Bromoxib causes changes in mitochondrial morphology, resulting in fission that is induced by OMA1-mediated processing of OPA1

Next, we focused on mitochondria as the executioners of the intrinsic apoptosis pathway. Therefore, we assessed the impact of bromoxib on the mitochondrial membrane potential (ΔΨm) in Ramos cells. As depicted in Fig. 4A, bromoxib induced a rapid breakdown of the mitochondrial membrane potential (ΔΨm) within 1–2 min, which was as fast as the protonophore carbonyl cyanide m-chlorophenyl hydrazone (CCCP), which served as a positive control.

Open in a separate window

Fig. 4

Bromoxib causes a rapid dissipation of the mitochondrial membrane potential (ΔΨm) and changes in mitochondrial morphology, resulting in fission that is regulated by OPA1 and OMA1. (A) Within a few minutes of bromoxib treatment, the mitochondrial membrane potential (ΔΨm) decreases rapidly. ΔΨm was monitored in Ramos cells after the addition of DMSO (0.1% v/v), bromoxib (10 µM), or CCCP (protonophore and mitochondrial uncoupler; 10 µM) by flow-cytometry measurement of TMRE fluorescence. (B) Upper panel: The kinetics of bromoxib-induced OPA1 cleavage were determined by immunoblotting in Ramos cells treated with 10 µM bromoxib or 10 µM CCCP (as positive control for OPA1 cleavage). Co-treatment with the pan-caspase-inhibitor QVD (10 µM) was conducted to ensure independence from apoptotic signaling. Lower panel: To study the recovery of long forms of OPA1, Ramos cells were treated with 10 µM bromoxib or 10 µM CCCP for 30 min, followed by substance removal through centrifugation and a recovery period of up to 6 h. (C) The effect of bromoxib (10 µM; upper panel) and of the protonophore CCCP (10 µM; lower panel) on OPA1 processing was determined in murine embryonic fibroblast (MEF) cells deficient for the OPA1 proteases OMA1 (OMA1-KO) or YME1L1 (YMEL-KO) and double knockout of OMA1 and YME1L1 (OMA1-YMEL-DKO). (B, C) Numbers under OPA1 immunoblots indicate densitometric analyses of the ratio of the long forms of OPA1 (L1 plus L2) to total OPA1 (L1 plus L2 plus S3-5). (D) Mitochondrial fission in HeLa cells stably expressing mito-DsRed targeted to the outer mitochondrial membrane after 30 and 120 min of treatment with DMSO (0.1% v/v), bromoxib (10 µM), or CCCP (10 µM). Fragmentation was observed via live cell imaging. Quantification of the mitochondrial morphology was assessed by spinning disc confocal microscopy of the same HeLa cells, categorizing 40 to 60 cells per condition in two independent experiments into tubular, intermediate, or fragmented. The bars show the mean, error bars show the range; statistics: two-way ANOVA with Dunnett’s multiple comparison test (**** = p≤0.0001). (E) Live measurement of the effect of bromoxib (10 µM) on total Ca²⁺ mobilization in Ramos cells using DMSO (0.1% v/v) as vehicle control and ionomycin (2 µM) as positive control. (F) Left graph: Effect of bromoxib on intracellular Ca²⁺ mobilization in Ramos cells. For the measurement of intracellular Ca²⁺ mobilization, cells were washed with Krebs-Ringer buffer containing EGTA as Ca²⁺-chelating agent and treated with bromoxib (10 µM), DMSO (0.1% v/v; diluent control; black line) or thapsigargin (10 µM; positive control). Right graph: Live measurement of the effect of thapsigargin (10 µM; red line) followed by bromoxib (10 µM; green line) treatment on intracellular Ca²⁺ mobilization in Ramos cells

In addition to the breakdown of ΔΨm, the mitochondrial death pathway is characterized by mitochondrial fragmentation (fission). Mitochondrial fission is partly regulated by the proteolytic cleavage of the dynamin-like GTPase OPA1 by OMA1 or YME1L1, thereby maintaining the equilibrium between mitochondrial fusion and fission. Consequently, OPA1 serves a pivotal role in mitochondrial homeostasis by sensing mitochondrial stress, such as the breakdown of the ΔΨm, and coordinating mitochondrial quality control and intrinsic apoptosis. Stress-induced cleavage of long isoforms of OPA1 (L-OPA1) and the subsequent formation of short isoforms (S-OPA1) tilts the balance towards mitochondrial fragmentation and increased sensitivity to proapoptotic stimuli [20, 21].

Therefore, we studied the impact of bromoxib on OPA1 processing and observed that bromoxib induced the rapid cleavage of L-OPA1 in Ramos cells within 5 min, which was not affected by caspase-inhibition via QVD (upper immunoblot of Fig. 4B). The recovery of long OPA1 forms upon processing after drug removal serves as an indicator of the recovery of the fusion machinery. To measure L-OPA1 recovery, Ramos cells were treated with bromoxib or CCCP for 30 min, followed by drug removal. Similar to CCCP, the continuous incubation with bromoxib led to L-OPA1 depletion. However, the removal of bromoxib and CCCP allowed the recovery of the long OPA1 forms (Fig. 4B; lower panel), indicating that the effect on mitochondrial fusion is reversible. Usually, S-OPA1 arises through proteolytic cleavage of L-OPA1 by proteases such as OMA1 or YME1L1 [22, 23]. To address which protease (OMA1 or YME1L1) catalyzes the cleavage of L-OPA1 into S-OPA1 forms, mouse embryonic fibroblasts (MEFs) lacking either OMA1, YME1L1, or both were treated with bromoxib or CCCP. The short and long OPA1 forms were then analyzed by immunoblotting (Fig. 4C). The findings indicated that OMA1 is necessary for cleaving L-OPA1 into S-OPA1, as cleavage was absent in OMA1 knockouts. Conversely, YME1L1 knockout did not affect OPA1 cleavage (Fig. 4C). Thus, these data indicate that OMA1 is the responsible protease for bromoxib (and CCCP) induced OPA1 cleavage. In addition, we observed that L-OPA1 depletion coincided with the rapid and pronounced fragmentation of the mitochondrial network within 30 min of treatment of HeLa cells with bromoxib or CCCP (used as positive control for mitochondrial fragmentation; Fig. 4D).

In addition, mitochondrial depolarization associated with a sustained increase in cytosolic Ca²⁺ can also induce mitochondrial fission through activation of the Ca²⁺-dependent phosphatase calcineurin and subsequent dephosphorylation of dynamin-related protein 1 (DRP1) [24, 25]. Therefore, we investigated whether bromoxib was able to induce the mobilization of Ca²⁺ in Ramos cells. As a positive control for Ca²⁺ influx, we used the ionophore ionomycin, which also induces mitochondrial permeability transition opening [26, 27]. Intriguingly, bromoxib induced an increase in cytosolic Ca²⁺ (Fig. 4E) even when extracellular Ca²⁺ was chelated by EGTA (Fig. 4F; left graph). In addition, we used thapsigargin as a sarcoplasmic/endoplasmic reticulum Ca²⁺- ATPase (SERCA) inhibitor to fully deplete the Ca²⁺ stores of the endoplasmic reticulum (ER). To determine the origin of the released Ca²⁺, thapsigargin was initially applied to release all Ca²⁺ stored within the ER. However, subsequent application of bromoxib was still able to mediate a residual release of intracellular Ca²⁺ (Fig. 4F; right graph), indicating that the Ca²⁺ mobilized by bromoxib originates from the ER and additionally from the mitochondria.

As severe mitochondrial damage and oxidative stress are often interdependent, we investigated the effect of bromoxib on cellular reactive oxygen species (ROS) generation. However, measurements with the fluorescent dye H₂-DCF-DA revealed no increase in ROS production (Suppl. Figure 1 A,B). Accordingly, treatment with the antioxidant N-acetylcysteine (NAC) did not affect bromoxib-induced toxicity (Suppl. Figure 1 C).

Bromoxib inhibits glycolysis and mitochondrial metabolism

Given that one of the primary functions of mitochondria is the supply of cellular ATP, we examined the extent to which bromoxib affects cellular ATP levels. To distinguish the effects on glycolysis from OXPHOS, cells were provided with either glucose or galactose as the sole sugar source within the medium. The glycolytic consumption of galactose instead of glucose does not result in a net ATP gain, thereby forcing the cell to rely exclusively on OXPHOS for ATP production. This renders the cell particularly susceptible to inhibitors of the ETC [28]. As shown in Fig. 5A, bromoxib induced a marked reduction in ATP levels under both glucose and galactose conditions, which was distinct from all other ETC inhibitors tested. To investigate whether bromoxib could directly inhibit one of the five electron ETC complexes, we analyzed the activity of each complex in purified mitochondria following bromoxib treatment. Intriguingly, bromoxib exhibited marked inhibitory activity and selectively targeted ETC complexes II, III, and the ATP-synthase (complex V) (Fig. 5B).

Open in a separate window

Fig. 5

Bromoxib specifically targets mitochondrial metabolism, leading to the inhibition of both OXPHOS and glycolysis. (A) The impact of bromoxib (10 µM) and several known mitotoxins on the ATP levels in Ramos cells. The cells were treated with the specified agents in a complete growth medium containing either glucose or galactose as the sole available sugar. The following complex-specific inhibitors of the ETC were used: complex I: 10 µM rotenone; complex II: 10 µM thenoyltrifluoroacetone (TTFA); complex III: 10 µM antimycin A; complex IV: 1 mM sodium azide (NaN₃); complex V: 10 µM oligomycin; mitochondrial uncoupler and protonophore: 1 µM CCCP and 10 µM bromoxib as well as DMSO (0.1% v/v) as vehicle control. ATP-levels were measured using the luminescence-based mitochondrial ToxGloTM assay (Promega). The depicted values were normalized to cells treated with DMSO in glucose containing growth medium (set to 100%). Error bars=mean±SD of three independent experiments performed in triplicates; p-values were calculated by two-way ANOVA with the Holm-Sidak post-test; (**** = p≤0.0001). (B) Bromoxib inhibits ETC complexes II, III, and ATP-synthase (complex V). The activities of the individual complexes of the respiratory chain were measured after treatment with bromoxib (10 µM) or the respective complex inhibitors (complex I: 10 µM rotenone; complex II: 10 mM thenoyltrifluoroacetone (TTFA) or 10 µM 3-bromopyruvate (3-BP); complex III: 10 µM antimycin A; complex IV: 1 mM potassium cyanide (KCN); complex V: 10 µM oligomycin) for 15 min using the MitoCheck^® kit (Cayman Chemical; utilizing mitochondria isolated from bovine heart). Depicted activities were normalized to cells treated with DMSO (0.1% v/v). Statistical analysis: Unpaired t test; two-tailed (**** = p≤0.0001). (C) The effect of bromoxib on the mitochondrial oxygen consumption rate (OCR; pmol O₂/min) was determined in a Seahorse XFe96 Extracellular Flux Analyzer with the Mito Stress Test Kit. HeLa cells (15000 cells per well) were subjected to acute injection with bromoxib (10 µM) or DMSO (0.1% v/v) as a solvent control, and the oxygen consumption rate was monitored over a 100-min duration. Error bars=mean±SD of three independent experiments. The following complex-specific inhibitors of the ETC were used: complex I: 0.5 µM rotenone; complex III: 0.5 µM antimycin A; complex V: 1 µM oligomycin; mitochondrial uncoupler and protonophore: 1 µM FCCP. (D) Various mitochondrial respiration parameters obtained from the Mito Stress Test were compared in HeLa cells treated with bromoxib (10 µM) and DMSO (0.1% v/v). Error bars=mean±SD of three independent experiments. (E) Effect of different bromoxib concentrations on OCR. For the Mito Stress Test HeLa cells (15000 cells per well) were acutely treated with various concentrations of bromoxib (ranging from 0.5 µM to 10 µM) or DMSO (0.1% v/v) as the solvent control. The OCR) was then monitored over a period of 90 min. The following complex-specific inhibitors of the ETC were used: complex I: 0.5 µM rotenone; complex III: 0.5 µM antimycin A; complex V: 1 µM oligomycin; mitochondrial uncoupler and protonophore: 1 µM FCCP

To confirm the role of bromoxib as an inhibitor of mitochondrial respiration, HeLa cells were treated with bromoxib and mitochondrial respiration was monitored by oxygen consumption rate (OCR) using the Agilent Seahorse system with the Mito Stress Test protocol. As shown in Fig. 5C, treatment with 10 µM bromoxib immediately caused a transient increase in OCR before mitochondrial respiration dropped to almost zero. Interestingly, subsequent addition of FCCP did not result in an increase in OCR, consistent with the observation that bromoxib impaired ETC complexes II and III. Subsequently, the non-mitochondrial respiration, basal respiration, maximal respiration, spare respiratory capacity (SRC), ATP production and proton leak were calculated from the OCR measurements (Fig. 5D). Thus, bromoxib induced a response and triggered a pronounced reduction of maximal respiration as well as spare respiratory capacity (Fig. 5D). In addition, bromoxib provoked a proton leak and completely abrogated the ATP production (Fig. 5A, D). These findings support the notion that bromoxib, like CCCP, might function as an uncoupler and in addition blocks OXPHOS complexes, thereby impairing mitochondrial morphology and respiratory metabolism.

Bromoxib acts as a mitochondrial uncoupler

In general, a protonophore can be defined as a proton translocator or ionophore that facilitates proton transfer across membranes. Prominent examples of protonophores are compounds like CCCP, 2,4-dinitrophenol, or triphenylphosphonium, which are known for their cytotoxic potential in various cancer cell types, including breast cancer, lung cancer, neuroblastoma, and glioblastoma cells [29–31]. These agents function by disrupting the normal proton gradient across the inner mitochondrial membrane, thereby uncoupling OXPHOS which results in increased oxygen consumption rates, yet diminished ATP production.

Since bromoxib displayed several features of the uncoupler CCCP, we compared bromoxib with other protonophores concerning OCR, cytotoxicity, caspase-activation and apoptosis-induction. Using the Seahorse Mito Stress Test we analyzed the concentration-dependent mitochondrial uncoupling of bromoxib. For the protonophore FCCP it has been shown that it induces a concentration-dependent metabolic demand, leading to rapid oxidation of substrates such as sugars, lipids and amino acids [32]. In the case of bromoxib, we observed a striking similarity when comparing the concentration-dependent, plateau-like elevation of OCR in bromoxib-treated cells with the corresponding increase induced by FCCP-treatment, suggesting a similar cellular response. Thus, bromoxib appears to uncouple mitochondrial respiration in a similar way to FCCP by disrupting the proton gradient, leading to a rapid increase in OCR even at low concentrations (Fig. 5E). The observed subsequent reduction in OCR with a kinetic delay (which cannot be induced even by the addition of FCCP) is fully consistent with the additional inhibition of complexes II, III, and V after prolonged incubation with brmoxib (Fig. 5C, E). Likewise, bromoxib displayed a similar potential as the protonophore CCCP in terms of cytotoxicity, caspase-activation, and apoptosis rate (Fig. 6A, B and C). Remarkably, both bromoxib and CCCP displayed similar cytotoxic IC₅₀ values in Ramos cells (bromoxib: 4.98 µM; CCCP: 4.96 µM; Fig. 6A). In addition, bromoxib and CCCP initiated the activation of caspases with similar kinetics (within 8 h) and to a similar extent in terms of caspase-activation and apoptosis-induction (Fig. 6B, C). In summary, we demonstrated that bromoxib functions as a mitochondrial uncoupler and induces apoptosis similar to the protonophore CCCP in Ramos cells.

Open in a separate window

Fig. 6

Bromoxib and CCCP display similar cytotoxic features. (A) To assess cytotoxicity in Ramos cells, bromoxib, and CCCP were administered and cell viability was determined after 24 h using the AlamarBlue^® viability assay. Shown in each graph is the mean±SD of one representative experiment performed in triplicates. Respective IC₅₀ values are shown. (B) Ramos cells were subjected to various treatments, including DMSO (0.1% v/v) as a solvent control, bromoxib (10 µM), CCCP (10 µM), or STS (2.5 µM), for a maximum duration of 8 h. To assess caspase 3 activation, DEVDase activity was measured using a spectrofluorometer. The fluorescence increase was recorded for each indicated timepoint over a 2-hour period, and the slope of the linear range of fluorescence increase was used as a parameter to quantify DEVDase activity. The DMSO control values were set to 100 and the normalized relative fold induction, shown as % of control was calculated. Shown in each graph is the mean only of one representative experiment performed in duplicates. (C) Ramos cells were exposed to escalating concentrations of bromoxib or CCCP, while DMSO (0.1% v/v) served as solvent control and STS (2.5 µM) as positive control. After 24 h of treatment, apoptosis was assessed by measuring apoptotic nuclei via flow cytometry. Error bars=mean±SD of three independent experiments performed in triplicates. Statistical analysis: 1-way ANOVA with Dunnett’s multiple comparisons test; (**** = p≤0.0001)

In thermal proteome profiling (TPP), bromoxib caused the stabilization of tubulin and proteins involved in fatty acid oxidation (FAO)

In order to identify proteins targeted by bromoxib, we conducted an unbiased proteomic analysis employing thermal proteome profiling (TPP). This method detects drug-protein interactions in living cells by identifying proteins whose thermal stabilities are altered, directly or indirectly, upon drug binding. Through mass spectrometry, the thermal profiles of these affected proteins can be determined, thereby enabling the successful identification of respective drug targets [33, 34]. Using the TPP-approach, 31 stabilized proteins could be identified in Ramos cells which were stabilized upon treatment with 40 µM bromoxib for 30 min (Fig. 7A and Suppl. Table 1). These proteins could be further subdivided into two main hit-clusters: (i) the COPI-mediated transport system, which is composed of tubulin isoforms and (ii) a protein-cluster involved in fatty acid metabolic processes (Fig. 7A, B and Suppl. Table 1).

Open in a separate window

Fig. 7

Thermal proteome profiling (TPP) revealed bromoxib-mediated stabilization of multiple proteins. Mass spectrometry based TPP was used to analyze the thermal stabilization and destabilization of proteins upon treatment with bromoxib. For this, Ramos cells were treated with 40 µM bromoxib or a diluent control (DMSO, 0.4% v/v) for a duration of 30 min. (A) Volcano-like plot of the statistical significance (expressed as the negative decadic logarithm of the adjusted NPARC p-value, -lg(adj. p, NPARC)) versus the bromoxib mediated melting point shift (difference of the means of the melting points with and without bromoxib treatment, ΔT_m) for each protein. Stabilized proteins (cutoffs: -lg(adj. p, NPARC)>6 and ΔT_m > 0.5°C) are labeled and color coded as in panel B. (B) A functional protein association network (based on a STRING database enrichment analysis, https://string-db.org, v11.5) of the selected 31 stabilized proteins from panel A. Proteins related to ‘fatty acid metabolic process’ (GO:0006631) are shown in red and to ‘COPI-mediated transport’ (‘COPI-mediated anterograde transport’: HSA-6807878 and ‘COPI-dependent Golgi-to-ER retrograde traffic’: HSA-6811434) in blue. The colors of the links indicate the type of interactions (light blue: known interactions from curated databases, purple: known experimentally determined interactions, green: predicted interactions by gene neighborhood, red: predicted interactions by gene fusions, predicted interactions by gene co-occurrence, yellow: text mining, black: co-expression, light purple: protein homology). (C) The five proteins belonging to the ‘fatty acid metabolism’ with their melting curves: (enoyl-CoA hydratase 1 (ECH1), acyl-CoA dehydrogenase very long chain (ACADVL), acyl-CoA synthetase long chain family member 4 (ACSL4), hydroxyacyl-CoA dehydrogenase trifunctional multienzyme complex subunit β (HADHB) and subunit α (HADHA)). Upon bromoxib treatment (40 µM bromoxib for 30 min), a thermal stabilization of these proteins was identified. (D) Exemplary immunoblots (CETSA; see (E) for quantitative analysis) against proteins of the ‘fatty acid metabolism’ (ECH1, ACADVL, ACSL4, and HADHB) employing one of the three independent biological replicates used for the TPP. Since HADHB and HADHA belong to the same protein complex whose subunits have different functions, only HADHB is shown. Upon 40 mM bromoxib treatment for 30 min, thermal stabilization of the proteins can be seen in the immunoblots. (E) Quantitative analysis of immunoblots (CETSA; see (D) for representative examples). The non-denatured fractions of the proteins ECH1, ACADVL, ACSL4, and HADHB from three independent biological replicates (n=3) were quantified, the normalized signal intensity is shown for DMSO in red and bromoxib in black. Error bars=mean±SD of three independent biological experiments

Tubulin molecules are composed of heterodimers, consisting of α- and β-chains, which arrange themselves in a head-to-tail manner to form protofilaments running along the length of the microtubule structure. Following bromoxib treatment, two α-tubulins, namely TUBA1B and TUBA1C, in addition to the two β-tubulins TUBB and TUBB4B, exhibited enhanced stability in the TPP-approach. Therefore, we investigated whether the cytotoxic mechanism of bromoxib might be due to a disruption of tubulin polymerization. However, using a tubulin polymerization assay we detected no significant influence on the polymerization rate of tubulin and no disturbance of cell cycle progression as would be expected for an antimitotic toxin (Suppl. Figure 2). Consequently, we excluded a disturbed tubulin polymerization as the major mechanism for bromoxib-mediated cytotoxicity.

Next, we focused on the protein cluster associated with fatty acid and lipid metabolism, including ECH1, ACADVL, ACSL4, and HADHA/B (Fig. 7B). Given that bromoxib predominantly targets mitochondria, it was expected that proteins localized within mitochondria would emerge as prominent candidates in the TPP results. As shown in Fig. 7C, these proteins were stabilized in TPP after only 30 min of exposure to bromoxib. To verify this finding, the bromoxib mediated stabilization of the candidate proteins (ECH1, ACADVL, ACSL4, and HADHB) was analyzed by cellular thermal shift assays (CETSA) [35] by subjecting the same temperature-treated samples analyzed by quantitative MS-based TPP to quantitative immunoblotting. CETSA confirmed that bromoxib treatment resulted in thermal stabilization of these proteins (Fig. 7D, E).

Reagents

Staurosporine (#9300) was obtained from LC Laboratories (Woburn, MA, USA), N-(2-quinolyl)-valyl-aspartyl-(2,6-difluorophenoxy) methyl ketone (QVD) (#S7311) from Selleckchem (Houston, TX, USA) and etoposide (#1043−100) from BioVision. SUPERKILLER TRAIL^® Protein (soluble) was obtained from human Enzo Life Sciences (#ALX-201-115-C010). Ionomycin (#10634), thapsigargin (#T9033), cyanide m-chlorophenyl hydrazine (CCCP) (#2759), rotenone (#45656), thenoyltrifluoroacetone (TTFA) (#88300), antimycin A (#8674), sodium azide (NaN₃) (#S2002), paclitaxel (#T7402), nocodazole (#M1404), vinblastine (VBL) (#V1377), and CD95-ligand (CD95L/Apo-1 L/FasL; #SRP3036) were obtained from Sigma (Munich, Germany), oligomycin from Toronto Research Chemicals (Toronto, Canada) (#O532970), potassium cyanide (KCN) was supplied in the MitoCheck^® kit (Cayman Chemicals; Ann Arbor, Michigan, USA), and the ETC complex II-inhibitor 3-bromopyruvate (3-BP) was obtained from Merck-Millipore (Darmstadt, Germany) (#376817). All other substances for which a manufacturer is not explicitly specified were obtained from Carl Roth (Karlsruhe, Germany).

Cell lines and cell culture

Ramos cells (human B cell Burkitt lymphoma) were kindly provided by Michael Engelke (Institute of Cellular and Molecular Immunology, University Hospital Göttingen, Germany). Stable transfectants of Jurkat cells (human T cell acute lymphoblastic leukemia) with Bcl-2 overexpression and corresponding empty vector control cells were kindly provided by Claus Belka [17, 83] (Department of Radiation Oncology, University Hospital, LMU Munich, Germany). Jurkat cells deficient for caspase 8 and the parental cell line A3 [13, 17] were kindly provided by John Blenis (Sandra and Edward Meyer Cancer Center, New York, NY, USA). Jurkat cells deficient for caspase 9 were kindly provided by Klaus Schulze-Osthoff (Interfaculty Institute for Biochemistry, University of Tübingen, Germany) [84] and retrovirally transduced with either empty pMSCVpuro (Clontech, Heidelberg, Germany) or pMSCVpuro containing cDNAs coding for untagged human wild-type caspase 9 as previously described [15, 17]. MCF7 (human breast carcinoma; #ACC-115), RT112 (human urinary bladder carcinoma; #ACC-418), HL60 (human acute myeloid leukemia; #ACC-3), HPBALL (human T cell acute lymphoblastic leukemia; #ACC-483), MOLT4 (human T cell acute lymphoblastic leukemia; #ACC-362), K562 (human chronic myeloid leukemia; #ACC-10) and SUPB15 (human B cell acute lymphoblastic leukemia; #ACC-389) were obtained from DSMZ and provided by Sanil Bhatia (Dept. of Pediatric Oncology, Hematology and Clinical Immunology, University Hospital Düsseldorf, Germany). KOPTK1 (human T cell acute lymphoblastic leukemia; CVCL-4965) was kindly provided by Oskar Haas (Children’s Cancer Research Institute, St. Anna Children’s Hospital, Vienna, Austria). HeLa cells stably expressing mito-DsRed were kindly provided by Aviva M. Tolkovsky (Department of Clinical Neurosciences, University of Cambridge, England, UK) and have been described previously [85]. HeLa cells with transient expression of Smac-mCherry and GFP-Bax were generated by lipofection at 70–80% confluence using Lipofectamine 2000 (Life Technologies, Darmstadt, Germany). Therefore, cells were incubated with 0.15 µL Lipofectamine 2000, 50 ng pcDNA3-Smac(1–60)mCherry (Addgene ID 40880; this plasmid was kindly provided by Stephen Tait (Beatson Institute, University of Glasgow, Scotland, UK) and has been described previously [86]) and 50 ng pGFP-Bax (kindly provided by Nathan R. Brady, Department of Molecular Microbiology and Immunology, Johns Hopkins University, Baltimore, MD, USA) per well on µ-Slide 8 wells (IBIDI, Planegg, Germany, #80826) for 16 h as described in [26]. MEF (mouse embryonic fibroblast) cells deficient for OMA1 and/or YME1L1 as well as the corresponding wild-type cells were generated by Ruchika Anand (Institute of Biochemistry and Molecular Biology I, University of Düsseldorf, Germany) and kindly provided by Thomas Langer (Institute for Genetics, University of Cologne, Germany) [21]. The glioblastoma cell line LN229 (#CRL-2611) was obtained from ATCC. The glioblastoma cell lines LN308 and TP365MG were kindly provided by Guido Reifenberger (Inst. of Neuropathology, University Hospital Düsseldorf, Germany). The glioblastoma cell line SJ-GBM2 was obtained from Children’s Oncology Group (Dept. of Hematology, Oncology and Clinical Immunology, University Hospital Düsseldorf, Germany) and YKG1 from Tebu-Bio (ref JCRB0746). All cells were cultivated at 5% CO₂ at 37 °C and stable humidity. The suspension cell lines (HL60, HPBALL, Jurkat, KOPTK1, K562, MOLT4, Ramos, SUPB15) and MCF7 were maintained in RPMI media with 10% FCS, 100 U/mL penicillin and 100 µg/mL streptomycin. MEF cells, RT112, HeLa, YKG1, TP365MG, LN229 and LN308 were maintained in high-glucose Dulbecco’s Modified Eagle’s medium (DMEM) supplemented with 10% FCS, 100 U/mL penicillin and 100 µg/mL streptomycin. SJ-GBM2 was cultivated in IMEM medium supplemented with 10% FCS, 100 U/mL penicillin and 100 µg/mL streptomycin.

Cytotoxicity measurements

To determine cytotoxicity, the AlamarBlue^® assay was performed as described in [87]. Briefly, suspension cells were seeded at a density of 1×10⁶ cells/mL and adherent cells were seeded at a density of 0.2×10⁶/mL. Suspension cells were immediately treated; adherent cells were treated the next day, with increasing concentration of the compound and after the specified incubation time, resazurin (Sigma, #R7017) was added to a final concentration of 40 µM. After 120 min of incubation, the fluorescence of resorufin (excitation (Ex): 535 nm, emission (Em); 590 nm) was measured with a microplate spectrophotometer (Synergy Mix plate reader). The viability of cells treated with DMSO (0.1% v/v) was set to 100% and the dose-response curves were then fitted with PRISM v7.01 (GraphPad Software, La Jolla, CA, USA).

Fluorometric caspase 3 activity assay

Cells were seeded at a density of 1×10⁶ cells/mL in a 96-well plate, treated with the compound for the depicted time (kinetics 0–8 h) and harvested by centrifugation at 600 g. The supernatant was removed and the cells were lysed with 50 µL of ice-cold lysis buffer (20 mM HEPES, 84 mM KCl, 10 mM MgCl₂, 200 µM EDTA, 200 µM EGTA, 0.5% NP40, 1 µg/mL leupeptin, 1 µg/mL pepstatin, 5 µg/mL aprotinin) on ice for 10 min. The lysates were transferred to a flat-bottom microplate and mixed with 150 µL ice-cold reaction buffer (50 mM HEPES, 100 mM NaCl), 10% sucrose, 0.1% CHAPS, 2 mM CaCl₂, 13.35 mM DTT) containing 70 µM of the profluorescent caspase substrate Ac-DEVD-AMC (Biomol GmbH, Hamburg, Germany, #ABD-13420). The kinetics of AMC release were monitored by measuring AMC fluorescence intensity (Ex: 360 nm; Em: 450 nm) at 37 °C every 2 min over a total of 120 min, using a Synergy Mix microplate reader. The slope of the linear range of the fluorescence curves (Δrfu/min) was considered as corresponding to caspase 3 activity.

FACS-based analysis of apoptotic cell death and cell cycle

Fragmented DNA leaking from apoptotic nuclei and the cell cycle was measured by the method of Nicoletti et al. [14]. Briefly, nuclei were prepared by lysing cells in a hypotonic lysis buffer (1% sodium citrate, 0.1% Triton X-100, 50 µg/mL propidium iodide) and subsequently analyzed by flowcytometry. For analysis of hypodiploid apoptotic nuclei (which are located to the left of the 2 N-peak), nuclei were measured in logarithmic mode. For cell cycle analysis, nuclei were measured in linear mode to distinguish G₁, S, G₂ phases and hypodiploid nuclei. All flowcytometry analyses were performed on an LSR-Fortessa™ (Becton, Dickinson, Heidelberg, Germany). Data analysis was performed using FlowJo_V10 (BD Biosciences). For statistical analysis of hypodiploid nuclei, 1-way ANOVA with Dunnett’s multiple comparison’s test was performed comparing treatments to DMSO (0.1% v/v). **** p<0.0001.

Immunoblotting

Cells were seeded at a density of 1×10⁶ cells/mL, treated as specified and harvested by centrifugation (3000 rpm, 5 min) followed by freezing in liquid nitrogen. The cell pellets were thawed on ice followed by three quick-freeze-cycles in liquid nitrogen. Subsequently, the pellets were lysed with lysis buffer (20 mM Tris-HCl, 150 mM NaCl, 1% v/v Triton X-100, 0.5 mM EDTA, 10 mM NaF, 2.5 mM Na₄P₂O₇, 0.5% sodium deoxycholate and protease inhibitors (Sigma, #P2714)) on ice for 30 min, accompanied by vortexing. The lysates were purified from cell debris by centrifugation (13300 rpm, 15 min). Protein concentration was determined with Bradford assay. Samples were diluted in lysis buffer and sample buffer, followed by SDS-PAGE and Western blotting according to standard workflows. Finally, target-specific antibodies (anti-PARP 1:2000 (Enzo, #BML-SA250); anti-GAPDH 1:5000 (Abcam, #ab8245), anti-α-tubulin 1:2000 (TUBA4A; Sigma, #T5168), anti-OPA1 1:1000 [88], anti-OXPHOS WB antibody cocktail 1:500 (Abcam, #ab110411, consisting of anti-NDUFB8 (NADH ubiquinone oxidoreductase subunit B8), anti-SDHB (succinate dehydrogenase subunit B), anti-UQCRC2 (ubiquinol-cytochrome c reductase core protein 2), anti-cox-2 (cytochrome c oxidase-2) and anti-ATP5A (ATP-synthase F1α)) and anti-vinculin 1:2000 (Sigma, #V9131); anti-caspase 8 (Cell Signaling Technology, #9746), anti-caspase 9 [15], anti-HADHA 1:1000 (Santa Cruz, # sc-374497), anti-HADHB 1:1000 (Santa Cruz, # sc-271495), anti-ACSL4 1:1000 (Santa Cruz, #sc-365230), anti-ECH11:1000 (Santa Cruz, #sc-515270) and anti-ACADVL 1:1000 (Santa Cruz, #sc-376239) were used for immunoblotting. Fluorescence-coupled secondary antibodies (LI-COR Biosciences) were used for the detection of target proteins on PVDF membrane using LI-COR Odyssey^® imaging system. Validation of bromoxib mediated thermal stabilization of ECH1, ACADVL, ACSL4, and HADHB by quantitative immunoblotting (CETSA) [35] was essentially performed as described [89]. A detailed description is provided in the Supplemental Information.

Densiometric analyses

Immunoblots were quantified using the LI-COR software Image Studio Lite Ver 5.2. The densiometric signals from the protein bands of interest were determined. The shown values represent the calculated ratio of the two long forms of OPA1 (L1 plus L2) to total OPA1 (L1 plus L2 plus S3-5).

Microscopy and quantification of mitochondrial fragmentation (fission)

Imaging of HeLa cells stably expressing mito-DsRed was performed using a Spinning Disc Confocal microscope (Perkin-Elmer, Waltham, MA, USA) equipped with a 60x/1.4 NA oil-immersion objective using a 561 nm laser line for excitation. Approximately 50 planes of 0.3 μm distance were recorded. The cells were seeded on glass bottom 3 cm dishes (MatTek, Ashland, MA, USA) and maintained in full growth medium supplemented with 10 mM HEPES during imaging. Images were taken for a duration of 30 min and 120 min in a chamber heated to 37 °C. Image analysis was performed using Fiji (V 1.53f51). 3D stacks of 50 planes in thickness were compressed with maximum intensity projection and brightness was adjusted for optimal visibility of mitochondria. Imaging files were blinded by renaming files with random numbers and all imaged cells were categorized into three categories (tubular, intermediate, and fragmented). Statistical significance was analyzed by 2-way ANOVA followed by Dunnett’s multiple comparison test. **** p≤0.0001.

Confocal microscopy – live imaging of Smac and Bax intracellular localization

Confocal microscopy live imaging of HeLa wild type cells transiently expressing GFP-Bax and Smac-mCherry was performed using an SP8 Leica microscope (Inverse DMI 6000 AFC, Leica) with a C-Apochromat × 63 N.A. 1.2 water DIC objective. Cells were seeded onto µ-Slide 8 wells (IBIDI, 80826) and transfected with GFP-Bax and Smac-mCherry as described in [26]. Cells were maintained in full growth medium at 37 °C and 5% CO₂ during imaging and recorded every 5–15 min during 120 min and processed with Fiji as described in [17]. For each individual time frame, we performed a maximum intensity projection, smoothing of the images and measured the standard deviation (SD) of the fluorescence intensity of GFP-Bax and Smac-mCherry channels, to analyze their subcellular distribution upon DMSO (10 µM) or bromoxib treatment (10 µM). To avoid downstream caspase-mediated effects on cell morphology (such as membrane blebbing), 10 µM of the caspase-inhibitor QVD was added prior to incubation with DMSO or bromoxib.

Live measurement of mitochondrial membrane potential

To evaluate changes in mitochondrial membrane potential (ΔΨm), Ramos cells (1×10⁶ c/mL) underwent treatment with the cell-permeable, positively charged dye tetramethylrhodamine ethyl ester (TMRE). This dye tends to accumulate in active mitochondria characterized by a negative net charge, while depolarized mitochondria fail to retain the dye. To perform this, cells were suspended in fresh medium with 10 mM HEPES and supplemented with 100 nM TMRE (AAT Bioquest, Sunnyvale, CA, USA; #22220). After a 15-min incubation at 37 °C, the cells were washed twice with RPMI (supplemented with HEPES) and then given an additional 15 min for recovery to ensure dye accumulation in active mitochondria. Subsequently, TMRE fluorescence was measured via flow cytometry (Ex: 488, Em: 575) using LSR-FortessaTM (Becton, Dickinson, Heidelberg, Germany). The initial fluorescence level was recorded for each sample over 2 min before introducing the test substance. The fluorescence level prior to substance introduction was set to 100%. As a positive control for inducing the complete breakdown of ΔΨm, treatment with the protonophore CCCP was employed.

Live measurement of total and intracellular calcium mobilization by Fluo-4-AM

For the detection of changes in Ca²⁺ homeostasis, Ramos cells (1×10⁶ c/mL) were stained by incubation in a growth medium containing 1 µM Fluo-4-AM (Life Technologies; #F14201), 0.005% w/v Pluronic F-127 (Sigma, #540025), 10 mM HEPES and 5% v/v FCS at 30 °C for 25 min. An equal amount of full-growth medium was added and the samples were incubated at 37 °C for 10 min. The cells were then pelleted at 3000 rpm for 3 min and resuspended in Krebs-Ringer buffer (10 mM HEPES pH 7.0, 140 mM NaCl, 4 mM KCl, 1 mM MgCl₂, 10 mM glucose) supplemented with 1 mM CaCl₂. The cells were kept in the dark until measurement at RT. Immediately prior to measurement, cells were washed and resuspended in Krebs-Ringer buffer alone for total Ca²⁺ mobilization. For the measurement of intracellular Ca²⁺ mobilization, cells were washed with Krebs-Ringer buffer supplemented with 0.5 mM EGTA. Fluo-4-AM fluorescence was measured live using flow-cytometer LSR-Fortessa™ (Becton, Dickinson, Heidelberg, Germany) recording fluorescence in the FITC channel (Ex: 488, Em: 530±30 nm). For each sample, after at least 30 s of baseline measurement, the stimulus was added and measurement was continued for the indicated time.

Measurement of cellular ROS level

To measure alterations in cellular reactive oxygen species (ROS) levels, Ramos cells were cultured in phenol red-free RPMI medium throughout the assay. The cells were loaded with 20 µM 2’,7’-dichlorodihydrofluorescein diacetate (H₂DCF-DA, Sigma, #D6883) and incubated for 1 h in the dark at 37 °C. Subsequently, the cells were centrifuged (600 g x 5 min) and washed with media, followed by the treatment of cells with desired compounds for 30 min at 37 °C. The cells were centrifuged (600 g; 5 min), washed with PBS, and transferred to FACS tubes. The fluorescence of DCF, which correlates with the cellular ROS level, was measured by flowcytometry at Ex: 488 nm, Em: 530 nm (FITC channel) at LSR-Fortessa™ (Becton, Dickinson, Heidelberg, Germany).

Measurement of cellular ATP-levels

For the detection of changes in cellular ATP-levels, the mitochondrial ToxGlo™ assay (Promega, Mannheim, Germany; #G8000) was performed according to the manufacturer’s instructions. To enhance the detection sensitivity of mitotoxins, cells were cultured in a medium containing either glucose or galactose as the sole available sugar source. The utilization of galactose through glycolysis does not result in a net gain of ATP, thereby forcing the cells to rely entirely on OXPHOS for ATP production. The statistical analysis was performed using 2-way ANOVA and Sidak’s multiple comparisons test. **** p≤0.0001. The ATP-assay was measured with a microplate spectrophotometer (Synergy Mix plate reader).

Seahorse analysis

The Seahorse XFe96 Extracellular Flux Analyzer (Agilent Technologies, Santa Clara, CA, USA) was employed alongside the Mito Stress Test Kit (Agilent Technologies, Santa Clara, CA, USA) following the manufacturer’s instructions. To induce inhibition of ATP-synthase (complex V) and restrict the electron flow along the ETC, oligomycin was applied, thereby leading to a decline in mitochondrial respiration, as indicated by the OCR. This oligomycin induced reduction in OCR is a surrogate marker for ATP production. Subsequently, a decoupling agent, FCCP (carbonyl cyanide-4 (trifluoromethoxy) phenylhydrazone; a CCCP analog), was administered. This agent induces the collapse of the proton gradient and disrupts the ΔΨm. As a result, electron flow through the ETC becomes unrestricted and oxygen consumption by complex IV reaches its maximum capacity, resulting in a plateau effect with DMSO between 60 and 80 min. The FCCP-augmented OCR was then used to calculate the spare respiratory capacity (SRC), defined as the difference between maximal and basal respiration. The SRC provides a measure of the cell’s ability to respond to escalated energy demands or stress conditions. The subsequent application of a combination of rotenone (inhibitor of complex I) and antimycin A (inhibitor of complex III) effectively stops mitochondrial respiration and facilitates the calculation of non-mitochondrial respiration driven by mechanisms outside the mitochondria. For each well, injection compounds were utilized at concentrations of 1 µM oligomycin, 1 µM FCCP, and 0.5 µM rotenone/antimycin A. 10,000 HeLa cells were seeded 24 h before Seahorse analysis. Acute injection of bromoxib (10 µM) or DMSO (0.1% v/v) vehicle preceded OCR measurement for three timepoints every 6 min, before progressing with oligomycin injection. In the assay involving various bromoxib concentrations (0.5–10 µM), four timepoints were assessed, each after 3 min. Non-mitochondrial respiration was designated as the lowest recorded OCR after interrupting respiration with rotenone/antimycin A. Mitochondrial respiration parameters were calculated according to Agilent Report generator user guide.

Lipidomics and measurement of fatty acid oxidation

Tubulin polymerization assay

For the detection of changes in tubulin polymerization, the OD-based tubulin polymerization assay kit (Cytoskeleton, Inc., Denver, CO, USA; #BK006P) was performed according to the manufacturer’s protocol. Based on the observation, that monomeric tubulin polymerization is accompanied by an increase in absorption (340 nm), this underlying reaction can be measured. The velocity of polymerization has been calculated using the formulas provided by the manufacturer. The respective controls for tubulin stabilization (paclitaxel; which increases the polymerization rate) and for tubulin destabilization (nocodazole; which reduces the polymerization rate) were used.

Thermal proteome profiling (TPP)

Thermal proteome profiling with temperature range (TPP-TR) was essentially performed as described [18, 33, 89, 94] employing 40 µM as final bromoxib concentration and 30 min treatment time on live Ramos cells in three biological replicates (n=3). In brief, aliquots of compound treated and control cells were shortly incubated side-by-side at ten discrete temperatures, equally distributed in a range between 37 °C and 68.5 °C, lysed, and the cell extracts containing the fraction of soluble, non-denatured proteins were obtained by centrifugation. Peptides were obtained by purification and tryptic digestion of the non-denatured proteins using a slightly modified SP3 protocol [95], labeled using TMT 10plex to code for the different temperatures, combined, offline high pH fractionated, and analyzed by LC-MS. In total, eight (high pH fractions per TMT set) times two (TMT sets per replicate) times three (replicates)=48 samples were analyzed using an Ultimate 3000 Rapid Separation Liquid Chromatography System and a nano-source ESI interface equipped Orbitrap Fusion Lumos Tribrid mass spectrometer operated in synchronous precursor selection (SPS) mode. MS data analysis was performed using MaxQuant version 2.0.3.0, and TPP-TR data were prepared from the TMT reporter ion intensities by a three-step normalization procedure essentially as described [89] using R (R version 4.1.2 on a x86_64-w64-mingw32/x64 (64-bit) platform). Statistical analysis of melting curves was performed essentially as described [18] using the TPP R package version 3.30.0 in R version 4.3.2. (https://bioconductor.org/packages/TPP). Functional protein association network analysis was performed using the STRING version 11.5 online tool (https://string-db.org [96]). A detailed description is provided in the Supplemental Information.

The authors thank Ruchika Anand (Inst. of Biochemistry and Molecular Biology I, University of Düsseldorf, Germany), Claus Belka (Ludwig-Maximilians University, Munich, Germany), John Blenis (Sandra and Edward Meyer Cancer Center, New York, NY, USA), Nathan R. Brady (Department of Molecular Microbiology and Immunology, Johns Hopkins University, Baltimore, MD, USA), Michael Engelke (Institute of Cellular and Molecular Immunology, University Hospital Göttingen, Göttingen, Germany), Oskar Haas (Children’s Cancer Research Institute, St. Anna Children’s Hospital, Vienna, Austria), Thomas Langer (Institute for Genetics, University of Cologne, Germany), Guido Reifenberger, (Inst. of Neuropathology, University Hospital Düsseldorf, Germany), Klaus Schulze-Osthoff (Interfaculty Institute for Biochemistry, University of Tübingen, Germany), Stephen Tait (Beatson Institute, University of Glasgow, Scotland, UK), and Aviva M. Tolkovsky (Department of Clinical Neurosciences, University of Cambridge, England, UK), for providing valuable cell lines and reagents. This work was supported by grants from the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation; 270650915/GRK 2158 (to B.S., P.P., S.B. and S.W.), 417677437/GRK 2578 (to A.S.R., B.S. and S.W.), and STO 864/4-3, STO 864/5-1, STO 864/6-1 (to B.S.)) and the Düsseldorf School of Oncology (funded by the Comprehensive Cancer Center Düsseldorf/Deutsche Krebshilfe and the Medical Faculty of the Heinrich Heine University Düsseldorf; to B.S. and S.W.). MCIU/AEI/FEDER, UE (PID2021-124425OB-I00) and Basque Government, Department of Education (IT1476-2) to P.A. H.F.-R. acknowledges support by the Ikerbasque Basque Foundation for Science, COFUND, H2020-MSCA-COFUND-2020-101034228-WOLFRAM2 and Spanish research ministry, Grant PID2022-141616OA-I00 funded by MICIU/AEI/10.13039/501100011033 and by ERDF a way of making Europe (ERDF/EU).