Human total RNA from different tissues used in this study was purchased to Takara Bio/Clontech (Shiga, Japan).

MCF7, T47D, BT474, MDA-MB-231, MDA-MB-453, MDA-MB-468, RCC4, 786-0, U2OS mTUNE, and human fibroblasts were cultured in DMEM low glucose medium (1g/L; Thermo Fisher Scientific, Waltham, Massachusetts) supplemented with 10% FBS no longer than 20 passages. HUVEC cells were purchased from Lonza (Basilea, Switzerland) and grown in EGM2 medium for maximum 7 passages. They were all mycoplasma tested every 3 months and authenticated during the course of this project. Cells were subjected to 1% or 0.1% hypoxia for the periods specified in each experiment using an InVivO₂ chamber (Baker, Sandford, Maine).

U2OS mTUNE glioblastoma cell lines were kindly donated by Dr Christian Frezza. Three isogenic cell lines with different levels of heteroplasmy of mutated mtDNA were used: M7, M45 and M80 (24).

143B and 143B Rho Zero (Rho Zero) cells were a gift of Dr Karl Morten. Rho Zero cells were generated by treating 143B cells (TK1 deficient) with 10µM 2’,3’-dideoxycytidine (ddC) for 10 days. Both cell lines were grown in high glucose DMEM (Gibco, Carlsbad, California) supplemented with 10% FBS. Rho Zero cell culture media was additionally supplemented with 50µg/mL uridine (A15227.06, Alfa Aesar, Haverhill, Massachusetts).

MCF7 cells were cultured in normoxia or 0.1% hypoxia and treated for 48h with the following mitochondria-targeting drugs: 5μM Vps34 inhibitor SAR405 (16979, Cayman Chemical, Ann Harbor, Michigan), 200μM chloramphenicol (C0378, Sigma-Aldrich, St Louis, Missouri), 100nM mubritinib (S-2216, Selleckchem, Houston, Texas), 5μM ABT-737 (sc-207242, Santa Cruz Biotechnology, Dallas, Texas), 2mM metformin (D150959, Sigma-Aldrich), 5μM gamitrinib-triphenylphosphonium (G-TPP, HY-102007, MedChemExpress, Monmouth Junction, New Jersey), 1μM/10μM IMT1 (HY-134539, MedChemExpress) or 1μM NBS-037 (Novintum Bioscience Ltd., London, England) (25).

BNIP3 siRNA transfection ( Supplementary Table 1 ) was performed in Optimem reduced serum medium at a final concentration of 5nM, the following day cells were exposed to 0.1% hypoxia for 48h. siRNA control was done in parallel and the following day was subjected to normoxia or 0.1% hypoxia for 48h. Oligofectamine (12252-011, Thermo Fisher Scientific) was used following the manufacturer’s instructions.

Whole cell lysates were prepared with RIPA buffer (R0278, Sigma-Aldrich) containing protease (cOmplete, 11697498001, Sigma-Aldrich) and phosphatase (phosSTOP, 4906845001, Sigma-Aldrich) inhibitors. Samples were subjected to SDS-PAGE and transferred onto PVDF membranes (IPVH00010, Millipore, Burlington, Massachusetts), after blocking, membranes were incubated overnight with primary antibodies ( Supplementary Table 2 ) at 4°C. They were later washed and incubated with HRP-anti-mouse/rabbit secondary antibodies (Gibco). Development was performed with Amersham ECL Prime Western Blotting Detection Reagent (GERPN2232, GE Healthcare Life Sciences, Chicago, Illinois) using ImageQuant™ LAS 4000. Stripping with Restore PLUS Western Blot Stripping Buffer (46430, Invitrogen, Carlsbad, California) was performed to blot different antibodies in the same membrane.

RNA was extracted using the Tri-Reagent protocol (T9424, Sigma-Aldrich) and 1µg was reverse transcribed with the High Capacity cDNA reverse transcription kit (44368813, Thermo Fisher Scientific) using random hexamer primers. The PCR reaction containing SensiMix™ SYBR Green^® No-ROX Kit (QT650-20, Bioline, London, UK) was run on a 7900 Real time PCR System (Applied Biosystems, Foster City, California) with standard cycling conditions: 10 minutes 95°C, and 40 cycles of 15 seconds 95°C followed by 1 minute 60°C. Gene expression was analysed with the Ct method using HPRT1 expression for normalization. The primers used are listed in Supplementary Table 3 .

HEK293T-P125 reporter cells (stably expressing the IFNβ promoter-Luciferase region (26)) were used to detect specifically immunostimulatory RNAs as they do not express cGAS and the expression of STING is very low. 4x10⁴ cells per well were seeded in 96-well plates. Next day, cells were pre-treated with 30U/mL of IFN-A/D (I4401, Sigma-Aldrich), and after 24h of incubation fresh medium was added and cells were transfected with 100ng of total RNAs from cell cultures or human tissues using Lipofectamine 2000^® (11668-019, Thermo Fisher Scientific). As positive controls, 1ng of IVT-RNA or V-EMCV-RNA was used (26). 24h post-transfection, cells were lysed and measured using OneGlo luciferase assay (E6120, Promega, Madison, Wisconsin) in a FluorOPTIMA luminometer.

Cells were plated on coverslips (VWR Collection, Radnor, Pennsylvania) and exposed to 0.1% O₂ hypoxia for 48h. Prior to fixation, mitochondria were stained for 1h with 200nM MitoTracker Deep Red (M22426, Thermo Fisher Scientific). After, cells were washed and fixed with 4% (v/v) paraformaldehyde (PFA) for 8 min at room temperature (RT). Then, cells were washed and permeabilized with 0.1% Triton X-100 for 20 min at RT. PFA was neutralized with 0.1M glycine for 10 min at RT. After washing three times, cells were incubated for 60 min with blocking solution [PBS containing 1% (w/v) BSA and 10% (v/v) normal goat serum (ab7481, Abcam, Cambridge, UK)]. Cells were incubated overnight at 4°C in a humidified chamber with J2 primary antibody (10010200, Scicons, Szirák, Hungary) at 1:200 or PKR (sc-6282, Insight Biotechnology Ltd, Wembley, UK) at 1:50, and rhodamine phalloidin (R415, Invitrogen) at 1:40 in block solution. Cells were washed three times and incubated with goat anti-mouse IgG Alexa Fluor 488 (R37120, Invitrogen) secondary antibody at 1:500 and Hoechst 33342 (H3570, Invitrogen) at 1:1000 concentration in block solution for 1h at RT. After washing three times with PBS, coverslips were mounted with Vectashield^® Mounting Medium (H-1000, Vector Labs, Burlingame, California) and sealed with nail polish.

Slides were imaged in a Zeiss 880 Inverted confocal microscope (Zeiss) using a 63x Plan-Apochromat objective. Laser properties, acquisition mode and detectors were manually adjusted for each experiment. Fiji Image J software was used for image analysis, using specific macros created by Dr Ulrike Schulze and Dr Dominic Waithe. A minimum of 35 cells per condition were analysed.

2x10⁴ MCF7 cells were seeded in 24-well glass-bottom SensoPlate™ (662892, Greiner) for 48h. 1h prior to imaging, media was removed and replaced with complete DMEM containing 50nM LysoTracker™ Red DND-99 (L7528, Invitrogen™) and 100nM MitoTracker™ Green FM (M7514, Invitrogen™). Then, media was removed and replaced with fresh media. Living cells were imaged in a Zeiss 880 Inverted confocal microscope (Zeiss) using a ZEN fluorescence microscopy system (Zeiss) and a 63x Plan-Apochromat objective. Laser properties, acquisition mode and detectors were manually adjusted for each experiment.

Mitochondria were isolated from MCF7 cells seeded in normoxia or 0.1% hypoxia for 48h using Mitochondria Isolation Kit for Cultured Cells (89874, Thermo Fisher Scientific) and following manufacturer’s instructions. Mitochondrial RNA was extracted using Tri Reagent, following the protocol previously explained. In step 8 of manufacturer’s instructions, the cytosolic fraction was kept and 200uL were used to extract RNA.

Protein G Dynabeads (10004D, Invitrogen) were washed and resuspended in NET-2 buffer. 5µg of J2 antibody or mouse IgG2 (400201, BioLegends, San Diego, California) were bound to 100μL of beads for 1h at RT on a thermoshaker. Conjugated beads were washed three times with NET-2 Buffer. 80–90% confluent MCF7 cells from 10cm² plate (×2) were washed with 10mL of cold PBS. Cells were scraped and transferred to a falcon and spun at 500g at 4°C, 5 min. Cell pellet from one 10cm² plate was lysed in 1mL of NP-40 lysis buffer and transferred to a tube and incubated on ice for 5 min. Following centrifugation at 17,000g at 4°C for 5 min, supernatant was carefully transferred to a new tube. Total RNA was harvested from 10% input lysate using Tri Reagent. For immunoprecipitation, lysate was supplemented with 10 units of RNase free TurboDNase (AM2238, Thermo Fisher Scientific) at 10mM MgCl₂ per 1mL of mix. 100μL of J2-Dynabeads was added to 1mL of above lysate and left for 1–2h at 4°C. Following magnetic separation, beads were washed twice with 1mL of high salt washing buffer (HSWB). Beads were transferred to a new tube with NET-2 buffer and washed twice with the same buffer. J2-bound dsRNA was extracted with Tri Reagent. The RNA samples were sent for sequencing. NET-2 buffer (50mM Tris-Cl, pH 7.4, 150mM NaCl, 1mM MgCl₂, 0.5% NP-40), NP-40 lysis buffer (50mM Tris-Cl pH 7.4, 150mM NaCl, 5mM EDTA, 0.5% NP-40), high salt wash buffer (50mM Tris-Cl pH 7.4, 1M NaCl, 1mM EDTA, 1% NP-40, 0.5% DOC, 0.1% SDS).

Libraries for paired end sequencing were prepared using standard Illumina protocol and sequencing was performed using Illumina NovoSeq 6000 sequencer at Wellcome Centre for Human Genetics. Raw reads were processed using FASTQC and Cutadapt and aligned to the genome using STAR. Normalised counts of nuclear encoded genes involved in mitochondrial function and mitochondrial encoded genes were used for generation of heatmaps. In order to estimate the proportion of counts that map to mitochondrial and non-mitochondrial genes; transcript per million (TPM) normalized value of transcripts encoded by mitochondrial and non mitochondrial genome in different replicates was calculated. Proportion of TPM counts in mitochondrial or non-mitochondrial fractions was calculated by dividing the sum of TPM counts in the fraction by total TPM counts.

GraphPad Prism 8.0 statistical analysis software (GraphPad Software) was used. If not otherwise specified, all the experiments were performed in 3 biological triplicates. ANOVA or ANOVA on ranks was normally used to study one variable in more than 2 groups depending on if they follow a normal distribution or not respectively. When two means were compared, t-test was performed if samples followed a normal distribution or Mann-Whitney if there was not a normal distribution. When analysing the βluence of two different independent variables on one dependent variable, 2-way ANOVA was applied. In the graphs, the error bars depict the standard error of the mean (SEM).

Given the immuno-suppressive role of hypoxia, we tested whether cancer cell lines cultured in normoxia or hypoxia contain different amounts of immunostimulatory RNA. We extracted total RNA from the breast cancer cell line MCF7 cultured for 48 hours in normoxia or in 1% or 0.1% hypoxia. We then transfected this RNA into an IFNβ promoter reporter cell line (26). RNA from cells grown in normoxia induced expression of the reporter, indicative of the presence of immunostimulatory RNA ( Figure 1A ). To determine the sensor for this endogenous RNA, we tested the response in reporter cells lacking MDA5, RIG-I or MAVS. This analysis showed that total RNA from MCF7 cells induced an MDA5-MAVS-dependent response ( Figure 1A ). Interestingly, RNA from hypoxic cells (hypoxic RNA) had a significantly reduced capacity to induce activation of the IFNβ promoter reporter compared with RNA from normoxic cells (normoxic RNA; Figure 1A ). Moreover, much like the response to normoxic RNA, residual reporter induction after hypoxic RNA transfection was MDA5-MAVS-dependent and RIG-I-independent ( Figure 1A ).

As 0.1% hypoxia had a greater effect than 1% hypoxia ( Figure 1A ), subsequent experiments were performed under 0.1% hypoxic conditions. A time course in hypoxia for 4h, 8h, 16h, 24h and 48h showed that 4h in hypoxia was enough to lower IFNβ promoter stimulation and it was maintained up till 48h ( Figure 1B left panel). The time course for recovery after reoxygenation following 48h hypoxia was evaluated at 15min, 30min, 1h, 2h, 4h, 8h, 16h and 24h. Reoxygenation caused a gradual recovery of IFNβ promoter stimulation reaching normoxic basal levels at 24h ( Figure 1B right panel).

A panel of breast cancer cell lines with different receptor status were used to rule out a cell line dependent effect of hypoxia in MCF7 cells. Hypoxic RNA was much less effective than normoxic RNA in stimulating IFNβ promoter in all cell lines ( Figure 1C ).

Normoxic and hypoxic RNA from 786-0 WT (HIF1α deficient and HIF2α upregulated due to VHL mutation) and 786-0 HIF2α KO (hereafter 786-0 KO, HIF1α and HIF2α deficient), and RCC4 WT (VHL mutation leading to HIF1α and HIF2α constitutive overexpression) and RCC4 VHL (VHL restored causing HIF1α and HIF2α downregulation) was tested. In both cell lines and all genotypes, hypoxic RNA triggered significantly lower IFNβ promoter stimulation ( Figure 1D ), highlighting the HIF-independent effect. However, 786-0 KO cells showed higher IFNβ promoter activation under normoxia compared to 786-0 WT, suggesting an effect of HIF2α in suppressing IFNβ induction, although minimal compared to the effect of hypoxia.

We also tested normal endothelial cells (HUVECS) and fibroblasts. Again, hypoxic RNA significantly reduced the activation of IFNβ promoter ( Figure 1E ) pointing to a general effect of hypoxia independently of cancer.

To analyse which RNA species from the total RNA were responsible for the IFNβ promoter induction, normoxic and hypoxic RNA from MCF7 cells was treated with different RNAses and EMCV (Encephalomyocarditis Virus), containing only dsRNA, was used as positive control. RNAse III (specific for dsRNA) treatment completely abolished normoxic, hypoxic and EMCV RNA induced IFNβ promoter activity, whereas RNAse A (specific for single-stranded RNA, ssRNA) treatment did not affect the luciferase signal ( Figure 1F ), showing that endogenous dsRNAs are responsible for the IFNβ promoter activation rather than ssRNA.

To visualise the downregulation of dsRNA levels in hypoxia, fluorescence microscopy was performed using the J2 antibody which is widely used to specifically detect dsRNA (22, 27). dsRNA staining was detected both inside ( Figure 2A white arrow) and outside ( Figure 2A blue arrow) mitochondria and although there was substantial variability of dsRNA intensity among individual cells, it was significantly lower in MCF7 hypoxic cells ( Figure 2A ). The downregulation was time-dependent and significant after 16h in hypoxia ( Figure 2B ). To confirm the HIF1α/HIF2α-independence observed in the IFNβ promoter assay, 786-0 WT and 786-0 KO cells were stained and both cell lines showed significantly lower dsRNA levels in hypoxia ( Figure 2C ).

As control for the antibody to dsRNA, MCF7 cells were treated with NBS-037, a drug inhibiting mitochondrial protein synthesis and secondarily mtRNA synthesis (25). NBS-037 treatment markedly decreased dsRNA levels ( Supplementary Figure 1 ).

It was recently shown that 99% of endogenous dsRNA originates during mtDNA transcription (22). Therefore, we tested cells lacking mtDNA (Rho Zero) and found significantly lower dsRNA staining in the Rho Zero cells than the parental 143B cell line ( Figure 3A ) and similar reduction in IFNβ promoter activation ( Figure 3B ). This strongly supports the specificity of the J2 antibody and the IFNβ promoter for dsRNA, and suggests that hypoxic downregulation of mtdsRNA relies on having functional mitochondria. Surprisingly, there was no difference between normoxia and hypoxia in the 143B parental cell line, either in the dsRNA staining or in the IFNβ promoter assay. This was the only cell line among all tested in which hypoxia did not downregulate the type I IFN pathway.

As hypoxia causes a shift to reductive carboxylation for glutamine utilisation by mitochondria (28), we considered that differences in metabolism in hypoxic mitochondria could contribute to mtdsRNA production. The mtDNA mutation m8993T>G affects ATP6, a key subunit of ATP synthase, leading to several serious diseases e.g. retinitis pigmentosa and fatal childhood maternally inherited Leigh’s syndrome. Isogenic cell lines with different levels of heteroplasmy of the mtDNA mutation m8993T>G were generated. As the percentage of heteroplasmy for m8993T>G increased from 7 to 80% (U2OS mTUNE M7, M45, M80), oxygen consumption decreased and extracellular acidification increased, allowing intra-mitochondrial metabolism to be modified without manipulation (24). We aimed to see if this switch in glucose utilisation by mitochondria could affect mtRNA synthesis. The basal level of cytoplasmic dsRNA was similar in M7 and M45 (data not shown), but higher in M80 cells. However, hypoxia downregulated dsRNA levels in all mTUNE cell lines independently of the mutation level and metabolic profile ( Figure 3C ) and caused significantly lower IFNβ promoter activation ( Figure 3D ).

Moreover, it was previously reported that mtdsRNA accumulated when the degrading enzymes PNPT1 and SUV3 were inhibited, and downregulation of PNPT1 triggered the type I IFN pathway (22). However, neither PNPT1 nor SUV3 protein levels were affected by 0.1% hypoxia for 48h ( Figure 3E ).

PKR, alongside RIG-I and MDA5, is a key member of the antiviral signaling hub activated by dsRNA molecules. It has been described to be located and activated in the mitochondria independently of PNPT1 deficiency (3). We therefore analysed the gene expression and protein levels of PKR in MCF7 cells and found that these were significantly downregulated in hypoxia both at mRNA ( Figure 4A left panel) and protein ( Figure 4A right panel) level. In addition, the downregulation was time-dependent ( Figure 4B ), starting to lower as soon as 4h of hypoxia exposure, but being significant after 16h, as previously seen for RIG-I and MDA5 (23). Furthermore, reoxygenation after hypoxia exposure recovered the expression levels of PKR ( Figure 4C ).

PKR locates in both the nucleus and the mitochondria, where it can interact with noncoding nuclear RNAs as well as mtRNA, respectively (3). Hypoxia increased the levels of PKR protein in the nucleus, staining specifically in the nucleolus, while decreasing the amount of PKR in the mitochondria. There were changes in mitochondria morphology as previously noted in hypoxia (29, 30) ( Figure 4D ).

We measured the expression of some mitochondrial encoded genes (12S, ND3, ATP6 and CYTB) and also nuclear encoded genes involved in mtDNA transcription [POLRMT (mitochondrial RNA polymerase that catalyses mtDNA transcription), TFAM (which stabilizes mtDNA, regulates mtDNA transcription, and is required for efficient promoter recognition by POLRMT) and TFB1M (mitochondrial dimethyladenosine transferase 1 whose interaction with POLRMT and TFAM is required for mtDNA transcription)]. Both sets of genes were significantly downregulated in MCF7 cells cultured under 0.1% hypoxia ( Figure 5A ). Most of the tested genes showed lower expression after only 4h under hypoxia (POLRMT, TFAM, ND3, ATP6 and CYTB) but it was significant for all when cultured for 16h in hypoxia. These results were confirmed by the general downregulation observed in mitochondrial encoded genes and nuclear encoded genes involved in mitochondrial function [from MitoCarta 2.0 (31)] using RNA-seq data from MCF7 cells cultured in normoxia or 0.1% hypoxia for 48h ( Figure 5B ). As expected, nuclear encoded genes involved in glycolytic metabolism such as pyruvate dehydrogenase kinase 1 (PDK1) or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were upregulated in hypoxia.

We quantified the mitochondria staining with MitoTracker in MCF7 and 786-0 cells exposed to normoxia or 0.1% hypoxia for 48h and no differences were observed ( Supplementary Figure 2A ).

Furthermore, we determined the expression of POLRMT, TFAM and TFB1M and some mitochondrial encoded genes in the parental 143B and Rho zero cells. 12S, ND3 and ATP6 mitochondrial genes were absent in Rho Zero cells whereas POLRMT, TFAM and TFB1M were expressed as previously reported (32). However, hypoxia did not show any effect in the parental 143B cell line ( Supplementary Figure 2B ). This could be potentially related to the cytosolic thymidine kinase (TK1) deficiency of 143B cells (33) and the consequent downregulation in the expression of a subset of genes, some of which may protect mitochondria from stress (34). As the cytosolic and mitochondrial thymidine triphosphate pools are in rapid exchange and mainly produced by TK1 (35), it is possible that there is a more steady state of mtDNA replication in the TK1 deficient cells and hence mtdsRNA synthesis, with a stable source of nucleotides from the mitochondrial compartment produced by the mitochondrial thymidine kinase (TK2), which is not cell cycle regulated (36).

To support the hypothesis that hypoxia induced downregulation in the genes involved in mtDNA transcription would ultimately lead to less mtdsRNA formation, MCF7 cells were cultured with IMT1, a specific and non-competitive POLRMT inhibitor, and stained for dsRNA molecules. IMT1 dose-dependently reduced J2 staining, achieving the same levels as 0.1% hypoxia after treating the cells with 10µM ( Figure 6 ).

Altogether, these data suggest that hypoxia leads to lower mtDNA transcription, without affecting the mitochondrial content, and thus lowers production of dsRNA available to trigger the type I IFN response.

It was also reported that hypoxia decreased protein expression of mitochondrial ribosomal proteins (MRPs) involved in mtRNA translation (37). We tested the expression of the mRNA coding for some of the MRPs under hypoxia in MCF7 cells and found that 16h of hypoxia significantly decreased their expression ( Figure 7A ), suggesting that less mtdsRNA in hypoxia could have a feedback mechanism on the expression of genes involved in mtRNA translation, as seen in bacteria (38). The same trend was observed in 786-0 WT and 786-0 KO cells, although it was not always statistically significant ( Figure 7B ), pointing to a HIF1α/HIF2α-independent mechanism.

To assess the role of mtdsRNA in inducing the type I IFN pathway, MCF7 cells were cultured in normoxia or 0.1% hypoxia for 48h and their intact mitochondria were isolated. RNA from the mitochondrial and cytosolic fractions was extracted. Firstly, we confirmed that mitochondria were successfully enriched in the mitochondrial fraction by analysing the expression of several mitochondrial encoded genes in both fractions ( Figure 8A ). Then, these RNAs were transfected to HEK-293T P125 reporter cells to measure IFNβ promoter activation. The mitochondrial hypoxic and normoxic RNA extracts include mtRNA and also ssRNA. However, the IFNβ promoter assay only detects dsRNA. mtRNA induced IFNβ promoter stimulation, whereas the luciferase signal using cytosolic RNA was hardly detected. Importantly, hypoxic mtRNA caused significantly lower stimulation of IFNβ promoter ( Figure 8B ).

To assess the composition of the dsRNA pool in hypoxia, dsRNA pull-down was performed using the J2 antibody in MCF7 cells exposed to normoxia or 0.1% hypoxia for 48h, and the resultant RNA was sequenced. Reads were normalised as transcript per million (TPM). Interestingly, the percentage of mitochondrial reads was significantly lower in hypoxia than in normoxia ( Figure 8C ), and 22 out of 37 mitochondrial encoded genes were significantly downregulated in hypoxia ( Supplementary Table 4 ). Density plots in normoxia and hypoxia showed that the non-mitochondrial genes pulled-down by J2 antibody had few reads which probably correspond to background noise (red peak, Figure 8D ).

To evaluate whether the downregulation of mtdsRNA in hypoxia was due to mitophagy, we knocked down one of the main genes involved in this process, BNIP3 (BCL2 Interacting Protein 3) in MCF7 exposed to 0.1% hypoxia for 48h (siBNIP3 and control siRNA, siCON). In the knock-down cells, BNIP3 showed no induction in hypoxia, either at mRNA or protein level ( Figure 9A ). The expression of various ISGs (DDX58, MX1, IFIT1, IFIT2, ADAR-p150 and ISG15) was tested. Hypoxia downregulated ISG expression (23), but no differences were found between siCON and siBNIP3 in hypoxia ( Figure 9B ), suggesting that hypoxic upregulation of BNIP3 does not interfere in the type I IFN signalling. This was further confirmed by the IFNβ promoter assay, in which hypoxia-induced inhibition on the IFNβ promoter activation was not recovered upon BNIP3 silencing ( Figure 9C ).

Live cell immunofluorescence was performed using mitochondrial and lysosomal markers in MCF7 cells cultured in normoxia or 0.1% hypoxia for 48h. As previously reported (29), mitochondria clearly exhibited morphological changes under low oxygen conditions and appeared more elongated and located closer to the nucleus ( Figure 9D ). However, no colocalization of the mitochondrial and lysosomal markers was observed in normoxia or hypoxia, suggesting that mitophagy is not increased in hypoxia and does not decrease mtdsRNA.

Next, the ability of different drugs targeting mitochondria to affect dsRNA levels was assessed in MCF7 cells. This was because drugs already approved or in preclinical trial may be detrimental to the normoxic response, or exacerbate the hypoxic suppression. Briefly, the drugs used were: chloramphenicol which inhibits mitochondrial protein synthesis via binding to the 50S subunit of the 70S mitochondrial ribosomes; ABT-737 induces cell apoptosis by inhibiting the anti-apoptotic molecule BCL2, as well as mitophagy; G-TPP accumulates in the mitochondria of tumour cells and by inhibiting the heat shock protein 90 (Hsp90) promotes cell apoptosis; mubritinib and metformin decrease mitochondrial respiration by inhibiting ETC complex I; and the Vps34 inhibitor SAR405 alters vesicle trafficking and inhibits autophagy by blocking autophagosome formation.

Mubritinib and G-TPP treatment decreased IFNβ promoter activation under normoxia by 4- and 2-fold, respectively ( Supplementary Figure 3 ). Hypoxia did not further inhibit IFNβ promoter activity in MCF7 cells treated with these two drugs.

Total RNA samples from different human tissues were obtained to assess their effects in the IFNβ promoter reporter assay. RNA from some tissues including brain, heart, kidney and testis strongly induced the IFNβ promoter reporter ( Figure 10A ). In contrast, RNA from other tissues, including skeletal muscle and pancreas, had little effect. ISG expression was also determined in RNA samples from some tissues. MX1, IFIT1 and IFNB1 exhibited much higher expression in testis and brain than in skeletal muscle ( Figure 10B ), confirming the results from the IFNβ promoter assay.