A scalable embryonic stem cell–based platform for efficient generation of mitochondrial DNA mutant mice

Significance

Mitochondrial DNA (mtDNA) accumulates mutations at a high rate, and more than 260 pathogenic germline mtDNA mutations have been identified in humans, producing diverse and often tissue-specific disorders. In addition, numerous population-specific mtDNA variants have arisen through evolution and may influence adaptation and disease susceptibility. However, the lack of animal models representing this diversity has limited mechanistic insight and therapeutic development. Our scalable embryonic stem (ES) cell–based platform enables efficient generation of mouse models carrying functional mtDNA mutations across a broad spectrum of physiological effects. This resource will facilitate systematic investigation of mtDNA variation in health, disease, and evolution, and accelerate efforts to develop treatments for mitochondrial disorders.

Abstract

Mitochondria are central to energy metabolism and cellular signaling, and mutations in mitochondrial DNA (mtDNA) can disrupt these processes and contribute to human disease. However, progress in defining how mtDNA variation influences adaptation, pathophysiology, and disease susceptibility has been limited by the lack of suitable animal models. Although recent base-editing approaches enable direct mtDNA modification, their low efficiency restricts the generation of diverse models reflecting human mtDNA variation. Here, we develop a scalable embryonic stem (ES) cell–based platform for efficient production of mtDNA mutant mice. Random mutagenesis using an error-prone mtDNA polymerase generates a broad spectrum of mtDNA mutations, which are transferred into ES cells via a multiplexed cybrid fusion strategy coupled with sensitive mutation detection. Optimized ES cell–embryo aggregation enables robust contribution of mtDNA mutant ES cells to host embryos, producing chimeric mice with germline transmission. Using this platform, we generate a library of 155 donor fibroblast lines carrying distinct homoplasmic single-nucleotide mtDNA mutations that produce diverse mitochondrial phenotypes, including impaired oxidative phosphorylation, increased reactive oxygen species, and altered mitochondrial membrane potential. We further generate 34 female C57BL/6 ES cell lines harboring 18 mtDNA mutations across a range of heteroplasmy levels, yielding multiple chimeric mice and achieving germline transmission for one mutation. These data reveal a strong correlation between mitochondrial function and early embryonic development, suggesting a minimal energetic threshold required for normal development. This scalable resource enables systematic investigation of mtDNA variation in physiology, adaptation, disease mechanisms, and therapeutic development.

Animal models are essential for understanding mitochondrial genetics and how mitochondrial DNA (mtDNA) mutations and variations contribute to human adaptation, disease susceptibility, and pathogenesis (1). Although multiple mtDNA mutant mouse models with stable germline transmission have been generated (2–7), their numbers remain critically limited compared with the vast spectrum of human mtDNA mutations and variations, which span all mitochondrially encoded genes (13 polypeptide-encoding genes, 2 rRNA, and 22 tRNA genes) and have distinct impacts on mitochondrial physiology. Consequently, there is a pressing need for a comprehensive collection of mtDNA mutant mouse models that carry mutations functionally analogous to human variants. Such a resource is crucial for dissecting the diverse and complex functional outcomes arising from the spectrum of human mtDNA mutations.

Transmitochondrial mouse models were originally produced using cybrid fusion between enucleated mtDNA mutant donor cells and either intact female mouse embryonic stem (fmES) cells (8) or fmES cells depleted of endogenous mtDNA using rhodamine-6G (R6G) (9). These models relied on selection for chloramphenicol-resistant mutations and initial mutations were too deleterious and rapidly lost in vivo (9). While this approach enabled the creation of most currently available mtDNA mutant mouse lines (2–6), it is inherently limited by low fusion efficiency, the need for extensive backcrossing of 129-derived fmES cells to C57BL/6 (B6), and its reliance on mtDNA mutations isolated through mutagenesis or toxin resistance selection, both of which are time-consuming and inefficient (10).

Recently, direct mtDNA editing has become possible through the development of mitochondria-targeted DNA base editors that use sequence-specific programmable DNA-binding proteins such as transcription activator–like effectors (TALEs) (11–15). Although this technology is powerful and capable of targeting many mtDNA positions, the selection of editable sites in mice has so far been restricted primarily to homologous human pathogenic mutations (13) or artificially introduced nonsense mutations (14). The former do not always reproduce the functional consequences observed in human disease (5, 16), whereas the latter are typically too deleterious and rapidly lost in the germline (3). Moreover, TALE design and specificity testing require extensive effort before phenotypic effects can be assessed. These constraints limit current mtDNA editing tools from generating a broad and functionally relevant collection of mtDNA mutant mouse models.

To address these limitations, we developed a scalable ES cell–based platform that integrates mtDNA mutagenesis, cybrid technology, high-sensitivity mutation detection, and optimized mouse transgenesis. Using this platform, we generated a library of 155 donor cell lines carrying distinct single homoplasmic mtDNA mutations with diverse mitochondrial functional effects and produced 34 B6 fmES cell lines harboring 18 different mtDNA mutations across a range of heteroplasmy levels. In addition, we optimized ES cell–embryo aggregation to enable efficient contribution of mutant fmES cells to developing embryos, creating chimeric mice with germline transmission. This streamlined platform provides an efficient path for generating mtDNA mutant mice and establishes a foundation for systematic investigation of the functional diversity of mtDNA mutations in evolution, physiology, and disease.

Results

Generation of a Library of MtDNA Mutant Cells Containing Diverse Functional Mutations.

A major limitation of the fmES–cybrid approach is its restricted ability to generate mtDNA mutations that are functionally relevant to human variants. To overcome this, we employed the error-prone mutant form of mtDNA polymerase γ (PolG), which is known to efficiently induce mtDNA mutations (17, 18). To generate a broad collection of mouse cell lines harboring single mtDNA mutations, we crossed wild-type (WT) female B6 mice with heterozygous male Polg-mutant mice, isolated embryonic mouse fibroblasts (MEFs) at E13.5 from heterozygous embryos, and cultured them for various times, allowing the mutant PolG to introduce limited mutagenesis into the WT mtDNA (SI Appendix, Fig. S1A). We sequenced mtDNA from PolG MEFs after 5, 10, and 20 days in vitro (DIV) and found that MEFs at 10-DIV exhibit a mutation rate of 0.7 mutations per 10 kb— roughly equivalent to our goal of one mutation per mtDNA molecule (SI Appendix, Fig. S1B).

We next performed cybrid fusion using enucleated 10-DIV MEFs and a mtDNA-depleted ρ° cell line derived from thymidine kinase–negative (TK⁻) LM cells, thereby removing the mutant PolG background and preventing further mutagenesis. Fused cells were treated with ethidium bromide (EtBr, 1 μg/mL) for 20 d to reduce mtDNA content to fewer than one copy per cell, followed by EtBr withdrawal and cloning in selective medium to eliminate any ρ° cells lacking mtDNA (Fig. 1A). In an initial trial following this procedure we generated 48 clones. Full-length mtDNA sequencing identified that 30 of these (62%) were mutant clones, including 15 (31%) that were expected to have functional consequences—single mutations in rRNA or tRNA genes or nonsynonymous (NS) changes in protein-coding genes (Fig. 1B and SI Appendix, Fig. S1 C and D).

Fig. 1.

Generation and characterization of mtDNA mutant cells. (A) Schematic overview of the workflow used to generate mtDNA mutant cells from enucleated Polg-mutant MEFs fused with a mtDNA-depleted ρ° cell line from thymidine kinase–negative (TK⁻) LM cells. (B) Percentage of isolated clones harboring different categories of mtDNA genotypes, including WT, heteroplasmic mutations, and homoplasmic mutations classified as single rRNA, single tRNA, single NS, single synonymous (S), single control-region (CR), or multiple mutations from two trials. (C) OXPHOS activity measured as OCR using the Seahorse Bioanalyzer, and (D) ROS levels quantified as relative fluorescence intensity (FI) of the MitoSOX superoxide indicator in the 15 mutant clones containing single functional homoplasmic mutations (rRNA, tRNA, and NS) from the 1st trial, shown in comparison to the WT control. n = 6 for OCR and 4 for MitoSOX. One-way ANOVA (multiple comparisons). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

To assess functional effects of these single homoplasmic mutations, we performed a series of phenotypic analyses, including Seahorse bioanalysis to evaluate oxidative phosphorylation (OXPHOS) activity through oxygen consumption rate (OCR) (Fig. 1C), MitoSOX-RED staining to quantify mitochondrial reactive oxygen species (ROS) production (Fig. 1D), and JC-10 staining to measure mitochondrial membrane potential (SI Appendix, Fig. S1E). These analyses revealed distinct patterns of functional alterations across clones, with 9 of the 15 mutants (60%) showing significant changes in at least one assay.

To expand the library and confirm the success of our methodology, we performed a second round of cybrid fusion and cloning. Among 384 clones isolated and sequenced in this trial, 140 (36%) carried homoplasmic single functional mutations—similar to the yield from the first trial (Fig. 1B and SI Appendix, Fig. S1C).

Optimization of Cybrid Fusion with B6 FmES Cells.

To improve the efficiency of the ES cell–cybrid fusion approach (Fig. 2A), we generated a B6 fmES cell line (fB4) (Fig. 2B) that, unlike 129-derived fmES cells used in earlier transmitochondrial studies, does not require extensive backcrossing. We derived the fB4 fmES cells from mTmG reporter mouse blastocysts, whose ubiquitous expression of the Tomato fluorescent protein enables convenient tracking of ES-cell contribution during embryonic and postnatal development. The fB4 line exhibited a 39, XO female karyotype—reflecting loss of one X chromosome during culture—consistent with the karyotype previously reported for 129-derived fmES cells (9) (Fig. 2B). The germline competency of the fB4 line was confirmed by its ability to generate fully ES-cell–derived pups (SI Appendix, Fig. S2A).

Fig. 2.

Optimized strategy to generate mtDNA mutant fmES cells. (A) Schematic overview of the workflow for generating mtDNA mutant fmES cells from LM(TK-) donor cybrids. (B) Bright-field, red-fluorescent, and merged images of a fmES-cell clone (fB4) derived from mT/mG reporter mice. The Bottom panel shows its 39, XO female karyotype. (C) Number of WT and mutant (G553A) fmES clones isolated from fusion experiments using different fusion reagents. The commercial reagent (Sigma P7306) contains 50% PEG (w/v) and 10% DMSO (w/v) in PBS, whereas the homemade reagents were prepared in DMEM. (D) Number of fmES clones (out of 24) with different mtDNA genotypes—WT and various levels of mutant (G553A)—derived from donor cells treated for 3 d with 0.6 to 1.0 µg/mL R6G. The fusion reagent used was 65% PEG + 10% DMSO.

Polyethylene glycol (PEG) is widely used for cell fusion, including in earlier transmitochondrial studies; however, commercial PEG formulations are typically optimized for hybridoma production and may not be ideal for ES cells. To better support ES cell fusion, we prepared homemade PEG solutions in DMEM at varying concentrations, with or without dimethyl sulfoxide (DMSO), which is known to enhance fusion efficiency. Fusion experiments using the G553A mutant clone and fB4 cells showed that in comparison to a commonly used commercial reagent (50% PEG and 10% DMSO in PBS) our homemade 65% (w/v) PEG solution was superior, with the addition of 10% (v/v) DMSO even further improving fusion efficiency as measured by the number of isolated clones carrying the G553A mutation (Fig. 2C).

With an optimized fusion reagent established specifically for ES cell fusion, we next evaluated conditions for R6G-induced depletion of endogenous mtDNA in fmES cells. fB4 cells were treated with different concentrations of R6G and fused with the G553A mutant cells. Mutation levels in 24 clones per condition were assessed by high-resolution melting analysis (SI Appendix, Fig. S2B). Higher R6G concentrations produced more efficient mtDNA depletion, yielding a greater proportion of clones containing mutant mtDNA and at higher heteroplasmy levels. Treatment with 0.85 or 1.0 μg/mL R6G resulted in complete elimination of WT clones, and 20 of 24 clones from the 1.0 μg/mL condition were entirely homoplasmic for the G553A mutation (Fig. 2D).

Multiplexed Fusion Enables Efficient Generation of MtDNA Mutant FmES Cells.

Although our optimized ES cell–cybrid fusion protocol is effective, it processes only one mtDNA mutant donor line per fusion, limiting throughput. To overcome this bottleneck, we developed a multiplexed fusion strategy that incorporates multiple donor cell lines in a single fusion experiment (Fig. 3A).

Fig. 3.

Efficient generation of mtDNA mutant fmES cells via multiplexed fusion. (A) Schematic overview of the workflow used to generate mtDNA mutant fmES cells using multiplexed fusion. (B) Number of mutant fmES clones isolated from fusion experiments using different donor cell lines, with or without HAT-medium selection. (C) List of nine selected mutant fmES clones harboring distinct mtDNA mutations for phenotypic characterization. (D) OXPHOS activity measured as OCR in the 9 mutant clones vs. a WT control. n = 6. **P < 0.01, ***P < 0.001, ****P < 0.0001.

To enable multiplexing, we first refined postfusion fmES cloning conditions by omitting hypoxanthine–aminopterin–thymidine (HAT) selection, which had been used in previous studies to eliminate residual TK⁻ donor cells. HAT medium, however, imposed variable stress on different mtDNA mutations and produced inconsistent yields of mtDNA mutant fmES clones across donor lines (Fig. 3B). Because ES-cell colonies have a distinct morphology and the donor cell fractions were fully enucleated, HAT selection was deemed unnecessary. Indeed, omitting HAT selection markedly increased transmitochondrial success rates with minimal variability among donor lines (Fig. 3B), establishing conditions suitable for multiplexed fusion.

Multiplexed fusion also requires an efficient and accurate method to detect mtDNA mutations in resulting fmES colonies. The high-resolution melting analysis (SI Appendix, Fig. S2B), though efficient, provides limited accuracy (~25% resolution). We therefore developed a high-throughput and high-accuracy mutation detection assay based on RNase H–dependent PCR (rhPCR) (19), which reliably detects mutant mtDNA at <1% abundance and enables straightforward quantification using standard curves generated from defined mixtures of WT and mutant oligonucleotides (SI Appendix, Fig. S3A).

With these improvements, we performed multiplexed fusions using up to 12 mutant donors simultaneously and successfully generated 34 fmES clones spanning 18 distinct mtDNA mutations and a range of heteroplasmy levels (SI Appendix, Table S1). These mutations were selected to capture a broad spectrum of mitochondrial functional impairment based on their effects in the fibroblast donor cells (Fig. 1). Several mutations were also chosen for their proximity to known pathogenic human mtDNA variants (SI Appendix, Table S2).

To evaluate the functional impact of these mutations, we selected nine representative fmES clones (Fig. 3C) for Seahorse bioanalysis. Eight exhibited significantly reduced OCR (Fig. 3D), and two of these also showed elevated extracellular acidification rate (SI Appendix, Fig. S3B), indicative of a compensatory increase in aerobic glycolysis in response to impaired OXPHOS activity. Despite these metabolic alterations, all mutant fmES clones maintained normal ES-cell morphology and expressed key pluripotency markers, including Nanog (SI Appendix, Fig. S3C) and Oct4 (SI Appendix, Fig. S3D).

Efficient ES Cell–Embryo Aggregation Using a Microwell-Based Platform.

The use of 129- rather than B6-derived fmES cells in previous transmitochondrial studies was largely driven by the higher transgenesis efficiency of 129 ES cells. Although our fB4 fmES line demonstrated robust germline competency (SI Appendix, Fig. S2A), multiple attempts using mtDNA mutant fmES clones failed to produce sufficient chimerism or germline transmission through conventional blastocyst injection. This limitation likely reflects the combined influence of the B6 genetic background and the metabolic defects associated with mtDNA mutations.

Previous work has shown that ES cell–embryo aggregation is highly effective for introducing B6 ES cells into the germline by aggregating ES cells with denuded 4- or 8-cell embryos (20, 21). However, these methods rely on manually generated depression wells made with an aggregation needle, a time-consuming approach that yields wells of inconsistent dimensions. To address this, we established a high-efficiency aggregation system using Aggrewell-400 microwell plates (Stemcell Tech), which contain thousands of uniformly shaped V-bottom microwells per well and provide ideal geometry for promoting ES cell–embryo aggregation (Fig. 4 A and B). These microwells enable rapid, reproducible setup and support much higher throughput than traditional depression wells.

Fig. 4.

ES cell–embryo aggregation and blastocyst development. (A) Illustration of the aggregation of fB4 mES cells with CD1 8-cell embryos using the Aggrewell400 microwell plate. (B) Images of embryos at different stages of ES cell–embryo aggregation (1 to 4: 8-cell; 5&6: blastocysts). (C) % of embryos (out of 24 per clone per trial; average of two trials) with different in vitro developmental outcomes using the 9 mtDNA mutant mES clones. (D) Correlations between mitochondrial OXPHOS function, measured by OCR, and normal (red) or arrested (blue) blastocyst development across the 9 clones. (E) Bright-field, red-fluorescent, and merged images of a developed blastocyst generated through aggregation with A7A mES cells.

Early Embryonic Developmental Defects Caused by MtDNA Mutations.

Using this platform, we generated chimeric embryos by aggregating denuded 8-cell CD-1 embryos with each of the nine mtDNA mutant fmES clones (Fig. 3C). Monitoring in vitro development to the blastocyst stage revealed increased frequencies of arrested development and reduced frequencies of normal development associated with mtDNA mutations, most notably in clones P19B and P19C (Fig. 4C), which exhibited the most severe metabolic defects (Fig. 3D and SI Appendix, Fig. S3B). These developmental defects were negatively correlated with mitochondrial OXPHOS function across the nine clones (Fig. 4D), suggesting an essential role for mitochondrial energy metabolism in early embryonic development.

The constitutive expression of Tomato from the fmES cells provided a qualitative readout of ES cell contribution in resulting blastocysts (Fig. 4E). To obtain quantitative measurements, we exploited a natural mtDNA sequence variant at nt9348 (A in CD-1, G in B6) and developed a rhPCR-based genotyping assay, enabling precise assessment of ES-cell contribution using a single universal assay. This analysis revealed significantly reduced ES-cell contribution in blastocysts derived from the P19B and P19C clones (SI Appendix, Fig. S4A), consistent with their severe metabolic defects and abnormal blastocyst development (Figs. 3D and 4C). Moreover, ES-cell contribution across all nine clones was negatively correlated with mitochondrial OXPHOS function (SI Appendix, Fig. S4B), further reinforcing the importance of mitochondrial energy metabolism in early embryonic development.

Chimeric Production and Germline Transmission.

Blastocysts derived from five mutant clones, including A7A, A7C, A2A, P19B, and P19C, were selected for transplantation into pseudopregnant CD-1 females. All clones except P19C, which exhibited the most severe OXPHOS defect (Fig. 3D), produced chimeric offspring, with A7A showing the highest contribution, including one female displaying an almost entirely black coat (Fig. 5A and SI Appendix, Table S3). While A7C and P19B also generated chimeric pups, those with greater than 50% contribution exhibited blunted growth and died prematurely, whereas surviving pups showed low levels of chimerism. Both clones are associated with relatively severe OXPHOS defects (Fig. 3D), indicating the existence of a minimal threshold of OXPHOS function required for normal development.

Fig. 5.

Chimera generation and germline transmission. (A) Photographs of chimeric mice generated from three heteroplasmic mtDNA mutant mES cell lines. The black coat color is contributed by the mutant fB4 mES cells, whereas the white coat color originates from CD1 host embryos. (B) Mutation levels of Ts1-6892 determined by tail genotyping of three A7A mtDNA mutant pups demonstrating successful germline transmission. (C) Distribution of Ts1-6892 mutation levels across 12 organs in A7A F1 mouse #2 at 3 mo of age.

Breeding of a high-chimerism A7A female with albino B6 males produced a litter of three all-black pups, demonstrating successful germline transmission. Tail genotyping confirmed the presence of the A7A mutation at Ts1-nt6892 in all three offspring, with heteroplasmy levels ranging from 55 to 92% (Fig. 5B). All three pups grew normally into adulthood, with no overt defects in growth or weight gain compared with age-matched B6 mice. At 3 mo of age, two male mice (55% and 80% tail heteroplasmy) were killed to evaluate tissue-specific distribution of the Ts1-nt6892 mutation. Both mice showed similar results, with all tissues examined exhibiting mutation levels comparable to those in tail biopsies (Fig. 5C and SI Appendix, Fig. S5), suggesting stable transmission and maintenance of this relatively mild mutation across tissues. This finding is consistent with those reported for a similar mtDNA tRNA (Ta-nt5024) mutant mouse model (22).

Discussion

Due to the lack of an efficient DNA repair system and the highly oxidative environment in which they reside, mtDNA exhibits a much higher mutation rate than nuclear DNA (nDNA) (23). Consequently, more than 260 pathogenic germline mtDNA mutations have been identified across the 16.5 kb human mitochondrial genome, causing diverse and often tissue-specific disorders. In addition to pathogenic variants, numerous population-specific mtDNA polymorphisms have arisen through evolution, potentially contributing to environmental adaptation and influencing disease susceptibility. Although several mouse models harboring mtDNA mutations have been generated (2–7), they fail to capture the breadth of human mtDNA variations and the diversity of associated functional outcomes. As such, there is a critical need for novel mtDNA mutant models and new approaches for expediting their development. To meet this critical need, we developed a scalable, efficient, and higher-throughput platform that overcomes long-standing barriers that have constrained the generation of mtDNA mutant mouse models.

A major challenge in the development of mtDNA mutant mouse models stems from the intrinsic difficulty of manipulating mtDNA (1). Conventional genome engineering methods for nDNA rely on DNA or RNA delivery, a capability not yet established for mitochondria. As a result, early approaches based on cytoplasmic transfer into fertilized eggs (24, 25) produced heteroplasmic animals rather than defined mtDNA mutations. The ES cell–cybrid fusion method enabled the first germline introduction of de novo mtDNA mutations (8, 9), but its usefulness has been constrained by the scarcity of available mutant mtDNAs and low throughput. Recent mitochondrial base-editing technologies represent a conceptual breakthrough for the field (11–15), however, these tools currently lack the efficiency and scalability needed to generate a broad collection of mtDNA mutant mouse models that captures the spectrum of human pathogenic and somatic mtDNA variation. To overcome these limitations, we developed a method to isolate and characterize mouse cell lines carrying distinct homoplasmic single mtDNA mutations. This resulted in a library of 155 donor lines, covering mutations in nearly every mitochondrial gene and displaying a wide range of functional defects. The library provides a unique and comprehensive resource for modeling the diversity of human mtDNA variation in vitro and can also be used to generate in vivo models through ES-cybrid technology.

A second limitation in the field comes from the reliance of the ES cell–cybrid approach on 129-strain fmES cells, which, despite high transgenesis efficiency, require lengthy backcrossing to the more commonly used B6 background. In addition, low fusion efficiency has prevented the establishment of a sufficiently large panel of fmES lines. To address these constraints, we generated a B6-derived fmES cell line incorporating a fluorescent marker to track ES-cell contribution. We further optimized the cybrid fusion workflow by developing a high-sensitivity rhPCR-based mutation detection assay and implementing a multiplexed fusion strategy capable of generating fmES cells from dozens of donor lines in a single experiment. This improvement resulted in the production of 34 mutant fmES cells containing 18 distinct mtDNA mutations that lead to diverse functional changes. The high efficiency of our approach would also allow rapid expansion of this library from additional donor mutant cells if resources permit.

One confounder in the use of B6 ES cells is their poor performance in conventional transgenesis—a limitation exacerbated by the metabolic defects associated with mtDNA mutations. This challenge was evident in our initial attempts at blastocyst injection. To circumvent these barriers, we developed a robust ES cell–embryo aggregation platform using microwell technology, enabling efficient chimerism and reliable germline transmission even for metabolically compromised mtDNA mutant B6 fmES lines. Using this platform, we successfully generated chimeric mice from three mutants and achieved germline transmission from one.

Beyond germline variation, mtDNA also accumulates somatic mutations throughout life—most prominently in long-lived, postmitotic tissues such as the brain (26). Many tumors likewise display substantially elevated mtDNA mutation burdens relative to their nontumorous counterparts (27, 28). Despite these associations, the extent to which somatic mtDNA mutations actively drive aging-related functional decline or tumorigenesis, as opposed to reflecting underlying metabolic stress or clonal expansion, remains poorly understood. Given that B6 is the preferred genetic background for aging and cancer studies, with a large number of knockouts and genetic tools available, our platform for generating mtDNA mutants has the potential to rapidly inform investigations on how somatic mutations contribute to aging and cancer.

Notably, the phenotypic effects of mtDNA mutations can be modified by variation in the nuclear genome (1, 29), including but not limited to genes involved in mitochondrial function. Our B6 fmES cells were derived from C57BL/6 mice obtained from The Jackson Laboratory (C57BL/6J). This commonly used B6 substrain carries a well-characterized loss-of-function mutation in the Nnt (nicotinamide nucleotide transhydrogenase) gene, which has been associated with impaired glucose homeostasis and altered mitochondrial redox balance (30, 31). Therefore, it is possible that some phenotypes observed in our mtDNA mutant mice will be influenced, at least in part, by Nnt deficiency. To more definitively assess the contribution of Nnt to mtDNA mutation-associated phenotypes, future studies will require crossing our mtDNA mutant mice onto a B6 background with WT Nnt, such as C57BL/6 N mice (e.g., from Charles River).

Together, the advances outlined in this study establish a powerful and generalizable platform for systematically modeling the functional diversity of human mtDNA mutations and polymorphisms in vivo. By enabling the generation of both pathogenic and physiologically relevant mtDNA variants—including those resembling somatic mutations associated with aging and cancer—this platform substantially expands the toolkit available to mitochondrial researchers. It opens the door to mechanistic dissection of how mtDNA variation influences metabolism, adaptation, and disease, and provides a strategically valuable foundation for accelerating therapeutic development through genetically precise mitochondrial disease models.

Materials and Methods

Animal Studies.

C57BL/6 (B6), PolG mutant (strain #017341) (32), and mTmG (strain #007676) (33) mice were obtained from The Jackson Laboratory. CD-1 mice, including vasectomized males, were purchased from Charles River Laboratories. All animals were housed in a temperature-controlled facility under a 12-h light/12-h dark cycle with ad libitum access to food and water. Cages were enriched with nestlets, and animal health, room temperature, and humidity were monitored daily by the Animal Resources Department. Mice were fed a standard laboratory diet (LabDiet 5001). All procedures were approved by the Salk Institute Institutional Animal Care and Use Committee and conducted in accordance with institutional and national guidelines.

Cell Culture.

PolG mutant MEFs were isolated from E13.5 embryos generated by mating 2-mo-old C57BL/6 females with heterozygous PolG mutant males, following a published protocol (34). LM(TK⁻) ρ° cells were produced by culturing LM(TK⁻) cells (ATCC) in 1 µg/mL ethidium bromide (EtBr) for 30 d, followed by cloning and genotyping to confirm complete mtDNA loss. All cells were maintained in DMEM containing high glucose, GlutaMAX (1×), pyruvate (1 mM), uridine (50 µg/mL), 10% FBS, and Anti-Anti (1×) at 37 °C with 5% CO₂.

Cybrid Fusion.

Cybrid fusion was performed as described previously (35). Briefly, 2 × 10⁷· 10-DIV PolG mutant MEFs were enucleated using cytochalasin B followed by Ficoll gradient centrifugation. The enucleated fraction was collected, washed, and fused with 2 × 10⁶ LM(TK⁻) ρ° cells using 1 mL of fusion reagent (50% PEG w/v, 10% DMSO; Sigma P7306). After 1 min, cells were washed and plated in selection medium lacking pyruvate and uridine for 2 wk. Resulting cells were treated with 1 µg/mL EtBr for 20 d, recovered for 3 d, cloned in selection medium, expanded, sequenced, and characterized.

ES Cell Culture.

Culture ware was coated with Matrigel (1:100, Corning) and plated with mitomycin C–treated DR4 feeder cells (ATCC) at 2 × 10⁶ cells per 25-cm². ES cells were maintained on feeders in 2i+LIF medium consisting of DMEM with high glucose, GlutaMAX, pyruvate (1 mM), uridine (50 µg/mL), 15% KnockOut Serum Replacement, 0.1 mM 2-mercaptoethanol, 200 U/mL ESGRO LIF, insulin (5 µg/mL), PD0325901 (1 µM), CHIR99021 (3 µM), and Anti-Anti (21). fB4 fmES cells were generated from blastocysts collected from 2-mo-old homozygous mTmG females mated with homozygous mTmG males, following a previously established protocol (21). Blastocysts were isolated at 3.5 dpc, cultured on feeder cells for 6 d, and resulting ES colonies were expanded, karyotyped, and tested for transgenesis.

ES Cell–Cybrid Fusion.

For ES cell–cybrid fusion, fB4 fmES cells were treated with Rhodamine 6G (R6G, 0.6 to 1 µg/mL) for 3 d to deplete endogenous mtDNA. Cell fusion was performed as described above using a range of fusion reagents, including homemade PEG 1,500 formulations (50% or 65% in DMEM, w/v), with or without 10% DMSO. For multiplexed fusion, 12 mutant donor cell lines were collected, mixed at equal ratios, and enucleated prior to fusion. Following fusion, ES cybrids were cultured on feeder cells for 7 d in complete ES cell medium with or without HAT to allow colony formation, after which individual colonies were picked, expanded, genotyped, and characterized. Clones were maintained on feeders, which were removed prior to Seahorse metabolic analysis.

Seahorse Bioanalysis.

Seahorse assays were performed according to the manufacturer’s instructions using the Seahorse XFe96 Analyzer (Agilent) (36). LM(TK⁻) cells were plated at 2 × 10⁴ cells per well and feeder-free ES cells at 4 × 10⁴ cells per well.

Genotyping.

Genomic DNA was isolated from mouse tail biopsies or cultured cells using the DNeasy Kit (Qiagen). For single-embryo mtDNA genotyping, individual embryos were transferred into 1 µL DirectPCR Lysis Reagent (Viagen), incubated for 5 min at room temperature, diluted 1:10 with nuclease-free water, and used directly for PCR. Two qPCR-based methods were used for mtDNA mutation quantification. High-resolution melting analysis was performed using the Bio-Rad Precision Melt Analysis platform. Primer pairs were designed to flank the mutation site and generate ~50 bp amplicons, maximizing melting resolution. PCR was carried out in 384-well format on a Bio-Rad real-time PCR system. To generate standard curves, defined mixtures of WT and mutant oligonucleotides were prepared at 25% incremental ratios. Mutation levels for experimental samples were determined by comparing their melt profiles to these standard curves. High-sensitivity genotyping was performed using rhPCR (Rnase H-dependent PCR) technology (IDT) with custom allele-specific primer sets designed for each mtDNA mutation. PCR and subsequent analysis were conducted using the Bio-Rad 384-well platform. Standard curves were generated using defined mixtures of WT and mutant oligonucleotides, with the lowest mutation fraction set at 1%. Mutation levels in experimental samples were calculated using the fitted equations corresponding to each segment of the curve (0 to 10%, 10 to 90%, and 90 to 100%) (SI Appendix, Fig. S3).

Sequencing.

For full-length mtDNA sequencing, we followed a published protocol with modifications (37). Briefly, cells were collected and lysed with a 0.2% NP-40 solution to selectively disrupt mitochondria while leaving nuclei intact. The mtDNA released into the supernatant was purified using AMPure XP beads (Beckman Coulter). To further enrich for mtDNA, samples were treated with a mixture of endonucleases and exonuclease (AgeI, DraIII, FspI, PshAI, PspXI, SexAI, XmaI, and ExoV) to selectively digest residual nDNA, followed by an additional AMPure XP purification step. DNA libraries were prepared from purified mtDNA using the Nextera XT Sample Preparation Kit (Illumina), incorporating Tn5-mediated tagmentation and IDT 384-mer indexed primer pairs. Libraries were amplified for 15 cycles, purified using AMPure XP beads, pooled according to concentration, and sequenced on an Illumina NextSeq platform using 2 × 100 bp paired-end reads.

Following sequencing, reads were quality-checked, trimmed, and aligned to the mouse mitochondrial reference genome using BWA-MEM. Duplicate reads were removed, and mitochondrial variants were identified using GATK Mutect2 in mitochondria mode, which is optimized for high-depth mtDNA sequencing. Variant calls were filtered based on read depth, base quality, and strand bias, and heteroplasmy levels were calculated as the proportion of mutant reads at each position. Identified variants were annotated relative to the mitochondrial reference genome, and all candidate mutations were visually inspected in IGV to confirm read support.

Embryo Isolation and Aggregation.

Embryo isolation and aggregation were performed following a published protocol with modifications (21). Briefly, 4-wk-old CD-1 females were superovulated with PMSG and hCG and subsequently mated with CD-1 males. Eight-cell embryos were collected from 2.5 dpc females by oviduct flushing and washed in Advanced KSOM medium (Sigma). To remove the zona pellucida, embryos were incubated in pronase (5 mg/mL) at 37 °C for 7 min, then washed and transferred into individual microwells of a 24-well AggreWell-400 plate precoated with antiadherent solution (STEMCELL Technologies) and prefilled with 500 µL embryo medium. Next, ES cells were detached from feeder layers by treatment with collagenase IV, followed by gentle digestion with Accutase (STEMCELL Technologies) to obtain small clumps of approximately 10 to 50 ES cells. After sequential washing in ES cell medium and embryo medium, ES cell clumps were added to each microwell containing zona-free embryos. Following overnight aggregation, embryos were assessed for development to the blastocyst stage and subsequently genotyped.

Transplantation.

Blastocyst transfer was performed using a nonsurgical embryo transfer (NSET) device (Paratechs) following the manufacturer’s instructions (38). Briefly, 2 to 6-mo-old CD-1 females were mated with vasectomized males 2 d prior to embryo aggregation, and only females confirmed by the presence of a vaginal plug were used as recipients. One day after aggregation, embryos were washed in embryo medium, and morphologically normal blastocysts were loaded into the NSET device and transferred into pseudopregnant females at a ratio of ~20 embryos per female. Pups were delivered naturally or retrieved by cesarean section if parturition had not occurred by 21 dpc. Chimerism was assessed by coat color. Chimeric females were subsequently mated with albino C57BL/6 (B6) males (Jackson Laboratory), and the presence of black offspring was considered evidence of germline transmission, which was further confirmed by mtDNA genotyping of tail biopsies.

Statistical Analysis.

Statistical significance was determined using one-way ANOVA, followed by multiple-comparison testing with WT designated as the control group.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Study data are included in the article and/or SI Appendix.