein genes, the expression pattern of AT3G15590 differed by showing a generally higher expression level at all the Arabidopsis developmental stages (Fig. 1e). Overall, this suggests that the function of AT3G15590 could be related to rPPR proteins, although it was not found to be part of the Arabidopsis mitoribosome 15,16 .

Dweorg1 plants show a dwarf phenotype.
To further study the role of AT3G15590 in Arabidopsis mitochondria, the GABI-Kat line GK_188B06 was analyzed. This line contains a T-DNA insertion in the third exon of AT3G15590, 974 bp downstream of its start codon (Fig. 1a). Progenies were tested for a homozygous T-DNA insertion via PCR amplification of the wild type allele and the T-DNA insertion in AT3G15590. Plants homozygous for the T-DNA insertion were exclusively used in this study and show a dwarf phenotype compared with the wild type (Fig. 2a). In reference to the reduced size of its mutant, the AT3G15590 was renamed DWEORG1 (which means "dwarf " in old English). In homozygous mutant plants, no full-length mRNA derived from dweorg1 could be detected via reverse-transcription PCR ( Supplementary Fig. S1), strongly suggesting that the used mutant line represents a null mutant. The mutant phenotype was reversed by transforming the mutant  www.nature.com/scientificreports/ plants with a functional copy of the DWEORG1 gene (Fig. 2b), confirming that the observed phenotype was caused by the absence of a functional DWEORG1 gene.
The translational efficiency in dweorg1 mitochondria is reduced. rPPR proteins are part of the mitochondrial ribosome in plants, and mutants of these rPPR proteins can show impaired mitochondrial translation 15 . The close similarity between DWEORG1 and rPPR expression patterns led us to investigate the impact of DWEORG1 loss on mitochondrial translation via ribosome profiling analysis 41 . Total ribosome footprints were prepared from flowers of wild type and dweorg1 mutant plants, mapped to the mitochondrial genome of Col-0 and then quantified in each mitochondrial gene (as previously indicated in Planchard et al. 25 . A statistical analysis of these footprints allows the determination of relative translation levels of the respective mRNAs, thereby allowing a comparison of translational efficiency of all the mitochondria-encoded genes between wild type and dweorg1 plants. To determine the translational efficiency, the abundance of each mitochondrial mRNA needed to be determined first since it is defined as the average number of ribosome footprints per mRNA 25 . The abundance of mRNAs was determined via RNA-seq using the input sample for the ribosome footprint preparation (before RNase I treatment) and is reported as reads per kilo base per million mapped reads (RPKM) values per mitochondrial gene in Fig. 3. RNA-seq revealed a slight elevation of all mRNAs in dweorg1 mitochondria except for atp4, ccmB, ccmC, ccmF, matR, mttB, nad2, rpl2, and rps7, which are equal or slightly less abundant in dweorg1 compared with the wild type (Fig. 3). The normalization of ribosome footprint number by mRNA abundance led us to calculate the translational efficiency for each mitochondrial transcript in wild type and dweorg1 plants (Fig. 3). A general decrease in ribosome loading per mRNA could be detected for almost all mitochondria-encoded mRNAs in dweorg1 plants. Most mitochondrial transcripts show a decrease in translational efficiency by a factor of 1.5 to 2 compared with the wild type. However, the translation of several mRNAs such as cox2, rps4, rpl5, and ccmF N2 is more pronounced, with a decrease of up to threefold in translation efficiency of the rps4 and rpl5 transcripts (Fig. 3).
The level of Cox2, Nad9 and Rpl16 is decreased in dweorg1 plants. As the overall translation in dweorg1 mitochondria appeared to be negatively affected, the abundance of several mitochondrial proteins was analyzed via western blot analysis. Antibodies against mitochondria-encoded proteins such as Nad6 and Nad9   Fig. S4). This analysis showed, that the Cox2 protein level was significantly reduced in dweorg1. In addition the levels of Nad9 and Rpl16 seem to be slightly reduced in dweorg1 while no significant differences in other protein levels could be detected (Fig. 4a, Supplementary Fig. S3 and S4). In the lines complemented with wild type DWEORG1, the level of Cox2 protein was restored to that of wild type ( Supplementary Fig. S3). Although the protein levels of Cox2 and Nad9 are decreased in dweorg1, BN-PAGE gel analysis did not reveal any visible reduction in complex I and IV accumulation compared with the wild type, as for other respiratory complexes ( Supplementary Fig. S5). As a control, the steady-state levels of the same proteins were measured in the rppr1 mutant and compared with that in dweorg1.
In the case of rppr1, a more drastic decrease in Cox2 protein accumulation as well as all other tested mitochondria-encoded proteins (i.e., Nad6, Nad9 and Cob) could be detected.
Mitochondrial mRNA processing is not negatively affected in dweorg1. To better understand the function of DWEORG1, its potential role in mitochondrial RNA processing was also investigated. As already indicated, the RNA-seq results did not reveal any loss of any mitochondria-encoded mRNA in mutant plants, suggesting that DWEORG1 does not carry any obvious stabilization activity toward certain mitochondrial mRNAs (Fig. 3). The splicing efficiency of each mitochondrial intron was next quantified via RT-quantitative PCR analysis, and indeed, some introns showed an improved splicing efficiency in dweorg1 plants compared with the wild type, ruling out any negative effect of the missing DWEORG1 on mitochondrial splicing (Supplementary Fig. S6). The editing status of mitochondrial transcripts was also investigated via the sequencing of RT-PCR products of the mitochondrial transcripts, which covered 180 editing sites, but no difference to the wild type was detected (Supplementary Table S2). In addition, a SNaPshot analysis [42][43][44] of 408 editing sites revealed no changes in RNA editing in the dweorg1 mutant compared with the wild type (Supplementary Table S3). As disturbed mRNA end formation could impact protein production in plant mitochondria 45 , the 5′ and 3′ termini of cox2, rps4, rpl5, and ccmF N2 transcripts were comparatively analyzed via circular RT-PCR (cRT-PCR) www.nature.com/scientificreports/ analysis in wild type and dweorg1 plants, since these transcripts' translational efficiency was most affected in dwe-org1. No difference in the size of the amplified products could be observed in the wild type and dweorg1 samples for ccmF N2 , cox2, and rpl5 (Fig. 4b). In the case of the rps4 transcript, different-sized products were detected in dweorg1 but not in the wild type. The generation of larger PCR products with this method was shown before by Forner et al. 24 . The different 5′ and 3′ termini for the cox2, rps4, rpl5, and ccmF N2 transcripts are shown in Fig. 4c. Mature cox2 mRNAs bear 5′ termini located either 205 or 151 nucleotides upstream of the start codon and a 3′ terminus positioned 361 nucleotides downstream of the stop codon. Mature rps4 mRNAs bear a 5′ termini located at position + 2 (+ 1 = the first base of the start codon) and a 3′ terminus positioned 498 nucleotides downstream of the stop codon. The dicistronic mature rpl5-cob mRNAs bear 5′ termini located either 459 or 406 nucleotides upstream of the start codon and a 3′ terminus positioned 58 nucleotides downstream of the stop codon. Mature ccmF N2 mRNAs bear 5′ termini located either 57 or 35 nucleotides upstream of the start codon and a 3′ terminus positioned 170 nucleotides downstream of the stop codon 24 . The sequencing of cRT-PCR amplification products showed that the major form of cox2 (− 151/+ 361) and ccmF N2 (− 35/+ 170) accumulate normally in dweorg1 and wild type plants (Fig. 4c). For rpl5-cob transcripts, sequencing confirmed a − 405/+ 58 form in both wild type and dweorg1 plants. The sequencing results for rps4 detected a + 2/+ 202 form in the wild type but a + 1/+ 498 form in dweorg1. Since the codon UUG, 39 nt downstream of the annotated rps4 start codon is likely used as the translation initiation start codon, this difference should have no effect on the translation of rps4 in dweorg1 46 .
To detect and quantify the − 205 5′ terminus of cox2, quantitative RT-PCR was employed using a forward PCR primer starting 205 nucleotides upstream of the cox2 start codon ( Supplementary Fig. S7). The relative expression value in dweorg1 showed no significant difference with the wild type regarding the 3′ terminus level . The dweorg1 mutant accumulates slightly reduced levels of the Cox2 and Nad9 proteins. 5′ and 3′ termini processing for cox2, rps4, rpl5, and ccmF N2 is not affected in dweorg1. (a) Immunoblot analysis comparing the steady-state levels of selected mitochondria-encoded (Nad9, Cox2, Rpl16, ATPβ) and nuclearencoded (RISP, rPPR1, CYTc) mitochondrial proteins in wild type, dweorg1, and rppr1 mutants. PORIN was used as a loading control; 25, 50, and 100 µg of total proteins were used in each lane, as indicated. The molecular weight of the respective proteins is given in kDa. Original, full-length blots are presented in Supplementary  Fig. S10. (b) cRT-PCR analysis amplifying the 5′ and 3′ termini of cox2, rps4, rpl5-cob, and ccmF N2 transcripts. Total RNAs were extracted from the indicated genotypes, circularized, and then used for cDNA synthesis prior to PCR amplification using divergent primers near the mRNA extremities. The obtained PCR amplification products were run on the shown agarose gel. H 2 O was used for PCR water control. Original, full-length gels are presented in Supplementary Fig. S11. (c) Illustration of the determined 5′ and 3′ termini of cox2, rps4, rpl5cob, and ccmF N2 in wild type and dweorg1 mutant plants via cRNA analysis, quantitative RT-PCR, and reverse transcription PCR. Depicted is the structure of cox2, rps4, rpl5-cob, and ccmF N2 mRNA, exons being shown as orange boxes, introns as dark blue boxes, intergenic regions (not to scale) as yellow boxes, and the 5′ and 3′ termini as light blue boxes. The 5′ and 3′ termini are shown as identified in 24 Supplementary Fig. S7), indicating intact 5′ and 3′ termini for the analyzed transcripts. Together with the results of the cRT-PCR analysis, these findings suggest that the decrease in translational efficiency of cox2, rps4, rpl5, and ccmF N2 is not caused by modified extremities of corresponding mature mRNAs in dweorg1 plants.
DWEORG1 sediments with ribosomes. Since the DWEORG1 gene shows an expression pattern similar to those of rPPR genes, an association of DWEORG1 with Arabidopsis mitoribosomes was analyzed via sucrose gradient sedimentation analysis. For this analysis dweorg1 plants expressing a C-terminal fusion of DWEORG1 with eGFP were used. Expression of the DWEORG1-eGFP complemented the dweorg1 phenotype (Supplementary Fig. S8). To this end, mitochondrial extracts prepared from plants expressing a DWEORG1-eGFP fusion were fractionated on a continuous 15%-55% sucrose gradient, and the eight recovered fractions were analyzed using immunoblot assays. To localize mitoribosomes within the gradients, antibodies to the mitoribosomal proteins Rps4 and Rpl16 were used, two proteins belonging to the small and large subunits of mitoribosomes, respectively. Rpl16 and Rps4 are both peaking in fractions three and four, respectively (Fig. 5a). DWEORG1 is peaking similarly to Rps4 and Rpl16 in fractions two to four, while being detected in fractions one to six.
To confirm an association of DWEORG1 with the mitoribosomes, the analysis was repeated under ribosomedestabilizing conditions by treating the mitochondrial extracts with 0.6 mM puromycin or 5 U RNase I before loading them on the gradients. Puromycin leads to a direct destabilization of the ribosomes, whereas RNase I treatment destabilizes ribosomes by degrading rRNAs 47 . In both cases, a shift of ribosomal proteins toward upper fractions was observed, as for DWEORG1 (Fig. 5a), further supporting an association of DWEORG1 with the mitochondrial ribosomes. When separating the mitochondria extract isolated from dweorg1 plants on the sucrose gradient, Rps4 was detected in fractions three to five (peaking in fraction three and four) and Rpl16 in fractions two to five (peaking in fraction four), while the signal in fraction five is very weak (Fig. 5b). This suggests less loading of mitoribosomes on mRNAs in dweorg1, contributing to a lower translation efficiency in the mutant, thus confirming the results shown in Fig. 3.
The accumulation of mitochondrial rRNAs is reduced in dweorg1 mutants. The rRNA level can be used as a proxy for ribosomal subunit accumulation since rRNAs are unstable when unassembled 48 . The www.nature.com/scientificreports/ levels of cytosolic 18S and 25S rRNAs, chloroplast 16S and 23S rRNAs, and mitochondrial 18S and 26S rRNAs were determined via quantitative RT-PCR analysis. As seen in Fig. 6a, no difference in rRNA levels could be detected for both cytosolic and chloroplast rRNAs; however, a significant reduction in mitochondrial rRNAs was observed in dweorg1 plants compared with the wild type. The mitochondrial 18S rRNA in dweorg1 was reduced to 30% of the wild type levels, while the mitochondrial 26S rRNA was only reduced to 65% of the wild type levels (Fig. 6a). Based on the respective rRNA level, the ratios among the cytosolic, chloroplast, and mitochondrial rRNAs of the LSU and SSU rRNAs were calculated (Fig. 6b). While the LSU/SSU ratio in the cytosol and chloroplast showed no significant difference in the wild type, a significant increase of this ratio was found in dweorg1, confirming an imbalance of SSU and LSU in dweorg1 mitochondria, while the cytosolic and chloroplast ribosomes are not affected. The decrease in 18S and 26S rRNA in the dweorg1 mutant was confirmed via Northern blot analysis (Fig. 6c).

Discussion
DWEORG1 is likely a novel mitochondrial rPPR protein. PPR proteins constitute a large family of plant proteins facilitating organellar RNA expression through processes such as RNA editing, splicing, 5′-and 3′-end maturation, RNA stability, and translation [5][6][7] . Additionally, PPR proteins were recently reported to be part of the plant mitoribosome. So far, 11 ribosomal PPR (rPPR) proteins were identified-rPPR1, rPPR2, rPPR3a, rPPR3b, and rPPR4-10 15,16,36 . Therefore, PPR proteins can affect the translation of mitochondrial proteins either directly or indirectly, depending on the process they are involved in. The evidence provided in this study suggests a direct role of DWEORG1 in mitochondrial translation through an association with the mitoribosome as a novel rPPR protein.
Most mutations in a PPR protein lead to a severe phenotype of the plant since they affect vital biological processes such as plant growth, development, and reproduction 5,6 . This is not true for the dweorg1 mutant; here, we only observe a mild growth delay (Fig. 2), just like rPPR mutants rppr1 or rppr3a, rppr3b, and rppr8, which also show a mild growth delay or do not show any macroscopic phenotype at all, respectively 15 . Another feature of DWEORG1 is its high expression during all developmental stages compared with other P-type PPR act2 was used as an internal control. The values shown are the mean of three biological and two technical repeats with ± SDs (**P < 0.01, student's t-test). (b) Ratios between rRNA levels of mitochondrial, chloroplastic, and cytosolic LSU and SSU. The values shown are the means of three biological and two technical repeats with ± SDs (*P < 0.05, **P < 0.01, student's t-test). (c) RNA gel blot analysis of wild type, dweorg1, and rppr1 mitochondrial ribosomal rRNAs. The total RNA was extracted from the indicated genotypes, separated on an agarose gel, and blotted onto a nylon membrane, and the membrane was hybridized with probes to the mitochondrial 26S, 18S, and 5S rRNAs. Original, full-length gels are presented in Supplementary Fig. S13 www.nature.com/scientificreports/ proteins with a much lower expression (Fig. 1e). The same expression pattern is seen for rPPR proteins (Fig. 1d). Furthermore, DWEORG1's copy number of 926 copies/mitochondrion is comparably high for a PPR protein; most other PPR proteins are only found at 1 and 20 copies/mitochondrion, while rPPR proteins are found at copy numbers similar to DWEORG1 37 . All this makes DWEORG1 a good candidate for a novel rPPR protein.
One of the identified rPPR proteins, rPPR1, was originally shown to be associated with polysomes and named PPR336 49 . When it was later found to be a core protein of the mitoribosome, it was renamed rPPR1, and an overall decrease in translational efficiency of mitochondrial transcripts was shown in rppr1 mutants 15 . Indeed, we could show that DWEORG1 does have a similar effect on mitochondrial translation as rPPR1 but to a lesser degree. We could show an overall decrease of the translational efficiency of mitochondrial transcripts, with cox2, rps4, rpl5, and ccmF N2 transcripts being the most affected (Fig. 3). While the translational deficiency in rppr1 is reflected at the mitochondrial protein level, it is different in dweorg1. For cox2 the decrease in translational efficiency could also be detected at the protein level, as well as for nad9 and rpl16 but not for rps4 or the other analyzed mitochondria encoded proteins ( Fig. 4 and Supplementary Fig. S3). This might be due to the minor decrease in translational efficiency for most of the mitochondrial transcripts requiring a more sensitive detection method than a western blot. Another reason for this could be a compensation of the decreased translation via a decreased protein degradation rate in dweorg1, especially in the case of important conserved proteins like Rps4. This has yet to be further analyzed.
The translation efficiency of mitochondrial transcripts can be affected by PPR proteins directly as some are part of the mitoribosomes and others act as transcript-specific translational activators or indirectly because of the defective RNA processing of specific transcripts. The data presented here points in the direction of a direct role of DWEORG1 by interacting with the mitoribosomes. If DWEORG1 acts as a transcript-specific translational activator (directly or indirectly) for cox2, rps4, rpl5, and ccmF N2 , the overall decrease in the translational efficiency of the other mitochondrial transcripts could be a secondary effect. Up to now, very few proteins have been found to control translation in plant mitochondria, all belonging to the PPR protein family. RF1 and PPR-B act as restorers of fertility in rice (Oryza sativa) and radish (Raphanus sativus), respectively, by negatively regulating the translation of cytoplasmic male sterility (CMS), causing transcripts orf138 in radish and orf79 in rice, respectively [50][51][52] . MPPR6 interacts with the 5′-UTR of rps3 mRNA in maize, and mppr6 mutants express 5′-extended rps3 transcripts whose translation is likely impaired, causing an overall reduction of mitochondrial translation 53 . MTL1, an Arabidopsis PPR protein, is directly required for translation initiation of nad7 mRNA. MTL1 does not affect the 5′-processing of nad7 mRNA just as Rf1 and MPPR6 do with their target RNA but rather directly controls nad7 translation through a mechanism that still needs to be understood 54 . In contrast to chloroplasts, mitochondrial transcripts do not possess a Shine-Dalgarno sequence or any other obvious ribosome binding site 21,22 . It is therefore believed that gene-specific trans-factors such as PPR proteins guide the ribosomes to the site of translation initiation. Binding to the mRNA's 5′-UTR could be one way of action for these trans-factors by altering its structure and thereby making it accessible for the mitoribosomes. This has been proposed for PPR10 in maize chloroplasts, the only PPR protein for which the molecular mode of action in translation initiation has been further analyzed. PPR10 binds to the 5′-UTR of atpH mRNA, preventing the formation of a stem-loop structure and thereby making the Shine-Dalgarno sequence accessible for the small ribosomal subunit 55 . A function similar to PPR10 seem unlikely for DWEORG1. DWEORG1 does not follow the postulated PPR-code ( Supplementary Fig. S9), hence no putative RNA binding site of DWEORG1 could be predicted. In addition, no sequence similarities are detected in the 5′ ends of cox2, rps4, rpl5, and ccmF N2 that could act as target sequences for DWEORG1. Hence, a sequence-specific translation activation through DWE-ORG1 is unlikely for these transcripts. In addition, no defect in RNA processing for any mitochondrial transcript was detected; all transcripts are transcribed and stable, show intact 5′-and 3′-UTRs, are edited correctly, and are normally spliced (see Fig. 4, Supplementary Fig. S4, S5, Supplementary Table S2 and S3).
These findings, in addition to the similarities to rppr1 mutants, lead to the conclusion that DWEORG1 could be a non-previously reported rPPR protein and also be part of the plant mitoribosome. This is supported by our finding that DWEORG1 containing particles fractionate with ribosomes (Fig. 5) as well as by the fact that in a study of plant mitoribosomes, DWEORG1 was also found to be co-sedimenting in sucrose gradients with mitoribosomal components 16 . Since DWEORG1 was co-precipitated with ribosomal components under stabilizing conditions but not under destabilizing conditions 15 , it seems more likely that DWEORG1 functions as a variable part of mitoribosomes rather than as a core protein of mitoribosomes. Defects in mitoribosomal proteins can have different effects on the mitochondrial translation. While rppr1 mutants show a more or less equal decrease in translational efficiency for all mitochondrial transcripts 15 , silencing the RPS10 gene in Arabidopsis leads to an increase in the translation of mRNAs coding for mitoribosomal proteins and a decrease in the translation of OXPHOS components 27 . In the case of dweorg1 mutants, the translational efficiency of all mitochondrial transcripts is slightly decreased, with four transcripts (cox2, rps4, rpl5, and ccmF N2 ) being more affected than the rest of the mitochondrial transcripts. These transcripts cannot be assigned to just one category since they are coding for mitoribosomal proteins as well as for an OXPHOS component and a component of the cytochrome c biogenesis. Why these four transcripts are more affected has yet to be further analyzed. Rps4 and Rpl5 are both conserved r-proteins 15,16 , a deficiency in their translational rate could cause the overall decrease in mitochondrial translation efficiency as seen for a rps3 mutation on maize plants showing an overall disturbed proteinsynthesis 53 . But as shown in rps10 plants, a low r-protein level usually leads to a general increase of translational rates of mitoribosomal proteins, which is not seen in dweorg1. Hence it is more likely that the decrease in translational efficiency seen in dweorg1 is a direct effect of the loss of DWEORG1. www.nature.com/scientificreports/ diverged strongly in their structure, rRNA, and protein content from the ancestral ribosomes during eukaryotic evolution 56,57 . Mitoribosomes are composed of two ribonucleoprotein complexes: the LSU and the SSU. These complexes show a distinct combination of rRNAs and proteins that is distinct for different eukaryotes 58 . The plant mitoribosome, with its many rRNA ESs and plant-specific ribosomal proteins (r-proteins), is one of the most complex mitoribosomes 15 . One distinct feature of plant mitoribosomes is their 11 rPPR proteins, a feature only shared by mammalian and trypanosomal mitoribosomes, containing six and two rPPRs, respectively 15,16,59,60 .
In animals, the rPPR protein mS39 is found to bind to leaderless mitochondrial mRNAs, leading them to the entrance channel of the ribosome 61 . While some rPPRs in trypanosomal mitoribosomes are only present as structural components of the ribosome and do not interact with the RNA at all 62 , other rPPRs in trypanosomes assemble around the shortened rRNAs, stabilize and position their functionally important regions in the ribosome 60 . A stabilizing function for rRNAs is also seen for plant rPPRs; the foot and head extensions of the plant mitoribosome SSU are stabilized by rPPR1, rPPR3a, or rPPR3b and rPPR6, respectively 36 . The same is seen for the extension segments of the 26S rRNA of the LSU.
Here, we have shown a decrease in the 26S and 18S rRNA levels in the dweorg1 mutant (Fig. 6), suggesting a role for DWEORG1 in rRNA stabilization. Reduced rRNA levels are a good indicator of ribosome assembly since non-assembled rRNAs are very unstable 48 . In dweorg1 mutants, we could observe a shift of ribosomal proteins Rps4 and Rpl16 to the upper fractions of a sucrose gradient (Fig. 5b). The upper fractions of the sucrose gradient represent monosomes attached to RNA, free and unassembled ribosomal subunits. This is in accordance with the decrease in translational efficiency in the mitochondria of dweorg1 plants and hints at a disturbed mitoribosome biogenesis in the mutant compared with the wild type.
A defect in mitoribosomal biogenesis is also seen in mitochondrial rps10 mutants 27 . The translation of mitochondrial transcripts is maintained in the rps10 mutant as well as in the dweorg1 mutant despite the disturbed mitoribosome biogenesis. In both cases, a change in translational efficiency of all mitochondrial transcripts can be observed, but the decrease or increase affects the various transcripts differently in the two mutants. Since two of the transcripts affected by DWEORG1 encode for ribosomal proteins-namely, Rps4 and Rpl5-there are two possibilities for the unstable ribosomes: (i) DWEORG1 itself stabilizes rRNAs and thereby mitoribosomes in an, up to now, unknown way; or (ii) the destabilization of the mitoribosomes is a secondary effect caused by the insufficiently translated mitoribosomal proteins Rps4 and Rpl5, leading to incompletely assembled/unstable mitoribosomes. In the latter case, one would expect an increase in the translation of other mitoribosomal protein mRNAs, as seen in the rps10 mutant 27 , but this is not the case in dweorg1 mitochondria. Thus, excluding scenario (ii). The data received in this study shows an imbalance of SSU and LSU in mitochondria, with an excess of LSU compared with SSU (Fig. 6b). Although the molecular details remain to be determined, our data indicate a stabilizing role of DWEORG1 for the small subunit 18S rRNA of mitoribosomes in A. thaliana in addition to an up to now unknown promoting effect on the translation of cox2, rps4, rpl5, and ccmF N2 .
In plants, the hypothesis of a heterogeneity of mitoribosomes exists and could allow a highly dynamic mitochondrial translation machinery composition 63 . Studies on LSU proteins in yeast showed that many r-proteins, conserved and specific, are not essential for functional mitoribosomes 64 , suggesting the possibility for differentially composed mitoribosomes. Existing evidence shows a different composition of mitoribosomes in different developmental stages and/or tissues in Arabidopsis. The mRNA levels of HUELLENLOS (HLL) and HUELLEN-LOS PARALOG (HLP), two paralogs of the mitoribosomal protein RPL14, differ in carpels and leaves. HLL is significantly more expressed in carpels than HLP, and the opposite is seen in leaves 65 . Since the two proteins are redundant in their function when expressed ectopically, it can be assumed that they are integrated into mitoribosomes in a tissue-specific manner. A changing mRNA level during leaf development is also observed for many other nuclear-encoded mitoribosomal proteins 63 . RPL14 is not the only mitoribosomal protein with different paralogues in Arabidopsis; for RPL12, four paralogues that are all part of the mitoribosome were described 66 . This could reflect a heterogeneous population of mitoribosomes, supporting the hypothesis of heterogeneity in mitoribosomes. A differential composition of mitoribosomes could allow for a regulative role of mitoribosomes in gene expression. DWEORG1 might also have a regulative role as part of the mitoribosome. The expression pattern of DWEORG1 resembles that of rPPR protein genes in all developmental stages, except the stem elongation phase (Fig. 1). In this phase, the expression of DWEORG1 is higher, than that of other rPPR genes, indicating a possible role of DWEORG1 during stem elongation as part of the mitoribosome. This could also explain the mild effect on translational efficiency seen here, since the data for translational efficiency in dweorg1 were collected from inflorescences of Arabidopsis that form after the stem elongation phase.
The ribosomal filter hypothesis postulates that ribosomes are not only the passive site for translation but also able to regulate translation and therefore gene expression by choosing specific mRNAs for translation as a response to different physiological conditions as well as different developmental stages [67][68][69] . This was shown for yeast mitoribosomes; the r-protein mS38 is specifically needed for the translation of cox1, cox2, and cox3 transcripts 70 . In mS38 mutants, a reduction of the overall translation rate of mitochondrial transcripts is observed, but still, all mitochondrially encoded proteins were synthesized in normal amounts except for the Cox1, Cox2, and Cox3 proteins, which are synthesized in a reduced amount. This seems to be similar in dweorg1 mutants (Fig. 4a). One process in plant mitochondria that is already proven to be regulated by translation is OXPHOS biogenesis 71 . This feature might enable plant mitoribosomes to translate transcripts without a distinct ribosome-binding site and to target specific mRNAs under different conditions. This theory would also explain why mitoribosomes can still translate mitochondrial transcripts when ribosomal proteins are not synthesized in a sufficient amount. If the heterogeneity of mitoribosomes exists, the composition of the ribosomes is variable to a certain degree, allowing for functional ribosomes even when one ribosomal protein is missing. The findings of this study support this theory.  74 . The flasks containing the seedlings were cultured, shaking at 80 rpm in the dark at 25 °C, and after two weeks, 50 µg/L of indole-3-acetic-acid was added to the medium to support root growth. After one more week of culture, the medium was renewed. One to two weeks after this renewal, the cultures were harvested and used for the isolation of mitochondria. Subcellular localization of DWEORG1. Protoplasts were prepared from ten-day-old dark grown A.
thaliana suspension cultures MM1 78 . The protoplasts were transformed with the DWEORG1-eGFP fusion construct using the polyethylene glycol (PEG) method 79,80 . After overnight incubation, the mitochondria of the protoplasts were stained with MitoTracker® Orange CM-H2TMRos (Invitrogen) at a final concentration of 500 nM and examined using a Leica confocal laser scanning microscope (Leica Microsystems).
Isolation of mitochondria and sucrose gradient analysis. The mitochondria of the Arabidopsis root cultures were isolated using a procedure previously described in Klein et al. 81 except that the Percoll gradients were replaced by sucrose gradients 82,83 . For sucrose gradient sedimentation analysis, 3 mg of mitochondria were first lysed in a lysis buffer (0.2 M Tris-HCl (pH 9.0), 0.2 M KCl, 35 mM MgCl 2 , 25 mM EGTA, 0.2 M sucrose, 1% (v/v) Triton X-100, 2% polyoxyethylene-10-tridecylether, 1 mg/ml heparin, 0.1 M β-mercaptoethanol, 100 µg/ml chloramphenicol, and 25 µg/ ml cycloheximide). A total of 800 µl of the obtained supernatant was layered onto an 11 ml 15%-55% sucrose gradient and centrifuged at 200,000 g for 120 min. Eight fractions of 1,400 µl were then manually collected and the proteins from each fraction extracted as described in Mašek et al. 84 . The obtained proteins were further analyzed via immunoblotting analysis. RNA analysis. Total RNA was extracted from flower buds using the TRIzol reagent (Thermo Fisher Scientific), following the manufacturer's instructions. To remove genomic DNA contamination from total RNA preparations, the DNase Max Kit (Qiagen) was used. For quantitative RT-PCR analysis, either 35 ng of total RNA was reverse-transcribed and amplified in a single tube with a volume of 15 µl using the QuantiTect SYBR Green RT-PCR Kit (Qiagen) or reverse transcription was performed separately on 1 μg of the total RNAs using a random hexamer. Quantitative RT-PCR for analyzing splicing efficiency in plant mitochondria was carried out, as previously described in Haili et al. 54 , using the described set of primers. Two biological repeats and three technical repeats were performed on each genotype. cRT-PCR was performed as described in Forner et al. 24 . RNA gel blot analysis was performed as previously indicated in Haili et al. 54 . Shortly, 10 µg of total RNA was separated on a 1.5% (w/v) agarose gel. After electrophoresis, the RNAs were transferred to nylon membranes and cross-linked in a UV Stratalinker. Hybridization was performed with specific DNA probes amplified by polymerase chain reaction (PCR) and radiolabeled by the Prime A Gene labeling kit (Promega), following the manufacturer's instructions.