Identification of mitochondrial ATP synthase as the cellular target of Ru-polypyridyl-β-carboline complexes by affinity-based protein profiling.

ABSTRACT Ruthenium polypyridyl complexes are promising anticancer candidates, while their cellular targets have rarely been identified, which limits their clinical application. Herein, we design a series of Ru(II) polypyridyl complexes containing bioactive β-carboline derivatives as ligands for anticancer evaluation, among which Ru5 shows suitable lipophilicity, high aqueous solubility, relatively high anticancer activity and cancer cell selectivity. The subsequent utilization of a photo-clickable probe, Ru5a , serves to validate the significance of ATP synthase as a crucial target for Ru5 through photoaffinity-based protein profiling. Ru5 accumulates in mitochondria, impairs mitochondrial functions and induces mitophagy and ferroptosis. Combined analysis of mitochondrial proteomics and RNA-sequencing shows that Ru5 significantly downregulates the expression of the chloride channel protein, and influences genes related to ferroptosis and epithelial-to-mesenchymal transition. Finally, we prove that Ru5 exhibits higher anticancer efficacy than cisplatin in vivo . We firstly identify the molecular targets of ruthenium polypyridyl complexes using a photo-click proteomic method coupled with a multiomics approach, which provides an innovative strategy to elucidate the anticancer mechanisms of metallo-anticancer candidates. ## INTRODUCTION INTRODUCTION Although platinum-based anticancer therapeutics have been widely applied in clinics, they have drawbacks including toxic side effects and drug resistance [ 1–4 ]. Among non-platinum metallo-anticancer agents, ruthenium complexes are the most prominent, and four ruthenium complexes have successfully advanced to clinical trials [ 5 , 6 ]. In particular, Ru(II) polypyridyl complexes have attracted increasing attention due to their convenient structural modifications [ 7 ], alternative anticancer mechanisms different from those of platinum drugs [ 8 ], potential phototherapeutic applications [ 9 , 10 ], multifunctionalities integrating imaging and therapy [ 11 , 12 ], good biocompatibility and selectivity, etc. [ 13 , 14 ]. However, to the best of our knowledge, very little effort has been put into the cellular targets validation of Ru(II) polypyridyl complexes to date. Although complexes containing large planar ligands, such as dipyrido[3,2-a:2′,3′-c]phenazine (DPPZ) derivatives, are considered to target DNA, the prerequisite is that they must accumulate in mitochondria or nuclei to bind with DNA [ 15 , 16 ]. Moreover, the variation of the ligands shows a great impact on the anticancer properties of these complexes, which may be attributed to their different molecular targets in cells. Activity-based protein profiling (ABPP) has been widely applied in medicinal chemistry and chemical/structural biology to identify drug targets, binding substrates or interaction sites [ 17–21 ]. Photoaffinity labeling (PAL) is a powerful tool to investigate non-covalent ligand–receptor interactions [ 22 ]. Typically, a covalent bond is formed between the photo-cross-linker and protein residues upon UV irradiation, which converts the interaction modes from reversible to irreversible binding [ 23–25 ]. Photoaffinity-based protein profiling (PA-BPP) has been widely used to identify the cellular targets of bioactive agents. Yao's group, as well as other groups, have put a lot of effort into developing highly efficient photo-cross-linkers with optimized stability and labeling specificity [ 26 , 27 ]. Among these linkers, the most interesting kind are the ‘minimalist’ bioorthogonal handle-containing photo-cross-linkers [ 27 ]. Without significantly affecting the structure and activity of the lead compounds, these linkers can improve labeling efficiency, and they are widely applied in the construction of irreversible inhibitors, proteome profiling, target identification, etc. [ 28 , 29 ]. β-Carboline alkaloids are a family of natural and synthetic products with a variety of biological activities [ 30 ]. They show potent anticancer activities through multiple mechanisms including intercalating into DNA [ 31 ], inhibiting topoisomerases I/II [ 32 , 33 ], impairing the cyclin-dependent kinases (CDKs) [ 34 ] and other oncoproteins [ 35 , 36 ]. Considering the easy coordination of nitrogen-containing heterocycles of β-carboline alkaloids with metal centers (e.g. ruthenium, iridium and copper), we and other groups have designed many metal-carboline complexes as anticancer agents in the past decade [ 37–40 ]. These complexes show interesting anticancer properties, including specific subcellular organelle localization and induction of different types of programmed cell death [ 41–43 ]. However, their molecular targets are still almost unknown, which is crucial for the understanding of their anticancer mechanisms and further applications. In this work, nine Ru(II) polypyridyl complexes (Scheme  1A ) containing β-carboline derivatives as ligands are synthesized, among which Ru5 shows the highest anticancer activity. By using Ru5a , a structural analog of Ru5 incorporating a photo-affinity tag, we identify that mitochondrial ATP synthase (ATPase) is the most important molecular target of Ru5. The findings from subsequent studies indicate that Ru5 functions as an ATPase inhibitor, resulting in mitochondrial dysfunction and subsequently leading to ferroptosis and epithelial-to-mesenchymal transition (EMT) inhibition in cancer cells (Scheme  1B ). The possible pathways through which Ru5 interferes with mitochondrial function by inhibiting ATPase are elucidated by combined analysis of mitochondrial proteomics and RNA-sequencing. Moreover, Ru5 can induce mitophagy and ferroptosis, and it also inhibits the EMT of cancer cells. Finally, in vivo evaluation shows that Ru5 possesses a higher anticancer potency than cisplatin. Our study identifies the molecular targets of ruthenium polypyridyl complexes by integrating multiomics information, which gives insights into the action mechanisms of these potential metallo-anticancer candidates. Scheme 1. (A) The chemical structures of Ru1 – Ru9 . (B) Workflow for molecular target identification and anticancer mechanism investigation of Ru5 using the photo-affinity probe Ru5a by PA-BPP. ## RESULTS AND DISCUSSION RESULTS AND DISCUSSION Synthesis and structure–activity relationship study Based on our previous work [ 43 ], we designed nine Ru(II) polypyridyl complexes by changing the auxiliary ligands (2,2'-bipyridine (bpy; Ru1, Ru4 and Ru7 ); 1,10-phenanthroline (phen; Ru2, Ru5 and Ru8 ); 4,7-diphenyl-1,10-phenanthroline (dip; Ru3, Ru6 and Ru9 )) to adjust the lipophilicity and the main β-carboline ligands (1-(quinolin-2-yl)-9H-pyrido[3,4-b]indole ( L1; Ru1 – Ru3 ); 1-(1H-imidazol-2-yl)-9H-pyrido[3,4-b]indole ( L2; Ru4 – Ru6 ); 1-(1H-benzo[d]imidazol-2-yl)-9H-pyrido[3,4-b]indole ( L3; Ru7 – Ru9 )) to modify the binding ability with the targeted biomolecules. The β-carboline ligands are varied by keeping the core structure and introducing different pharmacophores (quinoline, imidazole and benzimidazole) to the C1-position. L1 – L3 were obtained by reacting tryptamine with the corresponding aldehyde in dry anisole as we previously reported [ 44 , 45 ]. Ru1 – Ru9 were synthesized by refluxing 1 equiv of the β-carboline ligand and the corresponding precursor cis -[Ru(N–N) 2 Cl 2 ] followed by anion exchange with NH 4 PF 6 . The complexes were characterized by 1 H nuclear magnetic resonance (NMR), 13 C NMR spectroscopy, electrospray ionization mass spectrometry (ESI-MS) and elemental analysis (Supplementary Data, Figs S1 – S34 ). X-ray diffraction shows that Ru5 adopts a typical octahedral configuration ( Fig. S35 , Tables S1 and S2 ). Subsequently, the lipophilicity and aqueous solubility of Ru1 – Ru9 were investigated. The results indicate that the Log P o/w (oil-water partition coefficient) values of Ru1 – Ru9 range between −2.01 and 3.59 ( Table S3 ). Ru3, Ru5 and Ru7 exhibit suitable Log P o/w that can ensure good solubility and bioavailability [ 46 ]. By evaluating the linear relationship between concentration and absorbance, we found that the solubility of Ru1 – Ru9 in phosphate buffered saline (PBS) (with 1% dimethyl sulfoxide (DMSO)) was almost positively correlated with their lipophilicity. A linear relationship between the absorbance and the concentration is obtained for Ru1, Ru2, Ru4, Ru5, Ru7 and Ru8 up to at least 100 μM. The highest solubility of Ru3, Ru6 and Ru9 with dip as auxiliary ligands is lower than 40 μM under the same conditions ( Fig. S36 ). The antiproliferative activities of Ru1 – Ru9 were first evaluated on human cell lines including cervical carcinoma (HeLa), lung adenocarcinoma (A549), cisplatin-resistant A549R, prostate cancer (PC3), triple negative breast cancer (MDA-MB-231), malignant glioma (U87) and human lung fibroblast-like (HLF) cells (Table  1 and Table S4 ). The complexes containing β-carboline ligands with imidazole ( Ru4 – Ru6 ) and benzimidazole ( Ru7 – Ru9 ) as substituents have higher activity than those with quinoline ( Ru1 – Ru3 ) as substituents. Ru3 – Ru9 display relatively high anticancer activities towards the cancer cell lines tested and maintain good activity against cisplatin-resistant A549R cells. Ru3 – Ru9 also show a certain selectivity for cancer cells over normal cells, and Ru2, Ru3, Ru5 and Ru9 show higher selectivity index (SI) values than the other complexes. Table 1. Half maximal inhibitory concentration (IC 50 , μM, 72 h) values of tested compounds towards different cell lines. a Complexes HeLa A549 A549R PC3 MDA-MB-231 U87 HLF SI b Ru5 3.1 ± 0.7 2.1 ± 0.8 3.3 ± 0.7 2.7 ± 0.4 7.7 ± 0.2 14.7 ± 0.7 33.9 ± 0.2 16.1 Ru5a 3.8 ± 0.1 2.5 ± 0.4 4.2 ± 0.6 3.4 ± 0.6 8.4 ± 0.6 13.4 ± 0.4 23.2 ± 0.3 9.3 Cisplatin 10.3 ± 1.3 5.9 ± 1.0 28.1 ± 4.2 13.1 ± 2.7 9.0 ± 0.2 14.8 ± 1.2 15.4 ± 1.1 2.6 a Cells were incubated with the compounds for 72 h. Data are presented as the means ± standard deviations (SDs). b The SI (selectivity index) was calculated using the formula: SI = (IC 50 for HLF)/(IC 50 for A549). Overall, the anticancer activities and cellular uptake efficacy ( Fig. S37 ) of Ru1 – Ru9 are not completely positively related to their lipophilicity, represented as Log P o/w ( Table S3 ). Compared with the other complexes, Ru5 shows better water solubility, a higher cellular uptake level and relatively higher antiproliferative activity, which may be due to its suitable lipophilicity and its capability to form hydrogen bonds with the targeted molecule through the imidazole ring. Ru5 also shows a better selectivity for cancer cells, and its SI for cancerous A549 over normal HLF cells is 16.1 (Table  1 and Table S4 ), which makes Ru5 a potential candidate for further investigation. Verification of ATP5F1 as the molecular target of Ru5 by PA-BPP PA-BPP was then used to confirm the molecular targets of Ru5 by modifying it with a ‘minimalist’ biorthogonal handle-containing photo-cross-linker. Through our structure–activity relationship analysis and comparison with other ruthenium complexes utilizing phen as the auxiliary ligands with similar coordination structures (using cisplatin as a control) [ 43 , 47 , 48 ], we propose that the β-carboline ligands play an important role in antiproliferative activity. To maintain the binding characteristics of the β-carboline moiety with the targeted biomolecules, we linked the photo-cross-linker to one of the auxiliary ligands of Ru5 . Methylthiazolyldiphenyl-tetrazolium bromide (MTT) assay proves that Ru5a possesses similar antiproliferative activity to Ru5 (Table  1 ). Cell lysates of Ru5a -treated A549 cells were clicked onto the azide-conjugated biotin by UV irradiation, and the targeted proteins were enriched using the streptavidin-modified agarose resin. The samples treated with vehicle (1% DMSO) were used as controls. In A549 cells pretreated with Ru5 (a competitor), the Stable Isotope Labeling with Amino Acids in Cell Culture (SILAC)-based chemo-proteomic pull-down assay for Ru5a reveals no significant protein enrichment ( Fig. S38 ), which indicates that Ru5 and Ru5a share the same cellular targets. The peptides after trypsin digestion were isotopically labeled by reductive demethylation for liquid chromatography-mass spectrometry (LC-MS)/MS analysis. A total of 816 proteins are detected from Ru5a -treated and control samples, among which 13 proteins (Fig.  1A , Table S5 ) are differentially enriched in Ru5a -treated samples ( P  ≤ 0.05, log 2 (fold change) ≥ 1). Figure 1. Validation of ATP5F1 as the molecular target of Ru5 . (A) Proteins pulled down by streptavidin from A549 cells treated with Ru5a followed by PAL. The cells were treated with Ru5a (50 μM) for 12 h and irradiated with UV light for 20 min. (B) The top five peptides enriched by Ru5a -mediated PAL were identified with the assistance of SILAC-based chemoproteomic studies. H1F0: H1 histone family member 0; HIST4H4: histone cluster 4 H4; HIST3H2BB: histone cluster 3 H2B family member b; ATP5F1E: ATP synthase F1 subunit epsilon; HIST1H2AJ: histone cluster 1 H2A family member J. (C) Pull-down western blotting analysis of ATPase in A549 cells treated with Ru5a . (D) Cellular Ru content in cells treated with Ru5 / Ru5a (20 μM, 12 h) measured by ICP-MS. (E) Impact of Ru5 / Ru5a (10 μM) on the thermal stability of ATP5F1 measured by CETSA. (F) Impact on the anticancer activities of Ru5 / Ru5a against wild type and ATPase knocking down (by siRNA) A549 cells. The relative cellular viability was calculated using the cells without drug treatment in each group as the control group. (G) Inhibition of Ru5 / Ru5a on the enzymatic activity of ATPase. (-)-blebbistatin was used as the positive control. (H) Molecular docking of Ru5 with ATPase (PDB: 3ZIA). The red dashes indicate the hydrogen bonds formed between Ru5 and the amino acid residues (SER38 and GLN39). The yellow dashes indicate the hydrophobic interaction between Ru5 and the amino acid residues (ASN15 and GLN19). Data are presented as means ± SD from three independent experiments. The top five enriched peptides include H1F0 (H1.0 linker histone), HIST4H4 (histone cluster 4), HIST3H2BB (histone cluster 3 H2B family member b), ATP5F1E (ATP synthase F1 subunit epsilon) and HIST1H2AJ (H2A clustered histone 14; Fig.  1B ). ATP5F1E belongs to the F1 subunit of ATPase, a protein mainly localized in mitochondria [ 49 ]. The other four peptides are components of histone in nuclei, which may be due to the relatively high expression levels of these proteins leading to a high background enrichment signal. Moreover, another peptide (ATP5F1A, ATP synthase F1 subunit alpha) in the F1 subunit of ATPase is also differentially enriched in Ru5a -treated samples. The pull-down assays followed by immunoblotting show that Ru5a can efficiently pull down ATP5F1 (Fig.  1C ). Inductively coupled plasma-mass spectrometry (ICP-MS) experiments show that Ru5 / Ru5a are mainly located in mitochondria (Fig.  1D ). We propose that Ru5 may bind to ATPase in cells. The cellular thermal shift assay (CETSA) also proves that the thermal stability of ATP5F1 is increased upon Ru5 / Ru5a treatment (Fig.  1E and Fig. S39 ). Moreover, compared to the wild-type cells, siRNA-mediated knockdown of ATPase alleviates the anticancer effects of Ru5 and Ru5a (Fig.  1F ), while ATPase knockdown using siRNA has no significant impact on A549 cell proliferation ( Fig. S40 ). Both Ru5 and Ru5a display a dose-dependent inhibition on the enzymatic activity of ATPase in vitro (Fig.  1G ), and the inhibition constant ( K i) measured for Ru5 (6.2 ± 1.4 μM) and Ru5a (8.1 ± 1.5 μM) is comparable to that of the positive control (-)-blebbistatin (3.3 ± 0.5 μM). The estimated inhibition constant ( K i = 9.22 μM; Table S6 ) is close to the K i value obtained from the enzymatic inhibition assay. Molecular docking using the crystal structures of ATP synthase F1 (PDB: 3ZIA) and Ru5 shows that Ru5 binds to the interspace between the γ and ε subunit, and two hydrogen atoms on the indole and imidazole rings of the β-carboline ligand of Ru5 can form hydrogen-bonds (red dashes) with the residues SER38 and GLN39 of ATPase F1 ε subunit, respectively (Fig.  1H ). Moreover, the benzene ring of the β-carboline ligand of Ru5 can form hydrophobic interaction (yellow dashes) with ASN15/GLN19 on ATPase F1 ε subunit. The roles of SER38 and GLU39 were further validated by substituting them with compact-sized glycine, hydrophobic alanine or steric phenylalanine. The results demonstrate that mutations in either SER38 or GLU39 lead to disruption of the hydrogen bonds between Ru5 ( Fig. S41 ) and ATPase and result in a decrease in the binding affinity between Ru5 and ATPase ( Table S6 ). The docking results of the other complexes with ATPase indicate that Ru1, Ru2, Ru3, Ru4 and Ru7 cannot form hydrogen bonds with the protein ( Fig. S42 ). Ru8 and Ru9 can each form a hydrogen bond with residues GLU179 and ASP169 via the indole imine group on L3, respectively. Ru6 can form two hydrogen bonds with residue THR40 via the indole and imidazole imine groups on L2. The calculation of the binding energy shows that Ru5 possesses the most favorable affinity towards ATPase ( Table S7 ). Based on these results, we infer that ATP synthase is an important cellular target of Ru5 , and the β-carboline ligand plays an important role in target binding. Ru5 induces mitochondrial dysfunction As Ru5 is located in mitochondria and inhibits ATPase related to mitochondrial respiration, we then evaluated its impact on oxygen consumption rate (OCR) using the Seahorse XF analyzer. Ru5 inhibits ATP-linked respiration, basal respiration and maximal respiration capacity, and it also increases the proton leak and alters non-mitochondrial respiration (Fig.  2A and Fig. S43 ). Accordingly, Ru5 treatment leads to an increase in the opening of the mitochondrial permeability transition pore (MPTP) of A549 in a concentration-dependent manner (Fig.  2B ). 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolyl-carbocyanine iodide (JC-1) staining assay confirms the loss of MMP (mitochondrial membrane potential) in Ru5 -treated A549 cells (Fig.  2C ). Moreover, Ru5 elevates reactive oxygen species (ROS) levels dose-dependently in cells (Fig.  2D ). The colocalization of the fluorescence of dihydrorhodamine 123 (DHR-123) with MitoTracker Deep Red (MTDR) indicates ROS are mainly generated in mitochondria (Fig.  2E ). Figure 2. Ru5 causes mitochondrial dysfunction. (A) Impact of Ru5 (5 μM) on mitochondrial respiration of A549 cells measured by Seahorse. The OCR was measured under basal conditions, and after the sequential addition of oligomycin (1 μM), carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP, 0.8 μM) and a mixture of rotenone (0.5 μM) and antimycin A (0.5 μM, R&A). Data are presented as means ± SD from three independent experiments. (B) The impact of Ru5 on MPTP. The cells were treated with Ru5 for 6 h and analyzed by flow cytometry. λ ex = 488 nm, λ em  = 520 ± 20 nm. (C) Impact of Ru5 (5 μM) on MMP measured by JC-1 staining. The cells were treated with Ru5 for 6 h or 12 h, stained with JC-1 for 10 min, and analyzed by flow cytometry. λ ex = 488 nm, λ em  = 525 ± 30 nm (JC-1 monomers) or 585 ± 30 nm (JC-1 aggregates). (D) Effect of Ru5 on cellular ROS level. The cells were stained with 2',7'-dichlorodihydrofluorescein diacetate (DCFH-DA) and treated with Ru5 (5 μM). DCF: λ ex = 488 nm, λ em  = 520 ± 30 nm. (E) Impact of Ru5 on mitochondrial ROS generation. A549 cells were treated with Ru5 (10 μM) and stained with DHR-123 (5 μM) and MTDR (100 nM). Ru5 : λ ex = 488 nm; λ em  = 600 ± 20 nm. DHR-123: λ ex = 514 nm; λ em  = 540 ± 20 nm. MTDR: λ ex = 633 nm; λ em  = 700 ± 20 nm. Scale bar: 20 μm. Combined analysis of mitochondrial proteomics and RNA-sequencing Given Ru5 is mainly localized in mitochondria as an ATPase inhibitor, mitochondrial proteomics is used to further investigate its action mechanisms (Fig.  3A ). Of the 556 proteins identified ( Supplementary Data S1 ), 20 proteins (up-regulated: 1; down-regulated: 19) are found to be differentially accumulated proteins (DAPs; |fold change| ≥ 1.8, false discovery rate (FDR) ≤ 0.01; Fig.  3B and Table S8 ). These DAPs are mainly involved in ion channels, lipid oxidation and intracellular redox balance. The top five down-regulated proteins are GPX4 (glutathione peroxidase 4), CLIC4 (chloride intracellular channel 4), CLIC1 (chloride intracellular channel 1), HS1BP3 (heterochromatin protein 1 binding protein 3) and AGPS (alkylglycerone phosphate synthase, Fig.  3C ). CLIC4 and CLIC1 are intracellular chloride channel proteins mediating Cl − influx through 2Cl − /H + exchange [ 50 ]. As the counterion, Cl − provides the electrical shunt for proton pumping by ATPase [ 51 ]. Therefore, the decreased accumulation of CLIC4 and CLIC1 may be caused by ATPase inhibition [ 52 ]. Interestingly, GPX4, GPX1 and APEX1 are antioxidant selenoenzymes involved in ferroptosis, especially for GPX4 [ 53 ]. ALAS1 is involved in heme synthesis, and ALAS1 inhibition can impair iron utilization, resulting in iron overload and ferroptosis [ 54 ]. Intriguingly, AGPS (lipid biosynthesis), HSD17B4 (peroxisomal beta-oxidation pathway for fatty acids), GNPAT (lipid biosynthesis and metabolism), ACAT2 (lipid metabolism) and ACOX1 (fatty acid beta-oxidation) are all related to lipid metabolism/oxidation, which is consistent with the fact that ferroptosis is driven by lipid peroxidation [ 55 ]. HS1BP3 (formation of LC3-positive autophagosomes), Rab35 (membrane trafficking) and BAG5 (mitophagy regulating) are involved in the regulation of autophagy [ 56–58 ]. The reported functions of MAP2K2 (energetic metabolism) [ 59 ], CHCHD2 (hypoxia) [ 60 ] and PARP1 (DNA repair) [ 61 ] are all closely associated with mitochondrial functionalities. Finally, the only up-regulated protein, CYP24A1, is a cytochrome P450 oxidoreductase, which is reported to influence phospholipid peroxidation during ferroptosis [ 62 ]. Figure 3. Impact of Ru5 on mitochondrial proteomics and whole-cell transcriptome. (A) Schematic overview of the workflow of mitochondrial proteomics. (B) The volcano plot shows the DAPs in the mitochondria of A549 cells treated with Ru5 (10 μM, 12 h) compared with the control samples. (C) The relative expression of differentially expressed proteins (DEPs) identified by mitochondrial proteomic analysis. Data are presented as means ± SD from three independent experiments. (D) KEGG analysis of DEGs identified by RNA-seq of A549 cells after Ru5 (10 μM, 12 h) treatment. (E) GSEA analysis of the DEG genes in Ru5 -treated groups in the mitophagy pathway, ferroptosis and epithelial-mesenchymal transition pathway. RNA-seq was further performed to clarify the antitumor mechanisms of Ru5 . An average of 95.96% mappability and 46.1 million qualified fragments for each RNA-seq sample are obtained ( Table S9 ). High correlations ( R  > 0.95, Fig. S44 ) representing the reproducibility are obtained for parallel samples. As compared with the control samples, the total number of DEGs (differential expression genes; |Fold change| ≥ 2.0; FDR ≤ 0.05; Supplementary Data S2 ) for Ru5 -treated samples is 1517 (up-regulated: 673; down-regulated: 844). GO (gene ontology) term analysis shows that Ru5 treatment influences the cellular process, cell part and binding ( Fig. S45 ). KEGG (Kyoto Encyclopedia of Genes and Genomes) analysis shows that Ru5 influences pathways including PI3K-AKT (phosphoinositide 3-kinase and AKT), MAPK (mitogen-activated protein kinase), Rap1 (Ras-association proximate 1), FoxO (forkhead box O), p53 (tumor suppressor p53), TGF-β (transforming growth factor-β), biosynthesis of unsaturated fatty acids and steroid biosynthesis (Fig.  3D ). Gene set enrichment analysis (GSEA) shows that cell death induced by Ru5 is positively associated with autophagy of mitochondrion and ferroptosis. Besides, the GSEA analysis also shows the up-regulation of the mitophagy pathway and ferroptosis pathway in Ru5 -treated cells (Fig.  3E ). Moreover, gene signatures related to EMT are down-regulated in Ru5 -treated cells (Fig.  3E ). In all, these data show that Ru5 can mainly influence pathways related to mitochondrial redox homeostasis, ferroptosis, mitophagy and EMT. Ru5 induces mitophagy and ferroptosis, and inhibits EMT in vitro As shown by transmission electron microscopy (TEM) observation, in cells treated with Ru5 , some mitochondria are enclosed by a bilayer membrane with increased density, and many of them are wrapped in double-layered membrane structures (Fig.  4A ). At the same time, the morphology of the nucleus is unaffected. These features are typical morphological characteristics of autophagy and ferroptosis [ 63 ]. In A549 cells co-labeled with MitoTracker Green (MTG) and LysoTracker Deep Red (LTDR), Ru5 treatment causes a gradual morphological alternation in mitochondria from fibrous to dot-like. Prolonged incubation leads to a time-dependent overlap of the emission of MTG and LTDR (Fig.  4B and Fig. S46 ), which confirms the occurrence of mitophagy. In cells treated with Ru5 , western blotting shows the conversion of LC3I to LC3II, and an increased expression of PINK1 (PTEN-induced putative kinase 1) and Parkin, along with a decreased expression of GPX4 (Fig.  4C ). Figure 4. Ru5 induces mitophagy and ferroptosis and inhibits EMT in vitro . (A) TEM images of A549 cells treated with vehicle (1% DMSO) or Ru5 (10 μM, 12 h). Red rectangles represent the region enlarged. Scale bar: 5 μm and 500 nm (enlarged). (B) The morphological examination of mitochondria in A549 cells treated with Ru5 (10 μM) for the indicated time intervals. MTG: λ ex = 514 nm; λ em  = 540 ± 20 nm. LTDR: λ ex = 633 nm; λ em  = 700 ± 20 nm. Scale bar: 20 μm. (C) Western blotting analysis showing the impact of Ru5 (24 h) on the expression of proteins related to mitochondrial dysfunction and ferroptosis. (D and E) The impact of different inhibitors on antiproliferative activity of Ru5 (24 h). (F) RT-qPCR analysis of EMT-related genes in A549 cells after Ru5 (10 μM, 12 h) treatment. (G) The impact of Ru5 (5 μM) on the migration of A549 cells was measured by a wound healing assay. Scale bar: 100 μm. (H) Quantitative analysis of the impact of Ru5 on wound healing. (I) Selective images of the impact of Ru5 (5 μM, 48 h) on the migration of A549 cells detected by transwell assay. Scale bar: 100 μm. Data are presented as means ± SD from three independent experiments. * P  ≤ 0.05, ** P  ≤ 0.01, *** P  ≤ 0.001, by Student's t-test. We then used different specific inhibitors to further confirm the modes of cell death induced by Ru5 . These inhibitors act on autophagy (3-methyladenine, 3-MA, Fig.  4D ) [ 64 ], ferroptosis (liproxstatin-1, Lip-1, Fig.  4E ) [ 65 ], necroptosis (necrostatin-1, Nec-1, Fig. S47A ) [ 66 ] and apoptosis (Z-VAD-FMK, Fig. S47B ) [ 67 ]. The effect of 3-MA on the antiproliferative activity of Ru5 is concentration-dependent, with lower and higher concentrations enhancing and inhibiting the antiproliferative activity of Ru5 , respectively. Thus, Ru5 induces cytoprotective autophagy at lower concentrations, while autophagic cell death is induced at higher concentrations. As expected, Lip-1 can relieve the antiproliferative activity of Ru5 . However, Nec-1 and Z-VAD-FMK show no effect on the cell viability of Ru5 -treated cells. All these results show that inhibition of ATPase by Ru5 causes mitochondrial dysfunction and ROS accumulation, leading to mitophagy and ferroptosis. Based on the GSEA and KEGG analysis, we then evaluate the impact of Ru5 on EMT and angiogenesis. Firstly, real-time quantitative polymerase chain reaction (RT-qPCR) experiments (Fig.  4F ) confirm that Ru5 can down-regulate genes including CDH1 (cadherin 1) [ 68 ], FN1 (fibronectin 1) [ 69 ] and MMP2 (matrix metallopeptidase 2) [ 70 ]. Wound healing assays (Fig.  4G and H ) and transwell migration assays (Fig.  4I ) verify that Ru5 can markedly decrease the migration rates of the A549 cells. All these results indicate that Ru5 can induce mitophagy and ferroptosis, and also simultaneously affects tumor invasive and metastatic processes including EMT and angiogenesis by inhibiting ATPase. Ru5 exhibits potent anticancer activities in vivo Lastly, the in vivo anticancer activity of Ru5 was assessed via a A549 xenograft nude mice model (Fig.  5A ). Mice were treated with PBS, cisplatin (5 mg kg −1 ) or Ru5 (5 mg kg −1 ) for 14 days, during which body weights and tumor volumes were monitored. Ru5 showed better tumor inhibition than cisplatin and PBS (Fig.  5B–D ; Fig. S48 ). Compared with the control groups, the tumor growth inhibitory rates of Ru5 and cisplatin reached ∼89.6% and ∼74.4%, respectively. Moreover, no significant body weight loss (Fig.  5E ) or organ damage (Fig.  5F ) was observed during therapy for Ru5 . All these results show that Ru5 has great potential to be developed as an anticancer candidate. Figure 5. Ru5 exhibits potent anticancer activities in vivo . (A) Schematic illustration of in vivo therapeutic protocol. (B) Graphs of tumor volumes of nude mice after treatment with Ru5 (5 mg kg −1 ), cisplatin (5 mg kg −1 ) and PBS. The intratumoral injections were performed every 7 days. Data are presented as means ± SD, n  = 5. (C) Immunohistochemical analysis of hematoxylin-eosin (H&E) staining in tumor sections from mice with different treatments. Scale bar: 50 μm. (D) Tumors separated from nude mice. (E) The average body weight of the mice in each group during the treatment process. Data are presented as means ± SD, n  = 5. (F) H&E staining of the main organs from mice with different treatments. Scale bar: 100 μm. ## Synthesis and structure–activity relationship study Synthesis and structure–activity relationship study Based on our previous work [ 43 ], we designed nine Ru(II) polypyridyl complexes by changing the auxiliary ligands (2,2'-bipyridine (bpy; Ru1, Ru4 and Ru7 ); 1,10-phenanthroline (phen; Ru2, Ru5 and Ru8 ); 4,7-diphenyl-1,10-phenanthroline (dip; Ru3, Ru6 and Ru9 )) to adjust the lipophilicity and the main β-carboline ligands (1-(quinolin-2-yl)-9H-pyrido[3,4-b]indole ( L1; Ru1 – Ru3 ); 1-(1H-imidazol-2-yl)-9H-pyrido[3,4-b]indole ( L2; Ru4 – Ru6 ); 1-(1H-benzo[d]imidazol-2-yl)-9H-pyrido[3,4-b]indole ( L3; Ru7 – Ru9 )) to modify the binding ability with the targeted biomolecules. The β-carboline ligands are varied by keeping the core structure and introducing different pharmacophores (quinoline, imidazole and benzimidazole) to the C1-position. L1 – L3 were obtained by reacting tryptamine with the corresponding aldehyde in dry anisole as we previously reported [ 44 , 45 ]. Ru1 – Ru9 were synthesized by refluxing 1 equiv of the β-carboline ligand and the corresponding precursor cis -[Ru(N–N) 2 Cl 2 ] followed by anion exchange with NH 4 PF 6 . The complexes were characterized by 1 H nuclear magnetic resonance (NMR), 13 C NMR spectroscopy, electrospray ionization mass spectrometry (ESI-MS) and elemental analysis (Supplementary Data, Figs S1 – S34 ). X-ray diffraction shows that Ru5 adopts a typical octahedral configuration ( Fig. S35 , Tables S1 and S2 ). Subsequently, the lipophilicity and aqueous solubility of Ru1 – Ru9 were investigated. The results indicate that the Log P o/w (oil-water partition coefficient) values of Ru1 – Ru9 range between −2.01 and 3.59 ( Table S3 ). Ru3, Ru5 and Ru7 exhibit suitable Log P o/w that can ensure good solubility and bioavailability [ 46 ]. By evaluating the linear relationship between concentration and absorbance, we found that the solubility of Ru1 – Ru9 in phosphate buffered saline (PBS) (with 1% dimethyl sulfoxide (DMSO)) was almost positively correlated with their lipophilicity. A linear relationship between the absorbance and the concentration is obtained for Ru1, Ru2, Ru4, Ru5, Ru7 and Ru8 up to at least 100 μM. The highest solubility of Ru3, Ru6 and Ru9 with dip as auxiliary ligands is lower than 40 μM under the same conditions ( Fig. S36 ). The antiproliferative activities of Ru1 – Ru9 were first evaluated on human cell lines including cervical carcinoma (HeLa), lung adenocarcinoma (A549), cisplatin-resistant A549R, prostate cancer (PC3), triple negative breast cancer (MDA-MB-231), malignant glioma (U87) and human lung fibroblast-like (HLF) cells (Table  1 and Table S4 ). The complexes containing β-carboline ligands with imidazole ( Ru4 – Ru6 ) and benzimidazole ( Ru7 – Ru9 ) as substituents have higher activity than those with quinoline ( Ru1 – Ru3 ) as substituents. Ru3 – Ru9 display relatively high anticancer activities towards the cancer cell lines tested and maintain good activity against cisplatin-resistant A549R cells. Ru3 – Ru9 also show a certain selectivity for cancer cells over normal cells, and Ru2, Ru3, Ru5 and Ru9 show higher selectivity index (SI) values than the other complexes. Table 1. Half maximal inhibitory concentration (IC 50 , μM, 72 h) values of tested compounds towards different cell lines. a Complexes HeLa A549 A549R PC3 MDA-MB-231 U87 HLF SI b Ru5 3.1 ± 0.7 2.1 ± 0.8 3.3 ± 0.7 2.7 ± 0.4 7.7 ± 0.2 14.7 ± 0.7 33.9 ± 0.2 16.1 Ru5a 3.8 ± 0.1 2.5 ± 0.4 4.2 ± 0.6 3.4 ± 0.6 8.4 ± 0.6 13.4 ± 0.4 23.2 ± 0.3 9.3 Cisplatin 10.3 ± 1.3 5.9 ± 1.0 28.1 ± 4.2 13.1 ± 2.7 9.0 ± 0.2 14.8 ± 1.2 15.4 ± 1.1 2.6 a Cells were incubated with the compounds for 72 h. Data are presented as the means ± standard deviations (SDs). b The SI (selectivity index) was calculated using the formula: SI = (IC 50 for HLF)/(IC 50 for A549). Overall, the anticancer activities and cellular uptake efficacy ( Fig. S37 ) of Ru1 – Ru9 are not completely positively related to their lipophilicity, represented as Log P o/w ( Table S3 ). Compared with the other complexes, Ru5 shows better water solubility, a higher cellular uptake level and relatively higher antiproliferative activity, which may be due to its suitable lipophilicity and its capability to form hydrogen bonds with the targeted molecule through the imidazole ring. Ru5 also shows a better selectivity for cancer cells, and its SI for cancerous A549 over normal HLF cells is 16.1 (Table  1 and Table S4 ), which makes Ru5 a potential candidate for further investigation. ## Verification of ATP5F1 as the molecular target of Ru5 by PA-BPP Verification of ATP5F1 as the molecular target of Ru5 by PA-BPP PA-BPP was then used to confirm the molecular targets of Ru5 by modifying it with a ‘minimalist’ biorthogonal handle-containing photo-cross-linker. Through our structure–activity relationship analysis and comparison with other ruthenium complexes utilizing phen as the auxiliary ligands with similar coordination structures (using cisplatin as a control) [ 43 , 47 , 48 ], we propose that the β-carboline ligands play an important role in antiproliferative activity. To maintain the binding characteristics of the β-carboline moiety with the targeted biomolecules, we linked the photo-cross-linker to one of the auxiliary ligands of Ru5 . Methylthiazolyldiphenyl-tetrazolium bromide (MTT) assay proves that Ru5a possesses similar antiproliferative activity to Ru5 (Table  1 ). Cell lysates of Ru5a -treated A549 cells were clicked onto the azide-conjugated biotin by UV irradiation, and the targeted proteins were enriched using the streptavidin-modified agarose resin. The samples treated with vehicle (1% DMSO) were used as controls. In A549 cells pretreated with Ru5 (a competitor), the Stable Isotope Labeling with Amino Acids in Cell Culture (SILAC)-based chemo-proteomic pull-down assay for Ru5a reveals no significant protein enrichment ( Fig. S38 ), which indicates that Ru5 and Ru5a share the same cellular targets. The peptides after trypsin digestion were isotopically labeled by reductive demethylation for liquid chromatography-mass spectrometry (LC-MS)/MS analysis. A total of 816 proteins are detected from Ru5a -treated and control samples, among which 13 proteins (Fig.  1A , Table S5 ) are differentially enriched in Ru5a -treated samples ( P  ≤ 0.05, log 2 (fold change) ≥ 1). Figure 1. Validation of ATP5F1 as the molecular target of Ru5 . (A) Proteins pulled down by streptavidin from A549 cells treated with Ru5a followed by PAL. The cells were treated with Ru5a (50 μM) for 12 h and irradiated with UV light for 20 min. (B) The top five peptides enriched by Ru5a -mediated PAL were identified with the assistance of SILAC-based chemoproteomic studies. H1F0: H1 histone family member 0; HIST4H4: histone cluster 4 H4; HIST3H2BB: histone cluster 3 H2B family member b; ATP5F1E: ATP synthase F1 subunit epsilon; HIST1H2AJ: histone cluster 1 H2A family member J. (C) Pull-down western blotting analysis of ATPase in A549 cells treated with Ru5a . (D) Cellular Ru content in cells treated with Ru5 / Ru5a (20 μM, 12 h) measured by ICP-MS. (E) Impact of Ru5 / Ru5a (10 μM) on the thermal stability of ATP5F1 measured by CETSA. (F) Impact on the anticancer activities of Ru5 / Ru5a against wild type and ATPase knocking down (by siRNA) A549 cells. The relative cellular viability was calculated using the cells without drug treatment in each group as the control group. (G) Inhibition of Ru5 / Ru5a on the enzymatic activity of ATPase. (-)-blebbistatin was used as the positive control. (H) Molecular docking of Ru5 with ATPase (PDB: 3ZIA). The red dashes indicate the hydrogen bonds formed between Ru5 and the amino acid residues (SER38 and GLN39). The yellow dashes indicate the hydrophobic interaction between Ru5 and the amino acid residues (ASN15 and GLN19). Data are presented as means ± SD from three independent experiments. The top five enriched peptides include H1F0 (H1.0 linker histone), HIST4H4 (histone cluster 4), HIST3H2BB (histone cluster 3 H2B family member b), ATP5F1E (ATP synthase F1 subunit epsilon) and HIST1H2AJ (H2A clustered histone 14; Fig.  1B ). ATP5F1E belongs to the F1 subunit of ATPase, a protein mainly localized in mitochondria [ 49 ]. The other four peptides are components of histone in nuclei, which may be due to the relatively high expression levels of these proteins leading to a high background enrichment signal. Moreover, another peptide (ATP5F1A, ATP synthase F1 subunit alpha) in the F1 subunit of ATPase is also differentially enriched in Ru5a -treated samples. The pull-down assays followed by immunoblotting show that Ru5a can efficiently pull down ATP5F1 (Fig.  1C ). Inductively coupled plasma-mass spectrometry (ICP-MS) experiments show that Ru5 / Ru5a are mainly located in mitochondria (Fig.  1D ). We propose that Ru5 may bind to ATPase in cells. The cellular thermal shift assay (CETSA) also proves that the thermal stability of ATP5F1 is increased upon Ru5 / Ru5a treatment (Fig.  1E and Fig. S39 ). Moreover, compared to the wild-type cells, siRNA-mediated knockdown of ATPase alleviates the anticancer effects of Ru5 and Ru5a (Fig.  1F ), while ATPase knockdown using siRNA has no significant impact on A549 cell proliferation ( Fig. S40 ). Both Ru5 and Ru5a display a dose-dependent inhibition on the enzymatic activity of ATPase in vitro (Fig.  1G ), and the inhibition constant ( K i) measured for Ru5 (6.2 ± 1.4 μM) and Ru5a (8.1 ± 1.5 μM) is comparable to that of the positive control (-)-blebbistatin (3.3 ± 0.5 μM). The estimated inhibition constant ( K i = 9.22 μM; Table S6 ) is close to the K i value obtained from the enzymatic inhibition assay. Molecular docking using the crystal structures of ATP synthase F1 (PDB: 3ZIA) and Ru5 shows that Ru5 binds to the interspace between the γ and ε subunit, and two hydrogen atoms on the indole and imidazole rings of the β-carboline ligand of Ru5 can form hydrogen-bonds (red dashes) with the residues SER38 and GLN39 of ATPase F1 ε subunit, respectively (Fig.  1H ). Moreover, the benzene ring of the β-carboline ligand of Ru5 can form hydrophobic interaction (yellow dashes) with ASN15/GLN19 on ATPase F1 ε subunit. The roles of SER38 and GLU39 were further validated by substituting them with compact-sized glycine, hydrophobic alanine or steric phenylalanine. The results demonstrate that mutations in either SER38 or GLU39 lead to disruption of the hydrogen bonds between Ru5 ( Fig. S41 ) and ATPase and result in a decrease in the binding affinity between Ru5 and ATPase ( Table S6 ). The docking results of the other complexes with ATPase indicate that Ru1, Ru2, Ru3, Ru4 and Ru7 cannot form hydrogen bonds with the protein ( Fig. S42 ). Ru8 and Ru9 can each form a hydrogen bond with residues GLU179 and ASP169 via the indole imine group on L3, respectively. Ru6 can form two hydrogen bonds with residue THR40 via the indole and imidazole imine groups on L2. The calculation of the binding energy shows that Ru5 possesses the most favorable affinity towards ATPase ( Table S7 ). Based on these results, we infer that ATP synthase is an important cellular target of Ru5 , and the β-carboline ligand plays an important role in target binding. ## Ru5 induces mitochondrial dysfunction Ru5 induces mitochondrial dysfunction As Ru5 is located in mitochondria and inhibits ATPase related to mitochondrial respiration, we then evaluated its impact on oxygen consumption rate (OCR) using the Seahorse XF analyzer. Ru5 inhibits ATP-linked respiration, basal respiration and maximal respiration capacity, and it also increases the proton leak and alters non-mitochondrial respiration (Fig.  2A and Fig. S43 ). Accordingly, Ru5 treatment leads to an increase in the opening of the mitochondrial permeability transition pore (MPTP) of A549 in a concentration-dependent manner (Fig.  2B ). 5,5′,6,6′-tetrachloro-1,1′,3,3′-tetraethylbenzimidazolyl-carbocyanine iodide (JC-1) staining assay confirms the loss of MMP (mitochondrial membrane potential) in Ru5 -treated A549 cells (Fig.  2C ). Moreover, Ru5 elevates reactive oxygen species (ROS) levels dose-dependently in cells (Fig.  2D ). The colocalization of the fluorescence of dihydrorhodamine 123 (DHR-123) with MitoTracker Deep Red (MTDR) indicates ROS are mainly generated in mitochondria (Fig.  2E ). Figure 2. Ru5 causes mitochondrial dysfunction. (A) Impact of Ru5 (5 μM) on mitochondrial respiration of A549 cells measured by Seahorse. The OCR was measured under basal conditions, and after the sequential addition of oligomycin (1 μM), carbonyl cyanide 4-(trifluoromethoxy)phenylhydrazone (FCCP, 0.8 μM) and a mixture of rotenone (0.5 μM) and antimycin A (0.5 μM, R&A). Data are presented as means ± SD from three independent experiments. (B) The impact of Ru5 on MPTP. The cells were treated with Ru5 for 6 h and analyzed by flow cytometry. λ ex = 488 nm, λ em  = 520 ± 20 nm. (C) Impact of Ru5 (5 μM) on MMP measured by JC-1 staining. The cells were treated with Ru5 for 6 h or 12 h, stained with JC-1 for 10 min, and analyzed by flow cytometry. λ ex = 488 nm, λ em  = 525 ± 30 nm (JC-1 monomers) or 585 ± 30 nm (JC-1 aggregates). (D) Effect of Ru5 on cellular ROS level. The cells were stained with 2',7'-dichlorodihydrofluorescein diacetate (DCFH-DA) and treated with Ru5 (5 μM). DCF: λ ex = 488 nm, λ em  = 520 ± 30 nm. (E) Impact of Ru5 on mitochondrial ROS generation. A549 cells were treated with Ru5 (10 μM) and stained with DHR-123 (5 μM) and MTDR (100 nM). Ru5 : λ ex = 488 nm; λ em  = 600 ± 20 nm. DHR-123: λ ex = 514 nm; λ em  = 540 ± 20 nm. MTDR: λ ex = 633 nm; λ em  = 700 ± 20 nm. Scale bar: 20 μm. ## Combined analysis of mitochondrial proteomics and RNA-sequencing Combined analysis of mitochondrial proteomics and RNA-sequencing Given Ru5 is mainly localized in mitochondria as an ATPase inhibitor, mitochondrial proteomics is used to further investigate its action mechanisms (Fig.  3A ). Of the 556 proteins identified ( Supplementary Data S1 ), 20 proteins (up-regulated: 1; down-regulated: 19) are found to be differentially accumulated proteins (DAPs; |fold change| ≥ 1.8, false discovery rate (FDR) ≤ 0.01; Fig.  3B and Table S8 ). These DAPs are mainly involved in ion channels, lipid oxidation and intracellular redox balance. The top five down-regulated proteins are GPX4 (glutathione peroxidase 4), CLIC4 (chloride intracellular channel 4), CLIC1 (chloride intracellular channel 1), HS1BP3 (heterochromatin protein 1 binding protein 3) and AGPS (alkylglycerone phosphate synthase, Fig.  3C ). CLIC4 and CLIC1 are intracellular chloride channel proteins mediating Cl − influx through 2Cl − /H + exchange [ 50 ]. As the counterion, Cl − provides the electrical shunt for proton pumping by ATPase [ 51 ]. Therefore, the decreased accumulation of CLIC4 and CLIC1 may be caused by ATPase inhibition [ 52 ]. Interestingly, GPX4, GPX1 and APEX1 are antioxidant selenoenzymes involved in ferroptosis, especially for GPX4 [ 53 ]. ALAS1 is involved in heme synthesis, and ALAS1 inhibition can impair iron utilization, resulting in iron overload and ferroptosis [ 54 ]. Intriguingly, AGPS (lipid biosynthesis), HSD17B4 (peroxisomal beta-oxidation pathway for fatty acids), GNPAT (lipid biosynthesis and metabolism), ACAT2 (lipid metabolism) and ACOX1 (fatty acid beta-oxidation) are all related to lipid metabolism/oxidation, which is consistent with the fact that ferroptosis is driven by lipid peroxidation [ 55 ]. HS1BP3 (formation of LC3-positive autophagosomes), Rab35 (membrane trafficking) and BAG5 (mitophagy regulating) are involved in the regulation of autophagy [ 56–58 ]. The reported functions of MAP2K2 (energetic metabolism) [ 59 ], CHCHD2 (hypoxia) [ 60 ] and PARP1 (DNA repair) [ 61 ] are all closely associated with mitochondrial functionalities. Finally, the only up-regulated protein, CYP24A1, is a cytochrome P450 oxidoreductase, which is reported to influence phospholipid peroxidation during ferroptosis [ 62 ]. Figure 3. Impact of Ru5 on mitochondrial proteomics and whole-cell transcriptome. (A) Schematic overview of the workflow of mitochondrial proteomics. (B) The volcano plot shows the DAPs in the mitochondria of A549 cells treated with Ru5 (10 μM, 12 h) compared with the control samples. (C) The relative expression of differentially expressed proteins (DEPs) identified by mitochondrial proteomic analysis. Data are presented as means ± SD from three independent experiments. (D) KEGG analysis of DEGs identified by RNA-seq of A549 cells after Ru5 (10 μM, 12 h) treatment. (E) GSEA analysis of the DEG genes in Ru5 -treated groups in the mitophagy pathway, ferroptosis and epithelial-mesenchymal transition pathway. RNA-seq was further performed to clarify the antitumor mechanisms of Ru5 . An average of 95.96% mappability and 46.1 million qualified fragments for each RNA-seq sample are obtained ( Table S9 ). High correlations ( R  > 0.95, Fig. S44 ) representing the reproducibility are obtained for parallel samples. As compared with the control samples, the total number of DEGs (differential expression genes; |Fold change| ≥ 2.0; FDR ≤ 0.05; Supplementary Data S2 ) for Ru5 -treated samples is 1517 (up-regulated: 673; down-regulated: 844). GO (gene ontology) term analysis shows that Ru5 treatment influences the cellular process, cell part and binding ( Fig. S45 ). KEGG (Kyoto Encyclopedia of Genes and Genomes) analysis shows that Ru5 influences pathways including PI3K-AKT (phosphoinositide 3-kinase and AKT), MAPK (mitogen-activated protein kinase), Rap1 (Ras-association proximate 1), FoxO (forkhead box O), p53 (tumor suppressor p53), TGF-β (transforming growth factor-β), biosynthesis of unsaturated fatty acids and steroid biosynthesis (Fig.  3D ). Gene set enrichment analysis (GSEA) shows that cell death induced by Ru5 is positively associated with autophagy of mitochondrion and ferroptosis. Besides, the GSEA analysis also shows the up-regulation of the mitophagy pathway and ferroptosis pathway in Ru5 -treated cells (Fig.  3E ). Moreover, gene signatures related to EMT are down-regulated in Ru5 -treated cells (Fig.  3E ). In all, these data show that Ru5 can mainly influence pathways related to mitochondrial redox homeostasis, ferroptosis, mitophagy and EMT. ## Ru5 induces mitophagy and ferroptosis, and inhibits EMT Ru5 induces mitophagy and ferroptosis, and inhibits EMT in vitro As shown by transmission electron microscopy (TEM) observation, in cells treated with Ru5 , some mitochondria are enclosed by a bilayer membrane with increased density, and many of them are wrapped in double-layered membrane structures (Fig.  4A ). At the same time, the morphology of the nucleus is unaffected. These features are typical morphological characteristics of autophagy and ferroptosis [ 63 ]. In A549 cells co-labeled with MitoTracker Green (MTG) and LysoTracker Deep Red (LTDR), Ru5 treatment causes a gradual morphological alternation in mitochondria from fibrous to dot-like. Prolonged incubation leads to a time-dependent overlap of the emission of MTG and LTDR (Fig.  4B and Fig. S46 ), which confirms the occurrence of mitophagy. In cells treated with Ru5 , western blotting shows the conversion of LC3I to LC3II, and an increased expression of PINK1 (PTEN-induced putative kinase 1) and Parkin, along with a decreased expression of GPX4 (Fig.  4C ). Figure 4. Ru5 induces mitophagy and ferroptosis and inhibits EMT in vitro . (A) TEM images of A549 cells treated with vehicle (1% DMSO) or Ru5 (10 μM, 12 h). Red rectangles represent the region enlarged. Scale bar: 5 μm and 500 nm (enlarged). (B) The morphological examination of mitochondria in A549 cells treated with Ru5 (10 μM) for the indicated time intervals. MTG: λ ex = 514 nm; λ em  = 540 ± 20 nm. LTDR: λ ex = 633 nm; λ em  = 700 ± 20 nm. Scale bar: 20 μm. (C) Western blotting analysis showing the impact of Ru5 (24 h) on the expression of proteins related to mitochondrial dysfunction and ferroptosis. (D and E) The impact of different inhibitors on antiproliferative activity of Ru5 (24 h). (F) RT-qPCR analysis of EMT-related genes in A549 cells after Ru5 (10 μM, 12 h) treatment. (G) The impact of Ru5 (5 μM) on the migration of A549 cells was measured by a wound healing assay. Scale bar: 100 μm. (H) Quantitative analysis of the impact of Ru5 on wound healing. (I) Selective images of the impact of Ru5 (5 μM, 48 h) on the migration of A549 cells detected by transwell assay. Scale bar: 100 μm. Data are presented as means ± SD from three independent experiments. * P  ≤ 0.05, ** P  ≤ 0.01, *** P  ≤ 0.001, by Student's t-test. We then used different specific inhibitors to further confirm the modes of cell death induced by Ru5 . These inhibitors act on autophagy (3-methyladenine, 3-MA, Fig.  4D ) [ 64 ], ferroptosis (liproxstatin-1, Lip-1, Fig.  4E ) [ 65 ], necroptosis (necrostatin-1, Nec-1, Fig. S47A ) [ 66 ] and apoptosis (Z-VAD-FMK, Fig. S47B ) [ 67 ]. The effect of 3-MA on the antiproliferative activity of Ru5 is concentration-dependent, with lower and higher concentrations enhancing and inhibiting the antiproliferative activity of Ru5 , respectively. Thus, Ru5 induces cytoprotective autophagy at lower concentrations, while autophagic cell death is induced at higher concentrations. As expected, Lip-1 can relieve the antiproliferative activity of Ru5 . However, Nec-1 and Z-VAD-FMK show no effect on the cell viability of Ru5 -treated cells. All these results show that inhibition of ATPase by Ru5 causes mitochondrial dysfunction and ROS accumulation, leading to mitophagy and ferroptosis. Based on the GSEA and KEGG analysis, we then evaluate the impact of Ru5 on EMT and angiogenesis. Firstly, real-time quantitative polymerase chain reaction (RT-qPCR) experiments (Fig.  4F ) confirm that Ru5 can down-regulate genes including CDH1 (cadherin 1) [ 68 ], FN1 (fibronectin 1) [ 69 ] and MMP2 (matrix metallopeptidase 2) [ 70 ]. Wound healing assays (Fig.  4G and H ) and transwell migration assays (Fig.  4I ) verify that Ru5 can markedly decrease the migration rates of the A549 cells. All these results indicate that Ru5 can induce mitophagy and ferroptosis, and also simultaneously affects tumor invasive and metastatic processes including EMT and angiogenesis by inhibiting ATPase. ## Ru5 exhibits potent anticancer activities Ru5 exhibits potent anticancer activities in vivo Lastly, the in vivo anticancer activity of Ru5 was assessed via a A549 xenograft nude mice model (Fig.  5A ). Mice were treated with PBS, cisplatin (5 mg kg −1 ) or Ru5 (5 mg kg −1 ) for 14 days, during which body weights and tumor volumes were monitored. Ru5 showed better tumor inhibition than cisplatin and PBS (Fig.  5B–D ; Fig. S48 ). Compared with the control groups, the tumor growth inhibitory rates of Ru5 and cisplatin reached ∼89.6% and ∼74.4%, respectively. Moreover, no significant body weight loss (Fig.  5E ) or organ damage (Fig.  5F ) was observed during therapy for Ru5 . All these results show that Ru5 has great potential to be developed as an anticancer candidate. Figure 5. Ru5 exhibits potent anticancer activities in vivo . (A) Schematic illustration of in vivo therapeutic protocol. (B) Graphs of tumor volumes of nude mice after treatment with Ru5 (5 mg kg −1 ), cisplatin (5 mg kg −1 ) and PBS. The intratumoral injections were performed every 7 days. Data are presented as means ± SD, n  = 5. (C) Immunohistochemical analysis of hematoxylin-eosin (H&E) staining in tumor sections from mice with different treatments. Scale bar: 50 μm. (D) Tumors separated from nude mice. (E) The average body weight of the mice in each group during the treatment process. Data are presented as means ± SD, n  = 5. (F) H&E staining of the main organs from mice with different treatments. Scale bar: 100 μm. ## CONCLUSION CONCLUSION Metallo-anticancer agents have gained significant attention in recent years due to their potential to overcome chemotherapy resistance and target specific pathways in cancer cells. In this work, we report nine Ru(II) polypyridyl complexes containing β-carboline alkaloid derivatives as ligands, among which Ru5 is identified as a potent anticancer candidate. PA-BPP confirms that mitochondrial ATPase is one of the most important targets of Ru5 . Using the combined analysis of mitochondrial proteomics, RNA-seq and cell-based validations, we show that Ru5 can induce mitochondrial dysfunction, mitophagy and ferroptosis. Moreover, Ru5 can inhibit EMT and angiogenesis. The findings shed light on the complex interplay between mitochondrial function, EMT and ferroptosis in cancer cells. Finally, in vivo experiments demonstrate that Ru5 possesses a better antitumor effect than cisplatin. In conclusion, we have identified the molecular targets of Ru(II) polypyridyl complexes by coupling PAL with a multiomics approach, which provides an innovative strategy to elucidate the anticancer mechanisms of metallo-anticancer agents. ## METHODS METHODS Detailed information about the materials and experimental procedures can be found in the supplementary data. All animal operations followed the guidelines of the Sun Yat-Sen University Animal Care and Use Committee. ## Supplementary Material Supplementary Material nwae234_Supplemental_Files