Integrated proteomics and metabolomics reveal phytosesquiterpene lactones inhibit TNBC cell activity by depleting ATP synthesis and reprogramming primary metabolism.

Introduction
Triple-negative breast cancer (TNBC) with a lack of hormone (estrogen/progesterone) receptors and human epidermal growth factor receptor 2 (HER2) overexpression accounts for around 15–20% of BC patients. TNBC is regarded as an aggressive BC subgroup due to several characteristics such as molecular heterogeneity, high metastasis rate, poor prognosis, high relapse rate, and shorter disease-free survival time after treatment compared to other BC subgroups1. In recent clinical trials for patients with TNBC, combinatorial regimens of chemotherapeutic agents plus targeted molecular inhibitors, such as PARP1, VEGF/VEGFR, PI3K/AKT/mTOR, and immune checkpoint, have been implemented clinically to improve treatment outcomes compared to monotherapy2. However, the chemotherapeutic regimen remains a mainstay therapeutic option for TNBC patients due to the relatively better sensitivity of TNBC to chemotherapy. Nevertheless, the consequences of drug resistance and side effects using chemotherapy are still thorny issues in clinical application3. Therefore, finding novel agents that have superior anti-TNBC activity is a priority.
The mitochondrion is a vital organelle in cells that is recognized as the “cell powerhouse” producing ATP by oxidative phosphorylation (OXPHOS)4. Mitochondria-related reactive oxygen species (mtROS), such as the superoxide anion (O2−•) and H2O2, are produced as byproducts during the process of OXPHOS5. Damaging OXPHOS components and/or ROS scavenging systems, such as enzymatic antioxidants and thiol systems in the mitochondria could result in ROS imbalance and further induce oxidative stress that may disrupt cell function and/or drive cell death6,7. Excess mtROS resulting from the inactivation or dysregulation of antioxidative enzymes may, in turn, affect mitochondrial structure and functions, leading to disordered mitochondrial metabolic function and/or metabolic reprogramming8,9. Disruption of mitochondrial metabolism homeostasis in cancer cells is a potential therapeutic strategy against cancers10,11. Therefore, the induction of imbalances in mtROS production, mitochondrial dysfunction, and altering mitochondrial metabolism are promising topics of study for anti-cancer therapies.
The application of natural products against various cancers has been extensively investigated and many compounds have been developed into anti-cancer lead compounds or drugs based on their multiple or polypharmacological mechanisms12–14. Adjuvant therapy using natural products also showed the potential to improve therapeutic effectiveness or reduce the side effects of anti-cancer drugs15,16. A sesquiterpene lactone deoxyelephantopin (DET) isolated from the medicinal plant genus Elephantopus has been demonstrated to inhibit different types of cancer cell lines, such as A549 lung cancer, HCT116 colorectal cancer, HepG2 hepatocellular carcinoma, and MCF-7 breast cancer cell lines in which inducing apoptosis via both extrinsic and intrinsic pathways, promoting DNA fragmentation, cell cycle arrest in the G2/M or S phases, and/or inhibiting NF-κB signaling were the common modes of action17–20. In our previous studies, DET showed superior activities to the chemotherapeutic drug paclitaxel against metastasis of TS/A murine mammary adenocarcinoma in vitro and in vivo through G2/M phase arrest, ubiquitin proteasome activity disruption, and cell motility inhibition21,22. Furthermore, DET and DETD-35, a DET derivative obtained using a semi-synthesis approach, suppressed tumor growth and the cellular activity of human A375 BRAFV600E mutant melanoma, vemurafenib-resistant A375-R melanoma, and a lung-seeking A375LM5IF4g/Luc BRAFV600E melanoma cell and xenograft mouse models in vitro and in vivo23,24. The underlying molecular mechanisms were through downregulation of signaling molecules (MEK-ERK, Akt, and STAT3) and metastasis-associated markers (N-cadherin, MMP2, vimentin, and integrin α4), as well as the induction of apoptosis and glutathione depletion. Both compounds were identified as novel glutathione peroxidase 4 (GPX4) inhibitors and ferroptosis inducers by which vemurafenib-resistant melanoma was effectively suppressed25. Moreover, DETD-35 exhibited better inhibitory effects than parental DET on MDA-MB-231 TNBC cell migration, invasion, and motility in vitro as well as tumor growth and lung metastasis in NOD/SCID mice in vivo26. DETD-35 and DET all induced oxidative stress-mediated paraptosis-like cell death and mitochondrial DNA damage in MDA-MB-231 cells27–29; however, how these sesquiterpene lactone derivatives affect mitochondrial proteome and OXPHOS metabolism in TNBC cells remains unclear.
In this study, LC/MS-based comparative mitochondrial proteomics and primary metabolomics were utilized to decipher the underlying mechanisms through which DET and DETD-35 inhibit TNBC cell activities. Several compound-regulated proteins and associated biological pathways/signaling networking were analyzed. The role of two proteins/genes, i.e., PRKCA and MCL1, up-regulated by either compound treatment in TNBC cells was validated using a gene knockdown approach. The molecular mechanisms through which DET and DETD-35 deregulate proteins engaged in mitochondrial function and disrupt metabolic homeostasis to achieve anti-TNBC cell activities were dissected.

Materials and methods

Chemicals and antibodies
The 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), dimethyl sulfoxide (DMSO), and N-acetylcysteine (NAC) were purchased from Sigma-Aldrich (USA). Dulbecco’s modified essential medium (DMEM) and fetal bovine serum (FBS) were from Invitrogen (USA). Primary antibodies against β-actin (Millipore, USA), PKC-α (Proteintech, USA), glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and calnexin (Cell Signaling, USA), voltage-dependent anion-selective channel 1 (VDAC1), α-tubulin, lamin A/C, Mcl-1, ATP5A1 and ATP synthase gamma (ATP5C1) (GeneTex, USA), and cytochrome c (Santa Cruz, USA) were used.

Preparation of DET and its derivative DETD-35
DET is a natural sesquiterpene lactone originally isolated from Elephantopus scaber L.. E. scaber L. plants used in this study were collected from Yang-Ming Mountain area, Taipei, Taiwan. The authenticity of the plant material was validated by Dr. Yen-Hsueh Tseng, Department of Forestry, National Chung Hsing University, Taiwan. A voucher specimen ES001 was deposited in the Agricultural Biotechnology Research Center, Academia Sinica, Taiwan. The isolation of DET and semi-synthesis of DETD-35 from DET were performed following previously reported procedures21,26. One kilogram of dried E. scaber plants could isolate and purify ~ 0.015–0.02% DET in dry weight using reverse phase column chromatography. The yield of DETD-35 derived from the parental compound DET was about 50%26. Generally speaking, there is no concern about the scalability of either of these compounds for future studies or applications. The purities of the DET and DETD-35 compounds were > 99%, as determined by HPLC and NMR spectrometry. Both compounds are relatively stable when stored at − 20 °C, protected from light for several years; there was no observation of degradation during the study course.

Cell line and cell culture
Human TNBC MDA-MB-231 cells and normal human mammary epithelial MCF-10A cells were obtained from American Type Culture Collection (ATCC, USA) and grown in the manufacturers’ suggested medium supplemented with 10% FBS, 100 units/mL penicillin, 100 μg/mL streptomycin, and 1 mM sodium pyruvate (Gibco, USA) in a humidified 5% CO2 incubator at 37 °C.

Measurement of mitochondrial superoxide levels
MDA-MB-231 cells were seeded and incubated overnight to allow cell adhesion. The cells were pretreated with or without 5 mM ROS scavenger NAC for 1 h and then treated with vehicle (0.5% DMSO), DET (11 μM), and DETD-35 (4 μM) containing 1 μM MitoSOX (Thermo Fisher Scientific, USA) in phenol red-free DMEM medium (Gibco, USA) for 1 h at 37 °C. NAC served as a control to reverse the specific compound’s effect on the mitochondrial ROS induction and oxidative stress in TNBC cells. The labeled cells were washed with PBS and then harvested using trypsinization. After washing with PBS, the cells were resuspended in PBS and detected using BD LSR II flow cytometer (BD Bioscience, USA).

Mitochondrial permeability transition pore (mPTP) opening assay
TNBC cells were pretreated with or without 5 mM ROS scavenger NAC for 1 h and then treated with vehicle (0.5% DMSO), DET (11 μM) or DETD-35 (4 μM) for 2 and 4 h. NAC served as a control to reverse the specific compound activity on inducing mPTP opening. The mPTP opening was determined by MitoProbe Transition Pore Assay Kit (Thermo Fisher Scientific, USA) according to the manufacturer’s protocol. Briefly, the cells were first washed with Hanks’ Balanced Salt Solution (HBSS) (Gibco, USA), and 1 μM calcein AM and 2 mM CoCl2 were then used to label the treated cells for 30 min at 37 °C. The labeled cells were washed with HBSS again to remove the excess fluorescence dye, and the fluorescence-labeled cells were imaged using a fluorescence microscope (Zeiss Axiovert 200 M, Germany). The fluorescence intensity of calcein in over 1 × 103cells in each assay sample was quantified using ImageJ software.

Determination of intracellular ATP levels
MDA-MB-231 cells were treated with vehicle (0.5% DMSO), DET (11 μM) and DETD-35 (4 μM) for 2 and 4 h. The intracellular ATP levels were determined by ATPlite Luminescence ATP Detection Assay System (PerkinElmer, USA) according to the manufacturer’s protocol. Briefly, after treatment, the cells were incubated with PBS containing lysis solution. The plate was shaken for 5 min in an orbital shaker (Eppendorf Thermomixer Comfort, Germany) at 600 rpm to lyse the cells and stabilize the ATP. Next, the substrate solution was added and the plate was shaken again for 5 min. The plate was kept in the dark to stabilize luminescence and the intracellular ATP levels were measured by the microplate reader SpectraMax i3x (Molecular Devices, USA).

Cell viability assay
MDA-MB-231 cells (5 × 103 cells/well) were seeded in 96-well culture plates and incubated overnight. The cells were treated with vehicle, DET, and DETD-35 at the indicated concentrations for 24 h, and the viability of compound-treated cells was measured using MTT-based colorimetric assay according to a previously report30.

Cell mito stress test assay
TNBC cells (1.75 × 104 cells/well) were seeded in a testing plate obtained from Seahorse XFp Cell Mito Stress Test Kit (Agilent Technologies, USA). After overnight incubation, the DMEM medium was replaced with Seahorse XF base medium supplemented with 2% FBS, 10 mM glucose, 1 mM sodium pyruvate, and 2 mM glutamine. The cells were treated with vehicle (0.5% DMSO), DET (11 μM), or DETD-35 (4 μM) for 4 and 8 h at 37 °C in a CO2-free incubator. The mitochondrial respiratory activity of the treated cells was measured following the manufacturer’s instructions. The ETC complex inhibitors, including oligomycin (1 μM) and rotenone/antimycin A mixture (0.5 μM each), and the uncoupling agent FCCP (1 μM), were loaded into the plate following the injection order: oligomycin, FCCP, and rotenone/antimycin A. Oligomycin, rotenone, and antimycin A served as specific control inhibitors of Complex V (ATP synthase), Complex I, and Complex III, respectively. The real-time oxygen consumption rate was calculated by Seahorse XFp analyzer (Agilent Technologies, USA). The level of basal/maximal respiration, proton leak, ATP production, and coupling efficiency were estimated according to the OCR measurement value with/without inhibitor addition. The OCR values were normalized to cell number.

Molecular docking analysis
The binding interactions between DET or DETD-35 with ATP synthase were conducted using a protein–ligand docking tool SwissDock. The 3D whole structure of ATP synthase was downloaded from Protein Data Bank (PDB ID: 8KI3). The 3D structure of DET and DETD-35 as ligands was also prepared and converted as a submission form such as simplified molecular input line entry system (SMILES) notation using the Meeko python package. The docking regions of DET or DETD-35 with ATP synthase were selected based on known or predicted binding regions, including subunits α, β, γ, c, and a in ATP synthase. Multiple docking runs were conducted to predict and find possible binding sites with lower binding energy. The best-ranked pose was chosen for further analysis, focusing on important interactions, such as hydrogen bonding, hydrophobic interactions, and coordination with metal ions. The docking result of DET or DETD-35 with ATP synthase was visualized using the molecular graphic tool PyMOL, allowing for the checking of the interactions that contribute to binding affinity.

Isolation of mitochondria from TNBC cells
MDA-MB-231 cells were incubated overnight and treated with vehicle (0.5% DMSO), DET (11 μM) and DETD-35 (4 μM) for 1 and 4 h. Mitochondria from the treated cells were collected using a protocol adapted from Frezza et al.31. Briefly, the treated cells scraped off the dish and their culture media were centrifuged, and the pellet was washed with ice-cold PBS and resuspended in the ice-cold isolation buffer (IBc) containing 10 mM Tris–MOPS, 1 mM EGTA/Tris, 200 mM sucrose, protease inhibitor cocktail (pH 7.4). Suspended cells were homogenized for 40 passages through a syringe, and the homogenate was then centrifuged. The supernatant was collected as the mitochondrial fraction, while the pellet was resuspended in IBc and homogenized again. After centrifugation at 600 × g for 10 min, the twice supernatant was combined and centrifuged again. The supernatant was collected and further centrifuged at 10,000 × g for 20 min. The pellet was resuspended, washed with IBc and subjected to an additional centrifugation at 10,000 × g for 20 min at 4 °C to yield the pellet as crude mitochondria. The pure mitochondria were obtained using sucrose gradient ultracentrifugation. Briefly, the crude mitochondria samples were resuspended in IBc and loaded onto a discontinuous sucrose buffer containing 1.0 M sucrose over 1.5 M sucrose in 10 mM Tris–MOPS and 1 mM EGTA/Tris buffer. After ultracentrifugation at 95,000 × g for 60 min, the layer between 1.0 M and 1.5 M sucrose was collected and diluted with dilution buffer containing 10 mM Tris–MOPS and 1 mM EGTA/Tris. The solution was then centrifuged at 15,000 × g for 15 min to receive the pure mitochondria in the pellet. The purity of mitochondria was assessed by western blotting using the mitochondrial marker proteins (VDAC1 and cytochrome c), cytosolic marker proteins (α-tubulin and GAPDH), nuclear marker protein (lamin A/C), and an endoplasmic reticulum marker protein (calnexin).

Measurement of ATP synthase activity
ATP synthase activity present in the mitochondria of TNBC cells was determined using the ATP synthase enzyme activity assay kit (Abcam, USA) following the manufacturer’s instructions. Briefly, after 8 h of treatment with vehicle, DET, DETD-35, or oligomycin (5 μM) as a positive control, the crude mitochondrial fraction in the treated TNBC cells was isolated following the protocol described in the section “Isolation of mitochondria from TNBC cells” above. The protein concentration of the crude mitochondrial fraction was measured by Pierce 660 nm protein assay (Thermo Fisher Scientific, USA) according to the manufacturer’s protocol. Ten micrograms of crude mitochondrial protein extracts in a working buffer, including the buffer and detergent provided by the ATP synthase enzyme activity assay kit, were used for the enzyme activity assay. First, ATP synthase in protein extracts was immunocaptured in the microplate wells, then the levels of ATP synthase activity were determined using the microplate reader SpectraMax i3x (Molecular Devices, USA). ATP synthase activity was normalized to mitochondrial protein content and reported as ΔOD340/min/μg of protein.

Western blot analysis
The concentration of mitochondrial protein samples was measured by Pierce 660 nm protein assay (Thermo Fisher Scientific, USA) according to the manufacturer’s protocol. Mitochondrial protein samples were separated by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and then subjected to transblotting. Briefly, the gel was transferred onto polyvinylidene difluoride (PVDF) membrane (Millipore, USA), which was then washed and blocked with Tris buffered saline (TBS) containing 0.1% Tween 20 (TBST) and 5% bovine serum albumin (BSA) at room temperature (RT). After washing with TBST, the primary antibodies were reacted with the PVDF membrane at 4℃ overnight. After washing with TBST, the membrane was incubated with horseradish peroxidase (HRP) conjugated secondary antibody at RT. The reactive protein band was visualized using enhanced chemiluminescent detection reagents (Thermo Fisher Scientific, USA) and quantified by ImageJ software (National Institutes of Health, USA).

Apoptosis assay
MDA-MB-231 cells in a 6 cm culture dish (2 × 105) were incubated overnight and then subjected/or not to DET (11 μM) and DETD-35 (4 μM) treatment for 8 h. The use of FITC-Annexin V Apoptosis Detection Kit (BD Pharmingen, USA) determined the treated cells following the manufacturer’s suggestions, and the apoptotic population was estimated using BD LSR II flow cytometer (BD Biosciences, USA).

Transfection of small hairpin RNA
Lentiviral-based small hairpin RNA (shRNA) clones purchased from the RNAi Core Facility, Academia Sinica, were used in this study for silencing specific gene expression in TNBC cells. The cells were incubated overnight and replaced with fresh media containing polybrene, a cationic polymer, and the lentiviral carrying shRNA targeting PRKCA, MCL1, or shLacZ (control), and then incubated overnight again. The PRKCA and MCL1 shRNA sequences are as follows: 5′-CCGGGAAGATGAAGACGAGCTATTTCTCGAGAAATAGCTCGTCTTCATCTTCTTTTTG-3′ for PRKCA; 5′-CCGGGCAGAAAGTATCACAGACGTTCTCGAGAACGTC TGTGATACTTTCTGCTTTTT-3′ for MCL1. The efficiently infected cells were selected and further incubated for 48 h at 37 °C using fresh media containing puromycin.

RNA preparation and qRT-PCR analysis
The total RNA of cells was extracted with TRIzol reagent. DNase from the TURBO DNAfree kit (Thermo Fisher Scientific, USA) was employed to treat the total RNA, and cDNAs were prepared by reversing transcription of RNA using SuperScript III Reverse Transcriptase (Thermo Fisher Scientific, USA). Primer sequences of genes were designed and are listed in Supplementary Table S1. Real-time PCR was carried out with SYBR Green PCR Master Mix (Thermo Fisher Scientific, USA) on the Applied Biosystems 7500 Fast Real-Time PCR system. GAPDH gene expression served as an internal control to normalize gene expressions, and the fold changes of genes compared with vehicle control were assessed.

Primary metabolite analysis using gas chromatography time-of-flight mass spectrometry (GC-TOF/MS)
The primary metabolites of the cell pellets were prepared with 1 mL of 80% MeOH containing 0.2 mg/mL ribitol as an internal standard (Sigma-Aldrich, USA) and then lysed through three freeze–thaw cycles. The lysed samples were centrifuged twice at 13,000 × g for 10 min at 4 °C, and the supernatants were collected and vacuum-dried. The samples were subjected to derivatized using bis(trimethylsilyl)-trifluoroacetamide (BSTFA) containing 1% trimethylchlorosilane (TMCS) and analyzed by GC-Q-TOF/MS with an electron ionization (EI) source (Agilent, USA). The samples were separated with Zorbax DB5- MS + 10 m Duragard Capillary Column (30 m × 0.25 mm × 0.25 mm, Agilent, USA), and the temperature gradient of GC was carried out at 60 °C for 1 min and then increased at 10 °C/min to 325 °C and held at 325 °C for 10 min. The acquired data was analyzed using MassHunter Workstation software. Mass spectra were further compared against the mass libraries of NIST 2017, Fiehn, and the Wiley Registry 11th Edition. Signal deconvolution and peak alignment were performed, respectively, by Agilent MassHunter Unknows Analysis software and Mass Profiler Professional software.

Mitochondrial protein preparation, iTRAQ labeling, and liquid chromatography-tandem mass spectrometry (LC–MS/MS) analysis
Pure mitochondria pellets from the cells were dissolved in 50 mM Tris buffer (pH 8.5) containing 8 M urea. The concentration of mitochondrial protein extracts was determined by Pierce 660 nm protein assay according to the manufacturer’s protocol. The mitochondrial protein extracts were reduced and alkylated, respectively, using dithiothreitol for 1 h and iodoacetamide for 30 min. The protein was digested into peptides with Lys-C for 4 h and then trypsin overnight. The digested samples were desalted by the Oasis HLB column (Waters Corporation, USA), and the eluted samples were dried by a SpeedVac. The proteolytic peptide samples were labeled with 8-plex isobaric tags for relative and absolute quantitation (iTRAQ) reagent (AB SCIEX, USA) following the manufacturer’s protocol. iTRAQ-labeled samples were combined and then dried by a SpeedVac. The dried samples were resuspended in 0.1% TFA solution and then separated into eight fractions using Pierce High pH Reversed-Phase Peptide Fractionation spin column (Thermo Fisher Scientific, USA). The collected fractions were concentrated by ZipTip Pipette tip (Millipore, USA) according to the manufacturer’s protocol and dried in a SpeedVac. The proteolytic peptides were analyzed using an LC-nESI-Q Exactive mass spectrometer (Thermo Fisher Scientific, USA) coupled with nanoUHPLC (Dionex UltiMate 3000 Binary RSLCnano, USA). The MS data acquisition was carried out in data-dependent mode with a full MS scan followed by 10 MS/MS scans based on the top 10 precursor ions from the MS scan.

Bioinformatics analysis of proteome and metabolome
The proteolytic peptides and their corresponding proteins were identified by Proteome Discover software (v.2.2, Thermo Fisher Scientific, USA) using SEQUEST and MASCOT search algorithms (Matrix Science, UK) against a Swiss-Prot human protein database of Human uniprot 51,829 entries. The relative expression of each protein in three biological replicates was measured according to the ratios of iTRAQ reporter ions (e.g., 114/113 and 115/113) originating from MS/MS spectra. The differentially expressed proteins were identified by a Z-score cutoff of ± 1.96σ (representing 95% confidence level) as a threshold. The Database for Annotation, Visualization and Integrated Discovery (DAVID) version 6.8 was applied to determine the role of differentially expressed proteins involved in Gene Ontology (GO) biological process annotations and molecular functions. The biological mechanisms associated with canonical pathways and toxicity in compound-responsive proteins were further analyzed by the Ingenuity Pathway Analysis (IPA) database (Ingenuity Systems, USA). The significantly enriched molecular functions, biological processes, canonical pathways, and toxicity lists of the compound-responsive proteome were selected based on the -log(p-value) value of 1.3 (p < 0.05) as a statistical significance threshold. The pathways involved in the compound-responsive metabolome were analyzed by MetaboAnalyst (version 6.0) using the Small Molecule Pathway Database (SMPDB). The -log(p-value) value of 1.3 (p < 0.05) was used to identify the enrichment metabolic pathways. The relationship between the compound-responsive proteome and metabolome from the treated cells was evaluated by Pearson’s correlation analysis using the RStudio tool, and the relative correlation networks of targeted compound-responsive proteins and metabolites were constructed through the Cytoscape tool.

Animal study and immunohistochemistry (IHC) staining
All animal procedures were approved by the Institutional Animal Care and Utilization Committee of Academia Sinica, Taiwan (Protocol ID: 16–12-1031), and conducted in accordance with the Guide for the Care and Use of Laboratory Animals (Ministry of Science and Technology, Taiwan) and the Taiwan Animal Protection Law. Mice were given a standard laboratory diet and distilled water ad libitum and kept on a 12 h light/dark cycle at 22 ± 2 °C. Briefly, 15 five-week-old female NOD-SCID mice were orthotopically injected with 5 × 10⁶ MDA-MB-231 cells suspended in 100 μL PBS mixed with 50 μL Matrigel into the mammary fat pads. After tumor formation in the mammary fat pad areas reaching about 55–100 mm3 mice were randomly assigned into three groups (n = 5), administered vehicle (5% DMSO), DET (20 mg/kg), and DETD-35 (10 mg/kg), respectively, via intraperitoneal injection every three days with up to ten doses of vehicle or compound treatment in total. An independent sham control group (n = 5) was administered an intraperitoneal injection of 5% DMSO without prior tumor induction. Mouse body weights and tumor growth were monitored by caliper measurements at 3-day intervals. At the experimental endpoint, tumor tissues were collected from MDA-MB-231 xenograft–bearing mice treated with vehicle, DET, or DETD-3526 and processed into formalin-fixed, paraffin-embedded (FFPE) tissue blocks for subsequent analyses. Tumor Sects. (4 μm) from FFPE tissue blocks were deparaffinized, rehydrated, and subjected to antigen retrieval using heat (95 °C) in Tris–EDTA buffer. Endogenous peroxidase is blocked with 3% H2O2, and non-specific binding is prevented with 5% BSA. Sections were incubated with anti-ATP5A1 (GTX101741, 1:200) or anti-ATP synthase gamma (ATP5C1) (GTX114275, 1:200) primary antibodies overnight at 4 °C, followed by an HRP-conjugated secondary antibody for 1 h. Staining was developed with DAB, counterstained with hematoxylin, then the slides were mounted. Images were captured by microscopy, and ATP5A1/ATP5C1 expressions were quantified using ImageJ plugin IHC profiler by deconvoluting DAB staining, and measured mean intensity or stained area percentage in several selected tumor regions.

Statistical analysis
All data are indicated as mean ± standard deviation (SD). Statistical analysis of experimental results was performed using the SAS program (SAS Institute), and the significant difference between treatments was defined by student’s t-test with p-values of less than 0.05.

Results

Effect of DET and DETD-35 on mitochondrial function in TNBC cells
Previously, we observed that DET and its derivative DETD-35 (Fig. 1A) rapidly and significantly induced intracellular ROS and superoxide production within 1–2 h in TNBC cells29. Moreover, both compounds had significant anti-TNBC cell activity that induced oxidative stress-mediated cytoplasmic vacuolation derived from the mitochondria and loss of mitochondrial membrane potential27,28. Therefore, in this study, we further delineated the effects of DET and DETD-35 on mitochondrial function in TNBC cells. The doses of DET (11 μM) and DETD-35 (4 μM) used in this study were the concentration of the compounds that reduced the viability of cells by 50%, determined in MDA-MB-231 TNBC cells treated for 24 h. We treated the cancer cells for 1 h with the compound concentration determined at the 24 h treatment time, and intended to examine the early effect of the compounds on mitochondrial superoxide formation in the cancer cells. As shown in Fig. 1B, 11 μM DET and 4 μM DETD-35 treatment for 1 h elevated 1.52- and 1.35-fold, respectively, relative to the vehicle treatment in mitochondrial superoxide production, as measured by mitochondrial superoxide indicator MitoSOX labeling assay. The effect was reversed by 1 h pretreatment with ROS scavenger N-acetylcysteine (NAC) from 1.52-fold to 1.23-fold with DET treatment, and from 1.35-fold to 1.12-fold with DETD-35 treatment. The superoxide anion is a short-lived and highly reactive radical that can be rapidly converted to hydrogen peroxide (H2O2) by superoxide dismutase (SOD). H2O2 is subsequently decomposed to water and molecular oxygen through diverse antioxidant enzymes such as catalase, glutathione peroxidases (GPXs), and peroxiredoxins (PRDXs)32. Therefore, next, the expressions of mitochondria-related antioxidant enzyme genes SOD2, GPX1, GPX4, and PRDX3 in TNBC cells were examined after treatment with either compound for 1 and 4 h. As shown in Fig. 1C, 1 h of treatment did not affect the expression of any of the four genes; however, after 4 h of treatment, SOD2 expression was significantly elevated by DETD-35 (1.51-fold increase, p < 0.01) or DET (1.33-fold increase, p < 0.05) compared with the vehicle control. The gene expression of antioxidant enzymes (SOD1, catalase, GPX7, and PRDX6) were also examined in the cytosol and ER. As shown in Fig. 1D, only 4-h treatment of DET and DETD-35 significantly raised SOD1 expression (1.24- and 1.31-fold increase, respectively, p < 0.05) in the treated TNBC cells; catalase, GPX7, and PRDX6 genes were not affected. These results suggest that within the 1 to 4 h treatment timeframe, oxidative stress due to both DET and DETD-35 was mainly associated with the actions of superoxide-converting enzyme SOD1/SOD2.
Increased ROS production and Ca2+ levels are regarded as potent activators regulating the mitochondrial permeability transition pore (mPTP) opening that is associated with mitochondrial cell death and dysfunction33. We previously observed that DET and DETD-35 increased ROS production and Ca2+ levels and programmed cell death in TNBC cells27,28. Here, whether DET and DETD-35 affected mPTP opening in TNBC cells was assessed using calcein acetoxymethyl ester coupled with cobalt chloride quenching. As shown in Fig. 1E, DET/DETD-35 significantly decreased the fluorescence intensity of calcein at 4 h treatment compared to vehicle-treated cells. The quantified mean fluorescence intensities (MFI) of green-fluorescent calcein in 4-h vehicle, DET, and DETD-35 treatment were 4833, 2988, and 3863, respectively, indicating a 38% and 20% decrease in DET and DETD-35 treatment relative to the vehicle control. Of note, the reduction of calcein fluorescence intensity by either compound treatment was reversed by pretreatment with NAC.
The abnormal and long-lasting opening of mPTP causes cell death known to trigger a series of events such as adenosine triphosphate (ATP) depletion, mitochondrial swelling, release of cytochrome c and pro-apoptotic molecules, loss of mitochondrial membrane potential, and stimulated oxidative stress34,35. A luminescence-based assay to examine changes in the intracellular ATP levels was employed in TNBC cells with or without DET/DETD-35 treatment. As shown in Fig. 1F, after 4 h treatment, the luminescence intensity in either of the compound-treated cells revealed a significant decrease compared to vehicle control cells, quantified as 81,847 (100%), 43,521 (53.2%), and 54,167 (66.2%) in the vehicle, DET, and DETD-35 treatment, respectively. Pretreatment with NAC before compound treatment in TNBC cells could prevent the decrease in intracellular ATP measured by the luminescent signal from ATP interacted with luciferase and luciferin. Together, these results indicate that DET − and DETD-35 − induced ROS play an important role in regulating mPTP opening and intracellular ATP homeostasis in TNBC cells.

Comparative mitochondrial proteome analysis of DET − and DETD-35 − treated TNBC cells
Our current and previous28,29 data revealed that DET/DETD-35 treatment has a negative impact on mitochondria-related activities in MDA-MB-231 TNBC cells. The mitochondrial proteome in MDA-MB-231 cells was thus investigated to elucidate the molecular mechanisms by which DET and DETD-35 affect mitochondrial function. Differential centrifugation and sucrose gradient ultracentrifugation were used to isolate a high-purity mitochondrial fraction from the tested TNBC cells. The purity of mitochondria isolated from MDA-MB-231 TNBC cells after sucrose-gradient purification was examined by immunoblotting using various organelle-specific marker proteins, including VDAC1 and cytochrome c for mitochondria, calnexin for ER, GAPDH and α-tubulin for cytosol, and lamin A/C for the nucleus. The level of mitochondrial marker VDAC1 and cytochrome c was significantly enriched in the purified mitochondrial fraction compared with the crude mitochondrial fraction and whole-cell protein extracts. Meanwhile, the levels of cytosolic and nuclear marker proteins such as GAPDH, α-tubulin, and lamin A/C were significantly decreased in the purified mitochondrial fraction from sucrose gradient ultracentrifugation (Supplementary Fig. S1A). However, detectable amounts of ER marker protein calnexin were found in the crude and purified mitochondrial fractions, possibly due to the natural interaction between mitochondria and the ER36. The proteins differentially expressed in purified mitochondria were analyzed using LC/MS-based proteomics. The purified mitochondrial fractions from the TNBC cells treated with DET or DETD-35 for 1 and 4 h were collected, and mitochondrial proteins were subjected to iTRAQ-based quantitative proteomics analysis. The mitochondrial proteins differentially expressed in response to treatment with either compound were identified. As shown in Supplementary Fig. S1B and Table S2, a total of 2994, 3563, and 3277 proteins were identified from the three batches of biological replicates of proteome analyses; among them, cross-analysis of proteins identified in at least two of the biological replicates (i.e., 2751, 3114, and 2958 proteins) was conducted. An overview of the mitochondria-related proteins identified in TNBC cells treated for 1 and 4 h with vehicle versus DET or DETD-35 is illustrated using the volcano plot as shown by log2 fold-change vs. log10
p-value (Fig. 2A). The Z-score cutoff of ± 1.96σ (representing 95% confidence level) was defined as the threshold for indicating differentially expressed proteins (DEPs) in the vehicle control versus DET or DETD-35 treatment from the three biological replicates. The DEPs that were seen in at least two biological replicates of DET/DETD-35 treatment were further identified and are labeled in Fig. 2A and Supplementary Table S3.
Next, the Gene Ontology (GO) molecular functions (MFs) and biological processes (BPs) of these DEPs were analyzed using a web-based bioinformatics tool − Database for Annotation, Visualization and Integrated Discovery (DAVID; https://david.ncifcrf.gov/). Significantly enriched MFs and BPs were selected, using -log(p-value) value of 1.3 (p < 0.05) as the threshold. As shown in Fig. 2B, the functions of DEPs in 4-h DET-treated cells were significantly associated with the groups “hydrolase activity” and “ATP-dependent activity”, such as ATP-dependent activity-related proteins 26S proteasome regulatory subunit 7 (PSMC2), subunit 6A (PSMC3), subunit 6B (PSMC4), and subunit 8 (PSMC5), mitochondrial disaggregase (CLPB), and twinkle mtDNA helicase (TWNK), which were found deregulated. The DEPs involved in “ATPase activity” from the 4-h DETD-35 − treated group included ATPase family AAA domain-containing protein 2 (ATAD2), ralA-binding protein 1 (RALBP1), and heat shock 70 kDa protein 1B (HSPA1B). The significantly enriched GO BPs were divided into several main groups indicated in Fig. 2C. The categories “cellular catabolic process” [(-log(p-value) value of 5.92] in DET − treated cells, and “regulation of cell death” [(-log(p-value) value of 4.91] in DETD-35 − treated cells were revealed as the top most-significant GO BPs. For instance, serine/threonine-protein kinase PAK 4 (PAK4), calcium and integrin-binding protein 1 (CIB1), cyclic AMP-responsive element-binding protein 1 (CREB1), HSPA1B, casein kinase II subunit alpha (CSNK2A1), beta-2-glycoprotein 1 (APOH), CCN family member 1 (CCN1 also called CYR61), heme oxygenase 1 (HMOX1), cytochrome P450 1B1 (CYP1B1), and induced myeloid leukemia cell differentiation protein Mcl-1 (MCL1), responding to DETD-35 treatment were classified into the category of “regulation of cell death”; among them, CYP1B1, MCL1, and heat shock protein beta-1 (HSPB1) were associated with oxidative stress-mediated cell death.
The biological mechanisms of DEPs, including the canonical pathways and toxicity list associated with DET/DETD-35 effects against TNBC cell activity, were further evaluated using Ingenuity Pathway Analysis (IPA), a web-based bioinformatics database. The canonical pathways with statistical significance of -log(p-value) value 1.3 (p < 0.05) were identified and separated into several representative groups shown in Fig. 3A. In the cellular homeostasis group, oxidative phosphorylation, thioredoxin pathway, and D-myo-inositol (1,4,5)-trisphosphate degradation were correlated with energy, redox, and Ca2+ homeostasis, respectively, which are known to be modulated by mitochondrial activity to maintain homeostatic balance in cells4,37,38. As shown in Supplementary Fig. S2, the expression profiling of identified electron transport chain (ETC) complex-related proteins responsive to both compound treatments showed that NADH dehydrogenase 1-alpha subcomplex subunit 13 (NDUFA13) (ETC complex I) and ATP synthase F(0) complex subunit C2 (ATP5MC2) (ETC complex V also known as ATP synthase) in the DET group, and ATP synthase subunit gamma (ATP5F1C) (ETC complex V) in the DETD-35 group were significantly down-regulated after 1 h treatment. Moreover, DEPs connected with the modulation of mitochondria-associated canonical pathways were also observed. For example, cyclic AMP-dependent transcription factor ATF-1 (ATF1) and cyclic AMP-responsive element-binding protein 1 (CREB1) in ATM signaling were significant in the DETD-35 treated group; and histone acetyltransferase type B catalytic subunit (HAT1) and thioredoxin-interacting protein (TXNIP) in the adipogenesis pathway, and acetyl-CoA acetyltransferase (ACAT2) in ketolysis & ketogenesis were significant in the DET treated group; phosphoglycerate kinase 1 (PGK1) in glycolysis I & gluconeogenesis I, and MCL1 and protein kinase C alpha (PRKCA) in apoptosis signaling were significant in both treatment groups (Fig. 3A). On the other hand, mitochondrial dysfunction as the consequence of responding to DET/DETD-35 treatment was also observed in the IPA toxicity list analysis, which may provide important clues about the pharmacological response and mechanism of action of compound treatment. As shown in Fig. 3B, DEPs NDUFA13/ATP5MC2 and MCL1 observed in 1-h or 4-h DET treatment, and ATP5F1C/CREB1 and HSPA1B/MCL1/CYR61 observed in 1-h or 4-h DETD-35 treatment were related to mitochondrial dysfunction and depolarization of the mitochondria and mitochondrial membrane.
Mitochondrial OXPHOS in living cells is one of the central hubs of intracellular ATP production and is regarded as the predominant route for ATP production39. The proteomics results revealed that both compounds deregulated expression of OXPHOS-related proteins such as NDUFA13, ATP5F1C, and ATP5MC2 in TNBC cells. Therefore, OXPHOS activity in TNBC cells with DET/DETD-35 treatment was further investigated by measuring oxygen consumption rate (OCR) in living cancer cells using seahorse XF cell Mito stress assay. Before adding ATP synthase inhibitor oligomycin, the level of OCR in TNBC cells at 4-h and 8-h treatment with either compound was decreased. The 8 h treatment group showed a particularly significant effect compared to vehicle-treated groups (Supplementary Fig. S3A,B). As shown in Supplementary Fig. S3C, the level of parameters related to mitochondrial bioenergetics by measuring OCR using different ETC complex inhibitors and uncoupling agent showed that basal/maximal respiration, ATP production, and coupling efficiency were significantly disrupted in the treated TNBC cells. Of note, the level of proton leak was significantly increased with 4 h DET treatment and slightly increased in the DETD-35 treated group compared to the vehicle control. These results suggest that DET/DETD-35 treatment disturbed the mitochondrial respiration and bioenergetics homeostasis in TNBC cells. The status of mitochondrial respiration and bioenergetics in normal MCF-10A mammary cells upon DET/DETD-35 treatment was also examined. The level of OCR and mitochondrial bioenergetics, including basal/maximal respiration, ATP levels, proton leak, and coupling efficiency in normal living MCF-10A cells did not differ after vehicle or either compound treatment for 8 h (Supplementary Fig. S4A,B), indicating that DET/DETD-35 did not have a detrimental effect on mitochondria in normal mammary cells. The seahorse XF cell Mito stress data, in part, support the observations from the IPA canonical pathway and toxicity list analyses that showed the deregulation of ETC complex V-associated proteins, which resulted in ATP depletion and mitochondrial dysfunction as a key mechanism of DET and DETD-35.

The role of mitochondria-related PRKCA and MCL1 genes in compound-treated TNBC cells
Among the DEPs associated with mitochondrial functions, PRKCA (also known as PKCα), a member of the classical PKC family proteins involved in cell death/survival and immunomodulatory activities, was upregulated by compound treatment in TNBC cells. This protein was thus selected for further investigation of its roles in the anti-TNBC cell activity of DET/DETD-35. A shRNA knockdown approach was used to create PRKCA gene knockdown MDA-MB-231 cell clones. The knockdown efficiency revealed a 0.4- and 0.7-fold decrease in the targeted molecule PRKCA in the cells compared to wild-type (WT) TNBC cells, as examined using immunoblotting (Fig. 4A). The PRKCA knockdown clone 2, with a 0.4-fold decrease in PRKCA expression, was chosen for the subsequent studies. DET showed similar cytotoxicity to WT and shLacZ control TNBC cells with a 50% inhibitory concentration (IC50) of 10.7 μM at 8 h treatment. DETD-35 also inhibited both WT and shLacZ control cells with similar IC50 of 4.5–4.7 μM (Fig. 4B). Of note, the knockdown of PRKCA gene in TNBC cells significantly attenuated the anti-proliferative effect of DET and DETD-35 by 1.47- and 1.5-fold compared to shLacZ cells, respectively. Next, we examined the parameters associated with mitochondrial respiration/bioenergetics in the PRKCA knockdown cells. In the shLacZ control cells, the basal/maximal respiration, ATP levels, and coupling efficiency were significantly decreased upon DET/DETD-35 treatment for 8 h. Interestingly, the significant increase in proton leak observed in the DET − treated group was reversed in the PRKCA knockdown cells, but not in DETD-35 − treated cells (Fig. 4C,D). Moreover, DET − /DETD-35 − induced OCR decrease and mitochondrial bioenergetic dysfunction were partially blockaded by silencing PRKCA gene expression. The effect of knocking-down PRKCA gene/protein levels on DET − or DETD-35 − induced apoptosis in TNBC cells was further investigated using FITC-Annexin V apoptosis detection assay. As shown in Fig. 4E, after treatment for 8 h, the population of apoptotic cells in the shLacZ control cells with vehicle, DET, and DETD-35 treatments was around 16%, 34%, and 36%, respectively. The apoptotic fraction in the vehicle − , DET − , and DETD-35 − treated PRKCA knockdown cells were 20%, 22%, and 22%, indicating compound-induced cancer cell apoptosis, which was reversed in the gene knockdown cells.
In a parallel experiment, apoptosis signaling-related protein MCL1, a member of the BCL-2 family proteins associated with apoptosis, cell proliferation, and Ca2+ homeostasis, was also investigated for its potential role in DET/DETD-35 repression of TNBC cell activity. As shown in Supplementary Fig. S5A, the MCL1 gene knockdown clone 1 showed a 0.5-fold expression level relative to the WT MDA-MB-231 cells, which was selected to assess the influence on DET/DETD-35 affected cell proliferation, apoptosis, and mitochondrial respiration/bioenergetics in cancer cells. Silencing MCL1 gene expression slightly changed cell proliferation in DET-treated shMCL1 cells; interestingly, this change was not observed in DETD-35 − treated cells (Supplementary Fig. S5B). Furthermore, silencing the MCL1 gene did not alter the decrease in OCR and basal/maximal respiration and ATP production in the cells with compound treatment compared to the shLacZ control cells. Of note, DET-stimulated proton leak was reversed in the shMCL1 cells (Supplementary Fig. S5C,D). Meanwhile, the percentages of cell apoptosis were, respectively, 18%, 35%, and 36% in the vehicle − , DET − , and DETD-35 − treated shLacZ cells, and 24%, 27%, and 36% in the treated shMCL1 cells (Supplementary Fig. S5E), indicating that the knockdown of MCL1 gene expression only mildly affected DET-induced apoptotic effect in TNBC cells, whereas MCL1 gene was not associated with the anti-TNBC cell effect of DETD-35.

DET and DETD-35 reprogrammed primary metabolism in TNBC cells
Next, we investigated the effect of DET and DETD-35 on primary metabolism in TNBC cells, and whether PRKCA was involved in the regulation of cancer cell metabolism by either compound. The shLacZ control and shPRKCA knockdown TNBC cells were treated with DET or DETD-35 for 8 h and subjected to GC/TOF–MS analysis, and the metabolite data were subjected to partial least squares-discriminant analysis (PLS-DA). Using Mass Profiler Professional software, 74 metabolites were identified in the shLacZ control and shPRKCA knockdown cells. PLS-DA was utilized to visualize the primary metabolite levels and dynamics in different groups. As illustrated by the score plot in Supplementary Fig. S6, the vehicle control − and DET − /DETD-35 − responsive primary metabolome in the shLacZ control and PRKCA knockdown cells can be separated into two groups. Several metabolite outliers were observed in the loading plot corresponding to different treatments in shLacZ or shPRKCA cells (Supplementary Fig. S6). The relative levels of primary metabolites in the shLacZ control and shPRKCA cells to those in vehicle control (V) or compound-treated cells were further calculated (designated the DET/V and DETD-35/V groups). In the heatmap (Fig. 5A), the metabolites detected are classified into groups based on their chemical functionality, including amino acids and derivatives, carbohydrates, fatty acids, nucleotides, organic acids, urea cycle, and other metabolites. Next, a volcano plot, using log2 fold-change vs. log10
p-value was used to overview the significant metabolites responsive to both compound treatments in shLacZ and shPRKCA TNBC cells (Fig. 5B, C). The fold-change (FC) ratio ≥ 2.14 or ≤ 0.47 (Log2 FC around ≥ 1.1 or ≤ -1.1) was set as the threshold for selection of differentially expressed metabolites (DEMs). Upon DET treatment, there were 30 and 20 DEMs in shLacZ and shPRKCA TNBC cells, respectively, and in DETD-35 treated cells, there were 26 and 31, respectively (Supplementary Table S4).
To further predict the role of the DEMs in TNBC cells regulated by either compound, pathway enrichment analysis using the web-based platform MetaboAnalyst 6.0 was performed with threshold -log(p-value) value of 1.3 (p < 0.05) from which the categories of significantly enriched metabolic pathways in the treated shLacZ control and shPRKCA knockdown cells were revealed. Homocysteine degradation, cardiolipin (CL) biosynthesis, phosphatidylethanolamine (PE) biosynthesis, fatty acid (FA) biosynthesis, purine metabolism, and methionine metabolism were identified in DET − treated shLacZ control cells, and glycerolipid metabolism appeared in DET − treated PRKCA knockdown cells (Fig. 5D). In DETD-35 − treated shLacZ control cells, homocysteine degradation, methionine metabolism, and pantothenate and CoA biosynthesis categories were observed, and in DETD-35 − treated PRKCA knockdown cells, purine metabolism, glycerolipid metabolism, and plasmalogen synthesis were observed (Fig. 5E). The FC of CL, PE, and FA biosynthesis-related metabolites, i.e., phosphatidic acid, glycerol-3-phosphate (G3P), phosphoethanolamine (PEtn), and hexanoic acid (HA) related to mitochondrial dynamic balance and mitochondrial respiration activity40–42, were further analyzed. As shown in Fig. 5F, the relative FCs of PE biosynthesis-related metabolite PEtn of the DET/V group in shLacZ cells and shPRKCA cells were 2.22 and 2.78, respectively, and 1.72 and 3.20, respectively, in the DETD-35/V group. The relative FCs of FA biosynthesis-associated metabolite HA of the DET/V group in shLacZ and shPRKCA cells were 2.49 and 2.05, respectively, and 2.87 and 2.13, respectively, in the DETD-35/V group. Similarly, the relative FCs of CL biosynthesis-associated metabolite G3P of the DET/V group in shLacZ cells and shPRKCA cells were 2.40 and 1.72, respectively, and 2.48 and 1.83 in the DETD-35/V group. In the PRKCA gene knockdown TNBC cells treated with either compound, these metabolites’ levels showed the opposite trend. Overall, the primary metabolome and associated pathway analysis results indicate that PRKCA plays a role in the anti-TNBC cell effect of DET and DETD-35 in modulating primary metabolism.

Correlation of the compound-regulated proteome and primary metabolome in TNBC cells
Next, we analyzed the correlation and relationship between DEPs and DEMs in TNBC cells treated with the compounds. First, a cross-analysis of the metabolomes of vehicle − treated WT and shLacZ control cells was performed and showed that the levels of identified primary metabolites were similar (p > 0.05) between the two sets of treated cells (data not shown). We further utilized Pearson’s correlation coefficient (r) analysis to build the correlation networks of DEPs and DEMs regulated by DET or DETD-35 in the WT and shLacZ TNBC cells. The significant threshold was based on r ≥ 0.8 or ≤  − 0.8 and p < 0.05. The positive and negative correlations between DEPs and DEMs are shown in Fig. 6A, B, presented using blue and red color icons, respectively, and the circle sizes indicate the measured correlation coefficient (r). Some targeted proteins and metabolites in both compound treatment groups were identified and labeled with a white asterisk. The correlation networks of targeted proteins and metabolites from the DET and DETD-35 treatment groups were further built and visualized using Cytoscape software (Fig. 6C, D). The high correlation between targeted proteins (1 and 4 h treatment) and metabolites (8 h treatment) in the DET treatment group was identified and divided into two groups based on their interaction networks. The two groups were protein ATP5MC2 and metabolites elaidic acid/pelargonic acid/phosphoric acid/adenosine, and protein PRKCA and metabolites inosine/uridine from 1-h DEPs and 8-h DEMs, and protein PSMC3 and metabolites M6P/oxalic acid/phosphoric acid, and protein HSPB1 and metabolites phenylalanine/PEtn from 4-h DEPs and 8-h DEMs (Fig. 6C). In the DETD-35 treatment group (Fig. 6D), protein ATP5F1C and metabolites palmitic acid/citric acid, and protein ATAD2 and metabolites G3P/pyroglutamic acid from 1-h DEPs and 8-h DEMs, and protein CYP1B1 and metabolite pantothenic acid, and protein MCL1 and metabolites G3P/inosine from 4 h DEPs and 8 h DEMs exhibited a strong relationship.
The biological functions of targeted proteins and metabolites identified from Pearson’s correlation analysis were further illustrated using IPA network analysis. The functions of several significant proteins and metabolites were mitochondria-related. For example, of the correlated proteins versus metabolites identified in the 1 h DET treatment group, ATP5MC2 and phosphate/palmitic acid in OXPHOS and mitochondrial dysfunction, TXLNG/FITM2 and pyrrolidonecarboxylic acid/adenosine in energy homeostasis, and PGK1 and palmitic acid/adenosine in synthesis of ATP were observed (Fig. 7A). While, in the correlated proteins versus metabolites identified in the 4 h compound treatment, MCL1/HSPB1/TWNK and phosphate/palmitic acid/adenosine; MCL1 and inosine; MCL1 and palmitic acid are related to the transmembrane potential of mitochondria, respiration of mitochondria, and depolarization of mitochondria, respectively. In parallel, as shown in Fig. 7B, the function of PGK1 and palmitic acid/adenosine/L-glutamic acid; ATP5F1C and palmitic acid; ARID1A and palmitic acid from 1 h DETD-35 treatment group were grouped to the categories of synthesis of ATP, mitochondrial dysfunction, and depolarization of mitochondria, respectively. MCL1 and L-glutamic acid in respiration of mitochondria and depolarization of mitochondria membrane; CYP1B1 and palmitic acid in synthesis of ROS and mitochondria DNA damage were found in the 4-h DETD-35 treatment group.

Effect of DET and DETD-35 on ATP synthase in TNBC cells
Based on our observation of the mitochondrial proteome, measurement of mitochondrial bioenergetics, and IPA network analysis, DET and DETD-35 treatments caused mitochondrial ATP depletion and deregulation of ATP synthase-related proteins, which showed a high correlation with ATP synthesis and energy homeostasis. We thus hypothesized that both compounds might inhibit ATP synthase activity in TNBC cells. We prepared the crude mitochondrial fraction from wild-type TNBC cells treated with vehicle, DET, DETD-35, or oligomycin as a positive control for 8 h, and the activity of mitochondrial ATP synthase was measured using an immunocapture method. As shown in Fig. 8A, treatment with DET, DETD-35, or oligomycin significantly decreased the activity of mitochondrial ATP synthase in TNBC cells by 28%, 35%, or 17%, respectively, compared to vehicle control (three independent experiments with p < 0.05).
We further carried out IHC staining to examine the ATPase expression levels in the xenograft MDA-MB-231 mammary tumors from mice with or without DET or DETD-35 treatment. The staining results in tumor tissues were quantified and classified using IHC Profiler plugin in ImageJ software expressed as percentages of negative, low positive, positive, and high positive. The results revealed that, in the tumors treated with either compound, there was a significant decrease in percentage (%) of high positive/positive staining, along with an increase in the percentage of negative staining for ATP synthase-related proteins, such as ATP synthase F1 subunit alpha and gamma (ATP5A1 and ATP5C1) compared to the non-treated tumor control (p < 0.05). For ATP5A1, the high positive/positive and low positive/negative staining percentages were 8.9% and 91.1% for DET, and 3.7% and 96.3% for DETD-35, respectively, compared to the tumor control 25.6% and 74.4%. A similar trend was observed for ATP5C1 expression in the tumor tissues, i.e., the tumor control showed 29.0% highly positive/positive and 71.0% low positive/negative staining, while DET and DETD-35 groups exhibited 6.2% and 93.8%; and 6.1% and 93.9%, respectively (Fig. 8B). These findings are consistent with our in vitro mitochondrial proteome analysis, which demonstrated the decreased expression of ATP synthase-related proteins ATP5F1C and ATP5MC2 in TNBC cells treated with DET or DETD-35.
Molecular docking analysis was further carried out to observe the potential interaction sites of DET or DETD-35 with ATP synthase. To explore the potential interaction regions in ATP synthase subunits, binding free energy (ΔGbinding) for protein–ligand interactions was calculated by a web-based protein–ligand docking tool, SwissDock. Two possible binding regions of DET or DETD-35 with subunits α/β and c/a interfaces of the ATP synthase were predicted to have the lowest binding energy, indicating the most favorable interaction compared to others (Supplementary Table S5). The microenvironments of DET binding or interacting with the pocket in the subunits α/β are shown in Fig. 8C. DET can bind to the pocket formed by subunits α and β, and the γ-lactone of DET interacts with the side chain guanidino group of Arg373 in subunit α by ionic interaction. Furthermore, DET forms hydrophobic interactions with subunit α residue Val371, and with subunit β residues Val167, Tyr348 and Phe427. In the subunits c (designated C1)/a interfaces of the enzyme, the DET molecule revealed hydrophobic interactions with Ile51, Leu52, and Ala55 of subunit c, and Val113, Leu149, and Tyr221 of subunit a. The γ-lactone group of DETD-35 may form a hydrogen bond with the hydroxyl (-OH) group of Ser48 in subunit c; and DETD-35 may form hydrophobic interactions with amino acid residues, such as Ile51 and Ala55 in subunit c, and Val113, Phe128, Pro130, Leu149, and Tyr221 in subunit a (Fig. 8D). In addition, Val371 in subunit α, and Val167, Tyr348, Phe421, and Ala424 in subunit β located at the pocket of subunits α and β interface were predicted to be hydrophobic interactions with DETD-35. Together, these findings from in vitro and in vivo TNBC models and the structural modeling results lay out important groundwork and provide mechanistic insights into the anti-TNBC activity of DET/DETD-35.

Discussion
A growing body of evidence suggests that targeting mitochondrial activities is a promising strategy for cancer therapy due to the crucial role of mitochondria in cell growth and apoptosis and cell metabolism among others43,44. In particular, modulation of OXPHOS or ROS production in mitochondria seems attractive for inhibiting cancer cell activity. Typically, elevated mtROS can diffuse into the cytosol through the mPTP and the inner membrane anion channel, where the cytoplasmic antioxidant system neutralizes it45,46. This study elucidates the novel modes of action of the DET and DETD-35 in TNBC cells, including inducing mitochondrial superoxide production and its catalytic enzyme SOD2 expression, stimulating mPTP opening, and elevating the main cytoplasmic SOD1 expression. Furthermore, DET − and DETD-35 − induced oxidative stress caused a negative impact on mitochondrial function, ATPase activity and ATP production, and cellular metabolism in the cancer cells. Other anti-cancer drugs, such as doxorubicin, trastuzumab, and sunitinib, have also been shown to induce ROS overproduction accompanied by mitochondrial dysfunction, including respiratory chain impairment and loss of membrane potential47.
The mPTP channel comprises the dimer or c-subunit ring (an oligomer of subunit c) of F1F0 ATP synthase (ETC complex V). ATP5G1, ATP5G2, and ATP5G3 form the c-subunit of F0 ATP synthase, which is combined with subunit a to drive ATP synthesis by converting an electrochemical gradient of H⁺ ions into rotational movement48–50. Meanwhile, the subunits α, β, γ, δ, and ε constitute the F1 domain of ATP synthase, which is responsible for ATP synthesis from ADP and inorganic phosphate51. Therefore, the catalytic domains in F0 or F1, including subunit α/β and c/a interfaces, and subunit c, are regarded as targets for suppressing ATP synthase activity and ATP synthesis. In previous studies, ATP synthase inhibitors aurovertins, efrapeptins, and tentoxin were found binding to the pockets formed at the subunit α and β interfaces, and other inhibitors oligomycin and bedaquiline interact with subunit c or the interface between the subunits c and a, in turn, blocking ATP synthesis and proton translocation by stalling the c-ring rotation with subunit a52,53. Our molecular modeling results suggest that DET and DETD-35 attenuated mitochondrial ATP synthase activity in the TNBC cells is via the binding or interacting with the interfaces between subunits α/β and c/a of ATP synthase. Analysis with the molecular visualization/modeling tool PyMOL revealed that the predicted binding sites of DET and DETD-35 within the hydrophobic pocket at the α/β subunit interfaces of ATP synthase are close to the binding sites of ATP synthase inhibitor aurovertin B when aligned with the structure of mitochondrial F1-ATPase complex with aurovertin B (PDB: 1cow)52. Furthermore, several hydrophobic interactions existed between the DET/DETD-35 and subunit c/a interfaces, and the OH group of Ser52 located in subunit c formed a hydrogen bond with the γ-lactone group of DETD-35. Along with inhibiting the enzymatic activity of mitochondrial ATPase in the cancer cells, DET and DETD-35 are inhibitors of ATPase that disrupt ATP production and mitochondrial bioenergetics in the TNBC cells.
Mitochondrial ATP depletion is now considered to be a therapeutic strategy for metastasis prophylaxis. Treatment with an FDA-approved antibiotic drug bedaquiline deregulated ATP5F1C expression accompanied by inhibiting mitochondrial ATP synthase and ATP production, cell growth, and metastasis of breast cancer cells activities in vitro and in vivo54. Our mitochondrial proteome analysis revealed that DETD-35 downregulated the expression of ATP synthase-related protein ATP5F1C (also known as ATP5C1). DET treatment significantly downregulated ATP5MC2 (also known as ATP5G2). IHC analysis further confirmed reduced expression of ATP5C1 and ATP5A1 in xenograft mammary tumors treated with DET or DETD-35. These in vitro and in vivo study results support the existence of a compound effect on mPTP disruption and/or ATP depletion in TNBC cells. Furthermore, Pearson’s correlation analysis of the primary metabolome and mitochondrial proteome revealed that ATP5F1C positively correlated with citric acid in DETD-35 − treated TNBC cells. Citrate is an important regulator of mitochondrial ATP production and its derivative α-ketoglutarate suppresses ATP synthase activity55. A strong relationship was observed between the deregulation of ATP5MC2 and phosphoric acid in response to DET treatment. ATP synthase utilizes ADP and phosphoric acid for ATP synthesis, playing a crucial role in proton translocation and subunit rotation within the complex V56. The primary metabolome and mitochondrial proteome results support the novel mode of action of both compounds in inhibiting TNBC cells by inhibiting mitochondrial ATPase and causing ATP depletion.
PRKCA or PKCα, a ubiquitously expressed serine/threonine kinase and a member of the classical PKC family, is translocated to different subcellular compartments in response to distinct stimuli and interacts with different signaling molecules to modulate varied biological functions, such as apoptosis, cellular immunity, and cell proliferation/differentiation57. Nowak et al. have demonstrated that PKCα is an important regulator for maintaining ATP synthase activity and ATP levels through interaction with the F1 domain of ATP synthase in renal proximal tubules after oxidant injury58. Meanwhile, an increase in the activity of PRKCA in the mitochondria could result in mitochondrial toxicity, including disrupted mitochondrial transmembrane potential, elevated mtROS production, and reduced activity of ETC complex I and pyruvate dehydrogenase59. Our compounds induced PRKCA levels in TNBC cells. Knockdown of this gene reversed cytotoxicity and apoptosis induced by DET or DETD-35 in the cells. Meanwhile, compound treatment increased phosphoric acid level and decreased ATP level in the shLacZ control or wild-type MDA-MB-231 cells, results which were reversed in compound-treated shPRKCA cells. On the other hand, according to our mitochondrial proteome and metabolome data, DET and DETD-35 treatment can increase M6P levels (FC ratio ≥ 2.4) and decrease pentose phosphate pathway-related enzyme expression, such as glucose-6-phosphate dehydrogenase and 6-phosphogluconolactonase, indicating that M6P elevation can act as a modulator of glucose metabolism in treated TNBC cells. Knocking down the PRKCA gene in TNBC cells also decreased the level of M6P, a mannose degradation-related metabolite recently recognized to have anti-cancer activity60.
MCL1, a member of the BCL-2 family, has different spliced variants that regulate cell survival and death. The full-length MCL1-L has an anti-apoptotic function, whereas the shorter MCL1-S (short isoform) and MCL1-ES (extra short isoform) have a pro-apoptotic function. An increase in MCL1-S expression can disrupt mitochondrial function, leading to hyperpolarization, apoptosis, and Ca2+ accumulation in cancer cells. In addition, an imbalance between MCL1-L and MCL1-S expressions could refer to disrupting mitochondrial dynamics, leading to mitochondrial fragmentation in HeLa cells61. We did observe an elevated MCL1-S/MCL1-L protein ratio in TNBC cells treated with DET and DETD-35 that may support the mitochondria dysfunction induced by both compounds (data not shown). However, when the MCL1 gene was knocked down, DET − induced apoptosis and mitochondrial proton leak, and DETD-35 − induced decrease in maximal respiration (an indicator of the mitochondrial ETC’s maximum capacity) were attenuated in knockdown cells. (Supplementary Fig. S5). Mitochondrial proton leak comprises basal leak from ETC complexes and inducible leak by uncoupling proteins/adenine nucleotide translocases. Superoxide radicals generated in the mitochondrial ETC (complexes I and III) arise from electron leakage to molecular oxygen62. Basal proton leak mainly originates from ETC complexes I and III, and stimulation of basal proton leak can indicate mitochondrial damage29,63. Our mitochondrial proteomic analysis showed that DET and DETD-35 treatment lead to dysregulation of several complex I-related proteins, such as NDUFA6, NDUFA11, and NDUFA13 in TNBC cells (Supplementary Fig. S2). Consistent with our observation, reduced expression of NDUFA6, NDUFA11, and NDUFA13 by siRNA approach has been reported to be linked to elevated ROS levels and disruption of complex I activity in OP9 and HEK293T cells64–66. A previous study showed that downregulation of NDUFA13 in cardiac-specific knockout mice increased cytosolic H2O2 levels by affecting ETC complex I integrity and causing electron leak64. However, whether the dysregulation of complex I-related proteins observed in our study contributes to compound-induced mitochondrial ROS production remains to be further investigated.
In the primary metabolome analysis, we found that several DEMs up-regulated by DET or DETD-35 may contribute to the anti-cancer effects. For example, excessive mannose-6-phosphate (M6P) may provoke the decrease in intracellular ATP and result in dNTP loss, impeding DNA replication and cell cycle progression67,68. In our current study, we observed that DET or DETD-35 treatment increased M6P level, decreased intracellular ATP levels, and induced cell-cycle arrest26 in TNBC cells. The function of threonine has been correlated to increase antioxidative enzyme SOD level in Caenorhabditis elegans in response to acute oxidative stress69. In DET-treated TNBC cells, elevated expression of SOD1 and SOD2 genes associated with superoxide production was demonstrated (Fig. 1). We also observed that DET increased threonine levels in TNBC cells. On the other hand, elevated glutamate (the ionized form of glutamic acid) levels due to glutamate treatment impair mitochondrial function in human neuroblastoma SH-SY5Y cells through increasing mitochondrial calcium, enhancing ROS production, dissipating mitochondrial membrane potential, and promoting mPTP opening70. DETD-35 treatment induced glutamic acid levels in TNBC cells, which may be attributed to the compound-induced mitochondrial dysfunction observed in this study and the increase in intracellular Ca2⁺ and ROS levels in the cancer cells observed in our previous study27,28. Most importantly, the increased levels of the DET − and DETD-35 − responsive DEMs in the control shLacZ cells, as mentioned above, were decreased (FC ratio < 2.1) in the treated shPRKCA cells.
Phospholipids such as cardiolipin (CL) and phosphatidylethanolamine (PE) are crucial for building inner mitochondrial membranes and mediating OXPHOS efficiency by interacting with ETC complex II-V71. The deregulation of CL and PE biosynthesis-related metabolites such as G3P and PEtn was identified in the primary metabolome of treated cancer cells. In FDA-approved anti-diabetic metformin-treated human prostate DU145 or lung A549 cancer cells, G3P levels were elevated, referred to as mitochondrial dysfunction, impairing respiration and ATP production40. We observed that DET − and DETD-35 − treated shLacZ control cells showed significantly increased G3P (FC ratio ≥ 2.4) and decreased glycerol levels (FC ratio ≤ 2). In contrast, the compound-treated shPRKCA cells exhibited significantly reduced G3P (FC ratio < 1.8) and increased glycerol (FC ratio > 2.2), indicating that the increased levels of G3P may contribute to DET − and DETD-35 − induced mitochondrial bioenergetics disruption which involved PRKCA function. PEtn, a crucial participator in the Kennedy pathway for PE and phosphatidylcholine synthesis, regulates mitochondrial activity by competing with succinate as a substrate for ETC complex II succinate dehydrogenase (SDH) that oxidizes succinate to prompt ROS production41. Exogenous PEtn treatment restores mitochondrial function by inhibiting SDH, normalizing ROS levels, and protecting DNA from ROS-induced damage. On the other hand, treatment with the FA biosynthesis-related metabolite HA (also known as caproic acid), a 6-carbon saturated fatty acid that enters mitochondrial membranes without a shuttle system, reduced the proliferative activity of HepG2 and Hep3B hepatoma cells and caused mitochondrial depolarization in A549 lung carcinoma cells72. In treated shPRKCA cells, elevated PEtn level (FC ratio ≥ 2.8), decreased HA (FC ratio < 2.1) and succinic acid levels (FC ratio ≤ 1.3) and restored SDH activity as measured by MTT assay were observed compared to compound-treated shLacZ cells, suggesting that PRKCA expression influences compound-induced metabolic reprogramming and anti-TNBC cell activity. Together, our results indicate that the interference in mitochondrial bioenergetics in TNBC cells by DET and DETD-35 is likely through regulation of PRKCA function.
Glycolysis typically functions as a compensatory ATP-producing pathway when ETC activity is impaired. Deregulated levels of glycolysis-related proteins and primary metabolites were observed in DET − /DETD-35 − treated TNBC cells, including decreased phosphoglycerate kinase 1 (PGK1) expression and increased dihydroxyacetone phosphate (DHAP) level. PGK1 catalyzes the conversion of 1,3-bisphosphoglycerate and ADP to 3-phosphoglycerate and ATP. PGK1 inhibition using siRNA has been reported to cause the accumulation of upstream glycolytic intermediates (including fructose 1,6-bisphosphate, DHAP, glyceraldehyde 3-phosphate, and 1,3-bisphosphoglycerate) in various cancer cell lines (HeLa, MGC80-3, RKO, SK-HEP-1, and A549) and decrease cell growth rate in HeLa cells. However, such perturbation did not affect glucose-to-lactate conversion significantly during glycolysis, which might be attributable to the kinetics and thermodynamics of glycolysis73. DHAP has been reported to form a spontaneously harmful compound, methylglyoxal, which represses tumor cell growth by inhibiting glycolysis and mitochondrial respiration in EAC cell-inoculated mice74. Although we observed several DET − or DETD-35 − responsive DEPs and DEMs in the present study, the exact pharmacological roles of some of the deregulated metabolites and proteins in the TNBC cells functionally associated with ETC complex I activity and glycolysis have not yet been validated, and further work will be necessary in the future. In addition, measuring glycolytic flux may be useful to provide further support for the effect of DET or DETD-35 on glycolytic ATP production. Furthermore, understanding compound-associated impairment of glycolytic flux and its associated proteins or metabolites would provide further in-depth information about how phytosesquiterpen lactone could markedly suppress TNBC activity via inducing mitochondrial dysfunctions. Therefore, whether compensatory ATP-generating pathways, particularly glycolysis, are activated in response to mitochondrial ATPase inhibition remains to be addressed. In addition, whether or not it can be observed with metabolic rerouting (expect increased glycolysis, not increased OXPHOS, upon OXPHOS impairment) using 2-deoxy-D-glucose or other chemicals is an interesting question to be addressed in future investigations.

Conclusion
This study elucidates the underlying mechanisms by which phytosesquiterpene lactone DET and its derivative DETD-35 suppress TNBC cell activity. The compounds not only induced metabolic reprogramming in TNBC cells but also triggered multiple layers of mitochondrial dysfunction, including increased mitochondrial superoxide production and mPTP opening, disruption of mitochondrial bioenergetic homeostasis, and reduced ATP synthase activity, ultimately leading to decreased ATP levels (Supplementary Fig. S7). This study also shows that PRKCA plays a role in the pharmacological activities of DET and DETD-35 by mediating anti-proliferation, apoptosis, mitochondrial bioenergetics disruption, and metabolic reprogramming in TNBC cells.

Supplementary Information
Below is the link to the electronic supplementary material.