Messenger RNA delivery to mitoribosomes – hints from a bacterial toxin

In mammalian mitochondria, messenger RNA is processed and matured from large primary transcripts in structures known as RNA granules. The identity of the factors and process transferring the matured mRNA to the mitoribosome for translation is unclear. Nascent mature transcripts are believed to associate initially with the small mitoribosomal subunit prior to recruitment of the large subunit to form the translationally active monosome. When the small subunit fails to assemble, however, the stability of mt‐mRNA is only marginally affected, and under these conditions, the LRPPRC/SLIRP RNA‐binding complex has been implicated in maintaining mt‐mRNA stability. Here, we exploit the activity of a bacterial ribotoxin, VapC20, to show that in the absence of the large mitoribosomal subunit, mt‐mRNA species are selectively lost. Further, if the small subunit is also depleted, the mt‐mRNA levels are recovered. As a consequence of these data, we suggest a natural pathway for loading processed mt‐mRNA onto the mitoribosome.


Introduction
The human mitochondrial genome is transcribed from single promoter regions on each of the two strands [1,2]. The resultant large polycistronic RNA units are processed into individual messenger (mt-mRNA), ribosomal (mt-rRNA) or transfer (mt-tRNA) RNAs and are subsequently matured. Many of these events are described as occurring in discrete RNA granules close to the mtDNA in the mitochondrial matrix [3]. The early stages of mitochondrial ribosome assembly also occur within or close to these granules [4]. Apart from synthesis and maturation, much of the life cycle of mt-mRNA is unclear [5]. Early in vitro studies support the notion that mt-mRNA is first loaded onto the mammalian mitoribosomal 28S small subunit (mt-SSU) [6,7] but how is the nascent mt-mRNA transported to the mt-SSU and how is it protected ? Is mt-mRNA protected by the mt-SSU alone, or is stability dependent on further association with the mitoribosomal 39S large subunit (mt-LSU) and associated factors? For example, the RNA-binding complex formed by LRPPRC/SLIRP has been implicated in stabilizing mt-mRNA that is not bound to the mitoribosome [8] and exposing its required sites for translation [9]. A recent report suggests SLIRP promotes correct association of mt-mRNA with the mitoribosome [10]. Degradation of the antisense or mirror mt-mRNAs clearly involves the mitochondrial PNPase in tandem with the hSUV3 helicase [11], but are these proteins also involved in sense mt-mRNA degradation, and what promotes this degradation event ?
In an attempt to address some of these questions, we set out to selectively remove the mt-LSU. To facilitate this, we have exploited the sequence specificity of mycobacterial endoribonuclease, VapC20 [12]. This PIN domain protein is part of the extensive toxin:antitoxin system that is found within Mycobacterium tuberculosis. It associates with the bacterial large ribosomal subunit and cleaves the 23S ribosomal RNA within the highly conserved sarcin:ricin loop (SRL), unless bound to its cognate VapB20 antitoxin counterpart [12,13]. The exquisite cleavage selectivity is believed to be due to the ribonuclease recognizing the environment of SRL within the context of the ribosome, as VapC20 will not cleave isolated ribosomal RNA, in vitro or in vivo [12]. The SRL is essential during the elongation step of protein synthesis allowing the formation of stabilizing contacts between the ternary complex EF-Tu/aa-tRNA/GTP and the ribosome [14]. Being close to their GTPase domains, this rRNA loop interacts with both EF-Tu and EF-G [15,16] and its cleavage differentially affects elongation factors binding to the bacterial ribosome [17]. The interaction of LSU rRNA with GTPases has also been described recently in human mitochondria, where the 16S mt-rRNA domain VI, including the SRL and helices H96-H97, interacts with human MTG1, a conserved GTPase that couples mt-LSU assembly with intersubunit bridge formation [18].
Analysis revealed a strong but incomplete sequence identity between the SRL of Escherichia coli 23S rRNA and the human mitochondrial 16S rRNA. By fusing the VapC20 protein downstream of a mitochondrial localization signal, we have been able to show that, on induction, mtVapC20 localizes specifically to the mitochondrial matrix where it cleaves uniquely the mitoribosomal 16S rRNA not in the SRL, but in helix 91 that is immediately adjacent to the SRL, in the folded 16S mt-rRNA. Expression of mtVapC20 caused the temporal and substantial depletion of the mt-LSU without a concomitant decrease of mt-SSU, resulting in the loss of almost all mt-mRNA species tested. In many other systems that have been manipulated, levels of the LRPPRC/SLIRP complex reflect the steadystate levels of most mt-mRNA species, consistent with a mutual interdependency for stability [10, [19][20][21]. Under conditions of mtVapC20 expression, however, even with vanishingly low levels of mt-mRNA, the LRPPRC/SLIRP complex is maintained. Notably, when mt-SSU is depleted in tandem with mt-LSU, mt-mRNA levels as measured by the levels of MTCO1 are partially recovered.
We hypothesize that monosome formation is required to promote the natural stability of the mt-mRNA following its transfer from LRPPRC/SLIRP complex.

mtVapC20 localizes to the human mitochondrial matrix
To determine whether VapC20 could cleave mitochondrial 16S rRNA, a Flp-In TM 293 T-REx cell line was engineered to inducibly express M. tuberculosis VapC20 fused in-frame with an N-terminal mitochondrial presequence from the Neurospora crassa ATP synthase subunit 9 (Su9; [22]) as described. Clones were isolated, and inducible mtVapC20 expression was confirmed. After 7 days of induction, the cell media acidified, consistent with an OXPHOS deficiency; however, following 9 days of induction cell growth was only mildly affected (Fig. 1A). To assess whether mitochondria-targeted VapC20 (mtVapC20) was successfully imported into the mitochondrial matrix, mitochondria were isolated from an induced clonal line and subfractionated. The mtVapC20 displayed clear colocalization with markers of the mitochondrial matrix, glutamate dehydrogenase and mitochondrial elongation factor Tu (Fig. 1B). The product size (approx. 15 kDa, upper panel) was compatible with the cleavage of the Su9 presequence. Precursor was undetectable, consistent with a high level of import of the expressed product into mitochondria. mtVapC20 selectively cleaves the 16S mt-rRNA Initial sequence and structure comparisons showed significant similarity between the E. coli 23S SRL and the mitochondrial 16S rRNA SRL ( Fig. 2A); however, the region is not entirely conserved. To determine whether mtVapC20 was able to cleave the 16S mt-rRNA, total cell RNA was isolated following a time course of induction. Northern blot analysis with probes covering 5 0 and 3 0 sequences revealed a unique product of approximately 200 nucleotides, detected specifically with the 3 0 probe after 24 h of induction (Fig. 2B,C). Although specific cleavage occurred, the exact product size was unexpected, as cleavage in the SRL should generate a shorter 3 0 product of 100 nucleotides. Trapping of the digestion product by 5 0 blocked ligation-mediated PCR and sequence analysis revealed a spectrum of products differing at their 5 0 termini by a few bases but clearly  2D, [23]). Interestingly, crystal structures of the bacterial ribosome show that helix 91 of the 23S rRNA establishes loop-loop tertiary interactions with SRL [24]. This tight conformation seems to be essential for both the assembly and the function of the large ribosomal subunit [25].
Assembly of the mt-SSU is independent of mt-LSU formation To determine the effect of 16S mt-rRNA cleavage, extracts were isolated at various stages postinduction and subjected to western or northern blot analysis. The mitochondrially encoded complex IV protein, COX2, was depleted by day 3 of induction and was markedly affected after 7 days (Fig. 3A). Similarly, NDUFB8, a nuclear-encoded protein but a sensitive marker of complex I assembly, was also significantly depleted. Metabolic labelling of mitochondrial gene products was also substantially affected after 2 days of induction (Fig. 3B). These data are consistent with reduced mitochondrial protein synthesis. To determine the steady-state levels of the mitochondrial ribosome, markers of the large (mt-LSU) and small (mt-SSU) subunit were assessed (Fig. 3A). Steady-state levels of mt-LSU components uL3m, uL11m and mL45 were depleted after 3 days but were markedly affected after 7 days, consistent with cleavage of the 16S mt-rRNA leading to loss of mt-LSU assembly. In contradistinction, mt-SSU components mS29 and mS22 were unaffected, suggesting that mt-SSU can assemble and remain stable in the absence of mt-LSU. Targeting of the nuclease-deficient D5A-mtVapC20 mutant had no measurable effect on the levels of OXPHOS components or mitoribosome (Fig. 3C). To confirm that mt-SSU assembly was unaffected, mtVapC20 was expressed for varying time periods and extracts were subjected to isokinetic sucrose gradient analysis. After 3 days of induction, fully assembled mt-LSU was almost undetectable (Fig. 4A, fractions 6/7) with unassembled subunits being rapidly degraded or unexpressed. The mt-SSU, however, is retained (compare signals for mS29 in fractions 4/5 of uninduced and 6 days mtVapC20 induction). To further confirm the specific loss of mt-LSU on mtVapC20 induction, RNA FISH was performed with probes specific to either the 16S or 12S mt-rRNA (Fig. 4B,C). After 6 days of induction, mtVapC20 led to a marked decrease in

Absence of mt-LSU negatively affects mt-mRNA levels
Previous data have shown that the depletion of mt-SSU induced by a lack of modification or stability of the 12S mt-rRNA in both mice models and in human cell lines does not promote increased degradation of most mt-mRNAs [26,27]. To assess whether mt-mRNA is similarly protected when the mt-LSU is absent, mtVapC20 was expressed for varying periods, after which RNA was isolated and northern blots performed (Fig. 5). Levels of analysed mt-tRNAs were unaffected or only slightly reduced (Fig. 5A), consistent with the lack of any appreciable effect on mtDNA content (with exception for 10 days expression where mtDNA levels were increased) or transcription ( Fig. 6). Further, 12S mt-rRNA levels were only minimally depleted ( Fig. 5B), reflecting the specificity of mtVapC20 cleavage and that mt-SSU assembly was unaffected. Noticeably, however, although the pattern of steady-state levels over the induction time course varied between mt-RNA species, all mt-mRNA species analysed showed decreases in steady-state level by day 10. The assessed mt-mRNA levels quantitatively reflected the decreasing levels shown for 16S mt-rRNA, the marker of assembled mt-LSU. MTND3 was only mildly affected ( Fig. 5A), even after 10 days of induction (approx. 40% of control). Thus, in the absence of mt-LSU the majority of mt-mRNA species appear to be selectively lost.
With the aim of analysing the association of mt-mRNA with mitoribosomes, mtVapC20 was expressed for 3 or 6 days, extracts subjected to isokinetic sucrose gradient separation, followed by isolation of RNA from each fraction. Northern blot analysis (Fig. 5C) confirmed the presence of 12S mt-rRNA that was normally associated with assembled mt-SSU (fractions 4/5), even after 6 days of toxin expression. The 16S mt-rRNA and mt-mRNA species were strongly depleted and mainly associated with the remaining fraction of fully assembled monosome (6 days, fractions 8/9). The only exception was MTND3 that showed a lower rate of decrease and a different distribution within the gradient after 3 days of induction, accumulating in the lighter fractions.

LRPPRC/SLIRP is unaffected by the loss of mt-LSU
Evidence from patients, various transgenic mouse and human cultured cell models have implicated the LRPPRC/SLIRP RNA-binding complex in the stability of mt-mRNA [10, [19][20][21]. A substantial pool of data now shows a correlation between steady-state levels of mt-mRNA species and levels of LRPPRC. Increased steady-state transcript levels have been shown to mirror increased LRPPRC levels as observed in mouse models depleted for the 12S mt-rRNA methylase NSUN4 [28]. Several research groups have also reported that LRPPRC depletion led to decreased stability of most mt-mRNA species in both cellular and animal models [8,20,29,30]. Thus, upon mtVapC20 induction, LRPPRC levels would be predicted to decrease, to reflect the mitochondrial transcript reduction. Western blotting of extracts was performed following induction and showed that there was no measurable effect on the steady-state levels of LRPPRC or SLIRP (Fig. 7A).
Recovery of steady-state levels of mt-mRNA upon tandem depletion of mt-LSU and mt-SSU One viable possibility to reconcile this large body of data with our observation is that in the absence of mt- Fig. 2. The sarcin:ricin loop and its cleavage by mtVapC20 in human mitochondrial 16S rRNA. (A) Sequences of the human mitochondrial 16S rRNA SRL (GenBank ID: NC_012920.1, 3115-3141) and bacterial 23S rRNA SRL (GenBank ID: NR_076322.1, 2648-2674) were aligned using the EMBL-EBI MView online tool (https://www.ebi.ac.uk/Tools/msa/mview/). Identical nucleotides are blocked in black. Structural superimposition of mitochondrial (red) and bacterial (amber) SRL RNA tertiary structures was performed with the web server MultiSETTER [42] uploading PDB files (human 16S mt-rRNA, PDB ID: 3J7Y; E. coli 23S SRL, PDB ID: 1Q9A) as input (S-distance 0.209, P-value 0.039). (B) Cleavage products generated following mtVapC20 expression were analysed by high-resolution northern blot as described. Blots were hybridized with radiolabelled probes specific to 16S mt-rRNA 3 0 end. Upper left panel: mtVapC20 cleavage product is indicated with an arrow. Upper right panel: RNA was stained by SYBR TM Gold (Thermo Fisher Scientific) to validate loading. Low Range ssRNA Ladder (New England Biolabs) sizes are indicated on the right. Lower panel: sequence analysis of 5 0 -end ligation-mediated RT-PCR product. Blue bar depicts the sequence of 16S mt-rRNA helix 91. Data are representative of two independent experiments. (C) Cleavage analysis of 16S mt-rRNA was performed with two different probes. Total RNA was extracted from uninduced (U) and mtVapC20 cells induced for 5 h, overnight (O.N.) and 3 days. The mtVapC20 cleavage product, only detected using probe #1, is indicated with an arrow. Data for each time point are representative of two independent experiments. 5S rRNA was used as loading control. (D) Magnification of the boxed area shows the mtVapC20 cleavage site and its proximity to the SRL. Structures of the assembled human mitoribosome (PDB ID: 3J9M; mt-LSU in dark green, mt-SSU in light green) and the cleavage site were obtained with PYMOL software (Open Source, Version 1.8.2.0, Schr€ odinger, New York, NY, USA).  , as facilitated by expression of mtVapC20, all mt-SSU may be freely available and capable to receive mature mt-mRNA from the LRPPRC/SLIRP complex but on transfer, there is no mt-LSU to form the monosome, resulting in loss of protection for the mt-mRNA leading to its rapid degradation. One prediction of this hypothesis is that concomitant loss of mt-SSU and mt-LSU should lead to restoration of transcript stability as the newly mature transcript would remain associated with the LRPPRC/SLIRP complex. To test this prediction, mt-SSU levels were partially reduced by depleting the 12S mt-rRNA chaperone, ERAL1. This Fig. 3. Effects of mtVapC20 expression on mitochondrial translation. (A) Cell lysates were prepared and aliquots (40 µg) subjected to immunoblotting as described. Steady-state levels of representative OXPHOS subunits (green) were determined with antibodies against Complex I (NDUFA9 and NDUFB8) and Complex IV (COX2). Steady-state levels of mt-LSU (blue) and mt-SSU (red) components were determined by western blot using antibodies against components of the mt-LSU [HUGO protein names: MRPL3 (uL3m); MRPL11 (uL11m); MRPL45 (mL45)] and mt-SSU [HUGO protein names: DAP3 (mS29); MRPS18B (mS40); MRPS22 (mS22)] as indicated. Asterisk indicates that the antibody possibly recognizes one of the two other mS40 isoforms, mL66 (MRPS18A), found in the mt-LSU. Expression of mtVapC20 was detected by anti-6XHis antibody. b-actin and VDAC, a mitochondrial outer membrane protein, were used as loading control and mitochondrial marker, respectively. Data for each time point are representative of one experiment. (B) Metabolic 35 S-met/cys labelling of mitochondrial translation products was performed in uninduced (U) and mtVapC20 cells induced for 2 and 6 days as described. Mitochondrially encoded polypeptides were assigned after Chomyn [43]. Data for each time point are representative of one experiment. Equivalent protein loading was confirmed by Coomassie blue staining (CBS). (C) Immunoblotting analysis showing the effect of mutant mtVapC20 (D5A) expression on steady-state levels of OXPHOS Complex IV, subunit 2 (COX2) and a constituent of the large mitoribosomal subunit (uL3m). Mutant mtVapC20 was detected with the anti-6XHis antibody. VDAC was used as loading control. Data for each time point are representative of one experiment. was achieved by specific siRNA downregulation prior to the induction of mtVapC20. As shown in Fig. 7B, ERAL1 is depleted, resulting in less mt-SSU as evidenced by the reduction in 12S mt-rRNA (Fig. 7C,  lane 3). Concomitantly, induction of mtVapC20 lowered the mt-LSU levels, evidenced by 16S mt-rRNA reduction (Fig. 7C, lanes 2-4). Although the reduction in mt-SSU is only partial, there is a recovery in the steady-state levels of MTCO1 transcript (Fig. 7C, lane 3).

Discussion
Targeting of the selective bacterial endoribonuclease VapC20 to human mitochondria results in the cleavage of 16S mt-rRNA with the resultant loss of mt-LSU. Unexpectedly, although the toxin has been shown previously to cleave E. coli 23S rRNA in exactly the same position in helix 95 as sarcin and other eukaryotic ribotoxins, mtVapC20 does not cleave the SRL of 16S mt-rRNA in helix 95 but at helix 91, which is in close apposition to the SRL in the mitoribosome. In vitro, VapC20 is unable to cleave purified 23S rRNA alone, suggesting that cleavage requires a particular environment provided by the assembled or assembling ribosome. It would appear that in the case of the 16S mt-rRNA, the SRL is recognized but actual cleavage occurs close to this site within the RNA. By targeting VapC20 to human mitochondria, we have shown that the mt-SSU can assemble in the absence of the mt-LSU. The work also suggests that the natural stability of mt-mRNA shows dependence on the mt-LSU, presumably due to it combining with the mt-SSU to form the monosome, thus protecting the bound transcript. Several previous reports have detailed the molecular consequences when mt-LSU assembly has been partially compromised. Intriguingly, in certain cases where levels of mt-mRNA were assessed, there was no evidence that mt-mRNA stability was reduced. Two recent publications report the effect of depletion of the DDX28 helicase [4,31], which partially affects mt-LSU biogenesis. The maximal depletion of 16S mt-rRNA, as a surrogate marker of assembled mt-LSU, in each report did not fall below 40% levels of control. In contrast, longer term shRNA-mediated depletion of the mt-LSU component mL44 led to mitochondrial 16S rRNA levels dropping below 10% of control levels with a concomitant reduction of most mt-mRNA [32], similar to our current report. One explanation for this apparent discrepancy is that depletion of DDX28 may cause aberrant mt-LSU assembly, whereby monosome formation is still possible but translation is affected. This would also infer that levels of mt-LSU were present in substantial excess over mt-mRNA. Notably, Fig. 4. Effects of mtVapC20 expression on mitoribosomal components. (A) Isokinetic sucrose gradient centrifugation was used to analyse mitochondrial ribosomal protein distribution in mtVapC20 cells uninduced and induced for time periods as indicated. The migration of mt-SSU (red box) and mt-LSU (blue box) was determined using antibodies against mS29 (mt-SSU), uL3m and uL11m (mt-LSU). HUGO protein names: DAP3 (mS29); MRPL3 (uL3m); and MRPL11 (uL11m). Data are representative of two independent experiments. (B) Changes to distribution of 16S mt-rRNA (in blue) and 12S mt-rRNA (in red) in uninduced and mtVapC20-expressing cells were determined by RNA FISH as described. For each sample, four representative confocal images are shown (scale bar: 10 µm). (C) Quantitative analysis of fluorescence intensities was performed using IMAGEJ and GRAPHPAD PRISM8 as detailed in Materials and methods. Results are presented as the mean AE SD (n = 9). Statistical analysis was performed using the Welch's T test (*P < 0.05; ****P < 0.0001). our data showed that MTND3 mRNA behaved in a different way. Even after prolonged mt-LSU depletion (10 days of induction), this transcript was still present. In SLIRP KO mice, MTND3 was the most affected transcript with respect to interactions with mitoribosomes [10]. Here, we report that in the absence of mt-LSU, but in the presence of SLIRP, MTND3 accumulated in the least dense fractions of the gradient, as did 12S mt-rRNA (Fig. 5C). These observations may reflect the fact that MTND3 is the shortest mt-mRNA. It could be more easily protected by the mt-SSU than longer transcripts by the LRPPRC/SLIRP complex as would be the case in mtVapC20 cells.
There is mounting evidence that RNA maturation occurs within transient liquid droplets termed RNA granules and that this is likely to be where mitoribosomes are assembled [3,4]. Many of the components required for processing and maturation of mammalian mt-mRNA are now known, and many have been localized to these submitochondrial compartments [33]. However, although numerous mitoribosomal subunits have also been found close to the nucleoid, it is still unclear how the mature mt-mRNA is loaded onto the mitoribosome or indeed where this occurs. Early in vitro studies supported the idea that mt-mRNA is first loaded onto the mt-SSU before monosome formation [6,7]. It has been also speculated that the monosome can form without mt-mRNA loaded onto it and that the mRNA is inserted later as with leaderless bacterial mRNAs [34,35]. This hypothesis implies that LRPPRC/SLIRP may load the mt-mRNA directly onto the monosome, not onto the mt-SSU. We believe our data are now consistent with a model whereby the LRPPRC/SLIRP complex associates with newly processed mt-mRNAs and ensures normal maturation through regulated polyadenylation also ensuring  5. Effect of mtVapC20 expression on mitochondrially encoded RNAs. Steady-state levels of mt-RNAs were assessed by highresolution (A) and standard (B) northern blot as described. Total RNA was extracted from uninduced (U) and mtVapC20 cells induced (Ind) for 3, 7 and 10 days prior to analysis of RNA (2 µg per lane, A; 10 µg, B). Probes to cytosolic 5S rRNA and 28S rRNA were used as loading controls. (C) Isokinetic sucrose gradients (10-30%) were performed to analyse mitochondrial RNA distribution in uninduced and mtVapC20 cells induced for 3 and 6 days. RNA extracted from gradient fractions was analysed by northern blot with different probes, as indicated. Cytosolic 28S rRNA and 18S rRNA were used as loading controls. All the northern blot data relative to each different time point are representative of one experiment. normal translation by transferring the mature mt-mRNA to the mt-SSU or to the monosome (Fig. 8).
It has previously been shown that POLRMT knockout causes decrease of de novo mitochondrial transcription with a concomitant reduction of LRPPRC levels [21]. Although our data do not completely rule out a negative effect of mt-LSU lack on mitochondrial transcription, the unchanged levels of LRPPRC in VapC20 cells suggest that the selective loss of mt-mRNAs depends on their degradation rather than their diminished synthesis as a result of mt-LSU depletion. Hence, correct stability and translation are driven by monosome formation with the mt-LSU (Fig. 8). In the absence of LRPPRC/SLIRP, mt-mRNA can be degraded before associating with mitoribosomal subunits, with the remainder being aberrantly polyadenylated or aberrantly loaded onto the mitoribosome, resulting in translational abnormalities.
Finally, the inducible VapC20 cell line reported here could provide a tool for studying processes related to mitochondrial gene expression in humans. It may have potential to clarify some central aspects of mt-mRNA turnover and the identification of RNases involved in quality control of untranslated transcripts.
The nontoxic D5A-mtVapC20 construct was obtained by site-directed mutagenesis, performed with the QuikChange II Site Relative quantification of mitochondrial DNA levels by qPCR was performed after 10 days of induction. The relative mtDNA content of the induced mtVapC20 line was compared to uninduced control (set as equal to 1), normalized to 18S rRNA gene (endogenous control). Results are presented as the mean AE SD (n = 3). Statistical analysis was performed using one-sample two-tailed Student's t test (**P < 0.01). (B) Southern blot analysis was performed on total DNA isolated from uninduced (U) and cells induced (mtVapC20) for 2, 5 and 10 days. DNA (5 µg) was digested with BamHI, fractionated on a 0.7% agarose gel and electroblotted to nylon membrane. The filter was hybridized with radiolabelled probe for mitochondrial MTCO2 gene. Total DNA was stained by GelRed â (Biotium, Fremont, CA, USA) as loading control. Data for each different time point are representative of one experiment. (C) In organello transcription assay was performed on uninduced (U) and mtVapC20 cells induced for 6 days. Total RNA was stained by SYBR TM Gold (Thermo Fisher Scientific) as loading control, and radiolabel detected as previously described. Data are representative of one experiment.
For the growth curve analyses, 293 cells that could inducibly express mtVapC20 were seeded onto 6-well plates, induced with 1 µgÁmL À1 tetracycline or uninduced as controls, and propagated with the IncuCyte ZOOM Ò System (Essen BioScience, Ann Arbor, MI, USA). Sixteen fields per well, with biological triplicates, were imaged every 3 h for 9 days. All images were analysed by the INCUCYTE ZOOM Software.

Northern blotting analyses
Northern blots were performed as described previously [37]. Briefly, cells expressing mtVapC20 were harvested at different time points (3, 7 and 10 days) and total RNA was extracted. Aliquots (10 µg) were electrophoresed through 1.2% agarose under denaturing conditions and transferred to GeneScreen Plus membrane (Perkin Elmer, Waltham, MA, USA) following the manufacturer's protocol. Radiolabelled probes were generated using random hexamers on PCR-generated templates corresponding to internal regions of the relevant genes. High-resolution northerns were carried out to analyse the 16S mt-rRNA cleavage product and mt-tRNAs. Total RNA (2 µg) was denatured in 40% formamide/5 mM EDTA at 95°C, 5 min before being separated through 15% urea polyacrylamide gels in 19 TBE (100 mM Tris base, 100 mM boric acid, 2 mM EDTA disodium salt). RNA was transferred to GeneScreen Plus membrane in 0.259 TBE at 150 mA, 15 min followed by 500 mA for 30 min after which the membrane was exposed to UV (optimal crosslink, Stratalinker, Stratagene). Hybridization conditions and probes were as described in [38]. Signals were detected by Typhoon FLA 9500 phosphorimager (GE Healthcare Life Sciences, Marlborough, MA, USA) and quantified by IMAGEQUANT TL 8.1 software (GE Healthcare Life Sciences). To detect the RNA fragment produced by mtVapC20 cleavage, the following primers were used that were specific to the 3 0 -end of 16S mt-rRNA: SRLdown-For 5 0 -TAAGGCCTACTTCACAAAGCG-3 0 ; SRLdown-Rev 5 0 -CTTGGGTGGGTGTGGGTATAA-3 0 .

De novo mitochondrial protein synthesis
Metabolic labelling of de novo mitochondrial protein synthesis in cultured cells was performed after 2 and 6 days of mtVapC20 induction and carried out as described by [40] incubating cells with [ 35 S] methionine/cysteine (Perkin Elmer) for 1 h. Aliquots (50 µg) of total cell protein were separated by 15% SDS/PAGE and radioactive signals were detected as described above.

In organello RNA synthesis
The transcription assay was performed on mitochondria isolated from uninduced and mtVapC20-expressing cells (6 days of induction) as previously described [41] with few modifications. Briefly, mitochondria (320 µg) were incubated with 10 µCi of [a-32 P]UTP (800 CiÁmmol À1 ; Perkin Elmer) for 1 h at 37°C with gentle rotation. Mitochondria were then washed twice in washing buffer (10 mM Tris/ HCl, pH 6.8, 0.15 mM MgCl 2 and 10% glycerol) and pelleted by centrifugation at 12 000 g for 3 min at 4°C. Pellets were resuspended in 1 mL of TRIzol Ò Reagent (Thermo Fisher Scientific) and RNA was extracted.
To analyse the products of the in organello transcription, RNA was fractionated on a 2.2 M formaldehyde/1.4% agarose gel and stained by SYBR TM Gold (Thermo Fisher Scientific) to validate loading. The gel was dried and exposed to a PhosphorImager screen, and radioactive signals were detected as described above.

Mitochondrial rRNA FISH
Cells capable of inducibly expressing mtVapC20 were seeded onto coverslips in a 6-well plate, induced with 1 µgÁmL À1 tetracycline or left uninduced as controls. At 20-30% confluency, cells were fixed using 4% PFA/0.2% glutaraldehyde (10 min, room temperature) and washed twice in 19 PBS. Cells were then permeabilized (70% ethanol, 1 h, room temperature), washed twice with 19 PBS and equilibrated in 1 mL of wash buffer (10% formamide, 29 SSC) for 5 min. 16S-and 12S mt-rRNA Stellaris Ò Probe Sets (Biosearch Technologies, Novato, CA, USA) were diluted in TE Buffer (10 mM Tris/HCl, 1 mM EDTA pH 7.5) to 12.5 µM, then added 1 : 100 to hybridization buffer (10% formamide, 10% dextran sulfate, 29 SSC). Droplets (100 µL) of probe-hybridization buffer solution were placed onto parafilm in a humidified chamber on to which coverslips were lowered and incubated overnight at 37°C in a humidified atmosphere. Coverslips were rinsed, incubated twice in wash buffer (30 min), then stored in 19 PBS until mounted in ProLong Ò Diamond Antifade Mountant. Confocal imaging was performed using a SP8 microscope (Leica, Wetzlar, Germany) fitted with a Leica HyD hybrid detector and IMAGEJ software (NIH, Bethesda, MD, USA) was used to quantify fluorescence intensity. Each channel was analysed independently from maximum intensity projections generated from acquired stacks. To reduce background interference during intensity calculations, an area was segmented around the high-intensity signal using thresholding. In all cases, the area was determined by thresholding the 12S mt-rRNA image.
IMAGEJ was then used to measure the mean intensity for both channels using this area. The measurements were analysed using GRAPHPAD PRISM8 software (San Diego, CA, USA).