Large heterogeneity of mitochondrial DNA transcription and initiation of replication exposed by single-cell imaging

Mitochondria are ATP-producing organelles whose function is directed not only by the nuclear genome but also by their own genome. Each mitochondrion carries several copies of a circular double-stranded DNA that is replicated and transcribed autonomously in the organelle (Bonawitz et al., 2006; Falkenberg et al., 2007; Scarpulla, 2008). Mitochondrial DNA (mtDNA) is arranged in nucleoprotein complexes, nucleoids, that include factors involved in replication and transcription as well as structural proteins required for mitochondrial maintenance (Chen and Butow, 2005; Spelbrink, 2010). These proteins include DNA polymerase γ (Polγ), the enzyme responsible for replication of mtDNA, and TFAM (also known as mtTFA), a protein implicated both in transcription of and in binding to the mtDNA, and whose levels are correlated with those of mtDNA (Falkenberg et al., 2007; Poulton et al., 1994; Shutt et al., 2010). TFAM has been recently found to act more as a transcription activator than as a core component of the transcription machinery in vitro (Shutt et al., 2010). Human mtDNA, a 16.5 kbp molecule, is organized with two rRNA, 22 tRNAs, and 13 protein-coding genes that are transcribed from the (heavy) H-strand (12 mRNA, two rRNA and 14 tRNAs) and from the (light) L-strand (one mRNA for ND6, and eight tRNAs) with production of polycistronic precursor RNAs. These primary transcripts are processed to produce the individual mRNA, rRNA and tRNA molecules (Ojala et al., 1981).

The prevalent view of mtDNA replication is that DNA synthesis starts from origin O_H where the nascent H strand frequently terminates 700 bp downstream giving rise to the 7S DNA, which produces a characteristic triple stranded structure, the D-loop (Chang and Clayton, 1985; Clayton, 1991). When leading strand synthesis has reached two thirds of the genome, it exposes another major origin, the origin of L-strand DNA replication (O_L), and lagging-strand DNA synthesis then initiates in the opposite direction. However, coupled O_H and O_L strand DNA synthesis has been described in only a subset of mtDNA molecules (Holt et al., 2000) suggesting that this model is not fully elucidated. Mitochondria display a variety of shapes ranging from highly interconnected tubular structures to individual small spherical units. These structures, which are highly dynamic, are regulated by mitochondrial fusion and fission, and they vary during cell growth (Chan, 2006; Lee et al., 2007; Mitra et al., 2009). Whether these different structures are related to the processing of mtDNA (transcription and replication) needs clarification.

The processing of mitochondrial DNA has been intensively analysed with biochemical approaches (reviewed by Falkenberg et al., 2007; Scarpulla, 2008) which essentially examined global cellular and mitochondrial populations. However, little is known about the dynamics and the regulation of mtDNA transcription and replication in a single cell, as well as the implication of these processes in cell function. Indeed, currently available fluorescence in situ hybridization (FISH) tools including recent improvements (Alán et al., 2010; Ozawa et al., 2007), are not fit to identify mitochondria engaged in DNA replication, and to discriminate distinct transcription profiles of the organelles. Conversely, incorporation of bromodeoxyuridine (BrdU) may label a large fraction of organelles (Echave et al., 2009), cumulating replication events that started at different times, and thereby limiting the informative potential on mtDNA replication dynamics within the mitochondrial population. Importantly, higher resolution studies have recently shown that replication of mtDNA occurs only in a subset of nucleoids (Kukat et al., 2011), underscoring the need of understanding the properties of mitochondrial replication dynamics in single cells. In this context, recent approaches based on super-resolution imaging involving mitochondrial proteins provided insightful information on the organization of nucleoids (Brown et al., 2011; Kukat et al., 2011). In spite of these advancements, and although sequential RNA and DNA labelling (Masny et al., 2010), as well as labelling of either RNA or DNA, and proteins (Arabi et al., 2005; de Planell-Saguer et al., 2010) have been performed, immunofluorescence is actually not directly coupled to FISH to simultaneously allowing the triple detection of proteins and mitochondrial DNA and RNA. Thus, proteins of interest can be hardly monitored during mtDNA transcription and replication. As a consequence, it remains unclear how mtDNA processing is coordinated among the many organelles present in each cell and to what extent this process is deregulated during disease.

Using a novel imaging protocol based on FISH, we identify mitochondrial subpopulations engaged in the initiation of mtDNA replication and in RNA processing, and assess their dynamics in single cells. Our findings reveal significant heterogeneities within single cells that have been missed previously, and show that these functions are altered upon mitochondrial dysfunction, including in diseases.

To gain insight into the dynamics of mtDNA and mtRNA within the organelle, we developed a novel protocol that labels simultaneously mtDNA and mtRNA in cells by improved FISH, and performed 3D confocal acquisitions. This approach, called Mitochondrial Transcription and Replication Imaging Protocol (mTRIP) is a combination of DNA FISH and RNA FISH techniques, and it limits the use of potentially damaging agents for macromolecules. Moreover, in contrast to other protocols, proteins are not destroyed during this treatment, as evidenced by epitope conservation of tested mitochondrial proteins during permeabilization and incubation with formamide (steps 1+2), denaturation in formamide (step 3), and the complete mTRIP procedure, Fig. 1A–D. Hence, FISH can be coupled to immunofluorescence, and it is possible to quantitatively monitor mitochondrial DNA, RNA and proteins simultaneously. The mitochondrial porin VDAC is frequently used for assessing the mitochondrial mass (Acquaviva et al., 2005), but its presence elsewhere than in mitochondria (De Pinto et al., 2010, see also supplementary material Fig. S1) led us to use TOM22 instead (Fig. 1E). TOM22, is a subunit of the mitochondrial outer membrane translocase (Yano et al., 2000) which is uniformly distributed in mitochondria, and was previously used as an indicator of mitochondrial mass (Latil et al., 2012). In this context, mitochondria are visualised as individual units or structured in the interconnected mitochondrial network. Co-labelling of mTRIP probes with TOM22 immunofluorescence was performed in this study to assess the distribution of FISH labelling in mitochondria.

We observed that mTOT, a mixture of 14 probes that cover the entire mitochondrial genome (Table 1; supplementary material Fig. S2, Table S2) labelled only a fraction of the mitochondrial network (TOM22 immunostaining) in human cells (Fig. 1F). Partial labelling of mitochondria was expected since mtDNA is normally present only in nucleoids, and mTOT indeed marked small structures within the mitochondrial mass, likely nucleoids. However, large portions of the mitochondrial network and entire TOM22-positive areas were not labelled with mTOT (mTOT/TOM22 colocalization is 23.99±1.73%), suggesting that not all nucleoids were marked with mTOT (see also below, co-labelling with nucleoid markers). Treatment of cells with either DNaseI, RNaseA or RNaseH, and combinations of these specific nucleases before hybridization with the mTOT probes showed that labelling was specific essentially for RNA molecules (Fig. 1G,H; supplementary material Fig. S3). Higher intensity of fluorescence was observed in the presence of RNaseH (1.4-fold), suggesting that removal of the RNA moiety from RNA/DNA hybrids renders DNA available for pairing with fluorescent probes. These hybrids probably correspond to transcripts bound to their DNA template. Treatment with RNaseH, then with DNaseI, restored fluorescence levels of untreated cells (Fig. 1G,H), confirming that the DNA portion of RNA/DNA hybrids does pair with the fluorescent probe, after disruption of the RNA moiety. The latter observation, and the apparent absence of effect of DNaseI treatment indicate that mtDNA is available in limited amounts for mTRIP, at least in the double-strand conformation. Treatment with proteinase K prior to mTRIP resulted in a large increase in the signal (154% for mTOT, 206% for mTRANS and 202% for mREP, Fig. 1I,J; the last two probes recognize RNA and DNA, respectively, see below) compared with untreated cells, indicating that some mtRNA and mtDNA were inaccessible to the probes because they could be bound to, or masked by proteins. In spite of this increase in signal intensity, proteinase K treatment was avoided here because mTRIP was frequently coupled to immunofluorescence for the detection of proteins.

We then performed mTRIP with each of the 14 probes defined as mTOT to examine mitochondrial transcription. Interestingly, we observed that each probe recognized a specific subset of structures (Fig. 2A), and that not only the intensity but also the distribution of the labelling varied as a function of the mtDNA region tested. Moreover, some probes (e.g. 1–4) essentially displayed a diffused labelling, whereas other probes also (e.g. 5 and 6) or prevalently (e.g. 9–12) labelled distinct foci. The foci size and number is heterogeneous, although the most intense foci on each sample are relatively constant in number (20.88±1.38 and 23.55±0.5 for probes 9 and 11, respectively). These data suggest that only a fraction of mitochondrial structures carries detectable amounts of the target nucleic acid, and that mitochondria may not be functionally equivalent in activity. Treatment with DNaseI or RNaseA showed that all of the probes recognized essentially RNA targets, with the exception of probes 4, 8 and 13, which also recognized DNA (Fig. 2B). The prevalent transcript labelled by mTRIP is 16S rRNA, which was confirmed as the most abundant mitochondrial transcript by RT-qPCR (supplementary material Fig. S4A). RT-qPCR also showed that transcript levels of mitochondrial genes were generally consistent with RNA levels identified by mTRIP (supplementary material Fig. S4A,B, Table S2), thus validating the FISH data.

To determine the relative proportions of each of the transcripts within the mitochondria, we co-labelled TOM22 by immunofluorescence with the individual probes by FISH. This analysis confirmed that 16S rRNA was present in a larger fraction of the mitochondrial network (supplementary material Fig. S5A,B), as well as it was the highest in signal intensity correlating with TOM22 labelling (supplementary material Fig. S5C), compared with the other transcripts. In addition, experiments of co-labelling with two or more probes indicated that mTRIP labels both processed and unprocessed polycistronic transcripts. Indeed, the extensive colocalization (Fig. 3A,B) and the high levels of mTRIP labelling with probe 2 and probe 3, as well as with probe 7 and probe 9 (Fig. 3D,E), suggest that the RNAs recognized by these probe pairs, which label successive genes on the H-strand, are located on the same submitochondrial structures, and might thus correspond to unprocessed transcripts. Conversely, probe 3 and probe 4 (Fig. 3A,B), as well as probe 9 and probe 10 (Fig. 3D,E), which also label successive genes on the H-strand, are essentially located on different submitochondrial structures, and likely correspond to distinct processed transcripts. In colocalization experiments, although confocal fluorescence microscopy does not allow defining whether the target nucleic acids are located on the same molecules (the limit set on the spatial resolution by diffraction of the light is 200 nm (Isobe et al., 2010), these data nevertheless show that the nucleic acids co-labelled by the selected probes are spatially and quantitatively linked. Interestingly, these co-labelling experiments also showed that about half of mitochondrial entities labelled with probe 14, which recognizes 12S present at beginning of H-strand, colocalize with probe 12, which recognizes CytB present at end of H-strand, indicating that these mitochondria might contain the complete H-strand transcript (Fig. 3D–F).

To analyse further mitochondrial transcripts, a new mixture of probes was used (mTRANS: probes 1, 6 and 11) that are distributed evenly along the circular genome (supplementary material Fig. S6A). These probes label 16SrRNAs and several tRNAs and mRNAs, and they do not recognize regions involved in the initiation of DNA replication (see below). FISH experiments with mTRANS, in the absence or presence of nucleases, confirmed that this probe set detects only RNA (see below). The nucleoid markers TFAM and Polγ label submitochondrial structures, as shown in supplementary material Fig. S6B,C. Extensive colocalization between immunostaining of either Polγ or TFAM and mTRANS (supplementary material Fig. S6D) indicated that transcripts detected by this probe mix were mostly confined to labelled mitochondrial nucleoids, as suggested above. The different levels of colocalization between FISH probes and nucleoid markers might be linked to heterogeneity of nucleoids (DNA and protein content).

We observed above that three probes (4, 8 and 13) detect not only RNA but also DNA. Interestingly, probes 13 and 4 include the regions of initiation of replication of the H- and the L-strand, respectively, suggesting that these probes detect DNA regions engaged in the initiation of replication. Probe 8 includes the ND4 region, where an additional origin of replication for the L-strand (O_ALT) has been reported using atomic force microscopy (Brown et al., 2005), and which is expected to be activated less frequently than the two major origins. These three origins are distributed evenly on the mitochondrial genome, as schematized in Fig. 4A. Labelling with probe mREP, which is a subset of probe 13 (Fig. 4B,C), was DNaseI-sensitive and RNaseA- and RNaseH-resistant (Fig. 4D) indicating that this probe labels DNA specifically. Conversely, Fig. 4D shows that probe mTRANS labels mRNA specifically.

Since the DNA region recognized by mREP is present in the genome of all mitochondria, we reasoned that the local DNA structure, likely the adjacent O_H origin of replication, rendered it accessible to the probe. To verify this point, we assessed whether mREP is associated with nucleoids that contain factors involved in DNA replication and transcription. mREP labelling coupled to immunofluorescence with Polγ or TFAM showed that this was the case (74.4±2.5% colocalization of mREP with Polγ, and 71.7±1.5% with TFAM, Fig. 4E), and also that most of the mREP foci were preferentially localized to Polγ and TFAM-rich areas (Fig. 4F).

To assess whether mREP was associated with mtDNA replication, we checked BrdU incorporation, an indicator of DNA replication, in mitochondria. Proliferating cells were treated with 10 µM BrdU for 30 min, 90 min, 3 h, 6 h, 9 h and 18 h prior to FISH labelling (supplementary material Fig. S7A). Long-time exposure of cells to BrdU (i.e. 18–24 h) is generally used to detect full mtDNA replication (Echave et al., 2009). Short-time exposure to the drug was tested here to assess the kinetics of mtDNA replication, and early replication events. We observed that in spite of the BrdU labelling increases with the time of treatment, as expected because of the accumulation of labelled mitochondrial and nuclear DNA (nDNA), the extent of mREP/BrdU colocalization remains relatively constant, at least until 9 h of treatment (supplementary material Fig. S7B). This result indicates that mREP does not co-label mitochondrial structures carrying progressively longer mtDNA chains but rather a specific subset of mtDNA replication events at each time point, which are compatible with replication initiation events. Most mREP-positive structures are not labelled with BrdU, compatibly with an open mtDNA structure that has not incorporated BrdU or that has incorporated BrdU below detection levels (see schema in supplementary material Fig. S7C). Interestingly, after 18 h treatment, at constant mREP labelling, 80.4±2% of mREP positive structures colocalize with BrdU, indicating that under these conditions most of the open mtDNA structure located nearby the replication origin are productively engaged in initiation of mtDNA replication. As for above, the remaining 19.6% of mREP foci are compatible with an open mtDNA structure that is engaged in replication where little or no DNA synthesis has occurred.

Importantly, at all time points, mREP is detected in only a subset of BrdU-positive foci, i.e. a subset of structures engaged in DNA replication, in agreement with the notion that this probe does not detect extensive or completed replication, but rather the initiation of DNA replication. In this context, the intensity of BrdU labelling, evaluated after 18 h of BrdU treatment and taking into account BrdU-positive areas located extranuclearly, was lower in mREP-positive compared with mREP-negative areas (see isolated mitochondria located outside the nuclear region, and quantification, supplementary material Fig. S7D), in agreement with the limited incorporation of a nucleotide analogue at the beginning of replication of the mitochondrial genome. Finally, we observed that mREP signal increased when the mtDNA content was reduced, and returned to control values when the original mtDNA content was restored (treatment with low levels of H₂O₂, supplementary material Fig. S7E). Taken together, these results, and the unique characteristics of the region of the mtDNA recognized by mREP, support the notion that mREP marks initiation of replication.

Whether DNA synthesis detected by mREP proceeds from O_H until the end of the H-strand, or terminates earlier, leading to the formation of the 7S strand and thereby of the D-loop (Chang and Clayton, 1985; Clayton, 1991), was not resolved by FISH labelling alone. To elucidate this point, we compared endogenous levels of mtDNA and of 7S DNA detected by real-time qPCR, as described (Antes et al., 2010) in cells with affected levels of mtDNA (treatment with low levels of H₂O₂, see above). We found that the variations observed in the mtDNA content after exposure to H₂O₂, evaluated in the 12S region, and associated with changes in mREP levels (supplementary material Fig. S7E) are compatible with variations of the mtDNA and not of 7S DNA (supplementary material Fig. S7F). Thus, although mREP may label both the productive and the abortive initiation of mtDNA replication (formation of the D-loop), variations in mREP are compatible with productive replication of the mtDNA rather than with the formation of the D-loop.

To assess the fraction and the distribution of mtDNA processing activities within the mitochondrial network we performed co-labelling with mREP, mTRANS, and TOM22 (Fig. 4G). Notably, 58.9±2.7% and 12.9%±1.3% of the mitochondrial mass (TOM22 immunolabelling) co-labelled with mTRANS and mREP, respectively, revealing that most mitochondrial structures carry detectable transcripts (red label, Fig. 4G), whereas only a small fraction of these structures is engaged in initiation of mtDNA replication (blue label, Fig. 4G). A significant fraction of the mitochondria were not labelled with either probe indicating that either they are not involved in the transcription of the tested genes and/or in the replication of mtDNA, or that the levels of the target molecules are not detectable with this approach (replication initiation silent and transcript-negative mitochondrial structures, Is-Tn, or mREP-negative and mTRANS-negative, green label, Fig. 4G).

Moreover, the majority (71.3±2.9%) of mitochondrial structures involved in the initiation of replication also carried mTRANS transcripts (replication initiation active and transcript-positive mitochondrial structures, Ia-Tp, or mREP-positive and mTRANS-positive, purple, labelling). On the contrary, the majority (91.5±0.8%) of mitochondrial structures that carried detectable transcripts were not involved in initiation of mtDNA replication in these cells (replication initiation silent and transcript-positive mitochondrial structures, Is-Tp, or mREP-negative and mTRANS-positive, orange, labelling). The proportion of these distinct structures within the mitochondrial network is shown on the right-hand panel in Fig. 4G. Thus, we identify at least three classes of mitochondrial structures with distinct mtDNA processing activities within a single cell.

We reasoned that since mTRIP identifies distinct mitochondrial structures within single cells according to the DNA engaged in initiation of replication and to the transcript content, it should also identify mitochondrial structures with distinct RNA and DNA labelling patterns in the regulatory region, which might be functional to the regulation of mtDNA itself. To assess this point, we performed FISH with three probes located in the D-loop region (probes PL-OH and 7S) and at promoters of the H-strand (P_H1 and P_H2; probe PH1–2), see schema Fig. 5A. We performed single labelling, and colocalization experiments with two probes, in the presence and in the absence of nucleases.

Quantitative analyses of fluorescence showed that probe PL-OH, located downstream of mREP in the direction of DNA replication, and which includes the P_L and the O_H regions, labels accessible DNA structures (DNaseI-sensitive labelling, Fig. 5B,C), which might be part of the P_L transcription bubble, or of the O_H replication bubble, or of both. This last possibility is in agreement with a large set of evidences indicating that transcription from P_L is coupled to H-strand replication (Scarpulla, 2008). Probe PL-OH also labels RNA in RNA/DNA hybrids (reduction of labelling in the presence of RNaseH, Fig. 5B,C). These RNA molecules might consist of R-loops, i.e. the RNA primers for DNA synthesis for the O_H origin (Brown et al., 2008), present as processed as well as unprocessed molecules of various lengths, or regular L-strand transcripts, or both. Interestingly, simultaneous treatment with DNaseI and RNaseH resulted in residual labelling (19.9%±1.4%) that disappeared when cell were also treated with RNaseA (Fig. 5C). This experiment indicates that probe PL-OH also labels RNaseH-resistant RNA/DNA hybrids, which become accessible to RNaseA after the DNA moiety is removed by the action of DNaseI. This finding is in agreement with the notion that the structure of the RNA/DNA hybrids in R-loops may confer resistance to RNaseH (Brown et al., 2008). Increase in the intensity of labelling in the presence of RNaseA indicates not only that RNA is not a significant target of PL-OH but also that RNA molecules to a certain extent inhibit the labelling of other targets.

The labelling of nucleic acids appears rather different in the downstream 7S region (probe 7S). Indeed, treatment with nucleases indicates that probe 7S recognizes prevalently RNA (Fig. 5B,C). Probe 7S also targets RNA/DNA hybrids, since treatment with RNaseA did not reduce the FISH signal to background levels, and simultaneous treatment with RNaseH and RNaseA resulted in significant decrease of the signal, compared with RNaseA alone (P = 0.0013).

More detailed information on the heterogeneity of the nucleic acids composition of the D-loop region in mitochondria was provided by the direct observation of foci. Indeed, PL-OH and 7S labelling consist of small foci with poor fluorescence intensity as well as of large foci of intense fluorescence, and distinct colocalization patterns with the other probe (arrows with numbers, Fig. 5B). Fluorescence intensity measurements in Fig. 5C take in to account all foci; large foci were separately enumerated and further examined. Analysis detailed in supplementary material Fig. S8 identifies the nucleic acid composition of several labelling patterns associated with large PL-OH and/or 7S foci, and shows the large heterogeneity of mTRIP labelling in this region. The existence of these labelling patterns, and the proportions of these events are indicative of qualitatively and quantitatively distinct mtDNA processing activities in the D-loop which occur simultaneously within single cells.

Although the limits of resolution of FISH do not discriminate whether labelled nucleic acids are present on the same molecules (or on the same structures), several observations support the notion that most large PL-OH and 7S foci (patterns 1, 3–5) label related entities. Indeed, these foci appear simultaneously affected by treatment with RNaseA alone, and combined treatment with RNaseA and RNaseH (Fig. 5B; supplementary material Fig. S8A). Additionally, large PL-OH and 7S foci disappeared after simultaneous treatment with RNaseH and DNaseI (whereas treatment with DNaseI alone had essentially no effect, and treatment with RNaseH affected only PL-OH foci). This experiment indicates that RNA/DNA hybrids, at least at the level of 7S, are essential to the formation of the large structures labelled not only by probe 7S but also by probe PL-OH, thereby further supporting the notion that the nucleic acids labelled by the two probes are linked. Moreover, after treatment with two nucleases, large foci were replaced by smaller foci of variable size, which in most cases did not colocalize (Fig. 5B), indicating that, differently from large foci, the target nucleic acids are no longer present in linked structures. Importantly, nucleic acids in this regulatory region were accessible in only a limited fraction of mitochondria (co-labelling of either PL-OH or 7S with TOM22, not shown), revealing a further heterogeneity of mitochondrial structures in single cells for the PL-OH and 7S target regions.

Finally, colocalization experiments revealed that not only PL-OH colocalizes by 99.44%±0.05% with 7S, but also that mREP colocalized with 7S by 99.33%±0.07% indicating that mREP, PL-OH and 7S likely label linked, although heterogeneous, nucleic acid structure(s) (results are integrated in Fig. 5G). Conversely, only a fraction of 7S colocalizes with mREP (59.8%±2.6%), and with PL-OH (69.7%±2.7%), indicating that 7S does not label only RNA involved in the replication bubble, compatibly with labelling of L-strand transcripts. In agreement with this notion, the intensity of PL-OH labelling was 2.3-fold higher than mREP labelling (Fig. 5G).

In conclusion, probes PL-OH and 7S identify a fraction of mitochondria carrying a variety of linked structures with distinct nucleic acid composition that appear associated at different levels with O_H DNA replication and L-strand transcription and that co-exist in single cells.

On the other side of mREP, probe PH1–2, which is located in the region of promoters P_H1 and P_H2, was fully sensitive to DNaseI, indicating that it labels essentially DNA (Fig. 5D). The accessible DNA in this region might result from its opening as a consequence of the nearby O_H replication bubble, or represent the P_H1/P_H2 transcription bubble. The overwhelming colocalization between probe PH1–2 and probe 1, which labels 16S RNA transcribed essentially from P_H1, (Fig. 5E), and the elevated intensity of fluorescence observed with both probes support the second notion. Treatment with either RNaseA or RNaseH does not significantly alter the efficiency of PH1–2 labelling, although foci display a different aspect compared with untreated controls (Fig. 5D), indicating that although RNA and RNA/DNA hybrids do not appear to bind probe PH1–2, they might affect the structure of the DNA to which the probe binds. A summary of the prevalent nucleic acids detected in the regulatory region of the mtDNA is shown in Fig. 5F.

A large difference in the extent of labelling among probes located in the D-loop was noted. PH1–2 fluorescence intensity was 6.9-fold higher than mREP, and probe 1 fluorescence intensity, which marks the 16S transcript, was 10-fold higher than mREP, compatibly with robust transcription of rRNAs (Fig. 5G). Simultaneous labelling with two probes reveals that 48.73%±3.44% of mREP foci colocalize with PH1–2 foci, indicating that accessible DNAs in the two regulatory regions might be linked, although in other cases these foci do not colocalize, compatibly with the notion that transcription of rRNA can be uncoupled from the formation of the DNA structure that promotes replication in the O_H region.

To assess whether mTRIP detects alterations of DNA processing in cells with mitochondrial perturbations we examined HeLa rho⁰ cells where mtDNA is lacking (Parfait et al., 1998), and HeLa cells treated with ethidium bromide (EtBr) for three days to reduce their mtDNA content (King and Attardi, 1996). Notably, HeLa rho⁰ cells contained about one third of the mitochondrial mass (TOM22 immunolabelling, Fig. 6A,C; see also supplementary material Fig. S1A,C,D) compared with regular HeLa cells, but no signal was detected with either mTRANS or mREP, confirming the absence of mtDNA transcription and initiation of replication in these cells (Fig. 6B). Rho0 cells did not show detectable BrdU signal in extranuclear areas (Fig. 6D), as expected because of the lack of mtDNA. In contrast, cells treated with EtBr, which had a reduced mtDNA content, maintained a regular mitochondrial mass and displayed a 9.3-fold and a 5.9-fold increase in the levels of mREP and mTRANS, respectively, compared with untreated cells. High levels of transcripts were confirmed by RT-qPCR of mitochondrial 16S rRNA and CytB. These data indicate that in spite of mtDNA depletion, EtBr-treated cells dramatically increased their mtDNA replication and transcription activities, likely to compensate for the low DNA content, in agreement with previous studies on transcription in these treated cells. (Seidel-Rogol and Shadel, 2002).

Finally, we analysed primary fibroblasts mutated in RRM2B, the p53-inducible ribonucleotide reductase subunit which is essential for mtDNA synthesis and is associated with mtDNA depletion syndrome (Bourdon et al., 2007). We found that Rrm2b fibroblasts in spite of a 44% reduction in the mitochondrial mass (TOM22; see also VDAC and MitoTracker® labelling; Fig. 7C,D; supplementary material Fig. S9) display a 4-fold reduction in mREP and a 3-fold reduction in mTRANS signals compared with normal fibroblasts (Fig. 7A,B). Moreover the number of foci with mREP labelling decreased from 28.44±0.96 in normal fibroblasts to 11.55±1.38 in mutant fibroblasts. Furthermore, although normal and Rrm2b fibroblasts maintain relatively constant mREP as well as mREP+/BrdU+ signal at increasing exposure to BrdU, the extent of both signals is dramatically reduced in mutant fibroblasts, indicating that Rrm2b cells carry a severely reduced fraction of mitochondrial structures engaged in initiation of mtDNA replication (Fig. 7E,F).

Taken together, the analysis of three different cell types showed that mREP and mTRANS labelling identify altered or loss of mtDNA processing, which affects mitochondrial function, thus validating mTRIP for monitoring disease states both qualitatively and quantitatively.

Mitochondria play a crucial role in eukaryotic cells, therefore understanding the dynamics of DNA transcription and replication within the mitochondrial network, and possibly within mitochondrial substructures, is important to assess mitochondrial function (Brown et al., 2011; Chen and Butow, 2005; Kukat et al., 2011; Spelbrink, 2010). We describe here a novel FISH protocol, mTRIP, which identifies an unexpected variety of mitochondrial structures and mitochondrial populations with distinct properties within single cells. These populations differ in their intracellular localization, in the relative amount of transcripts that they express, in the initiation of DNA replication, and in the intensity and type of signal in regulatory regions, indicating that mitochondria and mitochondrial structures exhibit a greater level of heterogeneity in DNA processing activities than reported previously, including mitochondrial dynamics during mtDNA synthesis (Davis and Clayton, 1996). Within the limits of resolution of this approach, which relies on confocal microscopy, and not on super-resolution microscopy (Brown et al., 2011; Kukat et al., 2011), labelling of mtDNAs and RNAs was also shown to be correlated with nucleoids, the mitochondrial substructures involved in mtDNA processing. We observed different levels of colocalization between FISH and nucleoid markers, in agreement with the different amounts of regulatory proteins found in nucleoids and which might have regulatory functions (Chen and Butow, 2005; Shutt et al., 2011; Spelbrink, 2010).

mTRIP detects mitochondria and mitochondrial substructures rich in a given transcript. In addition to processed transcripts, mTRIP also appears to detect polycistronic RNAs, including signals compatible with transcripts of the complete H-strand, as well as RNA bound to the DNA template, the latter likely resulting from ongoing transcription. In general, we found a good correlation between RNA levels detected with mTRIP and RT-qPCR, thus validating the FISH approach described here which then allows assessing mitochondrial transcripts within the mitochondrial network, and in individual cells.

Labelling of DNA by mTRIP appears limited to locally open structures, as in transcription complexes after disruption of the RNA moiety, as well as in the transcription bubble at promoters, and in DNA engaged in initiation of replication. To date, identification of mitochondrial initiation of replication in single cells has been elusive. Importantly, mTRIP identifies a unique DNA region, targeted by mREP, that is distinct from but associated with the principal origin of replication O_H in humans. mTRIP also recognizes regions containing accessible DNA at the level of the O_L origin as well as a third mitochondrial replication origin expected to be activated only occasionally (Brown et al., 2005).

Importantly, at least three classes of mitochondrial structures are detected: (i) replication initiation active and transcript-positive (Ia-Tp), (ii) replication initiation silent and transcript-positive (Is-Tp), and (iii) replication initiation silent and transcript-negative (Is-Tn). The presence of these distinct structures reveals a large heterogeneity of mitochondria in single cells. In proliferating HeLa cells, Ia-Tp represent less than 15%, Is-Tp more than one half, and Is-Tn about one third of detected mitochondrial structures. mTRIP analysis may help in elucidating the link between mitochondrial DNA (i.e. ratio of Ia-Tp versus Is-Tp, and versus Is-Tn) and mitochondrial function. Moreover, the proportion of mitochondrial structures engaged in specific mtDNA processing activities appear dynamic, as they vary under stress and certain physiological conditions (Chatre and Ricchetti, 2013) and they may represent a significant parameter to take into account when assessing mitochondrial function.

Our data also highlight the heterogeneity of nucleic acids present at regulatory loci in the D-loop. In this context, mTRIP analysis identified DNA, RNA and RNA/DNA hybrids at the expected locations according to current knowledge on global mitochondrial populations (Chang and Clayton, 1985; Clayton, 1991; Falkenberg et al., 2007; Scarpulla, 2008), thus further validating this approach, which however operates in single cells. Indeed colocalization between probes pairs revealed accessible DNA, upstream of, and within the replication origin as well as at the level of promoter P_L on one side, and of promoters P_H1 and P_H2 on the other side. Accessible DNA appears the exclusive target upstream of O_H (from the mREP region to PH_1–2). Conversely, RNA is the almost exclusive target in the 16S region, as expected for P_H1 transcripts, and a major target in the 7S region where it probably consists of L-strand transcripts. RNA/DNA hybrids are detected at P_L promoter compatibly with the formation of R-loops that provide the RNA primers for DNA replication, and also at a minor extent at 7S where they may represent L-transcripts bound to the DNA template. Importantly, the extent of colocalization among probes reveals comparable levels of labelling for the region mREP, that signals initiation of mtDNA replication, and the region that includes the O_H replication origin as well as the downstream 7S transcript. This observation is in agreement with the notion that L-strand transcription and replication are coupled (Scarpulla, 2008). Our data suggest that these two processes are not only temporally but also quantitatively linked.

In addition to these observations, our experiments provide novel information on the dynamics of some key regulatory regions of the mtDNA. In this context, accessible DNA in mREP, which is associated with O_H replication and L-strand transcription, appears also linked to H-strand transcription (for rRNAs) since about one half of mREP and PH1–2 signals colocalize. Importantly, however, lack of colocalization between the two probes indicates that in the other half of cases rRNA transcription is uncoupled to O_H replication. This result also underscores the region labelled by mREP, which is located close to key regulatory elements (P_L, P_H1–2, O_H) but does not contain significant regulatory elements itself. mREP nevertheless appears as a region whose accessibility acts as indicator of the dynamics of mtDNA replication and transcription of both the H- and the L-strand.

Moreover, transcription of the L-strand appears different in labelling from transcription of the H-strand (16S rRNA) in their respective promoter regions since RNA/DNA hybrids, DNA, and RNA are detected in the former, and essentially RNA in the latter. Such difference could be associated with the lower transcription rate of the L-strand compared with rRNAs, or with the presence of structures linked to O_H replication. It is also possible that RNA/DNA hybrids and/or single-strand RNAs are functional to the transitory 7S triple strand structure, or else that 7S RNA labelled here has links with the specific RNA detected in immortalized cells (Duncan et al., 2000). A further difference between the two strands is the almost exclusive detection of accessible DNA at promoter P_H1 whereas RNA/DNA hybrids are prevalently present at promoter P_L. Since the open DNA structure at P_H promoters is detected in the large majority of mitochondria, and this is also the case for the P_H1 transcript 16S rRNA, transcription of rRNA provides the most abundant source of accessible DNA in the mitochondrial genome, at least detected by FISH.

Notably, only a limited fraction of mitochondria are recognized by probes that target the D-loop regulatory region (with the exception of P_H1–2 promoters). Non-labelled mitochondria either do not carry accessible DNA and/or RNA/DNA hybrids and/or L-transcripts in the respective regions, or the levels of these nucleic acids are below the threshold of detection. In both cases mTRIP reveals broad mitochondrial heterogeneity for the nucleic acid composition of the regulatory region of the mtDNA, which might have regulatory function.

In summary, by mTRIP the dynamics of mtDNA transcription and initiation of replication are exposed with unprecedented resolution at the single-cell level, and further classes are revealed when the D-loop regulatory region is dissected. Thus, mTRIP analysis may help in elucidating the link between mitochondrial DNA and the energizing state of individual cells, in particular in the context of diseases, and this approach may be used for future investigations of mtDNA processing under physiological and pathological conditions.

In this context, we also examined cells depleted in mtDNA, as is the case for several diseases (Rötig and Poulton, 2009), for example, Rrm2b fibroblasts, carrying a mutation that is associated with an mtDNA depletion syndrome (Bourdon et al., 2007). Here, reduced mtDNA transcription and replication were observed using mTRIP. In addition, we noted dramatically increased mitochondrial transcription and replication signals in cells with depleted mtDNA content following treatment with EtBr. This situation is compatible with the normal amounts of mitochondrial transcripts observed in HeLa cells depleted of mtDNA by EtBr treatment (Seidel-Rogol and Shadel, 2002). Moreover, it is likely also representative of muscle from patients with a particularly severe mtDNA depletion, which nevertheless displayed steady-state levels of mitochondrial transcription and had a surprisingly slow progression of the disease compared with other mtDNA depletion syndromes (Barthélémy et al., 2001). Thus, mTRIP reveals qualitative and quantitative alterations, which provide additional tools for elucidating mitochondrial dysfunction in diseases.

Human HeLa and HeLa rho⁰ cells, IMR90 fibroblasts (purchased from ATCC), and Rrm2b fibroblasts were grown in MEM medium with 10% foetal bovine serum (FBS) at 37°C and in the presence of 5% CO₂. HeLa rho⁰ cells and Rrm2b fibroblasts were a kind gift from Dr Agnes Rötig (Genetics of mitochondrial disorders, Hôpital Necker, Paris). Cells under low oxidative stress were treated with 50 µM H₂O₂ for the time indicated. For mtDNA depletion, HeLa cells were treated with 25 µM EtBr for 3 days (King and Attardi, 1996).

Cells plated on slides were fixed with 2% PFA and permeabilized with 0.5% Triton X-100. The slides were incubated in blocking buffer (BSA 5% in PBS) for 1 h then with the primary antibody for 1 h. DNA was stained with 10 µg/ml Hoechst and the image analysis was carried out using Perkin-Elmer Ultraview RS Nipkow-spinning disk confocal microscope. BrdU, anti-TOM22 Atto488, Hoechst 33342, and ethidium bromide were purchased from Sigma; anti-BrdU antibody from BD Biosciences; MitoTracker® Green FM from Invitrogen, PCNA, TFAM, VDAC 1/2/3 (sc-98708), GADPH, and Polγ antibodies were purchased from Santa Cruz Biotechnology; goat anti-mouse and goat anti-rabbit Alexa® Fluor 488, Alexa® Fluor 555, and Cy5 conjugated secondary antibodies from Invitrogen.

DNA probes for FISH were labelled by nick translation of PCR products, incorporating Atto425-dUTP, or Atto488-dUTP, or Atto550-dUTP (Atto425/Atto488/Atto550 NT Labeling kit, Jena Bioscience). Forward and reverse primers for PCR amplification of DNA probes initiate at the position ‘start’ and ‘end’, respectively, in Table 1. Forty ng of labelled probes were mixed with 400 ng of sonicated salmon sperm DNA (Sigma) in hybridization buffer (50% formamide, 10% dextran sulfate, in 2×SSC pH 7.0), denatured at 80°C for 10 min and kept at 37°C for 30 min.

Cells plated on slides were fixed with 2% PFA and permeabilized with 0.5% Triton X-100. Cells were then incubated in 50% formamide (pH = 7.0)/2×SSC for 30 min at RT, and denatured in 70% formamide/2×SSC for 4 min at 75°C. Hybridization was performed with 40 ng of probe (single probe or mix) for 16 hrs at 37°C. After washing the slides in SSC, the DNA was stained with 10 µg/ml Hoechst. When required, fixed/permeabilized cells on slides were treated with RNaseA (100 µg/ml, Roche), or RNaseH (100 U/ml, NEB) or DNaseI (100 U/ml, Invitrogen) for 1 hr at 37°C, or with proteinase K (5 µg/ml, NEB) for 5 min at 37°C. When more than one nuclease were used, the enzymes were either added simultaneously or the second nuclease was added after incubation with the first nuclease, followed by three washes with PBS, and further incubation for 1 hr at 37°C. For mTRIP-coupled immunofluorescence, after hybridization and SSC wash, the immunofluorescence procedure (see above) was applied. The nuclease concentration for these experiments was tested with a dose-response test on several FISH probes (not shown).

Confocal acquisitions were performed using a spinning-disk Perkin-Elmer Ultraview RS Nipkow Disk, an inverted laser-scanning confocal microscope Zeiss Axiovert 200 M with an Apochromat 63×/1.4 NA oil objective and a Hamamatsu ORCA II ER camera (Imagopole, PFID, Institut Pasteur). Images were acquired using non-saturating settings, and the same imaging parameters were used for all samples. Optical z-slices in 200-nm steps covering the whole depth of the cell were collected, at resolution of 1.024/1.024 pixels. Three-dimensional reconstruction of all the z-stacks was achieved using the 3D-volume rendering of IMARIS software (Bitplane). Original 2D images generated from 3D volume rendering were used for fluorescence quantification using the Integrated Density measurement tool of ImageJ 1.38×software (post-acquisition analysis). Graphic settings were applied to 2D images for the purpose of visualization (panels in figures). For each condition, 30 cells were analysed from three independent experiments. In certain co-labelling experiments with mREP (or mTRANS) and immunofluorescent markers, measurement of fluorescence of POLγ, TFAM or BrdU was performed in mREP-positive and mREP-negative (or in mTRANS-positive and in mTRANS-negative) areas, defined by identical circles and either containing or not the FISH marker. For each condition, 300 samples of identical surface, from at least 30 cells in three independent experiments, were analysed. Intensity correlation coefficient-based (ICCB) tools use global statistic analysis of pixel intensity distributions for colocalization analysis. The ImageJ plug-in JACoP is a compilation of the most relevant ICCB tools (e.g. Pearson's, Mander's and Costes's coefficients), and allows comparison of the various methods among them (Bolte and Cordelières, 2006). Colocalization studies were performed with ImageJ JACoP plug-in. Scale bars represent 10 µm, unless otherwise indicated.

The significance of differences between data was determined using Student's t-test for unpaired observations. *P≤0.05; **P≤0.01; ***P≤0.001.

Total RNA was isolated from HeLa cells using the RNeasy Mini kit (Qiagen), treated with DNaseI (Qiagen), then reverse-transcribed using Superscript®III Reverse transcriptase (Invitrogen). Real-time quantitative PCR was performed using Power SYBR Green PCR Master Mix (Applied Biosystems) and the rate of dye incorporation was monitored using the StepOne™ Plus RealTime PCR system (Applied Biosystems). Three biological replicates were used for each condition. Data were analyzed by StepOne Plus RT PCR software v. 2.1 and Microsoft Excel. TBP transcript levels were used for normalisation of each target ( = ΔCT). Real-time PCR C_T values were analyzed using the 2^−ΔΔCt method to calculate the fold expression (Schmittgen and Livak, 2008). Custom primers were designed using the Primer3Plus online software (http://www.bioinformatics.nl/cgi-bin/primer3plus/primer3plus.cgi). Primers used for amplifications are listed in supplementary material Table S1, with the corresponding references.

The quantification of mtDNA was performed as described (Parone et al., 2008). Real-time PCR amplification was performed on 200 pg of total DNA using the StepOne™ Plus RealTime PCR system (Applied Biosystems) and Power SYBR Green PCR Master mix (ABI) following the manufacturer's instructions. Human mitochondrial DNA accession number NC012920 at [http://www.mitomap.org/bin/view.pl/MITOMAP/MitoSeqs]. The region tested on mtDNA was included in the 12S gene. The nuclear encoded 18S rRNA gene was used as an endogenous reference. The level of mtDNA was calculated using the ΔC_T of average C_T of mtDNA and nDNA (ΔC_T = C_T nDNA−C_T mtDNA) as 2^ΔCT (Fan et al., 2009). Primers used for amplification are listed in supplementary material Table S1.

We thank A. Rötig for the kind gift of HeLa rho⁰ cells and Rrm2b fibroblasts. We also thank L. Arbibe and collaborators for helpful discussion, F. Brandizzi, B. Dujon, S. Tajbakhsh and C. Zurzolo, for comments on the manuscript, and the Imagopole (PFID) of Institut Pasteur for advice. The mTRIP protocol is the subject of the International patent application number PCT/EP2012/054739.