R-loop control and mitochondrial genome stability require the 5′-3′ exonuclease/flap endonuclease OEX1

Abstract

Maintenance of the plant organelle genomes involves factors mostly inherited from their symbiotic ancestors. In bacteria, DNA polymerase I (Pol I) performs multiple replication and repair functions through its 5′-3′-exonuclease/flap-endonuclease domain. Plant organelles possess 2 DNA polymerases that are evolutionarily derived from Pol I but lack this key domain. ORGANELLAR EXONUCLEASES 1 and 2 (OEX1 and OEX2) compensate for this missing function and are targeted to mitochondria and chloroplasts, respectively, in Arabidopsis (Arabidopsis thaliana). Loss of OEX1 causes developmental and fertility defects that increase with increasing differential segregation of mitochondrial DNA (mtDNA) subgenomes generated by recombination. OEX1 activity is modulated by alternative splicing, which generates 2 isoforms that variably affect mtDNA stability and repair. OEX1 has 5′-3′-exonuclease and flap endonuclease activities, with a high affinity for RNA–DNA hybrids. It rapidly degrades RNA in Okazaki-like structures and R-loops. Consistent with a role in suppressing R-loops, oex1 mutant plants accumulate RNA–DNA hybrids in highly transcribed mtDNA regions. Taken together, our results identify OEX1 as an important factor that compensates for the missing activity of plant organellar polymerases, playing multiple important roles in the processing of replication and recombination intermediates, such as replication primers and R-loops, whose accumulation can lead to genome instability.

Exonuclease/flap-endonuclease OEX1 compensates for a missing DNA polymerase I activity to maintain mitochondrial genome stability by degrading RNA–DNA hybrids and resolving replication intermediates.

Introduction

Mitochondria and chloroplasts have their own genomes (mtDNA and cpDNA), which are remnants of their past autonomous prokaryotic lifestyle. The mtDNA contains a limited number of genes (37 in humans and 50 to 60 in plants), yet these genes are crucial for cell survival, as they code for essential subunits of oxidative phosphorylation complexes or are required for their expression (Knoop 2004; Chevigny et al. 2020). Although plant and animal mtDNA code for a similar number of genes, plant mitogenomes are considerably larger, with sizes ranging from 200 kb up to 11 Mb (Sloan et al. 2012), and have been observed in diverse structural and organizational forms, including linear, circular, and branched configurations (Oldenburg and Bendich 1996; Backert et al. 1997; Manchekar et al. 2006). This high variability in size and structure results from rearrangements through homologous recombination (HR) involving repeated sequences, which are abundant in plant mitogenomes (Kubo and Newton 2008). These repeated sequences, ranging in size from several kilobases to <100 bp, engage in HR with different frequencies according to their size and genomic position, thus leading to the highly dynamic structure of land plant mitogenomes (Small et al. 1989; Hanson and Bentolila 2004). Recombination is also the main driving force for the rapid evolution of the plant mtDNA organization, which can substantially differ between closely related species and even among different accessions of the same species (Mower et al. 2007; Fertet et al. 2021).

Despite its remarkable organizational plasticity, plant mtDNA protein-coding sequences remain remarkably conserved across most species, indicating the presence of highly effective repair mechanisms. Both base excision repair (BER) and HR-dependent repair processes occur in plant mitochondria (Boesch et al. 2009; Ferrando et al. 2018; Trasvina-Arenas et al. 2018). In particular, break-induced replication (BIR) and illegitimate recombination through microhomology-mediated BIR (MMBIR) have also been proposed as important processes driving replication, repair, and evolution of plant mitogenomes (Cappadocia et al. 2010; Christensen 2013; Gandini et al. 2023). Genetic mutations that impair HR efficiency or specificity can significantly affect plant fitness, resilience to stress, and fertility (Gualberto and Newton 2017).

Factors already identified as involved in plant mitochondrial HR include RecA-like recombinases, several different types of single-strand DNA (ssDNA)-binding proteins, a MutS-like protein (MSH1) involved in the rejection of recombination between homoeologous sequences, and the branch-migration helicases RECG and RADA (Chevigny et al. 2020). Importantly, all HR-dependent repair processes require a DNA polymerase, and while the animal mtDNA is replicated by the single enzyme DNA Pol γ, the maintenance of land plants mitochondrial and plastid genomes involves 2 multifunctional DNA polymerases derived from prokaryotic DNA polymerase I (Pol I) (POL1A and POL1B). Both are dually targeted to plastids and mitochondria (Carrie et al. 2009) and differentially participate in replication and repair processes (Parent et al. 2011; Peralta-Castro et al. 2020). However, many other factors that should play pivotal roles in plant mtDNA replication and repair, including key nucleases, have yet to be identified. During replication, such nucleases should be involved in the processing of Okazaki-like primers prior to the ligation of the newly synthesized DNA segments. Primer removal could occur by strand displacement by a DNA polymerase, via the generation of a flap structure to be processed by a flap endonuclease (FEN) (Zheng and Shen 2011; Ma et al. 2022), or by an RNase H (Randall et al. 2019). Both mechanisms can coexist, as in animal mitochondria, where the flap endonuclease FEN1, RNase H1, endonuclease DNA2, and the ssDNA-specific nuclease MGME1 play roles in potentially redundant pathways for primer removal (Uhler and Falkenberg 2015). Exonucleases and RNase H can additionally be involved in the control of R-loops, which are 3-stranded structures generated during transcription, in which an RNA segment is hybridized to one of the DNA strands (Roy and Lieber 2009; Ginno et al. 2012). If not regulated, R-loops and unprocessed Okazaki primers can lead to replication fork stalling, genome rearrangements, and increased recombination (Gan et al. 2011; Wimberly et al. 2013; Aguilera and Aguilera 2025). Additionally, a FEN or 5′-3′ exonuclease should be involved in plant mitochondrial repair pathways, for strand resection following double-strand breaks and the processing of recombination intermediates. A FEN may also play a role in long-patch BER, to process the flap formed by strand displacement during DNA synthesis (Ranalli et al. 2002). To date, no FEN or functional analogs of MGME1 or DNA2 have been identified in plant organelles. A source of exonuclease and FEN activities could be the 5′-3′ nuclease domain of a bacterial-type DNA polymerase Pol I, a domain essential for bacterial Pol I repair functions. However, plant organellar POL1A and POL1B lack a 5′-3′ exonuclease domain.

Here, we describe OEX1, an Arabidopsis (Arabidopsis thaliana) mitochondrial nuclease conserved in the entire plant lineage and phylogenetically related to the 5′-3′ exonuclease domain of prokaryotic DNA Pol I. OEX1 is essential for mtDNA stability, and its loss leads to severe plant growth defects. In vitro, OEX1 degrades dsDNA processively from 5′ to 3′ and shows efficient FEN activity, capable of processing substrates generated by the displacement activity of DNA polymerase. We show that OEX1 can degrade RNA in Okazaki-like RNA:DNA hybrids, and that oex1-1 mutants accumulate R-loops within highly transcribed mtDNA regions. OEX1 exists as 2 isoforms resulting from alternative splicing, which have similar substrate specificities in vitro, but differ in processivity, affinity to the 2 organellar DNA polymerases, and differentially affect mtDNA segregation and repair. Our results show that OEX1 is an exonuclease/FEN that potentially plays multiple roles in mtDNA maintenance.

Results

Two Arabidopsis genes code for potential organellar exonucleases

The plant organellar DNA polymerases POL1A and POL1B lack the N-terminal 5′-3′ exonuclease domain that in bacterial Pol I plays important roles in DNA repair and the maturation of Okazaki fragments during replication (Lovett 2011). In bacteria, the loss of this activity results in conditional lethality, whereas mutations just affecting the DNA polymerase or the 3′-5′ exonuclease domains of DNA Pol I are viable (Lehman and Uyemura 1976). A BLAST search of the Arabidopsis proteome for potential counterparts to the 5′-3′ exonuclease domain of DNA Pol I revealed 2 candidate genes, At3g52050 and At1g34380, which we preliminarily named OEX1 and OEX2, respectively, for ORGANELLAR EXONUCLEASE 1 and 2 (Fig. 1A). They correspond to the genes previously identified through similar searches conducted by other researchers (Sato et al. 2003; Moriyama and Sato 2014). The encoded proteins contain a distinctive PIN (PilT N-terminus domain)-like nuclease domain and an H3TH (helix-3-turn-helix) DNA-binding domain, a structure also found in the 5′-3′ exonuclease domain of type A DNA polymerases, in eukaryotic flap endonuclease FEN1 and in RNase H of bacteriophage T4 (Matelska et al. 2017). Phylogenetic analysis revealed that OEX1 is present in all groups of the green lineage, including land plants and green algae, as previously observed (Moriyama and Sato 2014). In contrast, OEX2 is found exclusively in land plants (Fig. 1B). The similarity between OEX1 and OEX2 is lower than that between OEX1 and the orthologous bacterial sequences (Fig. 1C). Additionally, none of the intron positions is shared between the OEX1 and OEX2 genes. This suggests that OEX2 is either a paralog of OEX1 that has undergone rapid divergence, or was acquired independently of OEX1 later in evolution, for example, through horizontal gene transfer.

Figure 1.

OEX1 and OEX2 are plant organellar orthologs of bacterial exonucleases. A) Plant organellar DNA polymerases POL1A and POL1B lack the 5′-3′ exonuclease domain found in bacterial DNA Pol I and ExoIX. The 2 Arabidopsis genes OEX1 and OEX2 encode putative organellar proteins with similarity to this motif. OTS, organellar targeting sequence; PIN, PIN-like nuclease domain; H3TH, helix-3-turn-helix DNA-binding domain. B) Phenogram showing that OEX1 orthologs are present across the green lineage, including land plants and green algae, while OEX2 orthologs are restricted to land plants. The scale bar represents the number of pairwise amino acid differences per site. Bootstrap values are indicated in red. Branches supported by values below 70% were collapsed. C) Heat-map representation of the similarities between OEX1 and OEX2 from representative species as compared with bacterial proteins. The data suggest that the 2 plant genes did not diverge from a common ancestor but instead may have distinct evolutionary origins. The color scale bar represents the proportion of pairwise amino acid identities (vertically) or similarities (horizontally) per site.

OEX1 is targeted to mitochondria and OEX2 to chloroplasts

Both OEX1 and OEX2 display N-terminal extensions compared with their bacterial homologs. Predictions suggest that these extensions are mitochondrial and chloroplast targeting sequences, respectively (https://services.healthtech.dtu.dk/services/TargetP-2.0/; https://urgi.versailles.inra.fr/predotar/). To confirm the in silico predictions, genetic constructs encoding the fusion proteins OEX1-GFP and OEX2-GFP were introduced in wild-type (WT) Arabidopsis via Agrobacterium-mediated transformation. More than 3 T1 independent transformants expressing OEX1-GFP were observed, and all plants revealed a typical mitochondrial localization, with no evidence of dual localization in chloroplasts or the nucleus (Fig. 2A). Mitochondrial localization was confirmed with an OEX1-mCherry construct introduced into a mitochondrial marker line (mt-GFP), by the colocalization of the mCherry and GFP signals. Interestingly, the mCherry fluorescence was not evenly distributed within the mitochondrial matrix but accumulated in distinct foci. DAPI (4′,6-diamidino-2-phenylindole)-staining showed that these foci colocalize with DNA, suggesting that OEX1 is recruited to mitochondrial nucleoids (Fig. 2B). Attempts to express OEX2-GFP in stably transformed Arabidopsis plants were unsuccessful, presumably because of transgene silencing. However, transient expression in leaf epidermal cells following agroinfiltration showed strict chloroplast localization in all cells observed, from 3 agroinfiltrated plants, in speckles characteristic of nucleoids (Fig. 2C). This was confirmed by coexpression and colocalization with the RAP-RFP nucleoid marker (Kleinknecht et al. 2014). The nucleoid localization is consistent with proteomic data from maize (Zea mays), where the OEX2 ortholog was found significantly enriched in the plastid nucleoid proteome (Majeran et al. 2012).

Figure 2.

OEX1 and OEX2 are targeted to mitochondria and chloroplasts, respectively. A) Mitochondrial targeting of OEX1-GFP stably expressed in transgenic Arabidopsis plants. B) OEX1 localizes to mitochondrial nucleoids, as revealed by colocalization of OEX1-mCherry with DAPI (4′,6-diamidino-2-phenylindole)-stained nucleoids. mT-GFP, mitochondrially targeted GFP serving as compartment-specific marker. White arrowheads indicate signals of OEX1-mCherry colocalizing with DAPI-stained nucleoids. C) Targeting of OEX2-GFP to chloroplast nucleoids in Nicotiana benthamiana leaf epidermal cells, following agroinfiltration. The magnified image of a chloroplast highlights the colocalization of OEX2-GFP with the nucleoid marker.

OEX1 is required for normal plant development and fertility

In our subsequent work, we focused on characterizing the mitochondrial OEX1. The developmental and spatial expression patterns of OEX1 were studied by expressing the β-glucuronidase (GUS) gene under the control of the OEX1 promoter. GUS staining was observed in cotyledons and very young developing leaves, as well as in the central vascular cylinder of roots, in young pistils and stamens (Supplementary Fig. S1). No expression was detected in gametophytic tissues such as pollen and ovules. These results are mostly consistent with published RNA-seq data, which show that OEX1 is predominantly expressed in young leaves, petals, and pistils (Klepikova et al. 2016).

The consequences of OEX1 loss were analyzed in a T-DNA insertion mutant. Searching publicly available collections, several candidate mutants were initially identified, of which we ultimately could validate a single line (GABI_911E05). This mutant line, containing an insertion within the gene, was designated oex1-1. The T-DNA insertion in oex1-1 was mapped 10 nucleotides downstream of the ninth exon of OEX1. RT-PCR showed that the last 6 exons are no longer transcribed, a result that was confirmed by RNA-seq analysis (Supplementary Fig. S2). The mutation disrupts the protein-coding sequence at the level of the PIN-like nuclease active site, and results in the loss of 152 out of 425 amino acids, including the entire H3TH DNA-binding domain, indicating that oex1-1 is a true knockout line. Homozygous oex1-1 plants are viable, but severely impaired in development (Fig. 3A). Compared with WT plants, they display stunted growth, distorted leaves and flowers, and have very low fertility. Although oex1-1 plants can produce inflorescences, flowers are abnormal, with curved pistils, and pollen production is visibly reduced. The small flower stalks carry small siliques that mainly contain aborted embryos (Fig. 3A), and of the few mature seeds obtained many failed to germinate. The observed phenotype is fully penetrant, affecting all homozygous oex1-1 plants, although with some variation in the severity among individual T2 plants (Fig. 3A). In subsequent generations, the phenotypes worsened, and third-generation homozygous mutants were very strongly affected in growth and unable to flower. Reciprocal backcross attempts with WT plants were unsuccessful, suggesting that both male and female oex1-1 gametes are affected. Similar phenotypes have been described for mutants of several genes involved in mtDNA maintenance. In particular, the oex1-1 phenotypes are reminiscent of those of radA plants, deficient in the RADA branch-migration helicase (Chevigny et al. 2022).

Figure 3.

Loss of OEX1 has severe effects on plant growth and development. A) Representative images of 4-wk-old WT and oex1-1 plants, showing growth retardation, distorted leaves with chlorotic sectors, flower deformities, and reduced fertility of the mutant. The severity of the phenotypes can vary significantly among individual plants within the first homozygous mutant generation (T2 plants), as highlighted in the box plots showing the high variation in rosette size in oex1-1. Center line, median; box limits, upper and lower quartiles as determined by R software; whiskers, 1.5× interquartile range; outliers are represented by dots. Flowers harbor only few pollen grains that adhere to the papillae of the stigma. Siliques are small and mainly contain aborted embryos. B) The developmental phenotypes of the oex1-1 mutant are fully complemented in oex1-1 plants expressing OEX1 under its native promoter (oex1-1 OEX1 plant) and in mutants expressing OEX1 targeted exclusively to mitochondria, as achieved by fusion to the targeting sequence of AOX1 (oex1-1 AOX1:ΔN-OEX1 plant). However, chloroplast-specific OEX1 expression by fusion to the targeting sequence of RBCS does not rescue the phenotype (oex1 RBCS:ΔN-OEX1 plant).

To explore possible changes in mitochondrial morphology, oex1-1 rosette leaf sections were examined by transmission electron microscopy (TEM). The TEM images showed that mitochondria in oex1-1 mesophyll cells were smaller and more electron-dense than in WT plants of similar size. In contrast, the morphology of chloroplasts was unaffected (Supplementary Fig. S3).

The deleterious mutant phenotypes were fully complemented by expression of the OEX1 cDNA under the native OEX1 promoter. Complemented plants (oex1 OEX1) were phenotypically normal and indistinguishable from WT plants (Fig. 3B).

To confirm that the deleterious phenotype was caused by lack of OEX1 activity in mitochondria only, and not by the possible presence of undetectably low protein levels in chloroplasts, hemicomplementation experiments were also conducted. To this end, oex1-1 plants were transformed with constructs designed to specifically target OEX1 to either mitochondria or chloroplasts (see Materials and methods). As shown in Fig. 3B, full complementation of the mutant phenotype was achieved when OEX1 was specifically targeted to mitochondria, whereas no complementation was observed when OEX1 was solely targeted to chloroplasts. Together, these results support the hypothesis that the mutant phenotype of oex1-1 plants is primarily due to defects in mitochondria.

OEX1 plays a critical role in maintaining mitochondrial genome stability

Mutants deficient in factors involved in mtDNA maintenance and repair exhibit genome instability, mostly due to the accumulation of crossover products resulting from recombination mediated by small repeated sequences (100 to 500 bp; Gualberto and Newton 2017). To test whether lack of OEX1 results in crossover product accumulation, we performed total DNA sequencing (Illumina) on oex1-1 vs. Col-0 plants. For this analysis, oex1-1 seedlings from the first homozygous mutant generation, segregating from a heterozygous line, were used. The relative abundance (stoichiometry) of the different genomic regions of the mtDNA and cpDNA were evaluated, according to the coverage profiles, to highlight enriched or depleted regions.

The profile of the plastid genome sequence coverage from oex1-1 was indistinguishable from that of the WT, thus providing no evidence of cpDNA instability (Fig. 4B). In stark contrast, the mtDNA profile of oex1-1 plants was significantly perturbed, showing positive and negative stoichiometric changes, in sharply demarcated, discrete regions of the genome (Fig. 4A). Several of those discrete genomic regions in the mtDNA are flanked by directly oriented repeats (e.g. repeats A [556 bp], L [249 bp], and EE [127 bp], according to the nomenclature of repeats annotated in the mtDNA genome of Col-0, accession BK010421). This suggests that the mtDNA stretches corresponding to these regions may represent subgenomic molecules generated by a “loop-out” mechanism during recombination, as schematically represented in Fig. 4A. Such mechanism of mitogenome instability was already observed in other mtDNA recombination mutants, such as those of the DNA helicases RECG1 and RADA (Wallet et al. 2015; Chevigny et al. 2022). To confirm this hypothesis, the accumulation of crossover products resulting from recombination involving repeats L, F, and EE was tested by qPCR, as described before (Miller-Messmer et al. 2012) and according to the cartoon at the top of Fig. 4C. These repeats were selected because they are consistently mobilized in other mtDNA repair mutants and are small enough for qPCR analysis. As expected, a significant increase in crossover products was observed compared with WT levels (Fig. 4C). An asymmetric accumulation of primarily 1 type of crossover was seen, consistent with previous findings in other mutants with mtDNA recombination defects (Gualberto and Newton 2017). This asymmetry is likely due to the involvement of the error-prone BIR repair pathway (Christensen 2013; Gualberto and Newton 2017; Kockler et al. 2021).

Figure 4.

Structural alterations of the mtDNA in oex1-1 mutants. A) Sequence coverage of the mtDNA in oex1-1 vs. WT, revealing regions with altered stoichiometry flanked by direct repeats. The coordinates are those of the Col-0 mtDNA sequence. The positions of the mtDNA repeats LR1 and LR2 are shown on top of the plotted area. Regions with changed stoichiometry are shadowed and annotated according to the repeat name (repeats A, F, L, X, and EE, of 556, 350, 249, 204, and 127 bp, respectively). The “looping-out” of the corresponding subgenomes by recombination is schematically represented. B) Plastid genome analysis as performed in (A), showing no effects of OEX1 loss on cpDNA stability. LSC, large single copy region; SSC, small single copy region; IR, inverted repeat (only 1 copy is included). C) qPCR confirmation of the accumulation in oex1-1 plants of recombination products across repeats L, F, and EE. The parental sequences 1/1 and 2/2, and the corresponding crossover products 1/2 and 2/1 were quantified by qPCR, as schematically represented on top of the figure for the sequences flanking a repeated sequence “R.” Error bars correspond to Sd values from 3 biological replicates. D and E) Analysis of the relative copy numbers of the different mtDNA regions in individual severely affected T4 generation oex1-1 plants. Sequences spaced ∼5 kb apart in the mtDNA were quantified by qPCR. Values are the average and Sd of 3 technical replicates. Regions with altered stoichiometry flanked by direct repeats are shadowed as in (A). In Plants #5 and #6, regions that are highly increased in copy number correspond to subgenomes resulting from recombination involving a microhomology of 14 bp and an imperfect repeat of 85 bp, respectively. The parental and recombined sequences are shown.

To assess the extent of mtDNA changes in individual oex1-1 plants, the relative copy number of different mtDNA regions was measured by qPCR in 6 oex1-1 T3-generation plants that displayed severe growth and developmental impairments (Fig. 4D). All plants showed marked changes in the stoichiometry of different mtDNA regions, though the extent of these changes varied greatly among plants (Fig. 4E). Most changes occurred in the regions flanked by repeats A, L, F, and EE (Plants #1 to #4 in Fig. 4E), as it had been seen in T2-generation plants. However, some plants (Plants #5 and #6) showed a particularly large increase in copy number of sequences not associated with any annotated repeat. Further investigation revealed that these regions corresponded to subgenomic molecules formed through illegitimate recombination (i.e. involving sequences of little or imperfect homology), with 1 involving an 85 bp imperfect repeat and the other a 14 bp microhomology (Fig. 4E). These results suggest that, in the absence of OEX1, illegitimate recombination activities are activated, which can recruit small repeats and microhomologies. Consequently, the mitochondrial genome stoichiometry is altered, as the newly formed subgenomes seem to behave as episomes and are preferentially amplified (through unknown mechanisms). Thus, the severe growth phenotype of oex1-1 plants correlates with defects in mtDNA maintenance, resulting in the accumulation of ectopic recombination products and significant alterations in the stoichiometry of the different mtDNA regions. However, the mtDNA regions with lower copy numbers in oex1-1 plants do not contain any identified gene(s) that could explain the profound developmental defects observed in these plants.

Two OEX1 isoforms are generated by alternative splicing

RT-PCR amplification and sequencing of OEX1 transcripts from Arabidopsis leaves revealed 2 mRNA variants caused by alternative splicing of the sixth exon. This splicing was found in 6 out of 11 cDNA clones sequenced, and is supported by publicly available EST sequences (https://www.arabidopsis.org). The 2 mRNA variants encode proteins that differ by 18 amino acids, which we named OEX1a and OEX1b (Fig. 5A). The sixth exon-encoded extension retained in the OEX1a isoform is conserved across all flowering plants at the DNA level, though its sequence conservation is low (Supplementary Fig. S4A). Corresponding transcript variants have been identified, among others, in soybean (Glycine max) (XM_006574959.4), grapevine (Vitis vinifera) (XM_010652395.3), maize (Z. mays) (XM_008660501.4), rice (Oryza sativa) (XM_015776708.3), and Amborella (Amborella trichopoda) (XM_020676298.1). However, this extension is absent from OEX2 and from the 5′-3′ exonuclease domain of bacterial DNA Pol I (Supplementary Fig. S4A). Structural modeling suggests that the extension is located near the active site and the H3TH DNA-binding domain (Supplementary Fig. S4B), potentially influencing OEX1 activity and/or substrate specificity. The relative abundance of the OEX1a and OEX1b transcripts in different plant tissues was assessed by RT-qPCR, using variant-specific primers (Supplementary Fig. S4C). Both mRNA forms were detected at similar levels in most tissues analyzed, except for flower buds, where the ratio of OEX1b to OEX1a was higher, suggesting that OEX1 may be developmentally regulated through alternative splicing (Supplementary Fig. S4C).

Figure 5.

Alternative splicing results in the expression of 2 OEX1 isoforms. A) Schematic representation of the 2 proteins resulting from alternative splicing of exon 6. B) RT-qPCR relative quantification of OEX1 transcripts in oex1-1 mutant plants complemented either with OEX1a or OEX1b coding sequences (cOEX1a and cOEX1b plants, respectively), under the control of the endogenous OEX1 promoter. RNA was extracted from pools of 10-d-old seedlings from 3 independent lines. Primers amplified sequences spanning exon1-exon2, exon10-exon12, and exon12-exon14 (e1-e2, e10-e112, and e12-e14, respectively). Values are the average and Sd of the 3 biological replicates. C) Both cOEX1a and cOEX1b constructs complemented the severe developmental phenotypes of oex1-1. WT, wild type. The scale bar is 1 cm. D) Analysis of the mtDNA of cOEX1a and cOEX1b plants by qPCR, as described in Fig. 4E, showing full complementation by cOEX1a, but retention of a slightly remodeled mtDNA in cOEX1b plants, with altered copy number of the subgenomes defined by large repeat LR2. Results represent the average and Sd of 3 biological replicates. E and F) Phenotypes of plants grown under genotoxic stress induced by 0.5 µm CIP. The scale bar is 5 mm. Plants complemented with cOEX1b were more sensitive, as assessed by the percentage of plants still able to develop normal first true leaves, at a CIP concentration that had only a mild effect on the WT. Asterisks indicate statistically significant differences according to the χ² test (*P < 0.05, ***P < 0.001). F) Effect of CIP treatment on the stability of the mtDNA in WT and cOEX1a and cOEX1b plants. Analysis by qPCR, as in (C). G) Relative accumulation of crossover products resulting from recombination involving pair of repeats L, as described in Fig. 4C, in CIP-treated plants when compared with the nontreated control. H) The crossover product L-1/2 accumulates in CIP-treated WT and complemented plants cOEX1a, while it is the reciprocal product L-2/1 that accumulates in cOEX1b. Values are the average and Sd of the 3 biological replicates.

The impact of the 2 OEX1 isoforms on mtDNA maintenance was further examined by complementing the oex1-1 mutant with constructs harboring either OEX1a or OEX1b cDNAs, under the control of the OEX1 promoter. Heterozygous oex1-1 plants were transformed with the constructs, self-pollinated, and homozygous oex1-1 plants containing the transgene were analyzed, each biological replicate segregated from an independent T1 transformant. The expression level of the complementation constructs was evaluated by RT-qPCR, targeting exons that are not expressed in the mutant line (as seen by RNA-seq, Supplementary Fig. S2). It was found that the complemented lines had higher expression levels of both OEX1a or OEX1b than WT plants, despite using the endogenous OEX1 promoter (Fig. 5B). This increased expression may be due to additional regulatory elements present in the expression vector. Nevertheless, no reduced expression levels were observed, and there were no significant differences in expression between the cOEX1a and cOEX1b complemented lines. Both isoforms successfully complemented the development defects of oex1-1, and the resulting plants (cOEX1a and cOEX1b) were indistinguishable from WT (Fig. 5C). However, at the molecular level, while stoichiometric mtDNA replication was fully restored to WT levels in cOEX1a plants, cOEX1b plants showed a slight imbalance in the relative abundance of 2 mtDNA subgenomes flanked by large repeat 2 (LR2, 4.2 kb in size; Fig. 5D). The complemented cOEX1a and cOEX1b plants were further challenged for repair capacity by growth under genotoxic stress induced by ciprofloxacin (CIP), as described previously (Schatz-Daas et al. 2022). At the diagnostic concentration of 0.5 µm CIP, most WT plants were still able to develop normal first true leaves, whereas cOEX1b plants and, to a lesser extent, cOEX1a plants either failed to develop their first true leaves or these leaves were mostly wrinkled and distorted (Fig. 5E and F). qPCR analysis of mtDNA copy number across the genome revealed that CIP treatment significantly affects WT plants, causing significant changes in the relative stoichiometry of different mtDNA regions. A similar effect was observed in CIP-treated cOEX1a-complemented plants. However, more pronounced changes were seen in cOEX1b plants (Fig. 5G). We further analyzed the accumulation of the crossover products from recombination involving repeat L, which previous studies showed to accumulate differently in CIP-treated WT and repair-deficient mutants (Schatz-Daas et al. 2022). As described, in CIP-treated WT plants, sequence L-1/2 accumulates, likely due to HR-mediated repair of DNA breaks. A similar effect was observed in cOEX1a plants, but it was much reduced in cOEX1b plants, where the crossover sequence L-2/1 accumulated instead (Fig. 5H). Thus, defects in the replication, repair, and segregation of mtDNA subgenomes were not fully restored in either of the complemented plants, particularly not in cOEX1b. This suggests that OEX1a and OEX1b are 2 isoforms that are only partially redundant with respect to their roles in mtDNA maintenance.

Loss of OEX1 has no apparent detrimental effect on mitochondrial transcripts

To assess whether mtDNA instability in oex1-1 affects mitochondrial gene expression, mitochondrial transcript abundances in oex1-1 and WT Col-0 were analyzed by RT-qPCR, as described previously (Baudry et al. 2022). For genes with introns, the integrity of the exon–exon junctions in the mRNAs was also examined. oex1-1 exhibited an increase in the abundance of most transcripts, with only a few transcripts showing a modest decrease. No splicing defects were detected (Fig. 6). Because the reshuffling of mtDNA sequences in oex1-1 could potentially impact transcript processing, the RNA-seq coverage profiles from WT Col-0 and oex1-1 were comparatively inspected. This analysis revealed no apparent changes in the processing of mitochondrial protein-coding transcripts (Supplementary Fig. S5). This finding suggests that the severe growth defects in oex1-1 plants are not caused by deficient expression of a specific mitochondrial gene. Interestingly, several gene transcripts, including atp1, rps4, rps7, ccmC, matR, cox3, ccmFN1, and ccmFN2, were more than 2-fold higher in oex1-1, according to RT-qPCR. In certain cases, these changes in relative transcripts abundance can be attributable to changes in gene copy number because of differential accumulation of subgenomes generated by recombination. For example, atp1 is encoded by the EE episome that is amplified at the DNA level in the mutant, and ccmC, ccmFN1, ccmFN2, and cox3 are clustered in the subgenome defined by repeat L. In plants complemented with either OEX1a or OEX1b, mitochondrial transcript abundances were restored to WT levels (Fig. 6).

Figure 6.

Accumulation of mitochondrial transcripts in oex1-1 mutant plants. RT-qPCR analysis of steady-state levels of mitochondrial protein-coding transcripts and rRNAs in oex1-1 plants and the complemented lines cOEX1a and cOEX1b. Results were normalized to a set of nuclear housekeeping genes. The exon–exon borders were quantified for genes containing introns, to assay for possible splicing defects. Results are displayed at log₂ scale and represent the average and Sd of 3 biological replicates.

OEX1 is a 5′-3′ exonuclease/FEN

To investigate the enzymatic activities of the OEX1 nuclease, recombinant OEX1a and OEX1b proteins were expressed in Escherichia coli (Supplementary Fig. S6) and purified. According to the known structure of PIN-like nucleases, it was predicted that residue D250 is part of the active site. Therefore, a putative noncatalytic mutant protein version (D250A) was also designed and purified (Supplementary Fig. S6B). Preliminary experiments indicated that recombinant OEX1a was capable of degrading 5′-labeled linear dsDNA, which was therefore used as a substrate to optimize reaction conditions for maximal OEX1a activity (Supplementary Fig. S7). OEX1a efficiently degraded a 25-mer dsDNA substrate, showing strict 5′-3′ exonuclease activity, without evidence of endonuclease activity, as no intermediate products were detected with 5′-labeled substrates (Fig. 7A). To confirm that this activity is attributable to OEX1a and does not result from bacterial proteins contaminating the purified protein, the D250A noncatalytic mutant version was purified and tested under identical conditions, and showed no activity. Under the same conditions, OEX1b also showed 5′-3′ exonuclease activity, but with less processivity than OEX1a, and a significant portion of the substrate remained intact after 30 min incubation (Fig. 7B). The degradation pattern of a 3′-labeled dsDNA substrate ([GTCA]₅GTCCC) suggested sequence-dependent distributive activity, with pausing at G:C base pairs (Fig. 7A and B). This was confirmed by testing substrates with 1, 2, or 3 consecutive Gs, which showed intermediate degradation products whose ends mapped to the G positions, with the strongest signal for the G₃ substrate (Supplementary Fig. S8A).

Figure 7.

Exonuclease and FEN activities of OEX1. A) Analysis of the exonuclease activity of the WT OEX1a protein and the noncatalytic mutant D250A on a dsDNA substrate, schematically represented above the gel images and labeled either at the 5′ or the 3′ end (asterisks). Sizes in nucleotides are indicated. Substrates and degradation products were analyzed by electrophoretic separation in denaturing polyacrylamide gels. Size of DNA markers are indicated in nucleotides. B) Analysis as in (A), but for isoform OEX1b. C) Analysis as in (A), but with a substrate mimicking flap structures. D) Analysis as in (C), but for OEX1b. E) Analysis as in (A), but for a substrate with a 10 nt gap, in the absence or presence of a DNA polymerase (Klenow enzyme). F) Summary of OEX1 activities on the various DNA substrates tested.

To test for FEN activity, a 5′ flap-like substrate was prepared and labeled either at the 5′ or 3′ end of the flap strand. The flap strand was specifically designed without sequence homology to the complementary DNA, to prevent the formation of alternative structures. Precise processing of the flap was expected to produce either an 18 or a 25 nucleotide-labeled product, depending on the labeling site. Both OEX1a and OEX1b, but not the noncatalytic mutant, could precisely process the flap (Fig. 7C and D). Similar to the 5′-3′ exonuclease activity, OEX1a showed higher processivity on the flap substrate than OEX1b.

Following flap cleavage by OEX1a, the resulting 5′-labeled 18-mer was further degraded, implying that OEX1 can also degrade ssDNA. This prompted us to test OEX1 activity on ssDNA. Both OEX1a and OEX1b showed 5′-3′ exonuclease activities, as evidenced by using a 3′-labeled substrate (Supplementary Fig. S8B). However, these activities were much weaker than those observed on dsDNA, and most of the ssDNA substrate remained intact after 30 min. As for dsDNA, OEX1b showed much weaker activity than OEX1a. Testing a 5′-labeled ssDNA substrate further revealed that OEX1a also has weak 3′-5′ activity, removing 1 to 2 nucleotides before the label was removed by the 5′-3′ activity (Supplementary Fig. S8B). This low activity on ssDNA in vitro may not be biologically relevant, or alternatively, could play a role in the surveillance of unscheduled replication triggered by small ssDNA fragments.

Next, we tested whether OEX1 could play a role in BER, namely whether it can recognize single-strand gaps on dsDNA (mimicking BER intermediates) and extend them by removing a few nucleotides. On dsDNA substrates with gaps of 1, 3, or 10 nucleotides, OEX1a showed minimal exonuclease activity. Although it could extend the gaps regardless of their size (Supplementary Fig. S8C), this occurred with much lower processivity than the activity at the 5′ end of dsDNA, suggesting that OEX1 alone is inefficient in extending single-strand gaps. To test if gap extension is improved by strand displacement by a DNA polymerase, the experiment was repeated with the Klenow enzyme added after OEX1a. Under these conditions, gaps were processed significantly faster, indicating that OEX1a activity on gapped dsDNA is enhanced by strand displacement (Fig. 7E). No significant activity was observed with Klenow polymerase and the D250A noncatalytic mutant, confirming that the results cannot be explained by a contaminant or the 3′-5′ exonuclease activity of the Klenow enzyme. In summary, our in vitro assays show that OEX1 has 5′-3′ exonuclease activity on dsDNA, FEN activity, and gap-extension capability in cooperation with the strand-displacement activity of a DNA polymerase (Fig. 7F).

OEX1a readily degrades the RNA moiety in RNA:DNA hybrids and is involved in the processing of R-loops

Given the potential role of OEX1 in the processing of Okazaki fragments, its activity on RNA:DNA hybrids was tested. Of a 3′-labeled RNA paired to a complementary DNA primer (RNA:DNA substrate), OEX1a rapidly degraded the RNA strand (Fig. 8A), with apparently much higher processivity than on dsDNA substrates (as seen in the 3′-labeled product in Fig. 7A). On a substrate mimicking an Okazaki-like RNA primer on the lagging DNA strand, consisting of a dsDNA region followed by a small gap and an RNA:DNA duplex, OEX1a also rapidly degraded the RNA oligonucleotide (Fig. 8B). This contrasts with the weak activity of OEX1a on a similar substrate made of DNA duplexes only, where strand displacement by a DNA polymerase was needed to accelerate the reaction (Fig. 7E). Finally, when OEX1a was tested on single-stranded RNA, no significant nuclease activity was detected (Fig. 8C). In summary, the in vitro assays show that RNA:DNA hybrids are apparently the preferred substrates of OEX1. In vivo, such substrates may be formed during replication (e.g. Okazaki fragments) or might be by-products of transcriptional activities (e.g. R-loops). Both are known to compromise genome stability if unresolved (Crossley et al. 2019). We, therefore, explored the potential involvement of OEX1 in resolving R-loops. First, we tested the in vitro activity of OEX1a on a substrate mimicking an R-loop, with the 5′ dsDNA ends biotinylated to protect them from OEX1 activity. Upon incubation with OEX1a, the RNA component of this R-loop substrate was promptly degraded (Fig. 8D). Next, we assessed whether oex1-1 plants accumulate R-loops, by DNA:RNA immunoprecipitation (DRIP). To this end, nucleic acids from WT and oex1-1 plants were immunoprecipitated using the S9.6 antibody specific for RNA:DNA hybrids, and the precipitated DNA was analyzed by qPCR. As also reported by others (Sanz and Chedin 2019), the signal-to-background ratio in the DRIP experiments was low. Nevertheless, a clear and significant increase was observed in the oex1-1 samples compared with the WT (Fig. 8E). This overaccumulation of R-loops could be linked to the mtDNA instability observed in oex1-1 mutant plants.

Figure 8.

OEX1 exonuclease activity on RNA:DNA hybrids, and in vivo contribution to the clearance of R-loops in the mtDNA. A) Analysis of the exonuclease activity of WT OEX1a protein and the noncatalytic mutant D250A on a RNA:DNA hybrid (DNA in black, RNA in orange) labeled at the 3′ end of the RNA strand (red asterisk). Substrates are schematically represented above the gel images with sizes in nucleotides indicated in blue. Substrates and degradation products were analyzed by electrophoresis in denaturing polyacrylamide gels. B) Assay as in (A), but with a substrate representing an Okazaki-type RNA fragment. C) Assay as in (A), showing that OEX1 has no activity on a single-stranded RNA substrate. D) Assay as in (A), showing that OEX1 has high activity on the RNA moiety of an R-loop. E) DRIP-qPCR analysis of R-loops in oex1 plants compared with the WT. Results are the average of 3 biological replicates. The data show an increase of R-loops in the oex1 mutant, in the mtDNA region corresponding to the highly transcribed 18S and 26S rRNAs. Results were normalized against nontranscribed regions of the mtDNA. Asterisks represent statistically significant changes by Student's t-test (*P < 0.05, **P < 0.01, ***P < 0.001).

Discussion

 

Origin of the OEX1 and OEX2 genes and their relationship to bacterial counterparts

Due to its symbiotic origin, mtDNA replication and stability involve a combination of factors, many of which are derived from the prokaryotic ancestors of organelles. In eubacteria, a 5′-3′ exonuclease activity is essential for cell viability. It is provided by either Pol I or ExoIX orthologs (e.g. YpcP in Bacillus subtilis) (Fukushima et al. 2007). These enzymes have a structure-specific endonuclease activity that cleaves at the junction between a 5′ single strand and a duplex. The DNA Pol I activity at a nick displaces the DNA upstream of the enzyme, forming a 5′ flap that is subsequently cleaved by the 5′ endo/exonuclease (Anstey-Gilbert et al. 2013; Randall et al. 2019). In many bacterial species, the 2 enzymes coexist, but their respective contributions to DNA replication and genome stability remain unclear. A recent study found that they have different substrate affinities, hinting at specialized roles during replication (Lowder and Simmons 2023). Still, either enzyme is sufficient for cell survival, whereas double mutation leads to synthetic lethality (Fukushima et al. 2007). Thus, in bacteria lacking ExoIX activity, the 5′-3′ exonuclease activity of Pol I is indispensable (Diaz et al. 1992). Conversely, some bacterial genera have only an ExoIX ortholog and lack a Pol I-like repair polymerase. They correspond to Class 2 in Fukushima's classification, and they are all symbiotic or infectious bacteria, including the probable symbiotic ancestor of mitochondria. It is thus tempting to speculate that plant mitochondria fall into the Class 2 systems and that OEX1/2 originated from a bacterial ExoIX. However, the phylogenetic analysis (Fig. 1C) suggests that OEX1/2 are closer to the N-terminal domain of Pol I than to ExoIX, implying that plant OEX1/2 evolved from a Pol I that lost its polymerase domain. The phylogenetic analysis also indicates that OEX2 is not a paralog of OEX1, as their similarity to each other is lower than that to bacterial orthologs, suggesting distinct acquisition events.

Our subcellular localization analyses strongly suggest that OEX1 is exclusively mitochondrial and specifically localized in nucleoids, consistent with a role in mtDNA metabolism. In agreement with the confocal microscopy results, hemicomplementation of oex1-1 plants with a construct targeting OEX1 exclusively to mitochondria restored the WT phenotype, confirming that OEX1 is active only in mitochondria. Similarly, the exclusive chloroplast localization of OEX2-GFP is in line with the identification of the maize OEX2 ortholog in the plastid nucleoid proteome (Majeran et al. 2012). Taken together, these results suggest that OEX1 and OEX2 fulfill equivalent nuclease functions in mitochondria and plastids.

Major potential role of OEX1 in mtDNA replication

To date, details of how mtDNA replicates have been elucidated in only a few model systems. In animals, replication of the small and compact mtDNA is highly specialized, relying upon DNA polymerase γ. Like plant organellar DNA polymerases, it lacks a 5′-3′ exonuclease domain. It requires mitochondrial RNase H1 to generate RNA primers by cleaving longer transcripts within R-loops stabilized by ssDNA-binding protein mtSSB (Falkenberg et al. 2024). After replication, the RNA primers are removed by RNase H1, which cleaves at numerous sites between the ribonucleotides, leaving behind a few ribonucleotides attached to the 5′-end of the DNA, incompatible with ligation. A second nuclease is therefore required, analogous to the nuclear RNase H2/FEN1 pathway (Uhler and Falkenberg 2015). This could be nuclease MGME1, which cuts the ssDNA flaps formed by the displacement activity of mtDNA polymerase γ (Uhler and Falkenberg 2015). Other nucleases that are dually targeted to both the nucleus and mitochondria such as FEN1 and DNA2 may also be involved.

In plants, however, there is no MGME1 ortholog, and mtDNA replication and maintenance rely on many prokaryotic-like factors. The processing of RNA primers bound to the mtDNA during replication likely mirrors the bacterial system, with OEX1 emerging as the primary candidate enzyme for this function. The marked FEN and gap-extension activities of recombinant OEX1, facilitated by DNA polymerase, support the hypotheses that OEX1 has a central role in cleaving the small flaps formed by strand displacement when DNA polymerase reaches a downstream 5′-end. Consistent with this predominant role in replication, the recombinant protein efficiently recognized and processed the RNA strand of an RNA:DNA hybrid in vitro, as well as structures mimicking Okazaki fragments in a replication fork (Fig. 8). In these experiments, OEX1 was more processive on RNA than on DNA substrates, and was processive on gap-containing double-stranded DNA only in the presence of DNA polymerase activity. These results suggest that the elimination of RNA sequences hybridized to DNA is the main role of OEX1. Therefore, maturation of plant mitochondrial Okazaki fragments by OEX1 could occur either via exonucleolytic degradation and/or via FEN activity on the flaps previously formed by strand displacement by DNA polymerase. Interestingly, bacterial YpcP also showed its highest activity on RNA:DNA duplexes, recognizing Okazaki fragments and flap substrates (Randall et al. 2019). Organellar RNase H1 enzymes are believed to contribute to the removal of replication primers, explaining the vital importance of mammalian mitochondrial RNase H1 (Cheng et al. 2021; Misic et al. 2022) and the developmental and molecular phenotypes observed in Arabidopsis mutants deficient in plastidial AtRNH1C (Yang et al. 2017). However, the loss of mitochondrial RNase H1 (AtRNH1B) alone does not show adverse effects on plant development. It is therefore possible that plant mitochondria may have multiple pathways for primer removal. One could rely solely on OEX1, whose 5′-3′ exonuclease activity should be sufficient to functionally replace the need for an RNase H1. Another similar to that of animal mitochondria, involving mitochondrial RNase H1, but still requiring OEX1 to remove the remaining ribonucleotides left by RNase H1 at the 5′-end of the DNA.

Other potential roles of OEX1 in the mtDNA maintenance

HR is the primary mechanism for repair and maintenance of plant mitogenomes, and requires several nucleases. A 5′-3′ exonuclease is typically required for strand resection at double-strand breaks. In bacteria, this function is handled by the RecBCD complex or by RecJ in the RecFOR pathway, both of which are highly processive (Kowalczykowski 2015). OEX1 has 5′-3′ exonuclease activity on blunt-ended dsDNA, but the in vitro tests showed that its activity on these substrates is distributive rather than processive, suggesting that it is unlikely to be directly involved in strand resection. Nonetheless, it seems possible that OEX1 functions in vivo in conjunction with other factors such as helicases and ssDNA-binding proteins that enhance its processivity. Additionally, a FEN activity is required in HR pathways, particularly in single-strand annealing and the resolution of D-loops. Therefore, OEX1 is a primary candidate for resolving such recombination intermediates.

The FEN activity of OEX1 is likely also essential for BER, which is a principal mechanism for repairing damaged mtDNA in both plants and animals (Szczesny et al. 2008; Boesch et al. 2011; Ferrando et al. 2018). Okazaki fragment processing and long-patch BER share similar mechanisms, as both involve a DNA polymerase displacing the existing strand to form a flap, which is then cleaved by a FEN. Similar to BER, the FEN activity of OEX1 might also play a role in ribonucleotide excision repair (RER), a process that removes ribonucleotides misincorporated into DNA during DNA replication. RER requires an RNase H to cleave the strand containing the ribonucleotide, and the strand-displacement activity of a DNA polymerase to create a flap, which is then processed by a FEN, such as OEX1.

Finally, OEX1 might also play a role in the removal of R-loops, which can form in transcribed regions and are known to cause genomic instability (Gan et al. 2011; Wimberly et al. 2013). This conclusion is supported by both the strong in vitro activity of OEX1 in degrading the RNA strand of RNA:DNA hybrids, and by the DRIP-qPCR results that revealed accumulation of R-loops in the mitogenome of oex1-1, particularly in the highly transcribed regions encoding the rRNAs. The processing of R-loops may involve alternative pathways (Supplementary Fig. S9), similar to those discussed above for primal removal. Factors that might cooperate with OEX1 in the process include helicase RECG1, whose bacterial ortholog RecG participates in the dissociation of R-loops (Vincent et al. 1996), and RNase H1 (Al-Behadili et al. 2018).

OEX1 is critical for plant development and its absence causes mtDNA instability

The phenotype of oex1-1 plants confirmed the importance of OEX1 in plant growth, development, and fertility. By the third generation of homozygous mutants, plants could no longer produce offspring. This quasi-sterility can be attributed to severely deformed pistils and stigmas, along with reduced pollen viability, as evidenced by failed attempts to pollinate WT pistils with oex1-1 pollen. Although the overall phenotypic defects were qualitatively consistent among individual plants, their severity varied and correlated with the degree of mitochondrial genome instability. Typically, this instability resulted in changes in the relative abundances of mitochondrial genomic regions, as arising from preferential amplification of subgenomes created by ectopic recombination on small repeats (in the range of 100 to 500 bp). However, as seen in some oex1-1 plants, much smaller and imperfect repeats as well as microhomologies could also drive ectopic recombination in the mutant. These subgenomic mtDNA molecules may result from the accumulation of unprocessed DNA ends in the oex1-1 mutant, thus triggering strand invasion and replication re-initiation via BIR or MMBIR. As previously suggested, processes that generate mtDNA breaks potentially promote error-prone repair pathways, and likely explain the mtDNA instability and increased heteroplasmy observed in recombination surveillance mutants (Marechal and Brisson 2010; Gualberto and Newton 2017).

Whether the severity of the oex1-1 phenotypes is directly linked to the stoichiometric changes of a specific mitochondrial locus, is currently unknown. However, although changes in relative copy numbers of mtDNA regions can influence the mitochondrial transcriptome (Wallet et al. 2015), the regions with reduced copy numbers in oex1-1 plants do not contain any known genes. Furthermore, our transcriptomic data revealed no potentially detrimental changes in mitochondrial gene transcript expression or processing. On the contrary, several gene transcripts were more abundant, often correlating with the accumulation of subgenomes containing those genes. While we cannot exclude the possibility that toxic chimeric genes generated by cryptic recombination events are responsible for mitochondrial malfunction and the developmental defects in oex1-1 plants, the RNA-seq coverage profiles of mtDNA coding sequences failed to reveal any such putative transcriptome changes (Supplementary Fig. S5).

Thus, the harmful effects of the oex1 mutation on plant development is likely due to problems in mtDNA replication and/or the aberrant segregation of recombination-generated subgenomes. These defects seem to affect mitochondrial biogenesis, though not necessarily through obvious problems with gene expression. As revealed by expression of a promoter:GUS fusion, OEX1 is primarily expressed in young, rapidly dividing tissues, consistent with the observed defects in mitochondrial biogenesis and proliferation revealed by TEM imaging. Similar effects on mitochondrial size have also been observed in mutants deficient in the mitochondrial ssDNA-binding protein OSB1 (Zaegel et al. 2006). This suggests a possible link between mtDNA replication and/or segregation and mitochondrial division, and is reminiscent of the situation in mammalian cells, where nucleoids involved in active mtDNA replication localize to contact points with the endoplasmic reticulum, which form the sites of mitochondrial division (Lewis et al. 2016). However, such a functional connection remains to be demonstrated in plants. An additional component potentially contributing to the developmental defects in mutants with impaired mtDNA maintenance are retrograde responses that activate (or inactivate) the expression of nuclear genes and, in this way, suppress cell proliferation. Such responses have been observed in animal cells, and in plants, in chloroplast signaling and in the radA mutant (Tigano et al. 2021; Chevigny et al. 2022; Fu et al. 2023).

Differential roles for the 2 OEX1 isoforms generated by alternative splicing

Both our experimental data and publicly available transcriptomic datasets confirm that Arabidopsis has 2 OEX1 isoforms that are encoded by alternatively spliced transcripts. These isoforms seem conserved among flowering plants, as corresponding variants have been found in other species including monocots, thus pointing to their biological relevance. The extension found in mitochondrial OEX1a is neither found in chloroplast OEX2 nor in the bacterial orthologs, thus raising the question why mtDNA maintenance requires 2 isoforms. Our expression analysis suggests developmental regulation, with a higher OEX1b/OEX1a ratio in developing flowers. We also demonstrated that both proteins are functional, as they complement the oex1-1 developmental defects. However, they influence mtDNA segregation differently, and plants complemented with OEX1b maintained a slight imbalance in mtDNA subgenomes, which, by contrast, was fully corrected in plants complemented with OEX1a. This difference may suggest distinct roles or substrate preferences of the 2 isoforms. Structural modeling supports this assumption, by placing the OEX1a extension near the active site and the DNA-binding domain. Indeed, in vitro activity tests showed that, while both isoforms process the same substrates, they differ in processivity, with OEX1a having a higher specific activity on all tested substrates, including 5′-3′ exonuclease activity on DNA ends and FEN activity. Nevertheless, we cannot rule out the alternative possibility that OEX1b is specifically required to process certain specific recombination intermediates. The existence of 2 isoforms might reflect distinct interactions with other factors involved in mtDNA maintenance. OEX1 localization in nucleoids suggests that the protein may be part of larger nucleo-protein structures where mtDNA metabolism occurs. The requirement of DNA polymerase activity for efficient processing of DNA gaps, such as long-patch BER intermediates, supports the idea that OEX1a and OEX1b act in concert with other factors. Yet, these interactions might be transient, because published proteomic data revealed that OEX1 migrates as a single band in blue-native gels, of an apparent size that matches that of monomeric OEX1 (∼43 kDa) rather than that of a stable complex with other factors such as POL1A or POL1B, both of which are larger than 100 kDa (https://complexomemap.de/; Senkler et al. 2017).

Finally, our findings from plants complemented with either OEX1a or OEX1b under genotoxic stress provide support the idea that the 2 OEX1 isoforms have distinct roles in mtDNA repair. This parallels the distinct roles of POL1A and POL1B in mtDNA maintenance, which, while redundant for genome replication, differ in processivity, fidelity and involvement in DNA repair (Parent et al. 2011; Baruch-Torres and Brieba 2017; Ayala-Garcia et al. 2018; Garcia-Medel et al. 2019; Schatz-Daas et al. 2022). The picture emerging from this work and from other studies is that mtDNA maintenance relies on a fine-tuned interplay of multiple factors. These include RecA recombinases, SSB and OSB ssDNA-binding proteins, DNA polymerases, and OEX nucleases, which are present in different isoforms with specialized roles shaped by tissue-specific expression, activity profiles, and distinct functions in replication and repair pathways. These isoforms can arise from distinct paralogous genes, dual targeting of the same protein to both organelles, or, in the case of OEX1, alternative splicing.

Materials and methods

Plant material and constructions

Sequence data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: OEX1, At3g52050; OEX2, At1g34380. Arabidopsis (A. thaliana) T-DNA insertion line oex1-1 (GK-911E05) derives from accession Col-0 and was obtained from the Nottingham Arabidopsis Stock Centre. Plants were grown on soil or in vitro under a 16 h light/8 h dark photoperiod (65 to 85 μmol photons m⁻² s⁻¹) at 21 °C/18 °C. For in vitro cultures, surface-sterilized seeds were sown on agar plates containing 0.5× MS255 media (Duchefa) supplemented with 1% (w/v) sucrose and stratified 2 d at 4 °C in the dark. For DNA and RNA extraction, plant tissue frozen in liquid nitrogen was ground with a TissueLyser II (Qiagen). DNA was extracted using the cetyltrimethylammonium bromide method and RNA was extracted using TRI Reagent (Molecular Research Centre, Inc.). RNA was further extracted with phenol–chloroform before ethanol precipitation. For mutant complementation, the WT OEX1 gene and promoter sequences were cloned into binary vector pGWB613 and used to transform heterozygous oex1-1 plants. For complementation with either OEX1a or OEX1b sequences, the gene sequence including the promoter, 5′ UTR, and the first 2 exons and introns was fused to the OEX1a or OEX1b cDNA sequences. For the hemicomplementation constructs, the organellar targeting sequence (OTS) of OEX1 (86 codons) was replaced with either the OTS (49 codons) of mitochondrial alternative oxidase 1 from soybean (AOX1; NM_001249237) or the OTS (80 codons) of the chloroplast small subunit of ribulose bisphosphate carboxylase (RBCS; At1g67090). The constructs were assembled using MultiSite Gateway cloning in vector pB7m34GW (https://gatewayvectors.vib.be/), giving AOX1:OEX1 and RBCS:OEX1 constructs. These were used for Agrobacterium tumefaciens transformation of heterozygous oex1-1 plants by floral dip.

Phylogenetic analysis

For the building of phylogenetic trees, bacterial and plant orthologous sequences were aligned with T-Coffee implemented in the MacVector package with default parameters (Myers–Miller dynamic algorithm, gap open penalty of −50 and gap-extension penalty of −50) and a consensus maximum likelihood tree was built with IQ-TREE (http://iqtree.cibiv.univie.ac.at), from 1,000 bootstrap iterations. Graphical representation and edition of the tree was with TreeDyn (v198.3). The Arabidopsis OEX1 sequence was modeled on the structure of the 5′-3′ exonuclease domain of Thermus aquaticus DNA Pol I (Kim et al. 1995), using Modeler (Sali and Blundell 1993; http://salilab.org/modeller/about_modeller.html), and the modeled structures represented with PyMol (https://pymol.org/2/). Alignment and tree files are provided as Supplementary Files 1 and 2.

Subcellular localization and promoter-GUS fusion

For in vivo intracellular localization, the OEX1 and OEX2 gene sequences were cloned into the pUBC-paGFP-Dest vector, upstream and in frame with the GFP coding sequence, under the control of the UBQ10 promoter (Grefen et al. 2010). Leaf pavement cells of independent transformants were observed at the confocal microscope (Zeiss LSM700). For colocalization of OEX1 with mitochondria, the OEX1 cDNA was amplified and cloned into the pUBC-mCherry-Dest binary vector (Grefen et al. 2010), to obtain the pUbi::OEX1-mCherry construct. The resulting plasmid was used for Agrobacterium-mediated transformation of plants expressing the mitochondrial marker pCaMV35S::mTP-GFP, corresponding to the transit peptide of ATP citrate lyase (At2g20420) fused to GFP, under the control of the 35S promoter. The resulting seeds were selected on MS medium supplemented with phosphinotricin at a final concentration of 10 mg/L. To facilitate visualization of mitochondrial nucleoids, cells were observed under conditions that induce elongated mitochondria, as described (Jaipargas et al. 2015). To this end, Arabidopsis seedlings expressing both pCaMV35S::mTP-GFP and pUbi::OEX1-mCherry transgenes were grown in the dark for 8 d on 0.5× MS medium without sugar. DNA was stained with DAPI (4′,6-diamidino-2-phenylindole) solution (1 µg/mL) in phosphate-buffered saline for 30 min at 25 °C in the dark, and leaf pavement cells were imaged at a Leica TCS SP8 confocal laser-scanning microscope. Excitation of DAPI was at 405 nm (laser diode), GFP and chlorophyll at 488 nm and mCherry at 561 nm (argon lasers). Fluorescence emission was collected at 435 to 475 nm for DAPI, 500 to 545 nm for GFP, 595 to 645 nm for mCherry and beyond 650 nm for chlorophyll.

For promoter-GUS histochemical analysis, the 5′ upstream region of OEX1 (595 nucleotides comprising 161 nucleotides of OEX1 5′-UTR and a large part of upstream gene At3g52040) was cloned upstream of the GUS gene in the binary vector pMDC162 (Curtis and Grossniklaus 2003), and the construct was used to generate stable Arabidopsis transformants. Tissues from 6 independent lines were stained with 5-bromo-4-chloro-3-indolyl-β-D-glucuronic acid (X-Gluc; Biosynth) and observed by stereomicroscopy and transmission microscopy.

Transmission electron microscopy

Leaf tissue samples were fixed overnight in 3% glutaraldehyde, treated 2 h with 10% (w/v) picric acid, 2 h with 2% uranyl acetate and stained with 0.1% (v/v) osmium tetroxide in 150 mm phosphate buffer, pH 7.2. After dehydration through an ethanol series, samples were infiltrated with EPON812 medium-grade resin (Polysciences). Ultrathin 70 µm sections were collected on grids coated with formvar (Electron Microscopy Sciences) and visualized with a Hitachi H-600 electron microscope operating at 75 kV.

qPCR/RT-qPCR analysis

For RT-qPCR experiments, 5 μg of RNA were depleted from contaminating DNA by treatment with RQ1 RNase-free DNase (Promega) and reverse-transcribed with Superscript IV Reverse Transcriptase (Thermo Fisher Scientific), according to the manufacturer's protocol using random hexamers. The qPCR assays were performed with the LightCycler480 (Roche) on a reaction mix of 6 µL containing 1× LightCycler 480 SYBRGreen I MasterMix (Roche) and 0.5 μm of each primer. Three technical replicates were performed for each experiment and C_p values were determined from the second derivative maximum. Quantification of mtDNA and cpDNA copy numbers used a set of primer pairs located along the organellar genomes, as described previously (Wallet et al. 2015; Chevigny et al. 2022). Results were normalized against UBQ10 (At4g05320) and ACT1 (At2g37620) nuclear genes. The accumulation of ectopic recombination in mtDNA was quantified using primers flanking each repeat, as described (Miller-Messmer et al. 2012), and results normalized against the COX2 (AtMG00160) and 18S rRNA (AtMG01390) mitochondrial genes. RT-qPCR experiments were normalized using the GAPDH (At1g13440) and ACT2 (At3g18780) transcripts as standards.

Sequencing

Total leaf DNA of WT and oex1-1 plants was quantified with a QuBit Fluorometer (Life Technologies) and libraries prepared with the Nextera Flex Library kit according to the manufacturer instructions (Illumina), using 100 ng DNA. Final libraries were quantified, quality checked on a Bioanalyzer 2100 (Agilent) and sequenced on an Illumina MiSeq System (2 × 150 paired-end reads). For RNA-seq, total RNA was extracted from WT and oex1-1 10 d-old seedlings, tree biological replicates each, using NucleoSpin RNA Plant kit (Macherey-Nagel). RNAs were quantified by QuBit (Invitrogen) and quality checked on Bioanalyzer 2100. Five micrograms of total RNAs were ribodepleted using riboPOOL kit (siTOOLs Biotech). Ten nanograms of ribodepleted RNAs were used for library preparation with NEBNext Ultra II Directional RNA Library Prep, according to the manufacturer instructions. Libraries were sequenced (1 × 50 bases single read) on an Illumina NextSeq 2000. DNA-seq and RNA-seq reads were mapped to the mtDNA with Bowtie2, implemented in the MacVector Assembler package application. Because of sequences of chloroplast origin inserted in the mtDNA, reads mapping to the chloroplast genome were first removed from the analysis. Coverage profiles were generated with the MacVector Assembler package.

Expression and purification of recombinant proteins

The coding sequence of the AtOEX1 gene minus the first 85 codons was amplified from cDNA and both alternative cDNA sequences, OEX1a and OEX1b, were cloned in the pET28a expression vector (Novagen), to express recombinant proteins containing a N-terminal hexa-histidine tag. The recombinant proteins were soluble and were purified from the E. coli BL21(DE3) extracts by nickel ion affinity chromatography, followed by gel filtration (Supplementary Fig. S7). Briefly, protein expression was induced with 0.5 mm IPTG and after 2 h at 37 °C the bacterial cells were sedimented, resuspended in 150 mm Tris-HCl pH 8.0, 5% glycerol, 150 mm NaCl (Buffer A) supplemented with 1 mm PMSF, and lysed with a LM20 Digital Microfluidizer Processor (Microfluidics) under 1,200 PSI. The crude lysate was clarified by centrifugation at 17,000 × g for 20 min, adjusted to 30 mm imidazole. Proteins were purified by affinity on cOmplete His-Tag Purification Resin (Roche). Columns were washed with Buffer A followed by washing with 50 mm imidazole. The recombinant proteins were eluted with a 90 to 310 mm imidazole step gradient, followed by gel filtration on a Superdex 200 10/300 GL column equilibrated with Buffer A (Supplementary Fig. S5C). Protein aliquots were flash frozen in liquid nitrogen and stored at −80 °C. A catalytic mutant protein was equally expressed and purified, which contained a substitution of an essential aspartic acid with alanine in the active site (protein D250A, Supplementary Fig. S5B). Following purification, the recombinant proteins were free of detectable contaminants and purified as a single monodispersed peak by gel filtration, of a size consistent with a monomeric protein (Supplementary Fig. S7C). The optimal activity conditions of OEX1 on 5′-labeled dsDNA were tested (Supplementary Fig. S8). The protein requires Mg²⁺ ions that could be poorly substituted by Mn²⁺, but not by Zn²⁺ or Ca²⁺. It showed optimal activity at ∼37 °C but was inhibited at higher temperature. Finally, it worked in a broad pH range, between pH 7.4 and pH 9.0. Based on these results we established the assay conditions for further experiments as 2 mm Mg²⁺, pH 8.0 at 37 °C.

Nuclease assays

Structure-specific nuclease assays were performed using oligonucleotides synthesized by Integrated DNA Technologies. DNA oligonucleotides were either radiolabeled in 5′ using T4 polynucleotide kinase (Fermentas) and [γ-³²P]ATP, or in 3′, by template-directed addition of 1 to 3 cytosines by the Klenow fragment of E. coli DNA Pol I and [α-³²P]dCTP. RNA oligonucleotides were labeled in 3′ using poly(U) polymerase (NEB) and [α-³²P]UTP. When necessary, labeled products where annealed with cold oligonucleotides by incubation in 10 mm Tris-HCl pH 7.5, 1 mm EDTA, 100 mm NaCl (Buffer B) during 5 min at 95 °C, followed by slow cooling to room temperature. Oligonucleotide sequences used to form the different structures are listed in Supplementary Table S1. Substrates were purified by electrophoresis on nondenaturing 15% polyacrylamide gels and eluted in Buffer B, overnight at 4 °C under agitation. Standard nuclease reactions were performed with 50 fmol of labeled structure in reaction buffer (50 mm Tris-HCl pH 8.0, 1 mm DTT, 0.1 mg/mL BSA, and 4% glycerol) supplemented with 2 mm bivalent ion and 1.5 µm enzyme. The reaction was incubated at 37 °C and stopped by addition of an equal volume of stop-mix (95% formamide, 20 mm EDTA, 0.05% bromophenol blue, and 0.05% xylene cyanol). Reaction products were analyzed by denaturing PAGE (17% 19:1 acrylamide/bisacrylamide, 7 M urea, 1× TBE) and revealed using an Amersham Typhoon biomolecular imager (GE Healthcare Life Sciences).

DNA:RNA immunoprecipitation

Total DNA was extracted with Plant II Midi kit (Macherey-Nagel) and fragmented with EcoRI, BamHI, HindIII, and XbaI to cut the mtDNA into pieces between 200 and 5,000 bp. About 5 µg were either treated or not with Ribonuclease H (RNase H 5 U/μL) (Thermo Fisher Scientific), and incubated with 2 µL S9.6 antibody (Merck) specific for RNA:DNA hybrids, for 15 h at 4 °C under rotation (10 rpm), followed by incubation with 25 µL of Protein G Sepharose beads (Sigma-Aldrich) for 2 h at 4 °C (Yang et al. 2017). Immunoprecipitated fragments were detached from antibodies by proteinase K treatment (Thermo Fisher Scientific) and purified by phenol/chloroform extraction followed by ethanol precipitation. The isolated DNA was analyzed by qPCR using primer pairs targeting the highly transcribed mtDNA regions coding for the ribosomal RNAs, and results were normalized against the quantification of nontranscribed regions of the mtDNA.

Statistical analysis

Values are represented as mean ± Sd. Significant differences between 2 groups were tested by 2-tailed Student's t-test, using Microsoft Excel 2024 software. Significant differences between CIP-treated plants were evaluated using χ² test analysis, also using Microsoft Excel 2024 software. ImageJ software was used for the quantitative analysis of confocal images. Details of statistical analyses, including methods, sample sizes, and significance levels (P-values), can be found in the figures, figure legends, and Supplementary Data Set 1.

Accession numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: OEX1, At3g52050; OEX2, At1g34380; POL1A, At1g50840; POL1B, At3g20540; RADA, At5g50340; RBCS, At1g67090; Glycine max  AOX1, NM_001249237. RNA-seq data were deposited in the SRA database under reference PRJNA1237839.

Author contributions

D.S., A.L.B., F.G.M., K.P.C., E.V., P.M., R.K.I., F.W.-L., and J.M.G. performed research. D.S., F.W.-L., and J.M.G. designed the research and analyzed data. E.V. and J.M.G. analyzed the Illumina sequence data. D.S., F.W.-L., and J.M.G. wrote the paper. A.D. and R.B. reviewed it critically for intellectual content.

Supplementary data

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

Supplementary Figure S1. The widespread expression of OEX1 visualized using a ProOEX1:GUS fusion.

Supplementary Figure S2. Knockout of OEX1 in the oex1-1 line.

Supplementary Figure S3. Transmission electron micrographs showing smaller, more electron-dense mitochondria in oex1-1 compared with WT.

Supplementary Figure S4. Alternative splicing results in the expression of 2 OEX1 isoforms.

Supplementary Figure S5. RNA-seq analysis of the mitochondrial transcriptome.

Supplementary Figure S6. Expression and purification of recombinant OEX1 proteins.

Supplementary Figure S7. Conditions yielding optimal OEX1 activity.

Supplementary Figure S8. Sequence-dependent processivity and activity of OEX1 on ssDNA and DNA gaps.

Supplementary Figure S9. Possible pathways of R-loop resolution in plant mitochondria.

Supplementary Table S1. Primers.

Supplementary File 1. Alignment of plant OEX1 and OEX2 bacterial sequences.

Supplementary File 2. Phylogeny tree in Newick format.

Supplementary Data Set 1. Statistical data.

Funding

This work has been published under the framework of the LABEX (ANR-11-LABX-0057_MITOCROSS) and benefits from funding from the state managed by the French National Research Agency as part of the “Investments for the future” program.

Data availability

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

Dive Curated Terms

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

• MSH1 Gramene: AT3G24320

• MSH1 Araport: AT3G24320

• OSB1 Gramene: AT1G47720

• OSB1 Araport: AT1G47720