Replication competition drives the selective mtDNA inheritance in Drosophila ovary

SUMMARY

Purifying selection that limits the transmission of harmful mitochondrial DNA (mtDNA) mutations has been observed in both human and animal models. Yet the precise mechanism underlying this process remains undefined. Here, we present a highly specific and efficient in situ imaging method capable of visualizing mtDNA variants that differ by only a few nucleotides at single-molecule resolution in Drosophila ovaries. Using this method, we revealed that selection primarily occurs within a narrow developmental window during germline cysts differentiation. At this stage, the proportion of the deleterious mtDNA variant decreases without a reduction in its absolute copy number. Instead, the healthier mtDNA variant replicates more frequently, thereby outcompeting the co-existing deleterious variant. These findings provide direct evidence that mtDNA selection is driven by replication competition rather than active elimination processes, shedding light on a fundamental yet previously unresolved mechanism governing mitochondrial genome transmission.

INTRODUCTION

The mitochondrial genome (mtDNA) is prone to accumulating mutations due to its proximity to damaging free radicals and the lack of canonical DNA repair mechanisms.¹ Yet, the infrequency of deleterious mtDNA mutations in populations underscores the presence of effective mechanisms limiting their transmission across generations.² In nearly all animal species, mtDNA is exclusively transmitted through the maternal lineage.³ During oogenesis, the genome goes through a genetic bottleneck that promotes random segregation and rapid genetic drift of mitochondrial variants.^4,5 Oocytes with excessive mutations would, in principle, suffer from compromised energy metabolism and be eliminated during germ cell development. However, the frequency of spontaneous mtDNA mutations, estimated in the range of 10⁻⁵-10⁻⁶,⁶ is too low for an effective selection at the cellular level. In recent years, purifying selection, where the frequency of deleterious mutations even at low level is further reduced in female germline, has been observed in human and various animal models.^7–12 A clear reduction of mutation load on mtDNA occurs during human primordial germ cell specification and migration,¹³ and germline selection is considered a major force shaping mtDNA variant characteristics in human populations.¹¹ In mouse models, strong purifying selection against pathogenic mtDNA mutations, presumably at the organelle level, occurs during folliculogenesis.¹⁴ Nonetheless, the cellular mechanisms underlying these selections are unclear.

The selective inheritance against deleterious mtDNA mutations has also been demonstrated in Drosophila melanogaster,^8–10 allowing the detailed dissection of mtDNA selection in this genetically tractable model organism.^15–17 Previous studies have roughly mapped the timing of mtDNA selection to a developmental window spanning the late germarium to early stages of egg chambers.^8,9 Drosophila oogenesis begins with the asymmetric division of a germline stem cell (GSC) at the anterior tip of a germarium, producing a cystoblast that undergoes four rounds of division with incomplete cytokinesis, forming a 16-cell cyst in germarium region 2A (Figure 1A). Moving posteriorly, the 16-cell cysts will be encased and compressed by migrating somatic cells into the characteristic lens shape in the 2B region (Figure 1A). At this stage, germ cells begin to differentiate into one future oocyte and 15 nurse cells. Once fully enveloped by follicle cells, the cyst buds off from the posterior end of the germarium and continues to develop as an egg chamber.^18–21

Figure 1. CAR assay effectively detects mitochondrial nucleoids in the Drosophila germarium.

(A) Diagram of a Drosophila germarium illustrating successive developmental stages from anterior (A) to posterior (P). Germline stem cells (GSCs), dividing cysts (DC), regions 2A and 2B of 16-cell cysts, budding egg chambers, mitochondria (blue) and fusome (red) are depicted. (B) Schematic representation of the CAR assay for detecting mtDNA in situ. The wild-type D. melanogaster mitochondrial genome contains a single BglII and a single XhoI site. Endogenous mtDNA is digested by XhoI or BglII restriction enzyme in tissue, followed by Lambda exonuclease (λ Exo) treatment to generate single-stranded DNA (ssDNA) with a 3’-OH group. The padlock probes are designed with two key components: a DNA sequence complementary to the ssDNA adjacent to XhoI or BglII restriction site and a unique sequence tag for subsequent detection. Upon annealing, circular ligation, and rolling circle amplification (RCA), a long, periodic ssDNA product covalently linked to the mtDNA molecule at the restriction enzyme site is synthesized. The CAR products are visualized by fluorescence in situ hybridization using probes specific to the unique sequence tags on each padlock probe. (C) CAR signals detected using XhoI (left) or BglII (right) probes colocalize with TFAM-labeled mitochondrial nucleoids in the germarium. Drosophila ovaries endogenously expressing TFAM-mNeonGreen were hybridized with either XhoI (left) or BglII (right) probes. Images are projections of 10 z-stacks (0.3 μm/stack). Scale bar, 10 μm. (D) Frequency distribution plots of distances between TFAM and CAR signals. Centers of TFAM, XhoI and BglII puncta in dividing cysts were detected using Imaris. The shortest distances between the centers of TFAM and CAR puncta were then measured. Each data point represents a pair of TFAM-XhoI or TFAM-BglII spots (XhoI probe, n=3789; BglII probe, n=4147). Bin size, 0.1 μm. (E) Quantification of CAR assay efficiency in detecting mitochondrial nucleoids. The percentage of CAR signals relative to the total number of nucleoids in a germarium is shown. Each data point represents one germarium (n = 5). Data are presented as mean± 95% confidence interval.

In the germarium, a series of developmentally orchestrated mitochondrial processes that ensure the transmission of healthy mtDNA have been revealed.^16,17 In dividing germline cysts, mtDNA transcription is largely inactive.²² mtDNA replication, which depends on transcription for initiation,²³ is also quiescent.⁸ Meanwhile, mitochondria undergo active fission at this stage,^10,16 which, together with the lack of mtDNA replication, promotes mtDNA segregation. By region 2A, each mitochondrion in 16-cell cysts contains, on average, a single mtDNA nucleoid.¹⁶ This links the function of individual mitochondria to the quality of genome they carry, allowing potential mtDNA selection at the organelle level. From GSCs to rounded 16-cell cysts in region 2A, mitochondrial respiration is inactive,²² presumably due to the lack of mtDNA transcription that generates core components of electron transport chain complexes. In region 2B, the mechanical stress from surrounding follicle cells activates mitochondrial respiration through a Myc-mediated transcriptional boost in differentiating 16-cell cysts.²⁴ Interestingly, mtDNA replication commences concurrently with, and appears depending on the activation of, mitochondrial respiration, as mtDNA replication is severely impaired in region 2B cysts carrying homoplasmic mt:CoI^T300I, a temperature sensitive lethal mtDNA mutation that severely disrupts cytochrome C oxidase.⁸ These observations have led to a proposal that wild-type mtDNA or healthy genomes would be replicated more frequently and thereby outcompete deleterious variants.¹⁶ In support of this idea, severe inhibition of mtDNA replication diminishes mtDNA selection.²⁵ While this model of replication competition is logically compelling, it has not been empirically demonstrated that deleterious mtDNA variants indeed replicate less than wild-type genome in the same germ cell. Additionally, a developmentally regulated mitophagy - programmed germline mitophagy (PGM) that begins at region 2A as germ cells enter meiosis,¹⁷ has also been shown to contribute to the selective inheritance.^10,17 However, given that PGM occurs regardless of the presence of mtDNA mutations,¹⁷ it is unclear whether PGM impacts mtDNA selection directly by eliminating defective mitochondria or indirectly through other processes such as mitochondrial protein turnover. Hence, the underlying mechanism of selective inheritance remains unsettled. Differentially visualizing co-existing mitochondrial genomes in ovaries is necessary to understand the timing and mechanisms of mtDNA selection.

Currently, most mtDNA variants in D. melanogaster were generated using a selection scheme based on mitochondrially targeted restriction enzymes.²⁶ These variants differ from the wild-type genome on merely a single nucleotide or small indel (insertion-deletion) on the corresponding enzyme sites.²⁶ Distinguishing single nucleotide polymorphisms (SNPs) on mtDNA in situ remains a technical challenge. Previous studies generated heteroplasmic fly models containing mt:CoI^T300I from D. melanogaster and wild-type Drosophila yakuba mtDNA.^10,27 Mitochondrial genomes from these two sibling species share ~93% sequence identity in coding regions, but the non-coding AT-rich regions, which contain the origins of replication, are highly divergent.²⁸ These polymorphisms enable the design of a pool of genome-specific probes required for single-molecule fluorescence in situ hybridization (smFISH), as well as a more advanced method recently developed termed smFISH-HCR, to visualize mtDNA and thereby study the underlying mechanisms of mtDNA selection.^10,29 By providing functional cytochrome C oxidase activities, the D. yakuba mtDNA can be maintained at lower percentage with mt:CoI^T300I at the restrictive temperature for many generations.³⁰ However, it is quickly outcompeted by wild-type D. melanogaster mtDNA in the D. melanogaster nuclear background.²⁷ Hence, the potential mito-nuclear incompatibility associated with the D. mel/yak heteroplasmic line complicates the selective pressures on mtDNA and raises concerns about the validity and physiological relevance of this interspecies heteroplasmic model for studying mtDNA selective inheritance. It is preferable to use heteroplasmic variants from the same species to minimize confounding factors for understanding mechanisms of mtDNA transmission.

In this study, we introduce a highly specific and efficient in situ imaging method that enables the visualization of mtDNA genomes differing by only two loci in tissues. Using this approach, we reveal a selective replication advantage of the healthy mtDNA variant in differentiating cysts, where it outcompetes the co-existing deleterious variant. This precisely timed developmental window acts as a mitochondrial quality control checkpoint, increasing healthy mtDNA levels before the onset of large-scale mitochondrial biogenesis in developing egg chambers. Our findings provide key insights into the mechanisms driving mtDNA selection and inheritance, establishing a powerful framework for future studies on mtDNA dynamics with spatial resolutions across different tissues and model systems.

RESULTS

CAR assay effectively and specifically detects mtDNA SNPs in situ

To distinguish mtDNA variants with single nucleotide polymorphisms, we first tested a CRISPR-Cas9-mediated proximity ligation (CasPLA) assay³¹ in Drosophila ovaries. However, the resulting signals were diffused within the mitochondrial network (Figure S1), lacking the spatial resolution needed to differentiate mtDNA genotypes in heteroplasmic ovaries. Currently, most mtDNA variants in D. melanogaster are generated using mitochondrially targeted restriction enzymes, which select for mutations that abolish enzymes’ cleavage sites.²⁶ Taking advantage of this molecular nature, we improved a target-primed rolling circle amplification (RCA) method³² to distinguish genomes that differ by two loci with high efficiency and specificity (Figure 1B; Figure S2C). In our method, referred to as “CAR” (Circular DNA Amplification at Restriction enzyme sites), endogenous mtDNA was digested with restriction enzymes and treated with Lambda exonuclease in fixed ovaries, generating a stretch of single-stranded DNA (ssDNA) with 3’-OH group. A padlock probe was designed to have a unique sequence tag, which bears no homology to either the nuclear or mitochondrial genomes, flanked by 5’ and 3’ sequences complementary to the ssDNA. The two ends of the linear probe were brought to juxtaposition by hybridizing to the ssDNA, which primes rolling circle amplification after the circularization of the probe. The rolling circle amplification produced an ultralong, periodic ssDNA covalently linked to the mtDNA molecule at the restriction enzyme site (Figure 1B; Figure S2C). We then performed fluorescence in situ hybridization targeting tag sequences to visualize the linked mtDNA therein.

D. melanogaster mtDNA contains a single XhoI site and a single BglII site.²⁶ We designed two different padlock probes complementary to the ssDNA derived from XhoI (XhoI probe) or BglII (BglII probe) cleavage (Figure S2C). We performed CAR assay on wild-type ovaries using either of these two probes, with TFAM-mNeonGreen—an endogenously tagged TFAM fusion protein—marking mtDNA nucleoids.³³ In ovaries, CAR signals of both probes and TFAM puncta all appeared as puncta (Figure 1C) with mean diameters around 0.8 μm. About 74% of XhoI puncta and 58% of BglII puncta were either overlapped or closely associated with (center to center distance less than 0.8 μm) TFAM-mNeonGreen (Figure 1C,1D). The CAR puncta that did not colocalize with TFAM (Figure1C, 1D) likely represent uncompacted mtDNA molecules that lack TFAM coating.³⁴ Overall, 46±2% of mitochondrial nucleoids were labeled by the XhoI probe (Figure 1E), while 36±3% of nucleoids were labeled by the BglII probe (Figure 1E), demonstrating comparable efficiency of these two probes in labeling mtDNA. When both probes were applied together to the wild-type ovaries, only a minor fraction of CAR puncta were double-labeled (Figure S2A, S2B). This result indicates that the two probes compete for the same mtDNA molecule in the CAR assay, reducing the co-labeling frequency.

To assess the specificity of the CAR assay, we applied both padlock probes to ovaries carrying either mt:ND2^ins or mt:CoI^T300I genome, which lacks the BglII site and the Xho site, respectively.²⁶ In mt:ND2^ins ovaries, the XhoI probe effectively labeled mitochondrial nucleoids, whereas CAR puncta from the BglII probe were rare, accounting for less than 2% of all puncta detected in the same ovariole (Figure 2A, 2E). Similarly, in mt:CoI^T300I ovaries, CAR puncta from the XhoI probe were less than 2% of all puncta in the same ovariole (Figure 2A, 2E). These results demonstrate that CAR is both effective and specific in distinguishing mtDNA variants with SNPs in situ.

Figure 2. Specificity and accuracy of the CAR assay in distinguishing mt:ND2^ins and mt:CoI^T300I mtDNA variants in situ.

(A) The CAR assay specifically detects different mtDNA variants. Ovaries from homoplasmic mt:ND2^ins (left) or mt:CoI^T300I (right) flies cultured at 18°C were hybridized with both XhoI and BglII padlock probes. The XhoI probe predominantly detects mt:ND2^ins, while BglII probe mainly detects mt:CoI^T300I. Phalloidin (blue) stains actin. Images are projections of 10 z-stacks (0.3 μm/stack). Scale bar, 10 μm. (B) The CAR assay distinguishes mt:ND2^ins (XhoI probe) and mt:CoI^T300I (BglII probe) in heteroplasmic flies (hetero-4.4). Phalloidin (blue) stains actin. Scale bar, 10 μm. (C) Magnified view of the boxed region in (B), showing mitochondrial nucleoids detected by both probes. White arrows indicate that signals from the two mtDNA variants are adjacent but non-overlapping. Scale bar, 5 μm. (D) Frequency distribution plot of distances between the two mtDNA variants in heteroplasmic flies. Centers of XhoI and BglII probe puncta in dividing cysts within the germarium were detected using Imaris, and the shortest distances between the CAR signals were measured. Each data point represents a pair of XhoI-BglII spots (n=743, blue). Bin size, 0.1 μm. (E) Quantification of CAR signals detected by XhoI and BglII probes in homoplasmic mt:ND2^ins (n=5), heteroplasmic (hetero-4.4) flies (n=6), and homoplasmic mt:CoI^T300I flies (n=5). Each data point represents a single germarium. Data are normalized to the detection efficiency of each probe and are presented as mean±95% confidence interval. (F) CAR-based heteroplasmy quantification aligns with PCR-based measurements. Scatter plot comparing heteroplasmy levels quantified from the same fly using CAR and PCR assays. Each data point represents the mean percentage of mt:CoI^T300I in each line.

CAR accurately detects the level of heteroplasmy

To evaluate the accuracy of CAR in assaying heteroplasmy, we generated a series of heteroplasmic lines carrying both mt:ND2^ins and mt:CoI^T300I genomes at varying ratios using germ plasm transplantation (Figure S2E). We performed CAR in ovaries of heteroplasmic flies cultured at 18°C using both XhoI and BglII probes, to visualize corresponding mtDNA variants simultaneously (Figure 2B, 2C, and Figure S2D). CAR signals of both genomes were evident, with each CAR punctum appearing spatially distinct from the other (Figure 2B, 2C, 2D). This observation is consistent with the notion that mtDNA from different nucleoids rarely intermix.³⁵ We quantified the number of CAR puncta for each genome and deduced the heteroplasmy levels by factoring in the detecting efficiency of each probe (Figure 2F). The heteroplasmy levels determined by quantifying CAR puncta closely matched those obtained from the quantification of digested PCR products (Figure 2F, and Figure S2E, S2F), validating the accuracy of CAR in assessing heteroplasmy levels.

Selection against deleterious mtDNA variants takes place in differentiating cysts

The homoplasmic mt:CoI^T300I flies exhibit temperature-sensitive lethality,^8,36 whereas homoplasmic mt:ND2^ins flies are largely healthy.⁹ To assess the mtDNA selection in heteroplasmic flies, we performed a temperature shifting assay,^8,9 and quantified the load of mt:CoI^T300I in progeny produced by a single heteroplasmic mother at either a restrictive temperature of 29°C or a permissive temperature of 18°C by PCR-digestion method. The load of mt:CoI^T300I was lower in progeny produced at 29°C compared to their siblings produced at 18°C (Figure 3A). In addition, over multiple generations, the proportion of mt:CoI^T300I in the population continued to decline at 29°C, although the rate of decline slowed as mt:CoI^T300I levels decreased (Figure 3B). These results resemble previous observations of mtDNA selection in heteroplasmic flies carrying both wild-type and mt:CoI^T300I genomes,^8,9 validating the use of this heteroplasmic line to investigate the mechanisms of selective inheritance.

Figure 3. Germline selection against the mt:CoI^T300I genome occurs at germarium region 2B.

(A) Temperature shifting assay showing the heteroplasmy level (% mt:CoI^T300I) in individual larvae produced at 18°C and 29°C by a single mother. Each data point represents a single progeny (n= 10). Data are presented as mean± 95% confidence interval. Statistical analysis was performed using unpaired t-test. ***P≤ 0.001. (B) Proportion of mt:CoI^T300I in heteroplasmic (hetero-4.4) flies maintained at 18°C or 29°C across generations. Data were quantified using PCR analysis from 20 flies per generation. (C) (D) Representative 3D rendering of CAR images from heteroplasmic flies (hetero-4.4) maintained at 18°C (C) or 29°C (D). The mt:ND2^ins (green) and mt:CoI^T300I (red) mtDNA were detected by XhoI and BglII probes, respectively. In bottom panels, 3D rendering of developing cysts and CAR signals of mt:ND2^ins (green) and mt:CoI^T300I (red) mtDNA within the cysts are shown. The 2-cell cyst (orange), 4-cell cyst (yellow), region 2A (blue) and region 2B (magenta) were segmented using Imaris. Scale bar, 10 μm. (E) (G) Quantification of mt:ND2^ins and mt:CoI^T300I mitochondrial DNA numbers at different germarium stages using the CAR assay in heteroplasmic flies maintained at 18°C (E) and 29°C (G). For each temperature condition (18°C (E) and 29°C (G)), the number of each mtDNA variant in germline stem cells (GSC, n=6, 11), dividing cysts (DC, n=25, 28), region 2A (2A, n=6, 11), region 2B (2B, n=11, 7) and region 3 (3, n=5, 6) are shown. Each data point represents a single cyst. Data are presented as mean± 95% confidence interval. (F) (H) Heteroplasmy levels of the mt:CoI^T300I genome at different germarium stages in heteroplasmic flies maintained at 18°C (F) and 29°C (H). The proportion of mt:CoI^T300I relative to total mtDNA was quantified using the CAR assay. Each data point represents a single cyst. Data are presented as mean± 95% confidence interval. Statistical analysis was performed using an unpaired t-test. ns, not significant; *P≤ 0.05.

To monitor the dynamics of two mtDNA variants in developing germ cells, we performed CAR in ovaries of heteroplasmic flies cultured at either 18°C or 29°C (Figure S3A, S3B). Ovaries were co-stained with fluorescently labeled phalloidin, which binds to actin filaments and thereby outlines germ cells,³⁷ to identify development stages. We segmented germline cysts at different development stages in reconstructed 3D volumes (Figure 3C, 3D; Movie 1, 2) and quantified the number of CAR puncta derived from each genome. At 18°C, the total number of nucleoids, including both genomes, was estimated to be approximately 150 in a GSC and 166 in a dividing cyst (2-, 4-, or 8-cell cyst) after factoring in the detection efficiency of each probe (Figure 3E). For 16-cell cysts, the copy number increased from ~186 in region 2A to ~560 in region 2B (Figure 3E). In region 3, the copy number reached ~1308 per budding egg chamber (Figure 3E). This data aligns with the previous finding that mtDNA replication is largely quiescent in dividing cysts but activated in differentiating cysts at region 2B.⁸ We also quantified the load of mt:CoI^T300I in each cyst or GSC, and found that the proportion of mt:CoI^T300I were relatively stable over all developmental stages in a germarium (Figure 3F), indicating a lack of selection at the permissive temperature of 18°C.

At 29°C, the total number of nucleoids was comparable to that at the same developmental stage at 18°C and remained stable in GSCs, dividing cysts, and 16-cell cysts in region 2A (Figure 3G). A significant increase in nucleoid numbers was also observed in the 16-cell cysts in region 2B (Figure 3D, 3G). Importantly, mt:ND2^ins increased to a greater extent than mt:CoI^T300I, resulting in a 13% reduction in mt:CoI^T300I load in region 2B, compared to earlier germarium stages (Figure 3G, 3H). This value closely matched the 12% reduction of mt:CoI^T300I observed over a single generation (Figure 3B). Additionally, the heteroplasmy level of mt:CoI^T300I in budding egg chambers (region 3) was comparable to that in region 2B (Figure 3H). Together, these findings indicate the cross-generational selection against this deleterious mtDNA variant mainly occurs in the differentiating cysts at germarium region 2B.

We noticed that the copy number of mt:ND2^ins genome in 2B region of heteroplasmic flies was higher at 29°C than at 18°C, while being comparable in earlier stages (Figure 3E, 3G). This difference is likely due to the higher temperature, as homoplasmic mt:ND2^ins flies also had more nucleoids in 2B region cysts at 29°C, compared to 18°C (Figure S4A, S4B).

The healthy mitochondrial genome is preferentially replicated over the deleterious variant

The reduced mutation load could be caused by a selective elimination of mutant genome through mitophagy or other quality control mechanisms. However, the absolute number of the mt:CoI^T300I genome in heteroplasmic ovaries was not reduced at any stages of developing germ cells in the germarium (Figure 3G), arguing against a model of active elimination. We next asked whether the faster increase of the healthy genome, mt:ND2^ins in region 2B reflects different replication frequencies between these two genomes. We combined CAR with EdU (5-ethynyl 2´-deoxyuridine, a thymidine analog) incorporation to assess the replication of mt:ND2^ins and mt:CoI^T300I in heteroplasmic ovaries. After a 2-hour EdU incubation, a proportion of mtDNA was labeled with EdU (Figure 4A, 4B, and Figure S3C), consistent with previous studies showing that mtDNA undergoes relaxed replication.^38,39 Most CAR signals partially overlapped with EdU signals, likely because CAR products extend away from mtDNA loci (Figure 1B). The proportion of replicating events, indicated by EdU-positive CAR signals, was similar between two genomes in both region 2B and budding egg chambers of heteroplasmic flies cultured at 18°C (Figure 4A, 4C, Figure S3C). However, at 29°C, mt:ND2^ins exhibited a markedly higher replication frequency than mt:CoI^T300I in region 2B, but not in region 3 (Figure 4B, 4D, Figure S3C). Together, these results provide direct evidence that the healthy mitochondrial genome in fact replicates at a higher frequency than the co-existing deleterious variant.

Figure 4. The mt:ND2^ins genome replicates more efficiently than the deleterious mt:CoI^T300I genome at germarium region 2B.

(A) (B) Representative images of simultaneous EdU staining and CAR assay in heteroplasmic ovaries of flies cultured at 18°C (A) or 29°C (B). The mt:ND2^ins (green) and mt:CoI^T300I (red) mtDNA are detected by XhoI and BglII probes, respectively. Replicating mtDNA was labeled with EdU (white). CAR signals enclosed by white lines indicate those colocalizing with EdU staining. Germarium region 2B (solid line) and region 3 (dashed line) are indicated. Scale bar, 5 μm. (C) (D) Quantification of replicating mtDNA for each mtDNA variant in heteroplasmic flies maintained at 18°C (C) and 29°C (D). The proportion of replicating mt:ND2^ins and mt:CoI^T300I genomes was determined by normalizing EdU-positive CAR signals to the total number of corresponding mtDNA signals under each temperature in germarium region 2B (n=7, 6) and budding egg chambers (region 3, n=7, 6). Each data point represents a single germarium. Data are presented as mean± 95% confidence interval. Statistical analysis was performed using an unpaired t-test. ns, not significant; ***P≤ 0.001.

DISCUSSION

A padlock probe-mediated rolling-circle amplification method was previously developed to detect mtDNA SNPs in cultured cells.³² In this approach, wild-type and mutant padlock probes compete for the same target sequence, and the long single-stranded (ssDNA) 3’ overhang generated after exonuclease treatment could potentially form secondary structures, further limiting its efficiency in detecting mtDNA SNPs in tissues. In Drosophila, most available mtDNA mutants affect certain restriction enzyme cleavage sites. We implement the CAR assay by leveraging this feature to generate distinct, short ssDNA stretches, derived from different enzyme cleavage sites on each genome. Compared to the 10% efficiency observed in cultured cells using the previous method,³² CAR displayed 4–5 fold increase in detection efficiency (Figure 1E). In addition, the rolling circle amplification in CAR is initiated at the unique sequences on each genome, minimizing non-specific amplification and thereby further enhancing the specificity. We also found that treating ovaries with Triton X-100, rather than ethanol or proteases, was sufficient for tissue permeabilization while better preserving tissue integrity and making CAR compatible with immunofluorescence antibody staining (Methods). Another approach integrating CRISPR-associated (Cas) proteins with RCA amplification requires complex DNA/RNA probes,³¹ and were ineffective in spatially distinguishing different mtDNA genotypes in Drosophila ovaries (Figure S1). Notably, CAR products are covalently linked to endogenous mtDNA, enabling single-molecule resolution of heteroplasmy within tissues. Most human disease-associated mtDNA mutations are point mutations.¹ While CAR assay targets SNPs at restriction enzyme sites, recent advances in CRISPR/Cas, ZFN and TALEN based DNA editing technologies hold the potential to expand this approach to visualize any point mutations on mtDNA or the nuclear genome. Hence, CAR would offer unprecedented spatial resolution to track mtDNA segregation and dynamics in tissues, which is crucial for understanding the pathogenic mechanisms of mtDNA diseases.

Using CAR, we demonstrated that the healthier genome mt:ND2^ins exhibited a higher frequency of replication than mt:CoI^T300I within the same differentiating germline cyst (Figure 4B, 4D). This data provides direct evidence that healthy genome has an advantage in replication at this stage and thereby outcompetes the deleterious variants. The reduction of mt:CoI^T300I load in germarium region 2B was comparable to its decline across a single generation (Figure 3B, 3H), indicating that the selection primarily occurs at this stage. Consistent with this notion, the frequency of replication was comparable between mt:ND2^ins and mt:CoI^T300I in budding egg chambers (Figure 4B, 4D), and no further reduction of mt:CoI^T300I was observed beyond region 2B (Figure 3H). In female germarium, mtDNA copy number is ~100 in a GSC and remains stable in dividing cysts.¹⁶ From region 2B onward, germ cells undergo 16.6 replication cycles, ultimately accumulating about 10 million copies of mtDNA in a mature oocyte.^40,41 Our data indicates that the healthier genome, mt:ND2^ins underwent ~2.1 replication cycles, compared to ~1.6 cycles for mt:CoI^T300I in region 2B (Figure 3G). This modest difference in replication frequency sufficiently accounts for the ~15% reduction of mt:CoI^T300I that is commonly observed across one generation in heteroplasmic flies.^8,9 This reasoning not only corroborates with the finding that replication competition occurs in a narrow developmental window but also underscores the robustness of the replication competition in limiting the transmission of deleterious mtDNA mutations.

The absolute number of mt:CoI^T300I was not reduced throughout germ cell development under the restrictive condition (Figure 3G). Additionally, the numbers of mt:CoI^T300I genome in dividing cysts were comparable between 18°C and 29°C (Figure 3E, 3G). These findings argue against the idea that changes in mtDNA heteroplasmy during cyst differentiation is directly driven by the selective removal of deleterious mtDNA variants via mitophagy or other quality control processes. Consistent with this notion, programed germline mitophagy in dividing cysts, which promotes mtDNA selective inheritance, occurs regardless of the presence mtDNA mutations.¹⁷ Noteworthy, knocking down Atg1 or BNIP3, essential factors in PGM, reduces wild-type mtDNA levels while compromising the selective inheritance.¹⁷ It would be interesting to test whether PGM might regulate mtDNA replication to indirectly influence mtDNA selection.

Prior to region 2B, mitochondrial fission allows effective mtDNA segregation to minimize complementation between wild-type and deleterious mtDNA alleles,^10,16 setting the stage for selection based on the functionality of individual genomes. At region 2B, mtDNA expression serves as a stress test for each genome’s integrity—wild-type genomes would produce functional respiratory chain complexes (RC), whereas deleterious mtDNA variants would lead to defective RC. In ovaries, many nuclear-encoded mitochondrial proteins, including key mtDNA replication factors, are synthesized on the mitochondrial outer membrane.⁴² This local translation, along with the import of preproteins into the mitochondrial matrix relies on the mitochondrial membrane potential generated by mitochondrial respiration.²⁵ Therefore, in heteroplasmic germ cells, mitochondria carrying the functional genome would have an advantage over those containing deleterious mutations in acquiring replication factors. As a result, wild-type or healthier genomes are more likely to replicate and thereby outcompete coexisting defective counterparts. Consistent with this model, partial reduction of replication factors such as PolG1 has been shown to intensify this competition, thereby enhancing the selective inheritance of functional mtDNA.^15,17 Conversely, when mtDNA replication is severely impaired across all mitochondria, the efficiency of purifying selection is diminished.²⁵ It would be interesting to test whether any of the mtDNA replication factors are more enriched in healthy organelles, which might be the molecular underpinnings of replication competition and selective inheritance.

Strong purifying selection against deleterious mtDNA mutations, potentially at the organelle or genome level, has been observed in both human studies and mouse models.^7,11–14,43 In general, the mutations in protein-coding genes are subject to stronger selection than those in non-coding regions.^12,43 Moreover, within protein-coding genes, mutations at the first two codon positions are more strongly selected against than those at the third position (except for ATP6 and ATP8),^12,43 suggesting that selection pressure correlates with the functional impact of the mutation. Interestingly, different from mtDNA mutations affecting respiratory chain complexes, mutations affecting ATP synthase genes often lead to elevated mitochondrial membrane potential.^44,45 Our model posits that membrane potential is a key indicator of mitochondrial functionality to signal the selection. This may explain the weaker selection observed against mutations in ATP6 and ATP8. Purifying selection coincides with an upregulation of genes involved in mtDNA replication and transcription during primordial germ cell migration in human,¹³ suggesting a potential link between mtDNA replication and the purifying selection. In C. elegans, a mtDNA variant carrying a large deletion gains a selfish replicative advantage by preferentially associating with the mtDNA replicase, allowing its stable transmission across generations.^46,47 Taken together, replication competition might represent a conserved mechanism that enforce the selective mtDNA inheritance and shape the evolution of mitochondrial genome in metazoan.

Limitations of the study

In Drosophila melanogaster, most available mtDNA mutations were generated using mitochondrially targeted restriction enzymes. Accordingly, CAR assay was specifically designed to detect mtDNA mutations within restriction enzymes’ cleavage sites. However, in other organisms, most mtDNA mutations are not located in restriction enzymes’ sites, limiting the broader applicability of CAR assay. Nonetheless, CAR could be improved by adopting CRISPR, ZFN, or TALEN based DNA editing technologies, potentially detecting any mutation on mtDNA or the nuclear genome. Our model of replication competition is based on one particular mtDNA mutation, mt:CoI^T300I that is conditionally lethal in Drosophila. In mammals, germline selection limits the transmission of a broad spectrum of mtDNA mutations, many of which presumably are less deleterious than mt:CoI ^T300I. Whether replication competition also contributes to the selection against mtDNA mutations in mammals, particularly these mild ones, remains to be determined.

RESOURCE AVAILABILITY

Lead contact

Further information and requests for reagents should be directed to and will be fulfilled by the Lead Contact, Hong Xu (Hong.Xu@nih.gov).

Materials availability

All Drosophila lines generated in this study are available from the Lead Contact.

Data and code availability

• Data reported in this paper including original microscopy images and Excel files for quantifications have been deposited at Mendeley Data and are publicly available as of the date of publication. Accession numbers are listed in the KEY RESOURCES TABLE.

• This study did not report original code.

• Any additional information required to reanalyze the data reported in this paper is available from the Lead Contact upon request.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Paraformaldehyde	Electron Microscopy Sciences	Cat# 15710; CAS: 30525-89-4
PBS, pH 7.4	Quality Biological	Cat# 114-058-131
Triton X-100	VWR Life Science	Cat# 0694-1L; CAS: 9036-19-5
Tris-HCl, 1M, pH 8.0	Quality Biological	Cat# 351-007-101
SDS, 20%	Quality Biological	Cat# 351-066-101; CAS: 151-21-3
NaCl, 5M	KD Medical	Cat# RGE-3270; CAS: 7647-14-5
EDTA, 0.5M	KD Medical	Cat# RGC-3130; CAS: 60-00-4
Sodium deoxycholate	Sigma	Cat# D6750; CAS: 302-95-4
Tween-20	VWR Life Science	Cat# 0777-1L CAS: 9005-64-5
Formamide (Deionized)	Invitrogen	Cat# AM9342; CAS: 75-12-7
Dextran Sulfate Sodium Salt	Thermo Scientific	Cat# J14489-09; CAS: 9011-18-1
Aphidicolin	MedChem Express	Cat# HY-N6733; CAS: 38966-21-1
Vectashield mounting medium with DAPI	Vector Laboratories	Cat# H-1200-10
Vectashield mounting medium	Vector Laboratories	Cat# H-1000-10
FBS	Gibco	Cat# A5670701
Penicillin–streptomycin	Gibco	Cat# 15140122
BSA	Sigma	Cat# A8806
Schneider’s Drosophila medium	Gibco	Cat# 21720
XhoI	New England Biolabs	Cat# R0146L
BglII	New England Biolabs	Cat# R0144M
Hi T4 DNA ligase	New England Biolabs	Cat# M2622L
T4 DNA ligase	New England Biolabs	Cat# M0202L
Lambda exonuclease	New England Biolabs	Cat# M0262L
20×SSC buffer	KD Medical	Cat# RGF-3240
Salmon sperm DNA	Invitrogen	Cat# 15-632-011
dNTP Solution Set	New England Biolabs	Cat# N0446S
Agarose	Lonza Bioscience	Cat# 50004
TAE Buffer	Quality Biological	Cat# 351-008-491
CF®405M Phalloidin	Biotium Inc	Cat# 00034
Critical commercial assays
EquiPhi29™ DNA Polymerase	Thermo Scientific	Cat# A39391
Click-iT™ EdU Cell Proliferation Kit	Invitrogen	Cat# C10337
DNeasy Blood & Tissue kit	QIAGEN	Cat# 69504
Taq DNA Polymerase High Fidelity	Invitrogen	Cat# 11304029
Experimental models: Organisms/strains
D. melanogaster: TFAM-mNeonGreen	Zhang et al.33	N/A
D. melanogaster: w1118 (mt:ND2ins)	Xu et al.26	N/A
D. melanogaster: w1118 (mt:CoIT300I)	Xu et al.26	N/A
D. melanogaster: hetero-1.1	This paper	N/A
D. melanogaster: hetero-4.3	This paper	N/A
D. melanogaster: hetero-4.4	This paper	N/A
Oligonucleotides
XhoI padlock probes: 5’Phos/CAACGGTAAATATATGCAACAAACTGGTGAAAGTCGCTATCCCGTTCAGGAAAAGTATCTACATCTATTC	This paper	N/A
BglII padlock probes: 5Phos/ATGATTAATTGAAGAAACGAAGTTTTCATAAGACACGGAGAGAGAGCACCAAAATTAATATTCACCCTAA	This paper	N/A
XhoI detection probes: 5’Cy5/CTGGTGAAAGTCGCTATCCCGTT	This paper	N/A
XhoI detection probes: 5’Alexa488/CTGGTGAAAGTCGCTATCCCGTT	This paper	N/A
XhoI detection probes: 5’Cy3/CTGGTGAAAGTCGCTATCCCGTT	This paper	N/A
BglII detection probes: 5’Cy3/TTCATAAGACACGGAGAGAGAG	This paper	N/A
PCR primers for quantifying mt:CoIT300I load TGGAGCTATTGGAGGACTAAATCAGCTCCTGTTAATGGTCATGGACT	Hill et al.8	N/A
Software and algorithms
Prism	GraphPad	https://www.graphpad.com/
Fiji	NIH	https://fiji.sc; https://doi.org/10.1186/s12859-017-1934-z
Imaris	Oxford Instruments	https://imaris.oxinst.com
Deposited data
Original microscopy images and Excel files	This paper; Mendeley Data	DOI: 10.17632/vtbvbwwfvt.1
Other
Nikon CSU-W1 SoRa confocal system	Nikon	N/A
GelDoc Go Gel Imaging System	BIO-RAD	N/A

STAR METHODS

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Fly stocks and husbandry

Drosophila melanogaster stocks were reared in a humidity-controlled incubator under a 12-hour light-dark cycle using standard cornmeal molasses agar media. Heteroplasmic Drosophila melanogaster strains were maintained at 18 °C unless otherwise indicated. The following fly stocks reared at 25 °C were used: TFAM-mNeonGreen³³, w¹¹¹⁸ (mt:CoI^T300I)²⁶, w¹¹¹⁸ (mt:ND2^ins)²⁶.

METHOD DETAILS

Generation of heteroplasmic flies

Heteroplasmic flies were generated using germplasm transplantation, as previously described.^8,9 Briefly, germplasm from homoplasmic mt:ND2^ins embryos were injected into the posterior end of homoplasmic mt:CoI^T300I recipient embryos. Hatched females from injected embryos were crossed to w¹¹¹⁸ males, and the progeny from this cross were subsequently mated en masse to w¹¹¹⁸ males at room temperature. To select for heteroplasmic flies, embryos were shifted to 29°C, and surviving progeny were identified as heteroplasmic flies. Female escapers were individually crossed with w¹¹¹⁸ males and maintained as founder lines. The proportion of mt:CoI^T300I in each founder line was quantified (see below). Three lines with low (hetero-1.1), medium (hetero-4.3), and high (hetero-4.4) levels of mt:CoI^T300I were chosen and maintained at 18°C.

Quantification of heteroplasmy using PCR

Total DNA was extracted from the indicated flies using the DNeasy Blood & Tissue kit (QIAGEN). The BglII site is located 1568 base pairs (bp) upstream of the XhoI site on mtDNA. A 4 kb mtDNA fragment spanning both XhoI and BglII sites was amplified from total DNA using PCR primers described previously.⁸ The PCR products were digested overnight with the XhoI restriction enzyme and separated by agarose gel electrophoresis. DNA band intensities were quantified using ImageJ software. The heteroplasmy level was determined as the proportion of XhoI-resistant DNA fragment (mt:CoI^T300I) relative to the total DNA.

mtDNA selection assay

The mtDNA selection assay was performed following previously established methods.⁸ To examine mtDNA selection across generations, progeny from the hetero-4.4 founder line were divided into two groups and maintained at 18°C and 29°C, respectively. In each generation, 50 flies (mixed male and female) were randomly selected for mating. After 5 days, these flies were collected and stored at −80°C until heteroplasmy levels were quantified. Flies were maintained for at least five generations at respective temperature.

To assess mtDNA selection during female germline development, a single heteroplasmic female was mated with multiple w¹¹¹⁸ males at 18°C for 3 days. The flies were then transferred to a new vial and acclimated at 29°C for 4 days. During this period, embryos were discarded to ensure that only germ cells that developed at 29°C were analyzed. Progeny produced on days 1–3 at 18°C and days 5–7 at 29°C were collected, and their heteroplasmy levels were quantified.

CAR assay in Drosophila ovaries

Flies with specified genotypes were maintained at 18°C, 25°C, or 29°C, depending on experiment requirements. 10–20 ovaries were dissected in Schneider’s Drosophila medium supplemented with 10% fetal bovine serum (FBS) and fixed with 1 ml 4% paraformaldehyde in 1.5 ml microcentrifuge tubes for 20 minutes. Following fixation, tissues were washed three times in 1 ml 1× PBST (1× PBS with 0.1% Tween-20) and permeabilized in 1 ml RIPA buffer (150 mM NaCl, 1% TritonX-100, 0.5% sodium deoxycholate, 0.1% SDS, 1 mM EDTA, 50 mM Tris-HCl pH 8.0) for 30 minutes at room temperature with gentle rotation. The permeabilization step was repeated three times using fresh RIPA buffer, followed by two washes in 1× PBST. To prevent over-treatment by RIPA, tissues were fixed again in 1 ml 4% paraformaldehyde for 30 min, followed by three additional washes in 1× PBST.

For restriction enzyme digestion, ovaries were incubated with indicated NEB restriction enzyme (XhoI, BglII or both) (1 units/μl restriction enzyme, 1× NEBuffer r3.1, 250 μl) at 37°C overnight. The next day, the restriction enzyme was removed by three washes in 1× PBST, and tissues were treated with Lambda exonuclease (0.15 units/μl Lambda exonuclease, 1× Reaction buffer, 300 μl) for 15 minutes at 37°C, followed by four additional washes in 1× PBST. Padlock probe hybridization was then performed by incubating tissues in 250 μl reaction (2 μM Padlock probe, 20 units/μl T4 DNA ligase, 1× T4 DNA ligase buffer) for 8–12 hours at 16°C. Samples were then washed twice in 1× PBST (pre-cold on ice) for 3 minutes each before proceeding to rolling cycle amplification (RCA). For RCA, tissues were incubated in 250 μl EquiPhi29 reaction mix (0.25 U/μl EquiPhi29, 1 mM DTT, 0.5 mM dNTPs, 1× EquiPhi29 reaction buffer) at 30°C for 4–6 hours with gentle rotation, followed by washing three times in 1× PBST for 5 minutes each.

To visualize the amplified signals, fluorescent probe in situ hybridization was performed. Tissues were first pre-incubated in 100 μl hybridization buffer (10% deionized formamide, 2× SSC, 10% dextran sulfate) at 37°C for 15 minutes, followed by incubation in 300 μl hybridization mix (10% deionized formamide, 2× SSC, 10% dextran sulfate, 83 μg/ml salmon sperm DNA, 200 nM fluorophore-labeled detection probes) at 37°C for at least 7 hours or overnight in the dark. The next day, tissues were sequentially washed at 37°C in 1 ml 2× SSC buffer (10% deionized formamide, 2× SSC), 1 ml 1× SSC buffer (10% deionized formamide, 1× SSC) and 1 ml 0.5× SSC buffer (10% deionized formamide, 0.5× SSC) for 15 minutes each, before being immersed in 1 ml PBS for 5 minutes. For phalloidin staining, tissues were incubated with 300 μl CF^®405M Phalloidin solution (0.01 U/μl CF^®405M Phalloidin in 1× PBS) for 0.5–1 hour at room temperature, followed by three washes in PBS for 7 minutes each. Finally, ovaries were mounted using VECTASHIELD antifade mounting medium and imaged using a Nikon CSU-W1 SoRa confocal system (Nikon SR Plan Apo IR 60×/1.27 oil lens; Nikon Element software; Yokogawa CSU-W1 SoRa Confocal Scanner Unit). Image deconvolution was performed using NIS Elements AR (6.02.03) software for further processing.

Simultaneous EdU staining and CAR assay

Newly eclosed female flies with the specified genotypes were reared for 5 days at the indicated temperature. 10–20 ovaries were dissected in Schneider’s Drosophila medium supplemented with 10% FBS and transferred to 1 ml fresh medium containing 7 μM aphidicolin, followed by incubation for 3 hours at the respective temperature. The medium was then replaced with 1 ml fresh medium containing 20 μM EdU and 7 μM aphidicolin, and incubation continued for an additional 2 hours at 18°C or 29°C. After two 5-minute washes in 1 ml Drosophila medium, ovaries were fixed in 1 ml 4% paraformaldehyde in PBS for 20 minutes at room temperature. The subsequent procedures, including permeabilization, second fixation, restriction enzyme digestion, and Lambda exonuclease treatment, were performed as described in the CAR assay protocol.

The Padlock probe ligation step was adapted for EdU staining by incubating tissues in 250 μl reaction (2 μM Padlock probe, 20 units/μl HI-T4 DNA ligase, 1×HI-T4 DNA ligase buffer) for 3 hours at 25°C, followed with two 3-minute washes with 1× PBST. After rolling circle amplification, tissues were immersed in 1 ml 3% BSA in 1× PBS for 5 minutes. EdU detection was performed according to the manufacturer’s instructions using the Click-iT EdU labeling kit (Alexa Fluor 488 dye). Following the Click-iT reaction, tissues were rinsed once with 3% BSA in PBS and then washed overnight in 1 ml 3% BSA in 1× PBS in the dark.

The subsequent steps followed the fluorescent probe in situ hybridization protocol of the CAR assay. Then ovaries were mounted on slides using VECTASHIELD antifade mounting medium with DAPI, and images were acquired using a Nikon CSU-W1 SoRa confocal system. Image deconvolution was performed using NIS Elements AR (6.02.03) software for further processing.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification of mtDNA variants in the germarium

To quantify the absolute numbers of two mtDNA variants at different developmental stages, z-stack images (0.3 μm/section) from the CAR assay were processed using Imaris (version 9.7.0). Mitochondrial nucleoids labeled with BglII and XhoI probes were detected using the “Spot” function in the “Cy3” and “Alexa488” channels, respectively. For each channel, spots were created using the “automatic creation” algorithm with the “different spot sizes (region growing)” option. The estimated XY and Z diameters were set to 0.4 μm and 0.8 μm, respectively. The “background subtraction” method was applied for spot detection, and spots with a quality score lower than 15 were filtered out.

To outline individual cysts, the “Surface” function was used. In the phalloidin channel, a surface was created using the “Add Contour” function, and cysts at different developmental stages were identified in 3D based on cell size and the number of ring canals stained with phalloidin. Once contours were drawn, the “Create Surface” function was applied to generate a surface for each cyst. The number of BglII or XhoI-labeled spots within a cyst was quantified using the “Shortest Distance to Surface” filter, with spots having a distance less than 0 considered inside the cyst. To quantify the distances between TFAM and BglII, TFAM and XhoI, or BglII and XhoI- labeled mitochondrial nucleoids in the germarium, spots were generated for each channel using the method described above. The “Shortest Distance to Spots” statistics were obtained under the “Statistics” function for each spot pair and used for calculation and plotting.

To calculate detection efficiency, we accounted for the fact that not all nucleoids were labeled by TFAM. Therefore, the total number of nucleoids was defined as the sum of TFAM-positive puncta and CAR puncta lacking TFAM. Detection efficiencies were calculated as follows:

$$\text{efficiency}\text{of}\text{XhoI}\text{probe}=\text{total}\#\text{XhoI}\text{puncta}/\left(\#\text{TFAM}\text{puncta}+\#\text{XhoI}\text{puncta}\text{lacking}\text{TFAM}\text{signal}\right)$$

$$\text{efficiency}\text{of}\text{BglII}\text{probe}=\text{total}\#\text{BglII}\text{puncta}/\left(\#\text{TFAM}\text{puncta}+\#\text{BglII}\text{puncta}\text{lacking}\text{TFAM}\text{signal}\right)$$

The number of each mtDNA variant was then calculated by normalizing the corresponding spot counts to the detection efficiency (detection efficiency of XhoI probe, 0.46; detection efficiency of BglII probe, 0.36), using the following formulas:

$${\text{copy}\#mt:ND2}^{ins}={\#mt:ND2}^{ins}\text{spots}/\text{detection}\text{efficiency}\text{of}\text{XhoI}\text{probe}$$

$${\text{copy}\#mt:CoI}^{T300I}={\#mt:CoI}^{T300I}\text{spots}/\text{detection}\text{efficiency}\text{of}\text{BglII}\text{probe}$$

Heteroplasmy was calculated using the following formula:

$$\text{heteroplasmy}={\text{copy}\#mt:CoI}^{T300I}/\left({\text{copy}\#mt:ND2}^{ins}+{\text{copy}\#mt:CoI}^{T300I}\right)$$

Quantification of EdU-positive CAR signals

EdU-positive CAR signals in the germarium were quantified using ImageJ (v2.16.0). Images were pre-processed using the “Deconvolution” and “Subtracting background” functions, then split into single channel images. The DAPI channel was used to delineate different developmental stages. For the BglII or XhoI probe channels, individual puncta were segmented using the “Adjust Threshold” function. Connected objects were separated using the “Watershed” and “Brush” tools. The “Analyze Particles” tool was then used to identify and analyze objects, with those smaller than 8.5 pixels filtered out. Next, segmented objects were superimposed onto the EdU channel, and objects without any overlap with the EdU signal were removed using the “ROI manager”. The ratio of EdU-positive CAR signals was calculated by determining the proportion of CAR signals that were either completely or partially overlapped with EdU relative to the total number of CAR signals, in the indicated germarium stages.

Statistical analyses

Two-tailed Student’s t-test was used for statistical analyses. The difference was considered statistically significant when p<0.05. Data are presented as mean± 95% confidence interval or mean±SD.

Supplementary Material

Movie 1

Supplemental video 1. 3D rendering of CAR images and germarium cysts from heteroplasmic flies maintained at 18°C (related to Figure 3)

The mt:ND2^ins (green) and mt:CoI^T300I (red) mtDNA were labelled with XhoI and BglII probes, respectively. The 2-cell cyst (orange), 4-cell cyst (yellow), region 2A (blue) and region 2B (magenta) are shown. Scale bar, 5 μm.

Movie 2

Supplemental video 2. 3D rendering of CAR images and germarium cysts from heteroplasmic flies maintained at 29°C (related to Figure 3)

The mt:ND2^ins (green) and mt:CoI^T300I (red) mtDNA were labelled with XhoI and BglII probes, respectively. The 2-cell cyst (orange), 4-cell cyst (yellow), region 2A (blue) and region 2B (magenta) are shown. Scale bar, 5 μm.