Inducing respiratory complex I impairment elicits an increase in PGC1α in ovarian cancer

Anticancer strategies aimed at inhibiting Complex I of the mitochondrial respiratory chain are increasingly being attempted in solid tumors, as functional oxidative phosphorylation is vital for cancer cells. Using ovarian cancer as a model, we show that a compensatory response to an energy crisis induced by Complex I genetic ablation or pharmacological inhibition is an increase in the mitochondrial biogenesis master regulator PGC1α, a pleiotropic coactivator of transcription regulating diverse biological processes within the cell. We associate this compensatory response to the increase in PGC1α target gene expression, setting the basis for the comprehension of the molecular pathways triggered by Complex I inhibition that may need attention as drawbacks before these approaches are implemented in ovarian cancer care.

To foresee potential compensatory mechanisms that may occur upon the use of CI inhibitors in OC patients, we exploited genetic ablation and pharmacological inhibition of this enzyme to identify whether PGC1α may play a role upon the induction of energetic crisis and in the regulation of the metabolic phenotype of OC cells.

Levels of PGC1α correlate with mitochondrial abundance in OC cells. OC is a highly heterogene-
ous neoplasm for which stratification based on the correlation between metabolism and response to standard chemotherapy allows the identification of low-versus high-OXPHOS tumors 19 . Thus, we assessed the oxygen consumption rate (OCR), the glycolytic indicator extracellular acidification rate (ECAR) and ATP production rate in two different high-grade serous OC cell lines, namely, OVSAHO and SKOV3 to validate in our experimental settings their metabolic features, which had been previously characterized 19 . Despite a comparable basal OCR, the maximal respiration rate of OVSAHO cells was twice that of SKOV3 cells (Fig. 1a-c). Moreover, OVSAHO showed a spare respiratory capacity and reduced ECAR (Fig. 1d,e), suggesting a major contribution of OXPHOS to energy metabolism. Surprisingly, the overall ATP production rate was reduced in OVSAHO cells (Fig. 1f), but this apparent paradox can be explained by the fact that ATP is mainly derived from OXPHOS in this cell model, while SKOV3 cells mostly rely on glycolysis as a source of ATP (Fig. 1g). These data correlated with a trend in increased respiratory complex activities, particularly evident for those of CII and CIV, in OVSAHO cells compared to SKOV3 cells (Fig. 1h). Altogether, these data allowed us to categorize SKOV3 as low-OXPHOS cells with respect to OVSAHO, in our experimental settings, which instead show higher reliance on oxidative metabolism. We next investigated whether such divergent metabolic assets may be related to differences in mitochondrial mass and biogenesis. Interestingly, OVSAHO displayed a higher activity of citrate synthase (CS), a well-known indicator of mitochondrial mass, as well as more intense MitoTracker Red staining (Fig. 2a,b). Moreover, the steady-state levels of most of the analyzed OXPHOS complex subunits were higher in OVSAHO cells (Fig. 2c), along with the relative mitochondrial DNA (mtDNA) abundance, which was indeed nearly 5-fold higher in OVSAHO than in SKOV3 cells (Fig. 2d). Overall, these data indicate a richer mitochondrial phenotype, possibly due to a more active mitochondrial biogenesis in OVSAHO cells compared to SKOV3 cells. The master regulator of mitochondrial biogenesis is the transcriptional coactivator PGC1α, which exerts pleiotropic transcriptional control of several downstream pathways. Hence, we investigated PGC1α gene expression and that of some of its responsive genes specifically selected to be representative of such diverse pathways in the two OC cell models. We analyzed the levels of COX5B, coding for cytochrome c oxidase subunit Vb and thus representative of OXPHOS, ESRRA , which encodes the Estrogen Related Receptor α (ERRα), a transcription factor known to regulate energy metabolism 20 and ACADM, producing medium-chain acyl-CoA dehydrogenase (MCAD), crucial for lipid metabolism, the main source of energy for OC omental metastases [21][22][23] . In agreement with their high OXPHOS status and the more abundant mitochondrial mass, OVSAHO showed a 12-fold higher expression of PGC1α and a significant increase in all its analyzed responsive genes compared to SKOV3 (Fig. 2e,f), suggesting that their elevated OXPHOS status may derive from upregulated mitochondrial biogenesis with a concurrent activation of other pathways regulated by PGC1α.
Genetic ablation of CI triggers PGC1α expression and activation under glucose restriction. We next proceeded to gauge whether a potent stimulus such as CI derangement may induce changes in PGC1α expression in OC cells like those observed in other cancer contexts, such as osteosarcoma 8 . To this aim, we generated syngenic cell lines devoid of the core CI subunit NDUFS3 via gene editing (SKOV3 −/− and OVSAHO −/− ), in which protein expression was completely abolished (Fig. 3a). Similar to our previous findings in other cell lines 8 , the lack of NDUFS3 induced almost complete CI disassembly, causing its dysfunction (Fig. 3b) and the consequent abolishment of mitochondrial respiration (Fig. 3c). Interestingly, while ECAR was unaffected by the lack of CI in glycolytic low-OXPHOS SKOV3 cells, its values increased when NDUFS3 was ablated in oxidative OVSAHO cells (Fig. 3d), suggesting the occurrence of an adaptive metabolic switch toward glycolysis in the latter model. Then, we measured PGC1α levels in resting conditions (25 mM glucose -high glucose; HG), in which no increase was observed between CI-null and CI-competent cells (Fig. 3e), in contrast to what we previously found in osteosarcoma 8 . We reasoned that OC cells may not respond by increasing PGC1α upon CI dysfunction if a metabolic stress is not induced. To this aim, we cultured cells under glucose deprivation (5 mM glucose-low glucose; LG) for 24 hours, which induced an energetic impairment in CI-null cells, as testified by the activation of the main energy sensor AMP activated kinase (AMPK) (Fig. 3f and Supplementary Fig. 1a). Under these conditions, a significant increase in PGC1α expression was observed in CI-null cells compared to their CI-competent counterparts in both OVSAHO and SKOV3 background (Fig. 3e), which was followed by the upregulation of two out of three PGC1α-responsive genes tested (Fig. 3g). Interestingly, the PGC1α increase upon energetic impairment was more prominent in SKOV3 than in OVSAHO cells (10-versus 5-fold, respectively; Fig. 3e and Supplementary Fig. 1b), likely since the former needs to foster mitochondrial biogenesis to efficiently reprogram metabolism, whereas the latter intrinsically has a higher mitochondrial respiration. To gauge whether the increase in PGC1α translated into an augmented mitochondrial biogenesis, we evaluated mtDNA copy number and revealed approximately a 3-fold increase in LG in both CI-null cell lines, suggesting this response was independent from the cell metabolic features (Fig. 3h). Overall, these data demonstrate that the energy crisis caused by glucose restriction in synergy with CI ablation triggers a PGC1α-mediated compensatory response regardless of the OXPHOS status of the cells.  www.nature.com/scientificreports/ Pharmacological inhibition of CI mimics its derangement and induces PGC1α compensatory expression and activation. Last, with the intent to test whether the mechanisms displayed by cells in which CI was genetically ablated were adaptive or generally activated when CI was pharmacologically inhibited, we exploited the recently synthesized selective CI inhibitor EVP-4593, which is known to bind the ubiquinone reduction site of CI and has been preliminarily tested as an anticancer molecule 24 . First, we found that 1 µM EVP-4593 was able to completely abolish mitochondrial respiration in both SKOV3 and OVSAHO cells (Fig. 4a). This was associated with a marked increase in AMPK phosphorylation at Thr172 starting from 24 h for OVSAHO and 48 h for SKOV3 when cells were grown in LG, suggesting that the energy crisis had occurred ( Fig. 4b and Supplementary Fig. 1c). Upon inhibition of CI with 1 µM EVP-4593 in LG, ECAR was increased in both cell lines (Fig. 4c), corroborating the idea that a metabolic switch toward glycolysis may occur. In agreement and similar to what was observed in genetically ablated SKOV3 −/− and OVSAHO −/− cells, CI inhibition by 1 µM EVP-4593 in LG induced a 10-fold increase in PGC1α expression in SKOV3 cells and nearly 5-fold in OVSAHO cells (Fig. 4d). Moreover, in both cell lines, the treatment caused a rise in the expression of two out of three PGC1α-responsive genes (Fig. 4e). Taken together, pharmacological targeting of CI triggered a PGC1αmediated compensatory mechanism, recapitulating the phenomenon described in NDUFS3 knockout cells and suggesting that this was both an acute and an adaptive response.

Discussion
In this work, we set the basis for the study of the pleiotropic transcriptional coactivator PGC1α in OC. In particular, we correlated higher PGC1α expression and activity with higher OXPHOS metabolic status in OC cells, indicating that this coactivator is involved in the maintenance of mitochondrial mass required for elevated respiration and may be used as a marker when defining the OXPHOS-related metabolic status of OC cells. In this context, it is interesting to note that, compared to previous findings 19 , OVSAHO cells here are characterized by higher levels of mitochondrial proteins, enzymes activity and maximal respiration compared to SKOV3 cells, suggesting they exhibit a higher OXPHOS metabolism, and prompting that plastic parameters such as those defining cell bioenergetics should be evaluated prior to any investigation. Moreover, we investigated the PGC1α-mediated response to genetic or pharmacologic respiratory CI inactivation in OC cell models. A comprehension of the PGC1α response to pharmacological stimuli is required in cancer, since it has been defined as an oncojanus gene 25 , whereby it behaves as an oncogene 26,27 or as a tumor suppressor gene 28 according to the context. PGC1α is widely recognized as the master regulator of mitochondrial biogenesis and thus controls oxidative metabolism by promoting the expression of OXPHOS complexes 29 . In this frame, PGC1α has been shown to contribute to the metabolic heterogeneity of different types of human cancers and to modulate their response to chemotherapy [30][31][32][33] . Moreover, since the discovery that the interaction with ERRα renders PGC1α a regulator of hypoxia-independent vasculogenesis via transcription of the Vascular Endothelial Growth Factor (VEGF) 17 , more attention has been paid to PGC1α in cancer. However, much remains to be unveiled on what stimuli are able to trigger a process that may importantly impinge on tumor progression toward malignancy. This is of relevance in OC, where ERRα is highly expressed 34 , and vascularization is a cogent issue so that the anti-VEGF drug bevacizumab has been introduced, not without severe side effects, in the therapeutic regimen of advanced cases along with standard therapy 35 . An increase in PGC1α was previously shown to occur following CI genetic ablation or in the presence of disruptive mtDNA mutations 8,10 . We show here that inhibition of CI triggers a compensatory response increasing PGC1α expression, which has not been shown thus far. Indeed, a pharmacological approach is quickly emerging as an effective anticancer therapy based on the notion that cancer cells must rely on a functional respiratory chain to thrive 36 . The most plausible mechanism explaining the PGC1α increase upon CI inhibition may be retrograde signaling triggered by mitochondrial energetic impairment, which was previously shown to be mediated by AMPK 37 . In particular, AMPK activates PGC1α at post-translational level inducing its expression in a reinforced positive feedback loop 37 . The increase in PGC1α, albeit at different levels, occurs in both high-and low-OXPHOS OC cells, indicating that CI inhibition is a potent stimulus for PGC1α increase regardless of the intrinsic oxidative capacity of the tumor. It is important to note that apart from its regulation of mitochondrial biogenesis, PGC1α has been associated with several other processes relevant in the context of cancer. Indeed, we report that an increase in PGC1α is associated with augmented levels of specific PGC1α target genes involved in different pathways, such as mitochondrial energy metabolism, angiogenesis and lipid catabolism. Interestingly, while COX5B does not appear to increase upon an acute stimulus, suggesting that a trigger of mitochondrial biogenesis requires coordination with the mitochondrial genome and may rather be an adaptive response, ESRRA is among the PGC1α-responsive genes to be upregulated. It is plausible that strong consequent angiogenesis may be fostered, therefore, by a joint increase in both constituents of the molecular complex responsible for VEGF transcription 17 , calling for attention when using CI inhibitors in OC. MCAD was also shown to increase in association with PGC1α and CI inhibition; this is also relevant in OC, as it is well known that omental metastases preferentially use lipid catabolism, exploiting the lipid-rich environment by producing fatty acid binding protein 4 to internalize fatty acids from adipocytes 38,39 .
In conclusion, our data begin to unravel the PGC1α-mediated compensatory responses that may be triggered in OC when CI is hampered as a therapeutic approach. As PGC1α increase occurs independently of the initial and basal metabolic status of the cell, CI inhibition may be a strong and horizontal strategy to induce an energy crisis in both high-and low-OXPHOS cancer cells, but its implementation warrants investigation to counteract or prevent adaptive responses. Genome Editing. The CRISPR/Cas9 system was used to insert a frameshift mutation in the NDUFS3 gene in SKOV3 and OVSAHO cell lines. Cas9 protein was transfected following the manufacturer's instructions using Lipofectamine CRISPRMAX Cas9 Transfection reagent (Invitrogen #CMAX00008) together with synthetic RNA guides designed and purchased from IDT. Exon 2 targeting guide TGT CAG ACC ACG GAA TGA TG was used. Non-homologous repair efficiency was evaluated by Sanger sequencing using KAPA2G Taq Polymerase (Kapa Biosystems #KK5601) and the Big Dye protocol (Life Technologies #4337451). PCR for NDUSF3 was performed using the primers forward 5'-TCT CAA GGT GCT TCA GGG AG-3' and reverse 5'-GAA ACA AGT CTG CCC ACT CC-3' . Clonal selection was carried out to select cells with frameshift NDUFS3 mutations. DNA extraction was performed following the manufacturer's instructions using 8 µL of lysis buffer (Sigma-Aldrich #L3289) and 80 µL of neutralization buffer (Sigma Aldrich #N97784) per sample in a 96-well plate.

Gene expression quantitative real-time PCR. SKOV3 (2 × 10 5 cells) and OVSAHO (4.5 × 10 5 cells)
cells were seeded in a 6-well plate and grown in 25 or 5 mM glucose for 24 hours. RNA was extracted from cell pellets using RNeasy mini kit (QIAGEN #74106) and quantified by NanoDrop TM 2000 Spectrophotometer (Thermo Scientific). Three hundred nanograms of total RNA was reverse-transcribed into cDNA using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems #4368814) with random hexamers. Quantitative real-time PCR (qRT-PCR) was performed using either the intercalating dye SYBR ® Green dye (Promega) or 5' nuclease probes PrimeTime™ qPCR Probes (TaqMan assay). For the SYBR Green assay, the primer sequences were designed using Primer3 software 40 . The presence of 3' intra/inter primer homology was excluded using the IDT OligAnalyzer tool (https:// eu. idtdna. com/ analy zer/ Appli catio ns/ Oligo Analy zer/), and the availability of the target sequence was estimated by predicting cDNA secondary structure by the Mfold web server 41 Supplementary Fig. 4a,b. The intensity of each band was quantified by densitometry and data (mean ± SEM) were expressed as fold of phosphorylated (T172) to total AMPKα ( Supplementary Fig. 1a).   Microrespirometry and extracellular acidification rate assessment. Oxygen consumption rate (OCR) and extracellular acidification rate (ECAR) were measured using the protocol described for the Seahorse XFe Cell Mito Stress Test Kit (Agilent #103015-100) following the manufacturer's instructions. A total of 1.2 × 10 4 cells/well (SKOV3) and 2 × 10 4 (OVSAHO) were seeded in 80 μL of DMEM into XFe96 cell culture plates and incubated for 24 hours at 37 °C and 5% CO 2 . Seeding was optimized before the assay. Complete growth medium was replaced with 180 μL of XFe medium (Agilent #103575-100) supplemented with 10 mM glucose, 1 mM sodium pyruvate, and 2 mM L-glutamine at pH 7.4. For temperature and pH equilibration, cells were incubated at 37 °C for 30 min. After three OCR baseline measurements, 1 μM oligomycin, 0.5 μM carbonyl cyanidep-trifluoromethoxyphenylhydrazone (FCCP), 1 μM rotenone, and 1 μM antimycin A were sequentially added to each well. In the case of inhibition with 1 µM EVP-4593, rotenone and antimycin A were added together. FCCP concentrations were optimized in each cell line by titration before the experiments. At the end of the assay, the medium was removed, and a sulforhodamine B (SRB) assay was performed to determine the protein content. Briefly, plates were incubated with 10% trichloroacetic acid (TCA) for 1 h at 4 °C to fix the cells. Five washes in water were carried out. Once the plates were dried, proteins were stained by incubation with 0.4% SRB for 30 min at RT. Then, SRB was solubilized with 10 mM Tris, and the absorbance at 560 nm was determined using a Victor2 plate reader (Perkin-Elmer). Each biological replicate experiment (n=3-4) included measurements from at least six wells. Data (pmol/min) were normalized to blank corrected SRB absorbance. ATP production rate was determined using the protocol described for the Seahorse XF Real-Time ATP Rate Assay Kit (Agilent #103592-100).
Mitochondrial network staining. Cells (1 × 10 5 cells/dish) were seeded on glass cover slides (Ø 10 mm) and incubated with 2 mL of culture medium. After 24 hours, the cells were incubated with 10 nM MitoTracker Red CMXRos (Invitrogen, #M7512) for 10 min at 37 °C. After incubation, the cells were washed with PBS, and the slide was placed in a specific metal grid with 1 mL of DMEM without red phenol supplemented with 25 mM HEPES (Gibco #21063029). The mitochondrial reticulum was visualized with a digital imaging system using an inverted epifluorescence microscope with a ×63/1.4 numerical aperture (NA) oil objective (Nikon Eclipse Ti-U; Nikon). Images were captured with a back-illuminated Photometrics Cascade CCD camera system (Roper Scientific) and elaborated with Metamorph Acquisition/Analysis Software (Universal Imaging Corp.). Fluorescence intensity analysis was performed using ImageJ 44 . Fluorescence intensity data for each image were normalized to the nuclei number.