To explore the cellular responses to mtDNA stress in the absence of oxidative phosphorylation deficiency, we employed a TFAM heterozygous knockout (Tfam+/−) mouse model. Cells and tissues from these animals exhibit modest or no significant differences in mtDNA-encoded transcripts and oxygen consumption rates, despite an approximately 50% depletion of mtDNA (Extended Data Fig. 1a–c)5,6. In addition to mtDNA depletion, Tfam+/− mouse embryonic fibroblasts (MEFs) have reduced oxidative mtDNA damage repair capacity and markedly altered mtDNA packaging, organization and distribution (Fig. 1a)6. Nucleoids in Tfam+/− MEFs were less numerous and exhibited a larger size distribution (Fig. 1a and Extended Data Fig. 1d). Thus, Tfam+/− cells provide a robust model to characterize cellular responses triggered by moderate mtDNA stress.

Figure 1: Tfam+/− cells exhibit mtDNA stress, elevated ISG expression and augmented type I interferon responses.
figure 1

a, Confocal microscopy images of MEFs stained with anti-DNA (DNA) and anti-HSP60 (Mito.) antibodies. b, Heat maps of microarray analyses. Genes in Tfam+/− MEFs exhibiting statistically significant (P < 0.05), twofold or greater increases over wild type (WT) are shown. c, d, Quantitative real-time-PCR (qRT–PCR) (c) and western blots (d) of basal ISG expression in two littermate wild-type and Tfam+/− MEF lines. e, qRT–PCR analysis of type I interferon expression in MEFs 9 h after cytosolic delivery of poly(I:C). Error bars indicate ±s.e.m. of triplicate technical replicates and are representative of three independent experiments. P < 0.001.

PowerPoint slide

Gene expression profiling of Tfam+/− MEFs revealed an unexpected enrichment of interferon-stimulated genes (ISGs) and antiviral signalling factors (Fig. 1b). Of the 45 most overexpressed genes, 39 were ISGs, including many with direct antiviral activity (Ifi44, Ifit1, Ifit3, Oasl2, Rtp4)7,8. We also observed increased expression of cytoplasmic RNA and DNA sensors, such as Ddx58 and Ifih1 and p200 family proteins Ifi203, Ifi204 and Ifi205, as well as transcription factors Irf7, Stat1 and Stat2, ISGs that function to positively reinforce the antiviral response. Direct measurement of basal ISG mRNA and protein expression in Tfam+/− MEFs validated the microarray results (Fig. 1c, d). Finally, Tfam+/− MEFs expressed three- to fourfold more Ifnb and Ifna4 upon transfection with the IFIH1 agonist poly(I:C) (Fig. 1e), consistent with enhanced type I interferon responses.

To ensure that the mtDNA stress and ISG expression phenotypes were not unique to Tfam+/− MEFs, we employed inducible TFAM depletion models (TFD). Analogous to Tfam+/− cells, TFD MEFs and bone-marrow-derived macrophages (BMDMs) displayed mtDNA stress phenotypes, augmented ISG expression, and heightened type I interferon responses to poly(I:C) (Extended Data Fig. 1d–i). Collectively, these data indicate that TFAM depletion induces mtDNA nucleoid stress that triggers antiviral ‘priming’, characterized by basally elevated ISG expression and potentiated type I interferon production.

Since mitochondrial stress can trigger the release of mtDNA into the cytosol to engage the NLRP3 inflammasome, we assayed for extra-mitochondrial mtDNA in TFD cells9,10. Analysis of pure cytosolic extracts revealed a three- to fourfold increase of specific mtDNA fragments from the D-loop regulatory region, indicating liberation of immunostimulatory mtDNA into the cytosol (Extended Data Fig. 2)11. Confocal and electron microscopy of TFD cells revealed significantly elongated, interconnected mitochondrial networks consistent with a hyperfused phenotype (Fig. 1a and Extended Data Figs 1e, g and 3a, b). Since mitochondrial fission facilitates proper nucleoid distribution and removal of damaged mtDNA, we examined whether mitochondrial hyperfusion in TFD cells governed mtDNA stress-induced ISG expression12,13. Knockdown of mitofusin 1 (Mfn1) induced fission and largely abrogated ISG expression in TFD MEFs (Extended Data Fig. 3c–e). Moreover, depletion of the mtDNA quality-control enzyme endo/exonuclease (5′–3′), endonuclease G-like (EXOG) exacerbated ISG expression in Tfam+/− MEFs (Extended Data Fig. 3f)14. Collectively, these data indicate that TFAM depletion promotes accumulation of aberrant mtDNA, which accesses the cytosol to engage innate immune signalling.

We next examined the involvement of the cytosolic DNA sensor cGAS in mtDNA stress signalling, as it mediates ISG expression in response to exogenous and endogenous immunostimulatory DNA species15,16,17. Knockdown of cGAS in Tfam+/− MEFs or TFAM depletion in cGas−/− MEFs largely abrogated ISG expression (Fig. 2a). Furthermore, ISG mRNAs in TFD cells were reduced 70–90% in the absence of STING, indicating cGAS–STING signalling is the predominant driver of mtDNA stress-induced ISG expression (Fig. 2b). STING signals via the TBK1–IRF3/7 axis to trigger antiviral gene expression, and knockdown of either TBK1 or IRF3 robustly blocked ISG expression in Tfam+/− MEFs (Fig. 2c, d)18,19. Consistent with IRF3 activating ISG transcription, we observed enhanced nuclear accumulation of IRF3 in TFD cells (Fig. 2e). Finally, using cGas−/− MEFs reconstituted with hemagglutinin (HA)-tagged, murine cGAS, we observed prominent re-localization of cGAS from nuclear and/or cytoplasmic pools to the vicinity of aberrant mtDNA nucleoids in TFD MEFs (Fig. 2f, g). Taken together, these results indicate that mtDNA stress facilitates cGAS-dependent sensing of cytoplasmic mtDNA, resulting in STING–TBK1–IRF3 signalling to trigger ISG expression.

Figure 2: mtDNA stress triggers ISG expression in a cGAS- and STING-dependent fashion.
figure 2

a, b, ISG expression in Tfam+/− MEFs transfected with the indicated short interfering RNAs (siRNAs; top panels), or wild-type (WT), cGas−/− (a), and Sting−/− (b) MEFs transfected with TFAM siRNAs (bottom panels). Ctrl, control. c, d, ISG expression in Tfam+/− MEFs transfected with the indicated siRNAs for 96 h. e, Western blots of whole-cell and nuclear extracts of wild-type and Tfam+/− MEFs or Tfamfl/fl ER-cre+/− (cre+/−) BMDMs exposed to 4-hydroxytamoxifen (4OHT) for 96 h. f, g, cGAS−/− MEFs reconstituted with cGAS–HA were transfected with the indicated siRNAs for 96 h, then stained with anti-DNA (DNA), anti-HSP60 (Mito.) and anti-HA (cGAS–HA) antibodies and imaged. cGAS co-localization scoring was performed as described in the Methods. Error bars indicate ±s.e.m. of triplicate technical (ad) or biological (g) replicates and are representative of three independent experiments. P < 0.05; P < 0.01; P < 0.001.

PowerPoint slide

To establish functional significance of mtDNA stress-induced antiviral priming, we challenged MEFs with herpes simplex virus 1 (HSV-1) or vesicular stomatitis virus (VSV) that express green fluorescent protein (GFP) for easy detection. In contrast to wild-type cells, which displayed robust viral GFP expression post-infection, Tfam+/− MEFs were markedly resistant to HSV-1 and VSV (Fig. 3a). In addition, Tfam+/− MEFs exhibited heightened type I interferon and ISG expression upon viral challenge, consistent with potentiated type I interferon responses in these cells (Extended Data Fig. 4a). Similar results were obtained upon challenge with the rodent gammaherpesvirus MHV-68 (Fig. 3b and Extended Data Fig. 4b). Furthermore, TFD BMDMs displayed augmented antiviral gene expression and markedly lower HSV-1- and VSV-encoded mRNA and protein 6–24 h post-infection (Extended Data Fig. 4c–f). Finally, we found that Tfam+/− mice exhibit basally elevated ISG expression, which confers resistance to acute infection by lymphocytic choriomeningitis virus (LCMV) Armstrong (Extended Data Fig. 5a and Fig. 3c).

Figure 3: mtDNA stress potentiates viral resistance.
figure 3

a, Viral GFP expression in MEFs infected with HSV-1-GFP or VSV-GFP at multiplicity of infection (MOI) 0.5 for 24 h. b, MHV68-GFP abundance and ISG expression in MEFs infected at MOI 0.5. A.U., arbitrary units. c, LCMV Armstrong glycoprotein (GP) and nucleoprotein (NP) gene expression 4 days after intraperitoneal (i.p.) infection of wild-type (WT) and Tfam+/− mice; n = 4. d, e, Nucleoid area (d) or ISG expression (e) of MEFs exposed to ddC for 96 h. f, ddC-exposed MEFs were infected with HSV-1-GFP or VSV-GFP at MOI 0.1 and imaged after 24 h. Error bars represent ±s.e.m. of triplicate technical (b, e) or quadruplicate biological (c) replicates and are representative of two independent experiments. P < 0.01; P < 0.001.

PowerPoint slide

To probe a direct requirement for mtDNA stress in antiviral priming in TFAM-deficient cells, we used dideoxycytidine (ddC), a deoxyribonucleoside analogue that specifically inhibits mtDNA replication and decreases mtDNA nucleoid size2,20. Treatment of wild-type MEFs with ddC resulted in reduced mtDNA copy number and decreased average nucleoid size without altering basal ISG expression (Extended Data Fig. 5b–d). In contrast, ddC drastically diminished mtDNA stress (that is, enlarged nucleoids measuring greater than 450 nm2) in Tfam+/− and TFD MEFs (Fig. 3d and Extended Data Fig. 5e), which was accompanied by attenuation of antiviral priming and basal ISG expression (Fig. 3e and Extended Data Fig. 5d, f). Moreover, ddC ablated the viral resistance phenotype of Tfam+/− MEFs (Fig. 3f). We observed similar decreases in type I interferon production and a reduction in the viral resistance phenotype in ddC-treated TFD BMDMs (Extended Data Fig. 5g, h, blue bars). These results demonstrate that mtDNA stress directly potentiates antiviral innate immunity.

The observation that ddC-treated wild-type BMDMs displayed reduced Ifnb and increased viral gene expression upon challenge with HSV-1, despite normal responses to cytosolic nucleic acids (Extended Data Fig. 5h, i, grey bars), indicates that virus-induced mtDNA stress may boost host antiviral responses, consistent with reports linking viral infection to mtDNA dysregulation21,22. The alphaherpesvirus protein UL12.5, encoded by HSV-1 and HSV-2, localizes to mitochondria and promotes rapid mtDNA depletion in human cells, which we confirmed in MEFs (Extended Data Fig. 6a)22,23,24. Since mtDNA depletion and nucleoid stress are often coupled, we explored nucleoid architecture and abundance kinetically during HSV-1 infection. Notably, 3 h after challenge with HSV-1, mtDNA stress was readily apparent, with nucleoids less evenly distributed and enlarged (Fig. 4a). After 6 h, 10% of nucleoids measured larger than 450 nm2, and there was a significant decrease in total nucleoid intensity (Fig. 4b). After 12 h, we observed pronounced mtDNA depletion. The mtDNA stress observed 3 to 6 h after HSV-1 challenge closely mirrored that of TFAM-deficient cells (Fig. 4b), as did TFAM protein levels (Fig. 4c). MHV-68 and HSV-2 triggered mtDNA stress similar to HSV-1, indicating that mtDNA stress is a common cellular perturbation during herpesvirus infection (Extended Data Fig. 6b, c). However, induction of mtDNA stress and TFAM depletion were not a general consequence of viral infection, as cells infected with VSV, influenza, LCMV or vaccinia possessed normal mtDNA architecture, TFAM expression and copy number (Fig. 4a–c and Extended Data Fig. 6c, d).

Figure 4: HSV-1 induces mtDNA stress and TFAM depletion sufficient to trigger ISG expression and necessary to fully engage antiviral immunity.
figure 4

ac, Wild-type (WT) MEFs were mock infected or infected with HSV-1-GFP or VSV-GFP at multiplicity of infection (MOI) 10 for the indicated times, and imaged after staining with anti-DNA (DNA), anti-HSP60 (Mito.), and anti-HSV (HSV-1) or GFP (VSV) antibodies (a). mtDNA nucleoid area was calculated as described in the Methods (b). Extracts were blotted as indicated (c). d, e, Wild-type MEFs were transduced with HSV-1 UL12 M185–FLAG-expressing or empty retroviruses (RV) and cells were stained with anti-DNA (DNA), anti-HSP60 (Mito.) and anti-FLAG antibody (UL12 M185) (d), and protein or ISG expression examined after 24 h (e). f, g, Protein and RNA expression in BMDMs infected with HSV-1 (UL12–FLAG) or UL12-deficient HSV-1 (ΔUL12 + UL98–FLAG) at MOI 2 for the indicated times. h, HSV-1 genome abundance in L929 cells that were infected as in f, g. A.U., arbitrary units. Error bars indicate ±s.e.m. of triplicate technical replicates and are representative of two independent experiments. P < 0.001; NS, not significant.

PowerPoint slide

Finally, we sought to determine whether HSV-1-induced mtDNA dysregulation is necessary to fully engage antiviral signalling. Transduction of MEFs and BMDMs with replication-incompetent retroviruses encoding only the mitochondria-targeted HSV-1 UL12 M185 gene product was sufficient to cause mitochondrial hyperfusion, nucleoid enlargement and mtDNA loss, indicative of mtDNA stress (Fig. 4d and Extended Data Fig. 7a)24. UL12 M185 expression was also sufficient to trigger TFAM depletion and antiviral priming (that is, augmented ISG mRNA and protein expression) (Fig. 4e and Extended Data Fig. 7a). To explore the effect of HSV-1-induced mtDNA stress on innate antiviral responses, we employed a recombinant, UL12-deficient HSV-1 strain (ΔUL12 + UL98–FLAG) that is severely impaired in its ability to induce mtDNA stress but replicates similarly to a matched UL12-sufficient strain (Extended Data Fig. 7b, c)25. Infection with ΔUL12 HSV-1 resulted in attenuated TBK1 phosphorylation and type I interferon and ISG expression between 3 to 6 h post-infection, despite comparable early HSV-1 gene expression (Fig. 4f, g). However, after 24 h, ΔUL12 HSV-1 genome abundance was roughly threefold higher compared to the UL12-sufficient control, consistent with impaired antiviral innate immunity (Fig. 4h). Finally, ΔUL12 HSV-1 elicited less robust antiviral innate immune responses in the vagina and more readily spread to dorsal root ganglia of wild-type mice due to a deficit in mtDNA stress-dependent antiviral priming (Extended Data Fig. 7d, e). These results reveal that herpesvirus-induced mtDNA stress is necessary to effectively engage ISG expression and antiviral priming, and suggest that cellular monitoring of mtDNA homeostasis represents an additional sensory mechanism to robustly engage antiviral innate immunity.

In closing, our work uncovers a novel cellular response to mtDNA stress that engages the antiviral innate immune response. Specifically, we show that mtDNA stress, induced by herpesvirus infection and mediated by loss of the mtDNA packaging protein TFAM, triggers a cGAS–STING–IRF3-dependent pathway to upregulate ISGs and potentiate type I interferon responses to viral infection (Extended Data Fig. 8). Our results support a model whereby viral-mediated disruption of mtDNA homeostasis serves as a cell-intrinsic indicator of infection that works in parallel with canonical virus sensing to enhance antiviral innate immunity. Conversely, pathologic type I interferon signatures promote autoimmune diseases such as systemic lupus erythematosus, and altered ISG expression correlates with radiation-resistant and metastatic phenotypes in some cancers26,27. Mitochondrial and mtDNA dysregulation have been noted in systemic lupus erythematosus, and perturbations in TFAM and/or mtDNA homeostasis are frequently observed in cancer28,29,30. Therefore, further investigation into this pathway will not only expand our knowledge of innate antiviral defence, but may also broaden our understanding of how mitochondria contribute to the pathogenesis of human diseases and ageing beyond their well characterized roles in metabolism, apoptosis and reactive oxygen species production.


Animal strains

Tfam+/− and Tfamfl/fl mice were previously described and maintained on a C57BL/6 background6,31. Tfamfl/fl mice were bred to Estrogen receptor (ER)–Cre transgenic mice from Jackson (stock no. 004682) for inducible, 4OHT-mediated deletion. All animal experiments were conducted in compliance with guidelines established by the Yale University Institutional Animal Care and Use Committee.

Antibodies and reagents

Rabbit anti-mouse TFAM polyclonal anti-sera was previously described6, rabbit anti-VSV polyclonal anti-sera was a gift from J. Rose at Yale University, mouse anti-Viperin was a gift from P. Cresswell at Yale University, and rabbit-anti IFIT3 was a gift from G. Sen at Cleveland Clinic. The following antibodies were obtained commercially: goat anti-HSP60 (N-20) and rabbit anti-calnexin (H-70) (Santa Cruz Biotechnology); mouse and rabbit anti-FLAG (F1804, F7425) (Sigma); mouse anti-DNA (CBL186) (Millipore); mouse anti-GFP (JL-8) (BD Biosciences); rabbit anti-HSV-1/2 (ab9533) and anti-histone H3 (ab1791) (Abcam); rat anti-HA-FITC (11988506001) (Roche); rabbit anti-NLRX1 (17215-1-AP) (Proteintech); mouse anti-α-tubulin (DM1A) (Neomarkers); mouse anti-GAPDH (6C5) (Ambion); and rabbit anti-DDX58 (D14G6), -IFIH1 (D74E4), -STAT1 (9172), -IRF3 (D83B9), -TBK1 (D1B4) and anti-phospho-TBK1 (D52C2) (Cell Signaling Technology). Mouse IFNα enzyme-linked immunosorbent assay (ELISA) and recombinant mouse IFNβ was from PBL Assay Science, and mouse IL-6 ELISA was from eBioscience. All primer sequences and siRNAs used are found in Extended Data Tables 1 and 2.

Cell culture

Primary wild-type, Tfam+/−, Sting−/− and cGas−/− MEFs were generated from E12.5–14.5 embryos, maintained in DMEM (Invitrogen) supplemented with 10% FBS (Atlanta Biological), and sub-cultured no more than five passages before experiments. Sting−/− MEFs were provided by G. Barber at the University of Miami32. L929 cells were obtained from ATCC and maintained in DMEM supplemented with 10% FBS. siRNA transfection of MEFs was performed with 25 nM siRNA duplexes (Integrated DNA Technologies) and Lipofectamine RNAiMAX reagent (Invitrogen) according to the manufacturer’s instructions. ddC (Sigma) was resuspended in PBS, added to MEFs or BMDMs at a final concentration of 10–20 μM, and replenished every 48 h. BMDMs were generated from bone marrow of 8–12-week-old littermate Tfamfl/flER-cre and Tfamfl/flER-cre+ mice and cultured on Petri plates in DMEM containing 10% FBS plus 30% L929 culture media. To induce Cre-mediated deletion, 1 μM 4OHT dissolved in DMSO (Sigma) was added to BMDM cultures on day 6 and incubated for an additional 2–3 days. Cells were then lifted from plates by incubating in cold PBS containing 1 mM EDTA, re-plated in fresh media containing 10% L929 conditioned media, and allowed to rest overnight before experimentation (for a total of 72 or 96 h of 4OHT exposure). Transfection of interferon-stimulatory DNA (ISD)33 and poly(I:C) (Sigma) into the cytosol of BMDMs was performed using Lipofectamine 2000 (Invitrogen). In brief, 1 × 106 BMDMs were seeded in 6-well dishes after 4OHT treatment, and transfected the next day with 4 μg ISD per well or 2.5 μg per well of poly(I:C) complexed at a ratio of 2:1 Lipofectamine 2000 to nucleic acid. Poly(I:C) transfection into the cytosol of MEFs was performed as described previously34.

Viral stocks and infections

VSV-G-GFP35, HSV-1-GFP36, MHV-68-GFP, HSV-237, vaccinia virus (strain WR) expressing bacteriophage T7 RNA polymerase38, influenza A PR8 NS1-GFP39, HSV-1 (UL12–FLAG) and HSV-1 (UL12Δ + UL98–FLAG)25 were maintained as described previously34,40,41. MEFs or BMDMs were infected at the indicated multiplicity of infection (MOI) in serum-free DMEM for 1 h, washed, and incubated for various times. Cells were then fixed and stained for microscopy, lysed for western blot, solubilized in buffer RLT Plus (Qiagen) for RNA extraction, or prepared for FACS analysis. FACS was performed by first trypsinizing MEFs, followed by labelling with LIVE/DEAD Fixable Far Red stain (Molecular Probes). Cells were then fixed with 4% paraformaldehyde, washed, and analysed on a FACSCalibur flow cytometry machine (BD). FACS plots were first gated on live cells before analysing viral GFP fluorescence. Viral gene expression in BMDMs was determined using qRT–PCR as described below, except that after values were normalized against GAPDH cDNA using the method, all data points were subtracted by one to centre on zero.

LCMV Armstrong infection of wild-type and Tfam+/− mice was performed as described previously42. In brief, 10-week-old female mice were infected with 2 × 105 plaque-forming units of virus intraperitoneally, and 4 days post-infection, mice were euthanized, tissues isolated, and total RNA prepared using RNeasy Plus kits (QIAGEN). After generating complementary DNA, samples were subjected to qPCR analysis as described below using published methods43,44.

In vivo HSV-1 infection, dorsal root ganglia isolation and viral titration

Six-week old female mice were purchased from Charles River Laboratories and treated with Depo Provera (GE Healthcare) 5 days before infection45. The vaginal canals of Depo Provera treated mice were swabbed with a Calginate swab (Fisher) and 106 plaque-forming units were delivered via pipette tip into the vagina. One day post-infection, vaginal tissue was isolated for RNA extraction. Infected mice were euthanized at indicated time points and dorsal root ganglia were dissected as previously described46. Dorsal root ganglia were homogenized using a motorized pestle and total DNA was isolated using the DNeasy Blood & Tissue Kit (Qiagen) according to the manufacturer’s instructions. Relative HSV-1 genome abundance was determined using primers specific for nuclear Tert and HSV-1 TK (thymidine kinase).

cGAS–HA and UL12 M185 cloning and retroviral expression

A plasmid encoding HSV-1 UL12 M185 SPA containing a 3× FLAG tag at the carboxy-terminus was described previously24. This construct, or a plasmid encoding murine cGAS–HA (Invivogen), was sub-cloned into the pMXs-IRES-Puro vector and replication incompetent retroviruses were packaged using plat-E cells according to the manufacturer’s instructions (Cell Biolabs). SV40 large T immortalized cGAS−/− MEFs were exposed to supernatants containing cGAS–HA retroviruses and incubated overnight. Two days post-transduction, 3 μg ml−1 puromycin was added to select a stable population of cells expressing cGAS–HA. Supernatants containing empty or UL12 M185 SPA retroviruses and 4 μg ml−1 polybrene were incubated with cells (5 × 104 MEFs or 2 × 105 BMDMs) in 12-well dishes for a period of 8 h. Viral supernatants were then washed off, fresh media was added to wells, and the cells were incubated for the duration of the experiment until lysis.

Quantitative PCR

To measure relative gene expression by qRT–PCR, total cellular RNA was isolated using RNeasy Plus RNA extraction kit (Qiagen). Approximately 400–2000 ng RNA was normalized across samples and cDNA was generated using the High Capacity cDNA RT kit (Applied Biosystems). cDNA was then subjected to qPCR using Fast SYBR Green Master Mix (Applied Biosystems) and primers as indicated on the ViiA7 Real Time PCR system (Life Technologies). Three technical replicates were performed for each biological sample, and expression values of each replicate were normalized against GAPDH cDNA using the method. For relative expression (fold), control samples were centred at 1; for relative expression (%), control samples were centred at 100%. Mitochondrial DNA copy number analysis was performed as described using primers specific to nuclear Tert and the D-loop region of mtDNA (listed in Extended Data Table 1)6. Relative HSV-1 genome abundance was determined using primers specific for nuclear Tert and HSV-1 UL30 or TK. Relative MHV68 genome abundance was determined using primers specific for nuclear Tert and MHV68 orf40. Relative vaccinia genome abundance was determined using primers specific for nuclear Tert and vaccinia virus DNA polymerase E9L.

Immunofluorescence microscopy

For all microscopy images containing mtDNA nucleoids and associated panels, cells were grown on coverslips and transfected, treated, and/or infected as described. After washing in PBS, cells were fixed with 4% paraformaldehyde for 20 min, permeabilized with 0.1% Triton X-100 in PBS for 5 min, blocked with PBS containing 10% FBS for 30 min, stained with primary antibodies for 60 min, and stained with secondary antibodies for 60 min. Cells were washed with PBS between each step. Coverslips were mounted with Prolong Gold anti-fade reagent containing DAPI (4′,6-diamidino-2-phenylindole; Molecular Probes). Cells were imaged on a Zeiss LSM 510 META with a 63× water-immersed objective. A digital scan zoom of 3.0 was used to enhance magnification. Images were pseudo-coloured and merged using ImageJ software (NIH). For microscopy images in Fig. 3, MEFs were infected as described and fixed with 4% paraformaldehyde for 20 min. Viral GFP fluorescence and phase contrast images were captured using an Olympus IX-71 inverted scope with a 10× (Fig. 3a) or 20× (Fig. 3f) objective. Viral GFP images were pseudo-coloured using ImageJ.

For nucleoid area quantification, approximately 10–15 unique fields of view from 10 distinct confocal images, comprising between 200–400 nucleoids, were captured at random. After incorporating scale information obtained from the LSM Image Browser (Zeiss), images were made binary and the area of each nucleoid was determined using the ‘Analyze Particles’ feature of ImageJ. Nucleoids were divided into the three size cut-offs: <200 nm2; 200–450 nm2; and >450 nm2, and the percentage of nucleoids falling within each category was plotted. The percentage of nucleoids >450 nm2 displaying cGAS co-localization was scored by calculating nucleoid area from 5 distinct images of siCtrl- and siTfam-transfected cGAS–HA reconstituted MEFs as described above. Nucleoids larger than 450 nm2 with a substantial cGAS co-localization signal were scored as positive.

Electron microscopy

MEFs grown in Petri dishes and on coverslips for orientation were fixed in 2.5% gluteraldehyde in 0.1 M sodium cacodylate buffer pH 7.4 for 1 h. The cells were rinsed in sodium cacodylate and those in Petri dishes were scraped and spun down in 2% agar. All samples were fixed in 1% osmium tetroxide for 1 h, stained en masse in 2% uranyl acetate in maleate buffer pH 5.2 for a further hour, rinsed and dehydrated in an ethanol series, and infiltrated with resin (Embed812 EMS) and baked over night at 60 °C. Hardened blocks were cut using a Leica UltraCut UCT. 60-nm sections were collected on formver/carbon-coated grids and contrast stained using 2% uranyl acetate and lead citrate. Samples were viewed on an FEI Tencai Biotwin TEM at 80 Kv. Images were taken using Morada CCD and iTEM (Olympus) software.

For mitochondrial perimeter quantification, approximately 10–15 unique electron microscopy images of each genotype were captured at random. After incorporating scale information from iTEM software, the perimeter along the outer membrane of each mitochondrion was traced and quantified using ImageJ. Mitochondria were divided into the three size cutoffs: <2 μm; 2–5 μm; and >5 μm, and the percentage of mitochondria falling within each category was plotted.

Oxygen consumption analysis

Cells were plated in XF96 plates (SeaHorse Biosciences) at 10,000 cells per well and the next day cellular O2 consumption was determined in a SeaHorse Bioscience XF96 extracellular flux analyser according to the manufacturer’s instructions. Cells were maintained at 37 °C in normal growth medium without serum.

Nuclear fractionation and western blotting

Whole-cell extracts were solubilized in SDS lysis buffer (20 mM Tris-HCl, 1% SDS, pH 7.5, containing protease and phosphatase inhibitors), boiled for 5 min, and DNA was sheared by sonication. For nuclear extraction, PBS-washed cell pellets were resuspended in 10 pellet volumes of RSB buffer (10 mM NaCl, 1.5 mM CaCl2, 10 mM Tris-HCl pH 7.5), swelled on ice for 10 min, homogenized with a motorized Teflon pestle, and the homogenate was centrifuged at 980g for 10 min to pellet nuclei. Pellets were washed five times in PBS, SDS was then added to a final concentration of 1%, and extracts were boiled for 5 min before sonicating to shear DNA and normalizing protein concentration. Western blotting was performed using standard protocols, and HSP60 was used as whole-cell extract loading controls, while histone H3 was probed as a nuclear loading control.

Detection of mtDNA in cytosolic extracts

Digitonin extracts from MEFs and BMDMs were generated largely as described previously47. Wild-type and Tfam+/− MEFs (7 × 106) or Tfamfl/fl ER-cre+/− BMDMs exposed to 4OHT for 72 h (1 × 107) were each divided into two equal aliquots, and one aliquot was resuspended in 500 μl of 50 μM NaOH and boiled for 30 min to solubilize DNA. 50 μl of 1 M Tris-HCl pH 8 was added to neutralize the pH, and these extracts served as normalization controls for total mtDNA. The second equal aliquots were resuspended in roughly 500 μl buffer containing 150 mM NaCl, 50 mM HEPES pH 7.4, and 15–25 μg ml−1 digitonin (EMD Chemicals). The homogenates were incubated end over end for 10 min to allow selective plasma membrane permeabilization, then centrifuged at 980g for 3 min three times to pellet intact cells. The first pellet was saved as the ‘Pel’ fraction for western blotting. The cytosolic supernatants were transferred to fresh tubes and spun at 17000g for 10 min to pellet any remaining cellular debris, yielding cytosolic preparations free of nuclear, mitochondrial and endoplasmic recticulum contamination. DNA was then isolated from these pure cytosolic fractions using QIAQuick Nucleotide Removal Columns (QIAGEN). Quantitative PCR was performed on both whole-cell extracts and cytosolic fractions using nuclear DNA primers (Tert) and mtDNA primers (Dloop1-3, Cytb, 16S and Nd4), and the CT values obtained for mtDNA abundance for whole-cell extracts served as normalization controls for the mtDNA values obtained from the cytosolic fractions. This allowed effective standardization among samples and controlled for any variations in the total amount of mtDNA in control and TFAM-deficient samples. Using this digitonin method, no nuclear Tert DNA was detected in the cytosolic fractions, indicating nuclear lysis did not occur.

Bioinformatic analyses

Total cellular RNA from wild-type and Tfam+/− littermate MEFs was prepared using RNeasy Plus RNA extraction kits (QIAGEN) and used for the expression microarray procedure in conjunction with the Emory University Integrated Genomics Core. RNA integrity was first verified by an Agilent Bioanalyzer and then amplified, labelled, and hybridized onto Mouse Gene 1.0 ST arrays (Affymetrix) using standard protocols recommended by the manufacturer, starting from 2 μg of total RNA. Data were normalized by the RMA method using GeneSpring software (Agilent) for each biological sample in duplicate. Student’s t-test was used to determine statistically significant changes in expression in Tfam+/− MEFs relative to wild type, with a cut-off P value of 0.0548. Heat maps were generated using MultiExperiment Viewer49.

Statistical analyses

Error bars displayed throughout the manuscript represent s.e.m. unless otherwise indicated, and were calculated from triplicate or quadruplicate technical replicates of each biological sample. For in vivo experiments, error bars were calculated from the average of triplicate technical replicates of 3–4 mice per point. Sample sizes were chosen by standard methods to ensure adequate power, and no randomization or blinding was used for animal studies. No statistical method was used to predetermine sample size. Statistical significance was determined using unpaired Student’s t-tests; *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant (P > 0.05). Data shown are representative of 2–3 independent experiments, including microscopy images, western blots and viral challenges.