Introduction

Discovering safe and effective therapeutic targets for cancer treatment remains a major challenge in biomedical research¹. Recently, targeted therapies against complex I of the respiratory chain have garnered considerable attention in pharmacology, offering a promising strategy for cancer treatment due to their potential to disrupt the energy metabolism of tumor cells^2-5. However, despite initial promises, clinical trials with potent complex I inhibitors, such as IACS-0107596, have been compromised by severe toxicities, raising concerns about their clinical viability. Given the critical role of mitochondria in cancer cells, there is an urgent need for more effective and safer alternatives to target them^6,7. In this context, medicinal biguanides such as metformin and the more lipophilic phenformin are emerging as attractive alternatives. These biguanides are considered moderate respiratory chain inhibitors and offer a potential avenue for targeting mitochondrial metabolism in cancer⁸.

Metformin has been used since the 1950s for its anti-hyperglycemic properties⁹. As such, it has a well-established safety profile and is commonly used for the treatment of type II diabetes¹⁰. Several epidemiological studies showed that prolonged use of metformin in diabetic patients reduced the incidence of several cancers^11,12, particularly pancreatic cancer^13,14. However, its clinical use has proven ineffective in patients with advanced pancreatic cancer¹⁵, suggesting that its antitumoral mechanism is more preventive than therapeutic^16,17. The more potent phenformin⁹, which was withdrawn from the market in the 1970s for its anti-hyperglycemic applications due to deaths attributed partly to lactic acidosis¹⁸, is currently undergoing clinical trials for oncology purposes¹⁹.

Unlike other inhibitors of the respiratory chain, the mechanism of action of medicinal biguanides remains elusive^20,21. However, compelling evidence indicates that energy metabolism plays a pivotal role in their effectiveness^22,23. At supra-pharmacological doses (1-10 mM), typically 100-1000 times higher doses than those used for anti-hyperglycaemic effects, metformin moderately inhibits the activity of complex I of the respiratory chain in several cellular models^24-26. This inhibition subsequently decreases OXPHOS activity, leading to reduced cellular respiration and energy production. In response, cells adapt by shifting metabolism to glycolysis and decreasing their anabolic activity through the activation of the energy sensor AMPK^20,22,24. Nonetheless, to observe a decrease in isolated complex I activity in in vitro models, metformin doses of up to 50 mM are required^25-27. A recent cryo-electron microscopy study revealed that a metformin analogue binds to complex I in a specific conformation with a binding site unique to this analogue²⁷. However, further work using in vivo models is necessary to validate the biological meaning of this interaction as a critical site of action of metformin^27,28.

Additional targets such as glycerol phosphate dehydrogenase²⁹, complex IV³⁰, PEN2³¹ and F₁F_o- ATP synthase²⁵ have also been suggested for metformin. Metformin may interact with various targets in a nonspecific manner^32-35 because of its structural similarity with guanidine, a potent chaotropic agent. These various targets potentially exert pleiotropic effects^20,21 that could act synergistically.

It has been shown that atrazine, a molecule derived from biguanides, binds and inhibits F₁F_o-ATP synthase³⁶. Additionally, a study from our group revealed that medicinal biguanides likely disrupt cristae organization³⁷, a crucial process mediated by the oligomerization of F₁F_o-ATP synthase and resulting in the formation of ʿʿonion-liked structuresʾʾ^38-41. It is proposed that complex I and F₁F_o-ATP synthase of the respiratory chain may be particularly sensitive to biguanide inhibition due to the conformational mobility of their catalytic interfaces²⁵. Disrupting the folding of the mitochondrial inner membrane could alter the organization and functions of other respiratory chain complexes⁴². It remains an open question whether biguanides act through multiple targets or a single primary target.

Here, we identify another potential target of biguanides: the e subunit of F₁F_o-ATP synthase (ATP5I), a transmembrane protein involved in the dimerization of this supramolecular complex^38-40. Although medicinal biguanides are known to accumulate in mitochondria and disrupt the respiratory chain by inhibiting complex I, no direct evidence of their interaction with a subunit of F₁F_o-ATP synthase has been reported until now. CRISPR-Cas9 mediated KO of ATP5I recapitulates many of the effects of metformin in pancreatic cancer cells and reduces metformin sensitivity. In addition, the genetic signature of a cells treated with metformin is more similar to that of cells treated with an F₁F_o-ATP synthase inhibitor than to a complex I inhibitor. Taken together this work adds the F₁F_o-ATP synthase and its subunit ATP5I as a bona fide target of biguanides

Results

Identification of ATP synthase subunit e (ATP5I) as a mitochondrial biguanide binding protein

To identify proteins that bind biguanides, we synthetized a Biotin Functionalized Biguanide or BFB (Figure 1A, Figure 1 – figure supplement 1A-B) to perform an affinity-based pull-down assay with streptavidin beads⁴³. We first assessed the biological activity of BFB comparing it to metformin in the ability to activate AMPK and inhibit cell proliferation in KP-4 pancreatic cancer cells. Similarly to metformin, BFB activated AMPK and its downstream target ACC as indicated by their increased phosphorylation states (Figure 1B) and inhibited KP-4 cell growth with an EC₅₀ of 1.0 ± 0.2 mM (Figure 1C). Additionally, immunofluorescence experiments with fluorophore-conjugated Streptavidin confirmed that BFB accumulates in mitochondria, as shown by colocalization with the mitochondrial protein TOMM20 (Figures 1D, 1E).

Figure 1.

Figure 1 – figure supplement 1.

After confirming that BFB exhibited activity comparable to that of metformin, we performed pull-down assays using streptavidin immobilized on sepharose beads on mitochondrial-enriched cell extracts to analyse interacting proteins by mass-spectrometry. Given the significant issue of nonspecific binding associated with this technique, we performed multiple parallel pull-down experiments. These included a control biotin-conjugated amine derivative (BFA) possessing the same linker as BFB (Figure 1 – figure supplement 1C). Mass spectrometry analysis results identified a total of 69 proteins. Among these, 31 proteins showed specific interaction with BFB under the elution condition with metformin (see Materials and methods). Many of these proteins were cytoskeletal, with several keratin isoforms potentially involved in nonspecific interactions with streptavidin beads. However, several proteins stood out, including glycerol-3-phosphate dehydrogenase, arginase isoform 2 and ATP synthase subunit e (ATP5I). Glycerol-3 phosphate dehydrogenase is a known target of metformin²⁹, validating our pulldown experiment. Additionally, arginase binds arginine which has a guanidinium group structurally related to metformin⁴⁴.

Biguanide pharmacophore interacts specifically with ATP5I

As ATP5I was not previously shown to bind biguanides, we first confirmed this interaction through immunoblot analysis in independent pull-downs experiments. The results indicate that BFB allows specific interaction of ATP5I in streptavidin pull-down while BFA failed to do so. Of note, we observed a nonspecific band (≃11 kDA) in all pull-down conditions with streptavidin beads, possibly associated with the denatured streptavidin monomer in heated SDS buffer (Figure 1F). We then used surface plasmon resonance (SPR) to characterize the binding of BFB and purified recombinant ATP5I. Our purification steps yielded a highly pure recombinant protein, mainly organized in dimeric form (Figure 1– figure supplement 1D, 1E). Although SPR is sensitive^45,46, the size difference between BFB and recombinant ATP5I prompted us to immobilize the small molecule BFB on the gold surface. Adding increasing concentrations of recombinant ATP5I allowed us to estimate an affinity constant (KD ≃ 11.10 μM; Rmax ≃ 39.40 RU) significantly lower than typical metformin concentrations that affect complex I in mM range (Figures 1G, 1H)

ATP5I knockout in pancreatic cancer cells alters the organization of mitochondrial networks

In yeast, inhibition of the subunit e’s equivalent impaired mitochondrial inner membrane folding⁴⁷. This structural deficiency in yeast correlates with decreased dimerization of F₁F_o-ATP synthase, highlighting the essential role of ATP5I in dimer stability. However, yeast’s subunit e is not essential for the catalytic activity of F₁F_o-ATP synthase^47-51. In human cells, ATP5I is also crucial for dimer stability⁴⁰ but lacks thorough characterization regarding its roles in cellular energy metabolism. To elucidate its function in cancer and its relevance to cells treated with medicinal biguanides, we developed CRISPR-based reagents to inhibit ATP5I expression in KP-4 pancreatic cancer cells.

Initially, two clones each of the two guide (sgATP5I #1 and sgATP5I #2) were isolated, and their mitochondrial phenotype characterized. First, we measured levels of several proteins from the respiratory complexes. The results show that all ATP5I knockout clones did not express ATP5I and exhibited decreased expression of complex I NDUFB8 protein and complex IV COX II protein compared to control clones expressing a guide against GFP (sgGFP) (Figure 2A). A moderate to no reduction in mRNA levels encoding these proteins were observed (Figure 2 – figure supplement 1A) cannot account for the downregulations at the protein level. Additionally, the mitochondrial/nuclear DNA ratio (Figure 2 – figure supplement 1B) indicates no reduction in mitochondrial number, suggesting ATP5I may play a role in maintaining the stability of certain subunits of complexes I and IV. Consistent with this finding, the absence of subunit e in yeast also controls the stability of F₁F_o-ATPase proteins subunit g and subunit k⁵¹.

Figure 2.

Figure 2 – figure supplement 1.

We then examined mitochondrial morphologies using fluorescence microscopy in ATP5I knockout (KO) cells, revealing a disruption in the mitochondrial network characterized by predominantly punctate mitochondria (Figure 2B-C and Figure 2 – figure supplement 1C), relative quantification showed that about 70% of ATP5I KO cells exhibited a punctate phenotype, around 25% showed a fragmented phenotype and only a few displayed a filamentous phenotype (Figure 2C). These findings confirm that ATP5I plays a crucial role in the organization of mitochondrial network in KP-4 cells. Treatment of the same cell line with medicinal biguanides resulted in similar alterations in mitochondrial network organization³⁷ reinforcing the hypothesis that ATP5I may be a target of biguanides.

ATP5I knockout desensitizes pancreatic cancer cells to biguanides

Given the uncertain role of ATP5I in the cellular energy metabolism but considering the changes in global mitochondrial physiology (Figure 2B, 2C), we investigated the effects of ATP5I knockout (KO) on mitochondrial bioenergetics in KP-4 cells. The results reveal that ATP5I deletion significantly decreases the NAD⁺/NADH ratio (Figure 3A), most significantly affecting NAD⁺ concentration (Figure 3A – figure supplement 1A, 1B). Our results also indicate a decrease in the oxygen consumption rate (OCR) to extracellular acidification rate (ECAR) ratio (Figure 3B and Figure 3 – figure supplement 1C, 1D) suggesting a reduced respiration associated with a compensatory increase in glycolysis. This metabolic reorganization in ATP5I KO cells renders them up to 6 times more sensitive to glycolysis inhibition with 2-D-deoxyglucose (Figure 3 – figure supplement 1E), similar to control cells treated with 2.5 to 5 mM metformin (Figure 3 – figure supplement 1F). These findings confirm that both ATP5I deletion and metformin treatment disrupt respiration conferring a dependency on glycolysis⁵².

Figure 3.

Figure 3 – figure supplement 1.

Moreover, the inhibition of the mitochondrial respiratory chain observed in ATP5I KO cells also induces an energy imbalance, resulting in AMPK activation (Figure 3C), similarly to control cells treated with metformin (Figure 3 – figure supplement 1G). However, the increased phosphorylation state of AMPK in ATP5I KO cells treated with metformin at these concentrations suggests that metformin can still activate AMPK via interactions with other targets such as PEN2³¹.

Energy stress resulting from ATP5I KO in KP-4 cells leads to a noticeable slowdown in cell growth (Figure 3D), although this effect can be reversed by supplementing the cell culture medium with pyruvate and uridine (Figure 3 – figure supplement 1H). These findings suggest that while ATP5I is not essential for growth in KP-4 cells, its deletion subtly affects energy metabolism, akin to the effects seen in KP-4 cells treated with medicinal biguanides³⁷.

Seahorse kinetic monitoring of control and ATP5I KO cells treated with 5 mM metformin reveals, as expected, that metformin causes a decrease in oxygen consumption (Figure 3E) and an increase in glycolysis (Figure 3F) over time. However, these effects are more pronounced in control cells than what is observed in ATP5I cells, particularly in terms of glycolysis activation. The OCR/ECAR ratio also decreases more in control cells upon metformin treatment than in ATP5I KO cells (Figure 3G), suggesting reduced sensitivity to metformin in the KO cells. This resistance is also evident in the relative growth of cells under increasing doses of metformin. Indeed, ATP5I KO cells exhibit EC₅₀ values up to 9 times higher than those of control cells (Figure 3H). Overall, these results suggest that metformin is normally acting partially through ATP5I as indicated by the reduced sensitivity of the KO cells.

Similar trends where obtained, when using 100 μM phenformin in seahorse kinetic with more pronounced changes in both respiratory activity decline (Figure 3I) and glycolysis activation (Figure 3J). These differences appear earlier but over a shorter time compared to metformin, possibly due to phenformin’s pharmacokinetic profile⁵³. Again, phenformin affects both control and ATP5I KO cells but it does so less efficiently in ATP5I KO cells. The difference in OCR/ECAR ratio between control cells and the two ATP5I KO cell lines at 3 hours is significantly less impacted (Figure 3K), indicating lesser sensitivity of ATP5I KO cells to phenformin. Resistance to phenformin is also evident in cell growth, with ATP5I KO cells showing EC₅₀ up to 4 times higher than those of control cells (Figure 3L). These results suggest that ATP5I mediates the metabolic and antiproliferative effects of biguanides in KP-4 cells.

Re-expression of ATP5I normalizes mitochondrial morphology, metabolic profile and resensitizes ATP5I knockout pancreatic cancer cells to biguanides

To validate our previous hypotheses regarding the mechanism of action of ATP5I, we reintroduced wild-type ATP5I in ATP5I KO KP-4 clones. Immunoblot analysis results confirmed that exogenous ATP5I could be re-expressed in ATP5I KO cells, albeit at slightly reduced levels compared to control cells. Reintroducing ATP5I also restored the protein levels of NDUFB8 and COX II in KP-4 ATP5I KO cells (Figure 4A). Furthermore, exogenous ATP5I colocalized with TOMM20, indicating its successful mitochondrial localization and led to a reorganization of the mitochondrial networks (Figure 4 – figure supplement 1). Indeed, quantitative analysis showed that about 60% of ATP5I KO cells with exoATP5I exhibited an intermediate fragmented phenotype between punctate and filamentous forms, while around 34% showed a filamentous phenotype and only few displayed a punctate phenotype (Figure 4C).

Figure 4.

Figure 4 – figure supplement 1.

Similarly, ATP5I KO cells expressing exogenous ATP5I showed increased NAD⁺/NADH ratios (Figure 5A) by restoring NAD⁺ concentration (Figure 5 – figure supplement 1A, 1B), increased OCR/ECAR ratio (Figure 5B), enhancing respiration and reducing compensatory glycolysis (Figure 5 – figure supplement 1C, 1D) to levels comparable to control cells. This metabolic reorganisation made these cells up to 3 times less sensitive to 2-D-deoxyglucose (Figure 5 – figure supplement 1E) compared to ATP5I KO cells, thereby alleviating energy stress as seen with less phosphorylated AMPK (Figure 5C) and enabling increased cell growth (Figure 5D). Restoring the ATP5I expression is also correlated with increased sensitivity to the metabolic effects induced by metformin and phenformin on mitochondrial respiration and glycolysis and decreased OCR/ECAR ratios (Figure 5 – figure supplement 2). The rescue also enhanced the antiproliferative effects of metformin, with EC₅₀ values up to 5-fold lower than those of ATP5I KO cells (Figure 5E), as well as those of phenformin, with EC₅₀ values up to 3-fold lower than those of ATP5I KO cells (Figure 5F).

Figure 5.

Figure 5 – figure supplement 1.

Figure 5 – figure supplement 2.

In conclusion, these findings collectively demonstrate that the re-expression of ATP5I in ATP5I KO clones rescues mitochondrial morphology, metabolic profile and sensitivity to biguanides, which supports the concept that ATP5I mediates the metabolic and antiproliferative effects of these compounds in KP-4 cells.

Chemogenomic screening of metformin reveals an imprint on F₁F_o-ATP synthase

To obtain an unbiased observation of biological processes affected by metformin, we performed a genome-wide pooled CRISPR/Cas9 KO screen in NALM-6 cells cultured in the presence of metformin at a concentration affecting growth (16 mM). Over the 8-day period of screening, cells had 1.47 population doublings whereas untreated controls had 7.5. Based on the sgRNA frequency changes observed afterward between treated vs untreated samples, we generated enrichment/depletion scores for each genes using the CRANKS algorithm. Many gene KO were then predicted together to potentiate (sgRNA depletion, KO creates sensitivity, negative CRANKS scores) or to suppress (sgRNA enrichment, KO creates resistance, positive CRANKS scores) metformin-induced growth inhibition (Figure 6A). Inactivation of the amino acid transporter SLC7A6 that plays a role in arginine export⁵⁴ enhanced metformin activity, likely because this protein may also export metformin. Indeed, the end of the R group is very similar chemically to metformin. In addition, disabling the transcription factors SOX4, TCF4, SPI1 and the histone H3K79 methyltransferase DOT1L also enhanced metformin activity perhaps because they promote gene expression programs that compensate the mitochondrial dysfunction associated to metformin action. For example, SOX4 and SPI1 can increase glycolysis by promoting HK2 expression^55,56 while TCF4 increases glycolysis by inducing the expression of the glucose transporter GLUT3⁵⁷. Moreover, metformin can increase DOT1L mediated H3K79 methylation promoting the expression of SIRT3, a mitochondrial deacetylase that stimulates mitochondrial health and biogenesis⁵⁸. On the other hand, many genes were found to be required for metformin-mediated growth inhibition. Of note, DDX3X, an RNA helicase required for the translation of interferon response genes⁵⁹, suggests a role for the interferon pathway in the antiproliferative activity of metformin. The actions of metformin were also suppressed by inactivation of the OMA1-DELE1-HRI pathway. OMA1, is a protease localized to the inner mitochondrial membrane activated by mitochondrial stress. DELE1 is a target of OMA1 that interacts with and activates EIF2AKI (also known as heme-regulated inhibitor of HRI). HRI catalyzes eIF2a phosphorylation which in turns activates the translation of ATF4⁶⁰ which is also stimulated by DDX3⁶¹. ATF4 together with ATF3 regulates the expression of prop-apoptotic genes in response to both ER and mitochondrial stress^60,62. Finally, inactivation of VDAC2, BAK1, HCCS (which makes Cytochrome C), Cytochrome C (CYCS) and APAF1 also conferred resistance to metformin, suggesting that metformin induces the classical mitochondria-related intrinsic apoptosis in treated cells. We compared the pattern of genetic interactions of metformin with that of rotenone, a complex I inhibitor (Figure 6B) and oligomycin A, a complex V inhibitor (Figure 6C). Interestingly, there were more similarities between oligomycin A and metformin (16 genes in common) than between rotenone and metformin (4 genes in common). Notably, inactivation of the mitochondrial apoptotic pathway through OMA1-DELE1-HRI as well as glycolytic pathway through SOX4 and TCF4 significantly affected both oligomycin A and metformin but not rotenone. On the other hand, DDX3X was required for the actions of all three drugs. Taken together, this genetic experiment shows that the OMA1-DEL1-HRI pathway mediates the antiproliferative activity of both biguanides and the F1ATPase inhibitor oligomycin (Figure 6D) while a compensatory glycolysis protects the cells.

Figure 6.

Discussion

Using a biologically active biotin functionalized biguanide (BFB), we identified and characterized subunit e of F₁F_o-ATP synthase (ATP5I) as a potential antineoplastic mitochondrial target of medicinal biguanides. Knockdown of ATP5I in KP-4 human pancreatic cancer cells specifically altered the expression of NDUFB8 (complex I) and COX II (complex IV) proteins of the mitochondrial electron transport chain without affecting their mRNA levels, suggesting ATP5I as a role in maintaining OXPHOS protein stability. Thus, metformin binding to ATP5I can directly alter complex V activity and indirectly affect other respiratory complexes.

Interestingly, NDUFB8, a protein we found poorly expressed in ATP5I KO cells, is a target of the mitochondrial protease ClpP, whose activation can inhibit pancreatic adenocarcinoma growth⁶³. Our work shows that another protease, OMA1, is required for the action of metformin, suggesting that mitochondrial proteases can have potent antitumor activities. Other studies, using mass spectrometry, revealed interactions between ATP5I and essential components of mitochondrial protein import, such as TIMMDC1, involved in complex I’s import and assembly^64,65. These findings suggest that ATP5I is integral to respirasome assembly and mitochondrial proteome stability. Moreover, recent evidence implicates transmembrane proteins TMEM70 and TMEM242 in complex I and F₁F_o-ATP synthase assembly or subunit expression regulation, linking the ATP synthase complex to complex I⁶⁶. Therefore, biguanides may inhibit complex I activity by binding to and disrupting either complex I or the ATP synthase complex.

Similar to medicinal biguanides effects on KP-4 cells³⁷, ATP5I depletion leads to mitochondrial structural changes such as mitochondrial fragmentation. Additionally, ATP5I deletion disrupts energy metabolism, decreasing complex I mediated NADH reoxidation and respiration, promoting glycolysis and AMPK activation, and impairing cell growth. Furthermore, ATP5I KO cells exhibit resistance to metformin and phenformin growth effects, indicating its pivotal role in mediating their metabolic and antiproliferative effects. Reintroducing wild-type ATP5I in ATP5I KO cells partially restored their sensitivity to biguanides, consistent with the proposed role of ATP5I being a biguanide target. Of note, the phenotypes observed in the absence of subunit g (ATP5L)⁶⁷, the direct partner of ATP5I, and subunit f (ATP5K)⁶⁸ in HeLa cells are similar to those of ATP5I mutants and metformin treated cells. Further work is needed to determine how metformin affects the interactions between ATP5I, ATP5L and ATP5K. Together these three proteins are located at the base of F₁F_o-ATP synthase peripheral stalk which interact with OSCP (oligomycin sensitivity conferring protein) located in the F1 portion of the enzyme ⁶⁹. ATP5I’s movement within F₁F_o-ATP synthase⁶⁹ may regulate and transfer information between the F1 and the F0 rotating complex ^42,64-66. These structural connections could underpin the results reported here that metformin and oligomycin share a common genetic signature in a CRISPR screening in NALM-6 cells. Our findings, combined with recent published results, strengthen the correlation between mitochondrial ultrastructure and respiratory activity⁷⁰, suggesting that medicinal biguanides partially affect mitochondrial respiratory chain activity by indirectly inhibiting complex I of OXPHOS through ATP5I binding, although the exact complete molecular mechanism remains unclear.

The upregulation of ATP5I observed in hypoxia⁷¹, DNA damage⁷², liver cancer⁷³, and low-fat diet responses^74,75 underscores a regulatory role still poorly understood. In another study, RNA-seq data form A549 cells also suggest ATP5I levels influence metformin sensitivity⁷⁶, highlighting its potential clinical relevance. Identifying ATP5I as a potential antineoplastic target for medicinal biguanides provides insights into the mechanisms underlying their anticancer effects and opens new avenues for targeting mitochondrial metabolism. These findings also underscore the need for further exploration of ATP5I’s functions within F₁F_o-ATP synthase and its connections with other OXPHOS complexes, potentially leading to novel therapeutic strategies for cancer treatment.

Acknowledgements

We wish to express our deepest gratitude to Professor Omichinski, for his expert guidance and unwavering support in the purification of proteins, and also to Professor Masson and Affinité Instrument for their generosity in providing access to the SPR system.

Author contributions

The manuscript was drafted by G.L., with all authors approving the final version. The project was initially launched by M.C.R. and F.M., who conducted the first pull-down experiment and BFB biosynthesis, respectively (Fig. 1F and supporting information). V.B. generated the stable ATP5I knockout cells and performed all qPCR analyses (Fig. 2 – figure supplement 1A, 1B). E.L. conducted the Seahorse experiments (Fig. 3B, 3E, 3F, 3G, 3I, 3J, 3K, 5B, Fig. 3 – figure supplement 1C, 1D, and Fig. 5 – figure supplement 2). T.B. and M.K. carried out the chemogenomic CRISPR/Cas9 knockout screen (Fig. 6). G.L. completed the remaining work. A.R.S., G.F., and S.P.G. supervised the research and finalized the manuscript. All authors declare no conflict of interest.

Materials and methods

Drugs

Metformin hydrochloride was purchased from Combi-Blocks (#ST-9194; San Diego, CA, USA), phenformin hydrochloride was purchased from Sigma Aldrich (#P7045; Oakville, On, Canada), 2-D-deoxyglucose was purchased from Bioshop (#DXG498.5; Burlington, On, Canada) and D-biotin was purchased from Invitrogen/Thermo Fisher Scientific (#B1595; Waltham, MA, USA).

Synthesis

All chemicals were purchased from Sigma Aldrich, Oakwood Chemicals (Estill, SC, USA) and Combi-Blocks in their highest purity and were used without further purification. Deuterated dimethylsulfoxide (DMSO-d6), deuterated methanol (MeOD) and deuterium oxide (D₂O) were purchased from CDN Isotopes (Pointe-Claire, Qc, Canada). NMR spectra were recorded on Bruker avance 400 and Bruker avance 500 spectrometers. Coupling constants (J) were reported in hertz (Hz), chemical shifts were reported in parts per million (ppm, δ) and multiplicities were reported as singlet (s), doublet (d), triplet (t) and multiplet (m). ¹H and ¹³C chemical shifts are relative to the solvents: δH 2.50 δC 39.5 for DMSO-d6, δH 3.35, 4.78 and δC (49.3) for MeOD and δH 4.65 for D₂O. Final compounds were purified using a preparative liquid chromatography (prepLC-MS-Quadrupole, Waters) equipped with a C18 reverse phase column (2.1 × 100 mm, 3 μm, Atlantis) with HPLC-grade solvents (Sigma Aldrich). High-resolution mass spectra (HRMS) and LC-MS purity analysis were collected using LC-TOF with ESI ionisation source (Agilent Technologies, Santa Clara, CA, USA) by the regional mass spectrometry center of University of Montreal. After prep HPLC, all final compounds were subsequently freeze-dried and aliquots of the monochloride salt powder were carefully weighed.

Cell culture

Human pancreatic ductal carcinoma cell line KP-4 was as a kind gift from Dr. N. Bardeesy (Massachusetts General Hospital, Boston). Human embryonic kidney HEK-293T cells were obtained from Thermo Fisher Scientific. Phoenix ampho packaging cells was as a kind gift from Dr. S.W. Lowe (MSK, New York). NALM6 cells were a gift from Stephen Elledge (Harvard, USA). Except for NALM-6 cells which were cultured in RPMI medium (#350-035; Wisent, All cells were cultured in Dulbecco’s modified Eagle medium (DMEM #319-015; Wisent, St-Jean-Batiste, Qc, Canada) supplemented with 10% fetal bovine serum (FBS, Wisent), 2 mmol/L L-glutamine (Wisent) and 1% penicillin/streptomycin sulfate (Wisent) in a humidified incubator at 37 °C with 5% CO₂.

Immunoblotting

For immunoblotting, proteins extracts were obtained from 6-well cell culture dishes at 80-90 % confluence. Cells were treated with the corresponding drugs or vehicle 16 hours after seeding. As previously described⁷⁷, cells were washed twice with ice-cold phosphate-buffered saline solution (PBS), then the residual PBS was removed by aspiration. Cells were lysed in 250 μL of 2x Laemmli buffer (4% SDS, 20% glycerol, 120 mM Tris-HCl pH 6.8), recovered using a cell scraper and transferred into Eppendorf tubes for a mild sonication of 20s followed by heating for 5 min at 95 °C. Protein extracts were cooled to room temperature (rt) and quantified by measuring absorbance at 280 nm using Nanodrop microvolume spectrophotometer (2000c, Thermo Fisher Scientific). Samples were then diluted to a final concentration (2 mg/mL) using a modified Laemmli Buffer (10% β-mercapthoethanol, 0.1% bromophenol, 2x Laemmli) and were kept at - 20 °C until use. A quantity of 25 to 40 μg of protein extracts were loaded into SDS-PAGE (12-15% for resolving gels, 4% for stacking gels) and run in Tris-Glycine SDS buffer (192 mM glycine, 0.1% SDS, 25 mM Tris-Base) at 90 V. SDS-PAGE were then transferred on a nitrocellulose membrane (0.45 µm, Bio-Rad, Mississauga, On, Canada) in Tris-Glycine buffer (96 mM glycine, 10 mM Tris-Base) for 1h30 at 120 V. Membranes were blocked in a modified Tris-buffered saline + 0.05% Tween-20 (TBST) solution with 5% skim milk for 1h at rt, washed with TBST 3x 5 min at rt and were then incubated with an appropriate diluted solution of primary antibody (in 0.1% BSA, 0.02% sodium azide, PBS pH 7.4) overnight (o/n) at 4°C. After primary antibody incubation, membranes were washed with TBST 3x 5 min and incubate with a diluted (1:3000 in 1% skim milk/TBST) secondary antibody coupled HRP for 1 hour at rt. After secondary antibody incubation, membranes were washed with TBST 3x 5 min. The signal was revealed using enhanced chemiluminescence reagent (ECL), (#RPN2106, GE Healthcare Life Sciences, Chicago, IL, USA) and images were acquired by exposition with autoradiographic films or using ChemiDoc Imaging Sytems (XRS+; Bio-Rad). FroggaBio BLUelf Prestained Protein ladder (FroggaBio, Concord, Om, Canada) was used to estimate protein molecular weight. Of note, most membranes were cut into pieces and were incubated with several antibodies. The following primary antibodies were used: anti-phospho-ACC S79 (Rabbit, 1:1000, #3661S, Cell signaling, Danvers, MA, USA), anti-AMPK (Rabbit, 1:1000, #2532, Cell signaling), anti-phospho-AMPK T172 (Rabbit, 1:1000, #2531, Cell signaling), Anti-ATP5I (Rabbit, 1:500, #16483-1-AP, Proteintech, Rosemont, IL, USA), anti OXPHOS cocktail human (Mouse, 1:750, #ab110411, Abcam, Toronto, ON, Canada), anti-GAPDH (Goat, 1:3000, #NB300-320, Novus Biologicals, Centennial, CO, USA), anti-α-Tubuline (Mouse, #T6074, Sigma Aldrich) and anti-β-Actin (Mouse, 1:10000, #3700, Cell signaling). The following secondary antibodies were used: anti-rabbit IgG conjugated to HRP (Goat, 1:3000, #170-6515, Bio-Rad), anti-mouse IgG conjugated to HRP (Goat, 1:3000, #170-6516, Bio-Rad) and anti-goat IgG conjugated to HRP (donkey, #Sc-2020, Santa Cruz Biotechnology, Dallas, TX, USA).

Immunofluorescence

For immunofluorescence experiments, 50,000 - 100,000 cells were seeded on 1.5 mm thickness coverslips and were incubated for 48 h. As previously described⁷⁸, cells were washed twice with ice cold PBS, then were fixed with a 4% paraformaldehyde solution for 10 min at 4 °C and were washed 3x 10 min with a solution of PBS + 0.1 M glycine at rt with agitation. At this moment coverslips were sometime stored in PBS + 0.2 % Sodium Azide at 4°C until use. If stored, 3x 10 min washing steps with PBS were performed to remove the azide. Cells were then permeabilized with PBS + 0.2 % Triton X-100 and 3% Bovine Serum Albumin (BSA) for 5 min at rt. Then, cells were washed in a blocking solution (3 % BSA in PBS pH 7.4) 3x 10 min at rt with agitation and were then incubated with an appropriate diluted solution of primary antibody (in 3% BSA, PBS pH 7.4) o/n at 4°C in a humidified chamber. After primary antibody incubation, cells were washed in blocking solution 3x 10 min and were incubated with an appropriate diluted solution (3% BSA, PBS pH 7.4) of secondary AlexaFluor antibody for 1h at rt in the dark. Cells were then washed with PBS 3x 10 min and coverslips were mounted on glass coverslip in Vectashield mounting media containing DAPI (H-1200-10, Vector Laboratories, Newark, CA, USA) and were sealed off with nail polish and kept for a minimum of 24 hours at 4°C. Images were collected with the Zeiss Axio Imager Z2 upright microscope equipped with a CoolSNAP FX camera (Photometrics), Axiocam camera and ZEN 2 blue edition software and were analyzed using ImageJ software. Colocalization was assessed using the Job Plot profile functionality of ImageJ by determining fluorescence intensity for each pixel of each channel across a line drawn. For quantifications and colocalizations, raw data were exported to Prism (10.2.2, GraphPad) software to generate final figures. The following primary antibodies were used: Anti-ATP5I (Rabbit, 1:100, #16483-1-AP, Proteintech) and TOMM20 (Mouse, 1:100, #sc-104216, Santa Cruz Biotechnology). The following secondary antibodies were used: anti-mouse AF488 (Goat, 1:2000, #A-11029, Invitrogen), anti-rabbit AF568 (Goat, 1:2000, #A-11011, Invitrogen) and Streptavidin AF488(#S11223, Invitrogen).

Mitochondrial protein isolation

Mitochondrial crude extracts were obtained using Mitochondria Isolation Kit for Cultured cells (ab110170, Abcam) according to manufacturer’s instructions. For each condition, 1x 15 cm cell culture dish of KP-4 cells was washed twice in ice-cold PBS, then the residual PBS was removed by aspiration. Cells were then scraped in ice-cold PBS and transferred in Eppendorf tubes. The following steps for mitochondrial isolation extraction were performed as described in the manufacturer’s instructions. The pellet containing mitochondrial proteins was resuspended in lauryl maltoside buffer [1% lauryl maltoside, cOmplete-EDTA free Protease Inhibitor Cocktail (Roche, Laval, Qc, Canada), PhosSTOP (Roche), PBS pH 7.8]. The concentrations of mitochondrial proteins were determined using the bicinchoninic acid assay (BCA) (23225, Pierce). Of note, at the end of isolation, we obtained ∼ 0.7-1 mg of mitochondrial proteins per 15 cm dish.

Pull-down assay

For pull-down assay, 1 mg of mitochondrial proteins were incubated with either 1 mM of biotin functionalized biguanide (BFB) or biotin functionalized amine (BFA) in Eppendorf tubes for 3h at rt with mild agitation. In parallel, Dynabeads MyOne streptavidin C1 (65001, Thermo Fisher Scientific) were washed and resuspended at least three times in lauryl maltoside (LM) buffer [1% lauryl maltoside, cOmplete-EDTA free Protease Inhibitor Cocktail (Roche), PhosSTOP (Roche), PBS pH 7.8]. After incubation with biotinylated molecules, beads (0.5 mg ∼ 50 μL) were added to mitochondrial lysates and incubated for 30 min at 4°C with agitation. Beads were respectively recovered with a magnet, washed in LM buffer 3x 30 min at 4°C. Elution of bound proteins from beads in BFB condition was performed by adding 30 μL of 50 mM metformin in LM buffer. After an incubation time of 30 min at rt, beads were vortexed, pelleted with a magnet and the supernatant was collected in a new tube. Proteins were then denatured by adding 30 μL of 6x SDS-loading buffer (30% glycerol, 10% SDS, 1% bromophenol blue, 15% β-mercaptoethanol, 0.5 M Tris-HCl pH 6.8) to the mixture and boiled 3x 5 min at 95°C. Elution of bound proteins for BFA condition and remaining proteins on beads for BFB condition were achieved by boiling and vortexing the corresponding beads in 60 μL of 6x SDS loading buffer 3x 5 min at 95°C. After denaturation of proteins, all samples were loaded into SDS–PAGE. The resulting gel was stained with Coomassie brilliant blue R-250 (#33445225GM, Thermo Fisher Scientific) then washed with a destaining solution (40% water, 50% methanol, 10% acetic acid). Gel segments were then excised, digested with trypsin and dried overnight in a cold trap (Labconco, Kansas City, MO, USA). Three samples for each elution condition were prepared in two technical replicates and were then analyzed by the Mass spectrometry platform at IRIC (Institute for Research on Immunology and Cancer). It’s worth mentioning that for pull-down validation, same experiments were performed using western-blot analysis with anti ATP5I antibody after protein denaturation steps (see immunoblotting section).

Constructs

For recombinant protein expression in bacteria, full-length of human ATP5I sequence was PCR-amplified and cloned using BamH1/EcoRI restriction sites into a pET-TEV vector (Addgene, Watertown, MA, USA) for the expression of a N-terminal 6x-His-tagged protein. This plasmid was obtained as a kind gift from Dr. J.G. Omichinski (University of Montreal, Montreal). For lentiviral mediated transduction in KP-4 cells, annealed and phosphorylated oligos for two RNA guides (sgRNAs) targeting two regions of ATP5I’s cDNA (sgATP51 #1 and sgATP5I #2) and one RNA guide targeting GFP (sgGFP) sequence were subcloned into BsmBI restriction site of lentiCRISPRv2 plasmid (#52961, Addgene). For re-expression of ATP5I in control (sgGFP) and ATP5I knockout (KO) (sgATP5I#1 and #2) cells, full-length human ATP5I sequence with modified codons (same amino acid but different nucleic acid sequence) was generated by PCR site-directed mutagenesis and cloned using BamH1/EcoRI restriction sites into MSCV retroviral vector. All constructs were confirmed by DNA sequencing.

Constructs primers used:

Protein expression and purification

N-terminal 6x-His-tagged ATP5I was expressed in Rosetta E. coli BL21 competent cells (#70954, Addgene) for the expression of eukaryotic proteins that contain rare codons. To overexpress recombinant protein as suggested⁷⁹, cells were grown at 37°C in terrific broth (TB) medium supplemented with 1% of ethanol (v/v), 100 mg/mL ampicillin and 50 mg/mL chloramphenicol to an OD_600nm of 0.8. Expression was induced for 4h at 30°C with 1 mM isopropyl β-D-1-thiogalactopyranoside then cells were harvested by centrifugation (15 min, 5000 rpm at 4°C). Bacterial pellets were then solubilized in buffer A (6 M guanidine, 500 mM NaCl, 20 mM NaH₂PO₄ pH 8) for 30 min at 4°C and lysed by sonication. Cell lysate was then harvested by centrifugation (30min, 13 000 rpm at 4°C) and after filtration (0.45μm) the supernatant was loaded in an immobilized metal ion affinity chromatography column (HisTrap FF, Cytiva) with buffer A. After removing non-specifically bound molecules (flow through), the column was washed with buffer B (8M urea, 500 mM NaCl, 20 mM NaH₂PO₄ pH 8) then with buffer C (20 mM imidazole, 500 mM NaCl, 20 mM NaH₂PO₄ pH 8) until UV absorbance returned to baseline. After washing, the column was eluted with buffer D (500 mM imidazole, 500 mM NaCl, 20 mM NaH₂PO₄ pH 8) and the resulting eluate was loaded in a desalting column (HiPrep 26/10, Cytiva) with a PBS containing 125 mM NaCl, 5 mM DTT, PBS pH 7.4) and concentrated in a 3 kDa centrifugal filter (Amicon) to 1.5 mg/mL. Of note, all purification steps were done with AKTA pure chromatography system.

Surface plasmon resonance (SPR)

Experiments were performed at 25°C with a portable SPR (P4SPR) device from Masson’s laboratory (WO/2010/130045, Affinité instruments) using a compatible prism coated with a 200 nm streptavidin (SA) derivatized carboxymethyldextran hydrogel (Xantec bioanalytics) for immobilisation of biotinylated ligands. For each solution, a volume of 300 μL was injected in the flow cells comprising a multifluidic channel cell and a reference cell. After having stabilized the baseline (∼30 min) with running buffer (125 mM NaCl, 5 mM DTT, PBS pH 7.4), biotin functionalized biguanide chloride salt was immobilized on SA sensor chip at 125 μM (concentration experimentally determined to saturate streptavidin sites). Then, the chip was washed with running buffer until stabilising the baseline. After that, several concentrations of purified recombinant ATP5I were injected until saturation of the signal. Of note, between each concentration added, washing steps were performed with running buffer to remove non-specifically bound molecules. Finally, the surface was regenerated with urea (2M) prepared in milli-Q-water. It’s worth nothing that due to the low dissociation constant of biotin/streptavidin interaction, biotinylated ligands are almost irreversibly bound that make it difficult to regenerate the chip. The sensorgrams were analysed with P4SPR Control v2.022 software and final figures were exported to Prism (10.2.2, GraphPad) using a pseudo-first order association kinetics equation for each concentration. For binding affinity curve, each steady state for each concentration were exported to Prism (10.2.2, GraphPad) software to generate the figure using a one site-specific binding equation. Of note, compounds were dissolved in the running buffer.

Generation of stable ATP5I knockout cells

Knocking out ATP5I was realized in KP-4 cell line using lentiviral CRISPR/Cas9 technology with sgRNAs against ATP5I or a control sgRNA (see construct section). First, 5 × 10⁶ 293T cells were seeded in 10 cm plates and grown for 16 hours. They were then transfected using the calcium-phosphate precipitation method as previously described⁸⁰ with 3 μg of lentiCRISPRv2 plasmid vector, 2 μg of the pCMV-dR8.2 plasmid and 1 μg of the VSV-G envelope protein expression plasmid (both king gift of Dr. N. Bardeesy). After 16 h, 10 mM sodium butyrate (Sigma Aldrich) was added for a minimum of 6 h, and then the medium was changed. In parallel, target cells were seeded to obtain 60% confluence the following day. Media from the transfected cells were collected the following day, filtered through a 0.45 μm filter. This viral soup was supplemented with 4 μg/mL polybrene (Sigma Aldrich) and 10% fresh media before being placed on KP-4 receiving cells. Viral soup was removed 24 hours later for fresh DMEM cell culture media (see cell culture section) supplemented with 100 μg.mL^-1 sodium pyruvate (#600-110-EL, Wisent) and 50 μg.mL^-1 uridine (#URD222.10, Bioshop) to help them growth in case mitochondrial functions were affected⁸¹. KP-4 infected cells were selected 6 hours after media change with 2ug/mL puromycin (#400-160-EM, Wisent). After two weeks in culture to ensure time alterations in the targeted regions with CRISP/CAS9, cells were replated highly diluted to isolate clones of cells. Many clones were grown and characterized by ATP5I immunoblots to select proper ATP5I knock-out.

Retroviral infections

ATP5I re-expression was performed by retroviral-mediated expression of wild-type ATP5I in control (sgGFP) and ATP5I KO cells (sgATP5I #1 and #2) with a MSCV-ATP5I vector (see constructs).

For retroviral infections, 5 × 10⁶ Phoenix-Ampho packaging cells were seeded in 10 cm cell culture dishes and grown for 24 hours. Then, cells were transfected with 20 μg of MSCV-ATP5I vector and 10 μg of Helper envelope protein expression plasmid using calcium phosphate method as previously described⁸⁰. The next day, sodium butyrate was added at 10mM to the dishes for 6 hours and then fresh medium was added. In parallel, target cells were seeded to obtain 60-70% on confluence the day of infection. Two days following transfection, the viral soup from one 10 cm dish was filtered through a 0.45μm filter, supplemented with 4 μg.ml⁻¹ polybrene (#H-9268, Sigma Aldrich) and 1/10 (v/v) of fresh media and then used to infect one 10 cm dish of target cells. After at least 12 h post infection, infected cells were seeded into a new dish for passage. After at least 4 hours of adherence, infected cells were selected using 50 μg/mL hygromycin (#400-141-UG, Wisent). Of note, the selection was maintained for the length of the experiments.

Quantitative PCR (qPCR)

RNA isolation and qPCR quantifications were performed exactly as previously described⁷⁷ using the following primers (see below, BioCorp, Pierrefonds, Qc, Canada). Mitochondrial to nuclear DNA ratio were performed similarly by qPCR but with primer targeting DNA and on 30 ng of genomic DNA purified from cultures KP-4 cells. For the isolation, one 6 cm plate of sub confluent KP-4 cells were rinsed twice in PBS, collected by scraping in 500μL of SNET lysis buffer (5 mM EDTA, 400 mM NaCl, 1% SDS, 400 ug/mL Proteinase K, 20mM Tris-HCL pH8.0 [Sigma Aldrich]) and incubated at 55°C overnight. Equal volume of phenol:chloroform (Sigma Aldrich) was added and left to agitate for 30 min at rt. After a 5-minute centrifugation (16,000g), the aqueous phase was transferred to a new tube and precipitated by an equal volume of isopropanol, mixed well and DNA was collected by centrifugation at maximum speed (16,000g) for 15 minutes at 4°C. Pellet was washed with 70% ethanol, air dried and resuspended in TE.

qPCR primers used:

NAD⁺/NADH quantification

For NAD⁺/NADH quantifications, 5×10⁵ cells were seeded in 10 cm cell culture dishes for technical duplicates and incubated for 48h. NAD⁺/NADH ratio were quantified using a fluorometric kit assay (#MAK460-1KT, Sigma Aldrich) following manufacturer’s instructions. For sample preparation, 9×10⁵ cells were pelleted and homogenized with corresponding extraction buffers to be in the linear range of the kit. Fluorescence intensities were measured at λ_Ex = 530 nm/λ_Em = 585 nm on SPARK 10 M (TECAN) and data were exported to Prism (10.2.2, GraphPad) software to generate final figures.

Seahorse experiments

Seahorse XFe96 cell culture microplates (Agilent) were coated with 50 µg/mL of Poly-D-Lysine (#A3890401, Thermo Fisher Scientific) in D-PBS (#311-425-CL, Wisent) for 1 hour at room temperature and extensively washed with sterile water. Seahorse XFe96 sensor cartridges (Agilent) were calibrated according to the manufacturer’s protocol. Assay media was prepared as follows: 828 g DMEM without L-Glutamine, phenol red, sodium pyruvate, and sodium bicarbonate (#219-060-XK, Wisent), 29.2 mg L-glutamine (#AC386032500, Fisher Chemical), 450.4 mg D-glucose (#D-16-500, Fisher Chemical), 100 µL HEPES 1M (#330-050, Wisent), completed at 100 mL with Milli-Q water and adjusted pH at 7.2. Cells were plated in Poly-D-Lysine-coated microplates at a density of 40,000 cells/well in 100 µL of assay media. Following a 30-minute incubation at room temperature, 75 µL of assay media was added to each well, and the plate was transferred to an XFe96 analyzer. Typical runs were composed of the following steps: 1) baseline measurements for 18 minutes (3 cycles, each of 3-minute mixing and 3-minute measurement), 2) single drug injection (5 mM metformin, 100 µM phenformin, or sterile water as vehicle control), and 3) measurements over a period of 10.5 hours (63 cycles, each of 6-minute mixing, 1-minute pause, and 3-minute measurement). Oxygen consumption rates (OCR) and extracellular acidification rates (ECAR) were extracted from the Wave software (Agilent) and all time points were normalized to the first measurement. Data were then exported to Prism (10.2.2, GraphPad) software to generate final figures.

Crystal violet cell number assay

The crystal violet staining retention assay⁸² was used to estimate cell growth and viability by measuring absorbance at 590 nm in a microplate reader (SpectraMax 190 microplates, Molecular Devices).

For the growth assay 5,000 cells were seeded in 4 different 6-well cell culture dishes in technical triplicates and incubated for 4 different time points up to 6 days. Of note, media was replaced every 48 hours. At indicated times, cells were washed twice in PBS and fixed in 1% glutaraldehyde solution (PBS) for 10 min at rt. After fixation, cells were washed twice in PBS. Once all time point were collected, plates were stained in a 0.05 % crystal violet solution (PBS) for 30 min with agitation. After staining, plates were washed five times in water and were dried for 24 hours. The next day, crystal violet retained in cells were solubilized in 10 % acetic acid solution for 15 min with agitation and transferred into a 96 well plates for absorbance measurement. Data were then exported to Prism (10.2.2, GraphPad) software to generate final figures.

For the cell viability assay, 1,000 cells were seeded in 96 well plates in technical triplicates. After 24 hours, cells were treated at different concentrations with the corresponding drugs and were grown for 72 hours. Crystal violet staining was performed as described for the growth assay. The half minimal effective concentrations were obtained using a fit a curve with non-linear regression log(inhibitor) vs. response -variable slope (four parameters) from the corresponding dose response curves with Prism (10.2.2, GraphPad) software. Importantly, because growth rate was affected in ATP5I KO cells generating variabilities, cell viability assays comparing ATP5I KO with control cells were done in media supplemented with pyruvate and uridine (see generation of stable ATP5I KO cells section). Of note, drugs were dissolved in the corresponding media.

Chemogenomic CRISPR/Cas9 KO screen in NALM-6 cells

The Genome-wide pooled CRISPR/Cas9 KO screens made in the presence of chemicals inhibiting growth were performed by the ChemoGenix platform (IRIC, Université de Montréal; https://chemogenix.iric.ca/) as previously described⁸³. Briefly, a NALM-6 clone bearing an integrated doxycycline-inducible Cas9 expression cassette generated by lentiviruses made from pCW-Cas9 (Addgene #50661) was transduced with the genome-wide KO EKO sgRNA library⁸³ (278,754 different sgRNAs). After thawing the library from liquid N₂ and letting it recover in 10% FBS RPMI for 1 day, KOs were induced for 7 days of culture with 2 ug/mL doxycycline. The pooled library was then split in different T-75 flasks (28×10⁶ cells per flask; a representation of 100 cells/sgRNA) in 70 mL at 4×10⁵ cells/mL. Cells were treated with 16 mM Metformin (using 1M stock solution prepared in media) or 70 nM rotenone (Sigma, R8875) or 2µM oligomycin A (Tocris Bioscience, 4110) (both from 1000X stock solutions prepared in DMSO) for 8 days with monitoring of growth every 2 days, diluting back to 4×10⁵ cells/mL and adding more compounds to maintain same final concentration whenever cells reached 8×10⁵ cells/mL. Over that period, treated cells had respectively 1.47, 2.93 & 4.14 population doublings whereas no solvent or DMSO controls had about 7.5. Cells were collected, genomic DNA extracted using the Gentra Puregene kit according to manufacturer’s instructions (QIAGEN), and sgRNA sequences PCR-amplified as described⁸³. SgRNA frequencies were obtained by next-generation sequencing (Illumina NextSeq 500). Reads were aligned using Bowtie2.2.5 in the forward direction only (norc option) with otherwise default parameters and total read counts per sgRNA tabulated. Context-dependent chemogenomic interaction scores were calculated from comparing sgRNA frequency changes against negative controls from using a modified version of the RANKS algorithm⁸³ which uses guides targeting similarly essential genes as controls to distinguish condition-specific chemogenomic interactions from non-specific fitness/essentiality phenotypes. Raw read counts are available upon request directly from the ChemoGenix platform.

Statistical analysis

Statistical analyses were performed using GraphPad Prism (10.2.2) software. Unpaired two-tailed Student’s t-test or Ordinary one-way ANOVA with compensation of multiple comparison with Sida’k were used to determine significance except for seahorse analysis where paired two-tailed Student’s t-test and RM one-way ANOVA were preferred. A value of p < 0.05 was considered significant. All specific statistical details can be found in the corresponding legends.