To investigate whether enhanced ROS production mediates the cytotoxic effects of DPI, we used MitoQ, a ROS scavenger that selectively accumulates in mitochondria (27, 28). First, we assessed the viability of the MNA IMR-5/75 cells in response to MitoQ. In these cells MYCN protein expression can be downregulated by ca. 50% by a doxycycline inducible shRNA (29) ( Supplementary Figure S1 ). We found that MitoQ was non-cytotoxic up to a concentration of 100 nM over a 24-hour treatment period ( Figure 1A ). Then, we tested the antioxidant effect of MitoQ when combined with DPI. Mitochondrial superoxide production was measured using the selective fluorogenic MitoSOX Red dye (30). DPI is known to increase mitochondrial superoxide production in a dose dependent manner in MNA cells (16). Unexpectedly, MitoQ on its own increased ROS concentrations, although the effect was not significant ( Figure 1B ). Combining DPI and MitoQ effectively suppressed DPI-induced superoxide production to almost baseline level ( Figure 1B ).

To determine whether cell death follows the pattern of superoxide levels recorded in DPI and MitoQ-treated IMR-5/75 cells, cell toxicity was measured using the CellTox Green Dye that detects modifications in membrane integrity resulting from cell death ( Figures 1A, C ). As previously shown, DPI induced progressive cell death over the three-day observation period, which was significantly mitigated by downregulation of MYCN (16) ( Figure 1C ). 100 nM MitoQ on its own was significantly cytotoxic at day 2 and 3, an effect which was also lessened by downregulation of MYCN ( Figures 1A, C ). However, MitoQ failed to reduce DPI-induced cell death at all tested time points ( Figure 1C ). Cell death was further assessed by counting nuclei and Propidium Iodide (PI) staining ( Figures 1D, E ). Nuclei counting of IMR-5/75 cells treated with the same conditions ( Figure 1D ) for 72h confirmed the cell toxicity results ( Figure 1C ). Combining MitoQ and DPI resulted in complete disappearance of nuclei, indicating extensive cell death. PI cannot penetrate intact cell membranes and thus only stains cells whose membrane integrity is compromised, such as apoptotic or necrotic cells. While 100 nM MitoQ only minimally increased DPI-induced cell death ( Figure 1C ), higher MitoQ concentrations significantly enhanced DPI-mediated cell toxicity as measured by an increased PI uptake after 24h, in a MYCN-dependent fashion ( Figure 1E ). Consistent with these findings, caspase-3/7 activity showed a modest, non-significant increase with MitoQ alone after 24h, but was elevated when MitoQ was combined with DPI. 1 μM DPI alone did not increase caspase activity, although higher doses are required to trigger caspase activation, as shown previously (16).

These results suggest that the cytotoxicity observed following treatment with MitoQ alone and in combination with DPI is not solely dependent on its ability to regulate mitochondrial superoxide production in MNA NB cells, which is in line with recent studies on MitoQ effects in other cancer cell types (31, 32). Interestingly, the cytotoxic effects of MitoQ were amplified by higher MYCN levels. Moreover, caspase-3/7 activation is consistent with engagement of the intrinsic, mitochondrial apoptotic pathway.

The arrested differentiation of neuroblasts is a characteristic of NB tumorigenesis and invasiveness (33). To assess whether MitoQ treatment can similarly overcome the differentiation block observed with DPI (16), MNA IMR-5/75 cells were treated every two days with either 50 nM or 100 nM MitoQ and monitored by microscopy ( Figure 1G ). Neurite outgrowth was visualized and quantified using Phalloidin staining, which specifically binds to actin filaments. 100 nM MitoQ treatment resulted in differentiation as evidenced by an increase in neurite outgrowth 7 days after treatment, compared to the non-treated control. 10 μM retinoic acid was used as a positive control for maturation towards a differentiated neuronal phenotype, as shown previously (34). These results suggest that MitoQ slows down cell growth, induces cell death and differentiation of MNA NB cells.

To assess whether MitoQ and DPI act synergistically in reducing cell viability, we treated two MNA NB cell lines (IMR-5/75 and Be(2)-C) with various concentrations of DPI and MitoQ. Synergy was assessed using SynergyFinder 3.0 (35), which computes synergy scores based on different mathematical models. Given that both MitoQ and DPI target mitochondrial function, we used the Loewe synergy model, which is particularly suited for evaluating the combined effects of drugs with similar mechanisms of action (36). The combination exhibited a pronounced synergistic effect in both MNA NB cell lines ( Figures 2A, B ). Although the concept of combination synergy is highly dependent on the underlying assumptions (37), according to SynergyFinder 3.0 the synergy scores can be interpreted as follows: from -10 to 10 the interaction is likely to be additive, larger than 10 the interaction is likely to be synergistic, and conversely, if the score in less than -10, then the interaction between the two drugs is most likely antagonistic. In IMR-5/75, the highest Loewe score is 43.2 and obtained at ~400 nM MitoQ combined with ~500 nM DPI ( Figure 2A ). The highest Loewe score in Be(2)-C is 45.3 and reached at ~1800 nM MitoQ combined with ~10 nM DPI ( Figure 2B ). The highest score was reached with a ~4.5fold lower concentration of MitoQ in IMR-5/75 than in Be(2)-C cells, which seem to be more resistant to treatment ( Figures 2A, B ). These results are in line with previous findings indicating that inhibition of mitochondrial complex I synergizes with mitochondrial antioxidants in impairing cancer cell fitness, supporting the notion that complex I is a dominant determinant of viability (28).

Finding synergistic drug combinations allows for combining drugs at lower concentrations, which is an important strategy for enhancing therapeutic efficacy while minimizing adverse effects. Vincristine (vin) is a commonly used chemotherapeutic agent for treating patients with NB. As genotoxic treatment is clinically very effective but also toxic, we tested whether DPI could reduce the dosage of vincristine required for efficient cancer cell killing. A DPI and vincristine combination was tested for synergy in the two MNA NB cell lines IMR-5/75 and Be(2)-C ( Figures 3A, B ). To evaluate this combination, we used the Zero Interaction Potency (ZIP) model, which accounts for independent drug effects and is particularly suited for combinations of agents with differing targets, such as DPI and vincristine (38). ZIP scores of >10 are considered to indicate synergy (35). The combination showed synergistic effects in reducing cell viability in IMR-5/75 and Be(2)-C cells. In IMR-5/75, the highest ZIP score is 11.3 and reached at ~0.75 nM vincristine combined with ~100 nM DPI ( Figure 3A ). The highest ZIP score in Be(2)-C is 13.7 and observed for ~15 nM vincristine combined with ~30 nM DPI ( Figure 3B ). These results indicate that vincristine and DPI have an additive/synergistic interaction in MNA cells for reducing cell viability at low concentrations that are clinically achievable.

Given the synergy of DPI with both MitoQ and vincristine, we combined all three drugs at reduced dosage. To assess synergy of the 3-drug combination, the Bliss Independence model was used (39) ( Figures 3C–E ). The model is based on the premise that the relative effect of a drug is independent of the presence of the other drug. When two or more independent drugs are combined, their effects add up, and the Bliss synergy score is 0 (39). Within the Bliss independence model calculation, synergy is assumed when the actual effect of the combination is greater than the expected effect, i.e., if the Bliss score is greater than 0 (40). For a 2-drug combination (MitoQ + DPI, (M+D)/2), the drugs were tested at their IC50 concentration divided by 2 or for a 3-drug combination (MitoQ + DPI + vin, (M+D+V)/3) divided by 3. For both MNA NB cell lines tested, the IC50/2 and IC50/3 combinations displayed similar or slightly greater effects in reducing cell viability than the IC50 concentration of each of the single drugs. Indeed, for all the combinations, the Bliss score was > 0 in the two MNA NB cell lines ( Figures 3D, E ). Therefore, the double and triple combinations at reduced doses are synergistic in reducing the cell viability of IMR-5/75 and Be(2)-C.

To characterize the interaction between MitoQ, DPI, and vincristine, we assessed their effects on cell death at IC50/2 and IC50/3 concentrations in four NB cell lines: Be(2)-C (MNA), IMR-32 (MNA), IMR-5/75 (MNA with MYCN downregulatable), and SH-SY5Y/MYCN (non-MNA, MYCN-inducible). The Be(2)-C cell line showed greater resistance with IC50s higher than other tested cell lines. Therefore, in order to assure that the drug sensitivity range is covered in full, the IC50 concentrations of Be(2)-C cells for the 3 drugs were taken as a basis to treat all other cell lines. Cell toxicity was measured using CellTox Green ( Figure 4 ). The double combination ((M+D)/2) significantly increased toxicity in IMR-32 and IMR-5/75 cells within 24 hours, persisting up to 72 hours. However, in Be(2)-C cells, it reduced ATP content by more than half ( Figure 3D ) without increasing cell death ( Figure 4A ), indicating a cytostatic rather than cytotoxic effect. In Be(2)-C, significant toxicity occurred only after 72 hours with the triple combination ((M+D+V)/3) ( Figure 4C ). In SH-SY5Y/MYCN cells, (M+D)/2 had no effect on cell death, while (M+D+V)/3 increased toxicity at 48 and 72 hours ( Figure 4A ). MYCN abundance modestly influenced sensitivity: MYCN downregulation in IMR-5/75 cells slightly reduced toxicity, while MYCN induction in SH-SY5Y/MYCN slightly increased it, though neither change reached significance. Although MYCN downregulation or overexpression are in the region of 50% or more, they are still within the natural range of variation observed in MNA and MYCN non-amplified cells, respectively (41, 42). This suggests that modest changes in MYCN alone are not sufficient to strongly alter sensitivity under our experimental conditions. Overall, the triple-drug combination was effective in inducing cell death across four high-risk NB cell lines. The double-drug combination promoted cell death in two MNA cell lines, but in the more resistant MNA Be(2)-C cells it primarily caused metabolic inactivity.

To further measure cell death upon treatment with MitoQ, DPI and vincristine, we measured caspase 3/7 activation using Caspase-Glo 3/7 (Promega) ( Figure 4D ). Caspases 3 and 7, key mediators of the mitochondrial apoptotic pathway, were significantly activated by (M+D)/2 after 24 hours in IMR-5/75 -Dox and IMR-32, but not in Be(2)-C and SH-SY5Y/MYCN cells. (M+D+V)/3 enhanced caspases 3/7 activation in all cell lines tested. Downregulating MYCN in IMR-5/75 decreased caspase activation for both treatments, while inducing MYCN expression in SH-SY5Y/MYCN increased caspase activation, although the differences were not statistically significant ( Figure 4D ). These data suggest that MNA NB cells exhibit great sensitivity to the triple drug combinations. The non-MNA cell line SH-SY5Y/MYCN showed no sensitivity to the double combination, similarly to the most resistant MNA cell line, Be(2)-C. Combining compounds having a synergistic interaction enables the use of reduced concentrations while still inducing cell death and caspase activation at a high level.

To assess the antitumor potential of the dual and triple drug combinations in a more physiological model than cell lines, we used a 3D spheroid model of a transgenic mouse that overexpresses MYCN under the control of the tyrosine hydroxylase (Th) promoter. This promoter is active in neural crest cells, and driving MYCN overexpression in these progenitor cells of NB causes the animals (Th-MYCN) to develop tumors that closely resemble human NB (43). Spheroid tumor models are 3D aggregates of cancer cells which mimic the architecture and cellular heterogeneity of tumor tissue in vitro. They are viewed as an intermediate model between 2D monolayers and animal models, as they preserve key features of tumors, especially the response to drug treatments (44).

Treating the Th-MYCN spheroids with the DPI-MitoQ combination showed a strong synergistic effect. The highest Loewe score was 45.7, achieved when ≈ 300 nM MitoQ is combined with ≈ 7.8 nM DPI ( Figure 5A ). The combination between DPI and vincristine was also synergistic but less potent. Peak synergy occurred when ≈ 2 nM vincristine was combined with 7.8 nM DPI, with an associated ZIP score of 13.6 ( Figure 5B ). Interestingly, the synergy maps for both MitoQ-DPI and vincristine-DPI are highly similar to those observed in Be(2)-C cells. Moreover, the IC50 values of MitoQ, DPI and vincristine are within the range observed for IMR-5/75 and Be(2)-C MNA cell lines ( Figures 2A, B ; Figures 3A, B ). These correlations show that 2D and 3D culture models of MNA behave very similarly in terms of drug responses, which validates drug screening in 2D cultures as not only as convenient but also meaningful method for finding new therapies for MNA.

In addition to assessing cell viability, DNA damage was visualized by PI (45, 46) and Hoechst staining of Th-MYCN spheres following exposure to M+D and M+D+V combinations ( Supplementary Figure S2 ). The drugs were used at their IC50 concentrations to maximize synergy and induce a high rate of cell death. After 24 hours, both the double and the triple combinations increased cell death. The visual increase in Hoechst fluorescence intensity following M+D and M+D+V treatments can be explained by accumulation of the dye due to highly condensed and fragmented DNA (46). This observation confirms that the corresponding increase in PI intensity is due to apoptotic processes ( Supplementary Figure S2 ).

To elucidate the synergistic mechanism of our double and triple drug combinations, we examined the proteomics and phosphoproteomics changes in MNA Be(2)-C cells treated with DPI+/-MitoQ+/-Vin for 24 hours by mass spectrometry (MS). The Be(2)-C cell line was selected due to its similar response profile as spheroids and delayed response to treatment compared to both MNA NB cell lines IMR-5/75 and IMR-32. Indeed, Be(2)-C cells only significantly apoptose after 72 hours of (M+D+V)/3 treatment ( Figure 4C ), allowing a broad time window to identify pathways that mediate apoptosis-associated signaling.

First, we assessed protein expression in Be(2)-C cells treated with either MitoQ IC50, DPI IC50, vin IC50 or the double and triple drug combinations, versus control ( Supplementary Figures S3 - S6 ; Supplementary Table S1 ). Protein enrichment analysis showed that DPI and MitoQ single treatments induced a similar profile of downregulated peptides, affecting mainly members of the complex I of the mitochondrial respiratory chain and cell cycle/division proteins ( Supplementary Figures S3 , S4 ; Supplementary Table S1 ). However, they exerted contrasting effects on proteins involved in histone H3-K27 and H3-K4 trimethylation, with DPI causing downregulation and MitoQ causing upregulation. MitoQ upregulated the expression of proteins involved in cellular amino acid synthesis, and hydrogen peroxide catabolic processes. On the other hand, DPI mainly upregulated members of the cholesterol biosynthetic process ( Supplementary Figure S4 ; Supplementary Table S1 ). At its IC50 concentration, vincristine induced significant downregulation of Ribonucleotide Reductase M2 (RRM2), involved in DNA damage repair and cell cycle regulation ( Supplementary Table S1 ). In agreement with their single-drug effects, both (M+D)/2 and (M+D+V)/3 downregulated proteins linked to mitochondrial ATP synthesis and cell cycle ( Supplementary Figures S5 , S6 ; Supplementary Table S1 ). (M+D)/2 upregulated TTC5 and HMGCR, involved in p53-dependent apoptosis and cholesterol metabolic process, respectively ( Supplementary Table S1 ). The triple combination upregulated UBE2C, TTC5, ELOVL5 and NDUFS8. These proteins are implicated in mitosis, fatty acid metabolism, p53-dependent apoptosis and complex I of the mitochondrial respiratory chain, respectively ( Supplementary Table S1 ). Interestingly, the DNA damage repair protein RRM2, often upregulated in NB to support oncogenicity, was upregulated across all conditions, and phosphorylation at the Ser-80 and Ser-86 sites was significantly downregulated by both combinations ( Supplementary Table S1 ) (47–49).

Based on the phosphoproteomic profiling we then inferred the upstream kinases responsible for the differential phosphorylation observed in each treatment versus control, using a kinase enrichment analysis ( Figure 6 ). Biological functions highly represented by the inferred kinases for both MitoQ and DPI single treatments include cell cycle regulation, DNA damage response and cell proliferation (DNAPK, CDK1, CDK2, RSK2, P7056K, UL97) ( Figures 6A, B ). UL97 is a viral protein kinase that shares substrates with CDK1 (50). DNA damage response and regulation of cytoskeleton organization were also prominently represented in MitoQ IC50 treatment compared to control (LIMK1, LIMK2, TESK1, CAMK4, CAMK2B, PKACA, ERK1) ( Figure 6B ). Moreover, MitoQ decreased MYCN phosphorylation at the Ser-156 residue ( Supplementary Table S2 ). No upstream kinases were predicted for vincristine on its own. For the double and triple combinations, the abundance in phosphorylation marks suggested that mainly kinases linked to cell cycle and cell proliferation are differentially regulated (UL97, PRKD1, CDK2, CDK4) ( Figures 6C, D ).

Together, proteomic and phosphoproteomic profiling highlighted cell cycle and proliferation pathways as key targets of the double and triple drug combinations ( Figures 6C, D ; Supplementary Figures S5 , S6 ). MitoQ, DPI and the double drug combination decreased the expression of cyclin D1 (G1/S-phase) and cyclin B1 (G2/M phase), suggesting an inhibitory effect on cell cycle progression. These findings were confirmed by Western blotting ( Supplementary Figure S7A ). Similarly, the (M+D+V)/3 combination decreased the expression of cyclin B1 and based on Western blot data, also downregulated cyclin D1. Vincristine caused a sharp increase in histone 3 phosphorylation at serine 10 (pS10-H3), likely due to its interference with mitotic spindle assembly, causing mitotic arrest ( Supplementary Figure S7A ). Since pS10-H3 is a marker for chromosome condensation linked to mitosis, cellular stress (51) and DNA damage (52), its elevation in the triple drug combination likely reflects both a proliferation block and a stress response to vincristine.

According to the MS results, MitoQ and to a lesser extent DPI single treatments reduced MYCN protein levels. These findings were confirmed by Western blotting ( Supplementary Figure S7 ). However, this effect was not observed in Be(2)-C cells treated with the double and triple combinations ( Supplementary Figure S7A ). In contrast, in the IMR-5/75 MNA cell line, (M+D)/2 reduced MYCN by half, while (M+D+V)/3 led to an even greater decrease ( Supplementary Figure S7B ). This discrepancy may be related to the finding that Be(2)-C cells are more resistant to treatment and undergo apoptosis at a later time point.

Overall, these results suggest that the effects of MitoQ, DPI, and vincristine are mediated mainly through OXPHOS inhibition, MYCN regulation, and disruptions in DNA damage response and cell cycle progression.

To identify the proteins and pathways modulated by the combination treatments versus the single treatments, we applied a linear regression model to the matched global proteomic and phosphoproteomic data ( Figure 7 ). The model identifies pathways responsible for the effect of the double and triple drug combinations, versus the effect of the addition of each single agent at its IC50 concentration. The differentially expressed proteins revealed that mainly the respirasome, mitochondrial ribosome, hydrogen peroxide and glutathione metabolic processes are responsible for the effect exerted by the double combination on Be(2)C cells ( Figure 7A ). Regarding the triple combination, the pathways are highly similar to those of (M+D)/2, with an additional enrichment in pathways associated with anion transmembrane transporter activity ( Figure 7B ). In terms of kinase enrichment, pathways related to cell cycle, DNA damage response, cell proliferation and regulation of cytoskeleton organization account for the effects of (M+D)/2 and (M+D+V)/3 compared to the addition of single agents (LIMK1, TESK1, P70S6K, CDK1, CDK2, CAMK4, CAMK2B) ( Figures 7C, D ). These findings suggest that the mechanism of synergy underlying the interaction between DPI and MitoQ is linked to their impact on mitochondrial metabolism and cell cycle regulation. Furthermore, the addition of vincristine does not significantly alter the proteomic profile linked to the activity of the other two compounds, while still causing more cell death when combined with MitoQ and DPI at a low dosage. The independent function of vincristine underscores the therapeutic potential for MitoQ and DPI to engage previously untargeted pathways involved in cancer progression.

In both the double and triple drug combinations, processes generating ROS (NADH dehydrogenase, NADH quinone oxidoreductase/complex I) and antioxidant activities are significantly dysregulated ( Figures 7A, B ). Therefore, we examined superoxide and H₂O₂ generation in MNA cells ( Figure 8 ). Specifically, we exposed Be(2)-C cells to conditions akin to those used for MS analysis, utilizing either DPI, MitoQ and vincristine individually or in combination. Indeed, superoxide production has previously been demonstrated to be linked to the susceptibility of MNA NB cells to DPI, resulting in cytochrome c release, caspase activation, cytochrome c release and cell death (16). MitoSOX red dye (Invitrogen) was used as a probe to monitor mitochondrial superoxide, and Abgreen indicator (Abcam) for H₂O₂. The Abgreen indicator is a cell-permeable fluorescent indicator which generates green fluorescence when reacting with H₂O₂. 10 μM DPI was used as a positive control for superoxide production, while 10 μM doxorubicin was used as a positive control for H₂O₂ generation. DPI, MitoQ, and vincristine at their IC50 concentrations individually increased mitochondrial superoxides over a 24-hour period ( Figure 8A ). Interestingly, MitoQ decreased H₂O₂ after 2 hours of treatment, but this effect was reversed after 24 hours ( Figure 8B ). Both (M+D)/2 and (M+D+V)/3 combinations increased superoxide production after 24 hours, as well as H₂O₂ ( Figures 8A, B ). Thus, the increase in H₂O₂ production for the combinations after 24 hours reflects the enrichment analysis of the proteomics data, which indicate enhanced ROS production and less ROS removal ( Figure 7B ). This also aligns with Figure 1B , where MitoQ partially reduced DPI-induced mitochondrial ROS but did not return levels to baseline; these residual ROS levels are sufficient to perturb metabolism and contribute to cell death alongside other mechanisms. These results further suggest that a balanced generation of ROS by mitochondrial complex I is crucial to maintain MNA cellular fitness.

To determine whether H₂O₂ contributes to cell death, cells were pre-treated with 0.1mM N-acetylcysteine (NAC), a thiol-containing antioxidant that supports intracellular redox buffering, and monitored over a 72-hour period ( Figure 8D ). NAC has been previously shown to rescue DPI-induced cell death (16), however, in the present experiments it failed to protect against the (M+D)/2 and (M+D+V)/3 combinations and instead further enhanced toxicity. While these results seem at odds with the common use of NAC as ROS scavenger, more recent literature shows that NAC also can enhance ROS generation under conditions, where cells express a certain pattern of enzymes that regulate ROS production (53), are co-treated with certain ROS generators (54), or suffer from mitochondrial stress (55). The latter situation is likely to occur in neuroblastoma cells co-treated with DPI and MitoQ and may explain the NAC mediated toxicity. The inability of NAC to rescue cell death in the combination treatments, despite its protective role in DPI-induced toxicity, underscores the complex interplay between antioxidants and ROS-generating agents in cancer cells.

Finally, to evaluate the efficacy of MitoQ and DPI in vivo, both drugs were administered individually and in combination to Th-MYCN transgenic NB mice. Despite the promising efficacy observed in Th-MYCN-derived 3D spheres, administration of the compounds resulted in gastrointestinal toxicity, including blocked intestine and stomach enlargement, necessitating early termination of the experiment. To address this limitation, the study was re-conducted using an IMR-5 (MNA) xenograft immunodeficient NB mouse model, where a reduced toxicity was observed ( Figures 9A, C ). In contrast to our in vitro results, treatment with MitoQ, DPI, or their combination did not produce notable effects on tumor size reduction, even at the highest tolerable doses ( Figure 9A ). While a trend toward increased survival probability was observed in animals treated with either compound or their combination, the effect was not statistically significant ( Figure 9B ). These findings suggest that MitoQ and DPI are more tolerated in an in vivo immunodeficient environment, but their limited efficacy in reducing tumor burden may arise from challenges linked to oral administration, including poor absorption and first-pass metabolism. Alternatively, these findings may point to a role of the immune system in the effect of these drugs in vivo. We are currently investigating possible reasons for this discrepancy.

Mitoquinone Mesylate (MitoQ, MQ) was purchased from MedChemExpress (#HY-100116A), and Diphenyleneiodonium chloride (DPI) from Enzo Life Sciences (#BML-CN240-0010). Dimethyl sulfoxide (DMSO, vehicle) (#D2650), Vincristine Sulfate (Vin) (#V8388), N-acetyl-cysteine (#A15409) and retinoic acid (#R2625) were from Merck. Doxorubicin was purchased from Selleck Chemicals (#S1225).

Cell toxicity was evaluated using the CellTox Green fluorescent assay (Promega). Typically, cells were seeded in 50 μL media per well in duplicates on a black 96-well plate with clear bottom (Thermo Scientific). Complete RPMI phenol red-free media (Gibco) was used and was beforehand completed with CellTox Green Dye at a 1:500 ratio (v/v). Six h after seeding (or 24 h after doxycycline induction for MYCN-regulatable cell systems), the cells were treated with 50 μL of drug(s) or vehicle and incubated at 37°C for 72 h. Fluorescence was measured at 4, 24, 48 and 72 h after treatment at 485 nm excitation and 530 nm emission on a SpectraMax M3 microplate reader (Molecular Devices).

Mitochondrial superoxide was measured with MitoSOX Red fluorescent indicator (Thermo Scientific). Propidum iodide (PI) (Thermo Scientific) was used to selectively stain apoptotic cells. Typically, cells were seeded in 100 μL RPMI phenol red free media per well in duplicates on a black 96-well plate with clear bottom (Thermo Scientific). To prevent IMR-5/75 from detaching, the plates were treated beforehand with 100 μg/mL Poly-D-Lysine (Sigma-Aldrich). Six h after seeding (or 24 h after doxycycline induction for IMR-5/75), the cells were treated with 100 μL of drug(s) or vehicle and incubated at 37°C. After the appropriate incubation time, the cells were washed once with DPBS Ca²⁺/Mg²⁺ (Invitrogen). MitoSOX Red dye (4 μM) or PI (1 μg/mL) were added together with Hoechst nuclear counterstain (1 μg/mL) (Bio-Sciences) to the cells in DPBS Ca²⁺/Mg²⁺ and incubated for 20 minutes at 37 °C protected from light. The cells were washed twice with DPBS Ca²⁺/Mg²⁺, and fresh RPMI phenol red-free media was added before live cell imaging. MitoSOX Red indicator and PI were detected using the TRITC filter (Ex/Em: 396/610 nm) and Texas Red (Ex/Em:560/630), respectively, while Hoechst was detected using HOECHST 33342 filter (Ex/Em: 353/483).

Confocal images of 4 optical planes interspaced by 1 μm (from -1.0 μm to 2.0 μm included) across 25 x 25 fields of view with a 10% overlap were acquired per well on an Opera Phenix High Content Screening System (PerkinElmer) using a 40x or 63x/1.15 NA water-immersion objective, with live cell parameters (37 °C and 5% CO₂). The acquired images were analyzed using an optimized pipeline for each cell line on Harmony^® v.4.9 high-content analysis software (PerkinElmer). The building blocks within the software were used to identify and calculate the mean dye parameters per well: number of nuclei, MitoSOX Red, or PI mean intensity/cell.

IMR-5/75 cells were plated in duplicates on a 6-well plate in 2 mL complete RPMI phenol red-free media (Gibco). One day after seeding, the cells were treated with either 50 nM MitoQ, 100 nM MitoQ, 10 μM retinoic acid or vehicle, and incubated 37°C for 7 days. Every two days, the cells were treated with the appropriate drug or vehicle, in 2 mL fresh media. After 7 days, the cells were fixed, permeabilized and neuronal differentiation was assessed by Alexa Fluor 488 Phalloidin staining (Bio-Sciences) according to the manufacturer’s instructions. Hoechst nuclear counterstain (1 μg/mL) (Bio-Sciences) was used for nuclei counting. Imaging was performed on an EVOS Digital Inverted Fluorescence Microscope (Invitrogen), using the FITC (green) and the DAPI (blue) filters. Neurite outgrowth (cells with neurites >= 2 cell body diameters) were counted per cell using the NeuriteJ plugin of ImageJ (v1.53e; Java 1.8.0_172).

Cell viability by loss in ATP content was measured using the CellTiter-Glo Luminescent and CellTiter-Glo 3D Luminescent Cell Viability Assays (Promega). Cells were seeded in duplicates in 100 μL media on a Nunc 96-Well White Microplate (Fisher Scientific). Six h after seeding, the cells were treated with 100 μL of drug(s) or vehicle and incubated at 37°C for 72 h. To assess cell viability of the neurospheres, cells were seeded after dissociation in 100 μL PrimNeuS media on a Nunclon Sphera 96-Well U-Shaped-Bottom Microplate (Thermo Scientific). Two days after seeding, the cells were treated with 100 μL of drug(s) or vehicle and incubated at 37°C for 72 h. CellTiter-Glo reagent was added to the wells according to the manufacturer’s instructions and luminescent signal was measured on a SpectraMax M3 microplate reader (Molecular Devices).

Caspases 3 and 7 activation were measured using Caspase-Glo 3/7 (Promega). Cells were seeded in duplicates in 100 μL media on a Nunc 96-Well White Microplate (Fisher Scientific). Six h after seeding, the cells were treated with 100 μL of drug(s) or vehicle and incubated at 37°C for 24 h. Caspase-Glo reagent was added to the wells according to the manufacturer’s instructions and luminescent signal was measured on a SpectraMax M3 microplate reader (Molecular Devices).

Six hours after seeding (or 24 h after doxycycline induction for IMR-5/75), the cells were treated with either MitoQ IC50, DPI IC50, Vin IC50, (M+D)/2, (M+D+V)/3 or vehicle and incubated at 37°C for 24 h. Cells were harvested using 110 μL ice cold lysis buffer containing 20 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM MgCl₂, 1% (v/v) Triton X-100 and 10% (v/v) of protease and phosphatase inhibitor cocktail (Roche). After sample denaturation and electrophoresis, membranes were blocked with 5% milk in TBS-T (120 mM Tris-HCl, pH 7.4, 150 mM NaCl, and 0.05% Tween 20) for 1 h at room temperature. Target proteins were detected using primary antibodies at 1:1000 dilution overnight at 4°C (MYCN #sc-53993, Santa-Cruz Technologies; Phospho-Histone H3 (Ser10) #3465, Cell Signaling Technology; Phospho-Histone H2AX (Ser139) #05-636, Sigma-Aldrich; Cyclin B1 #12231, Cell Signaling Technology; Cyclin D1 #2922, Cell Signaling Technology; GAPDH #2118, Cell Signaling Technology). Anti-mouse or anti-rabbit secondary antibodies (Cell Signaling Technology) HRP-conjugated at 1:5000 dilution were incubated with the membranes for 1 h at room temperature, and visualized using a chemiluminescence system. GAPDH was used as a loading control.

Synergy scores were calculated using SynergyFinder 3.0 for 2-drug combinations. Normalized cell viability values were imported to SynergyFinder 3.0 in a matrix format, and %viability was chosen as a readout. LL4 was selected for curve-fitting algorithm, detection of outliers was activated, and the Zero Interaction Potency (ZIP) or Loewe model were used to capture the drug interaction relationships.

To investigate the potential synergy of two or more drugs in combination at reduced doses, the Bliss Independence model was used. According to Rukhlenko et al. (39), the Bliss synergy score was defined as follows for reduced drug combinations:

According to the Bliss criterion, when two independent drugs are combined, their effects simply add up, and the Bliss synergy score is zero. In that case, if the combination is merely additive, the effect would be the following:

In this equation, E is the %viability effect obtained from the CellTiter-Glo output. E_A+B is the observed effect of the combined drugs, and E_A*E_B the expected effect when the drug effects are independent (viability values of the two single drugs). The Bliss synergy score was therefore calculated as follows:

In this context, a positive Bliss score means that the observed effect of the combination is lower than what would be anticipated if the drug effects were independent, thereby indicating synergy. In contrast, a negative Bliss score would suggest antagonism.

Be(2)-C cells were cultured in triplicates in 10 cm dishes and treated the following day either with MitoQ IC50, DPI IC50, Vin IC50, (M+D)/2, (M+D+V)/3, or vehicle for 24 h and washed once with PBS. Cells were lysed in 100 μL ice cold 8 M Urea/50 mM Tris HCl lysis buffer (Fisher Scientific) with 1:100 of Halt Cocktail Inhibitor, EDTA-free (100X) (Thermo Scientific). Samples were sonicated on ice twice for 9 seconds, at a power of 15% with a Skl-150IIDN Ultrasonic Homogenizer (Medical Supply Co. Ltd.), and protein concentration was measured using the Pierce BCA protein assay (Thermo Scientific). The same amount of protein (400 μg) were used for all samples, and 8 M Urea lysis buffer was used as a diluent if required. 8 mM of reducing agent DTT (Sigma) was then added to the samples and incubated in a thermomixer at 30°C for 30 minutes, followed by the addition of 20 mM iodoacetamide (Sigma) for protein carboxylation with incubation at 30°C in the dark for 30 minutes. Samples were diluted with 50 mM Tris-HCl to bring urea concentration down to below 2 M to prevent trypsin inhibition. Trypsin/Lys-C Mix, Mass Spec Grade (Promega) was added to the samples which were digested overnight at 37°C, on a thermomixer. Protein digestion was terminated by adding 100% formic acid to 1% final volume. Peptides were extracted from each sample using HyperSep C18 tips (Thermo Scientific), and 50 μg and 350 μg were kept for total proteomics and phosphoproteomics analysis, respectively.

Phospho-peptide enrichment was performed using TiO₂ (Titansphere Phos-TiO Bulk 10 um, GL Science), ratio of beads to sample 10:1, mixed for 30 minutes at room temperature in 80% acetonitrile, 6% trifluoroacetic acid, 5 mM monopotassium phosphate, 20 mg/mL 2.5-dihydroxybenzoic acid. The phosphopeptides were eluted by adding 80 μL of elution buffer directly (50% acetonitrile, 500 uL H₂O, 7% Nh4OH ammonium hydroxide). Eluents were evaporated on the CentriVap Concentrator (Labconco) at 45°C for an hour. Peptides and phosphopeptides were analyzed separately by mass spectrometry.

Samples were analyzed on a timsTof Pro Mass Spectrometer (Bruker) as described previously (86).

The raw data was searched against the Homo sapiens subset of the Uniprot Swissprot database (reviewed) using the Fragpipe proteomics pipeline 21.1 with specific parameters for TIMS data-dependent acquisition (TIMS DDA). The normalized protein intensity of each identified peptide was used for label-free quantitation using the LFQ-MBR workflow. Within MSFraggers, precursor mass tolerance was set to +/-20ppm, fragment mass tolerance was set to +/-20ppm, tryptic cleavage specificity was applied, as well as fixed methionine oxidation and fixed N-terminal acetylation. Phospho-proteomic analysis was carried out using the LFQ-phospho workflow which includes phosphorylation as a variable modification on serine, threonine, and tyrosine residues (87).

Analysis for treatment versus untreated control of peptides and phosphopeptides was performed using Perseus v2.0.9.0. Phosphoproteomics results were normalized to the total proteomics. The data file from the Fragpipe output was loaded as a generic matrix in the software, with the Max LFQ intensities as main columns. The rows were filtered to remove the reverse proteins, the proteins only identified by site and the contaminants. The values were log2 transformed into normally distributed data. Rows were filtered based on the number of valid values: a minimum of 4 valid values out of 6 replicates in at least one group. Then, missing values were imputed from normal distribution by calculating width and center of the distribution. Statistical analysis was done using a One-Sample t-Test (s0 (fold change) = 0.1, FDR = 0.05). Kinase Substrate Enrichment Analysis (KSEA) was performed using the fgsea R package. Kinase-substrate relationships were obtained from the PhosphoSitePlus database and provided as gene sets (PhosphoSitePlus). Further protein-protein interaction and pathway enrichment analysis were done by exporting the final matrix to STRING database.

Upregulated and downregulated protein lists from each treatment condition versus untreated control were exported to the STRING database (https://string-db.org) for functional enrichment analysis and PPI network, using the k-means clustering method.

The Fragpipe combined_proteins.txt file was filtered to remove known contaminants and protein groups containing no protein. Protein LFQ intensity values were log base 2 transformed, with zero values being assigned NaN to circumvent undefined values. Filtering was applied to keep only those proteins with at least four valid values in at least one drug treatment condition. NaN values were then reassigned to zero before median centering. Site- and condition-specific imputation was performed using the scImpute function in PhosR (88) if the protein was quantified in more than 50% of samples from the same drug treatment. Sample-wise tail-based imputation was implemented with the tImpute function (88). Linear regression through limma was used to model the relationship between protein abundance and drug treatment. The design matrix was coded as model.matrix(~treatment), with control samples representing the reference. Differential effects were identified by contrasting the coefficients from the combination drug treatment to the addition of each drug treatment in isolation. Differential protein abundance by empirical Bayes was performed to identify proteins reaching statistical significance. A signature of combination drug treatment synergy was created by selecting statically significant (adj p value < 0.05) and upregulated (LogFC > 0) genes from the above comparison. Gene set enrichment analysis (GSEA) was performed using the fgsea R package on the ontology gene sets retrieved from gsea-msigdb.org (https://www.gsea-msigdb.org/gsea/msigdb). An FDR threshold of 0.1 was implemented.

Cells were seeded per well in 100 μL PrimNeuS media on a Nunclon Sphera 96-Well U-Shaped-Bottom Microplate (Thermo Scientific). Four days after seeding, the cells were treated with MitoQ IC50 + DPI IC50, MitoQ IC50 + DPI IC50 + Vin IC50, or vehicle for 24 h. The spheres were washed once with PBS, then fixed in acetone, washed twice with PBS, and permeabilized using Triton X-100. The spheres were washed twice and stained with 1:500 PI solution (Bio-Sciences) supplemented with 250 μg/mL RNAse A (Qiagen) and Hoechst nuclear counterstain (2 μg/mL) (Bio-Sciences) for 30 min protected from light, at room temperature. Spheres were then washed once in PBS and imaged in fresh PBS on an Olympus IX83 Inverted Microscope (Mason Technology) using the TRITC (Ex/Em: 396/580 nm) and DAPI (Ex/Em: 353/483) filter sets. Image processing was performed on ImageJ (v1.53e; Java 1.8.0_172).

All normalized values were transferred to GraphPad Prism 8.4.3 for statistical analysis of biological replicates. All graphs are represented as mean ± SEM. Two-Way ANOVAs were used when using the IMR-5/75 and SH-SY5Y/MYCN -Dox/+Dox cell systems.