Silver Clusters of Five Atoms as Highly Selective Antitumoral Agents Through Irreversible Oxidation of Thiols

Low atomicity clusters present properties dependent on the size, due to the quantum confinement, with well‐defined electronic structures and high stability. Here it is shown that Ag5 clusters catalyze the complete oxidation of sulfur to S+6. Ag5 catalytic activity increases with different oxidant species in the order O2 ≪ H2O2 < OH•. Selective oxidation of thiols on the cysteine residues of glutathione and thioredoxin is the primary mechanism human cells have to maintain redox homeostasis. Contingent upon oxidant concentration, Ag5 catalyzes the irreversible oxidation of glutathione and thioredoxin, triggering apoptosis. Modification of the intracellular environment to a more oxidized state to mimic conditions within cancer cells through the expression of an activated oncogene (HRASG12V) or through ARID1A mutation, sensitizes cells to Ag5 mediated apoptosis. While cancers evolve to evade treatments designed to target pathways or genetic mutations that drive them, they cannot evade a treatment that takes advantage of aberrant redox homeostasis, which is essential for tumor progression and metastasis. Ag5 has antitumor activity in mice with orthotopic lung tumors reducing primary tumor size, and the burden of affected lymphatic nodes. The findings suggest the unique intracellular redox chemistry of Ag5 may lead to new redox‐based approaches to cancer therapy.


Introduction
The mitochondrial electron transport chain is indispensable for cell proliferation, [1][2][3][4] generating as by-product reactive oxygen species (ROS). [5,6] H 2 O 2 sensors measure H 2 O 2 concentration by a redox-based mechanism and set H 2 O 2 -specific scavengers' expression, maintaining the H 2 O 2 concentration below a toxic threshold. [7] Thiol redox status within mitochondria, is determined by the mitochondrial glutathione (GSH) and thioredoxin (Trx) systems, critical for cell viability and function. [8] Simultaneous and irreversible inhibition of these pathways could constitute an effective antitumor treatment. [9] Of interest, Au clusters with 5−10 atoms can catalyze the aerobic oxidation of thiophenol to disulfide (S −2 to S −1 ) at room temperature and 5 bar O 2 with similar catalytic activity to the sulfhydryl oxidase. [10] Metal clusters of just a few atoms can display unique properties that are not observed (see Characterization of Ag 5 and Figures S1 and S2, Supporting Information), UV-Vis, and fluorescence spectroscopy (see Characterization of Ag 5 and Figures S3 and S4, Supporting Information), electrospray ionization time-of-flight (ESI-TOF) mass spectrometry (see Characterization of Ag 5 and Figures S5 and S6, Supporting Information), X-ray absorption near edge structure (XANES) (see Characterization of Ag 5 and Figure S7, Supporting Information). All characterization techniques indicate that samples contain mainly Ag 5 , with a small percentage (≤10%) of noncatalytic smaller Ag 3 and Ag 2 clusters, because they do not bind S groups, as we reported before. [14,15] Using X-ray absorption near-edge structure at the sulfur K edge (S-K XANES), we assessed the effect of Ag 5 on the oxidation of the sulfur groups of free Cys, GSH, and recombinant Escherichia coli Trx. The R-SH groups of these three species were oxidized to S +6 (sulfate) in the presence of Ag 5 at very low concentration (1000 times less than R-SH, [R-SH]/[Ag 5 ] ≈1000) after 10 min of reaction (Figure 1a). The gradual increase in the S +6 peak relative to the S −2 peak with increasing Ag 5 concentration (Figure 1b) is indicative of the SH oxidation reaction dependence on the Ag 5 concentration (see XANES study of Ag 5 catalysis, Table S1 and Figures S8-S10, Supporting Information). Thus, Ag 5 catalyzes R-SH overoxidation until a biological irreversible state (S +5 or S +6 ). [18] When the concentration of Ag 5 is reduced ([R-SH]/[Ag 5 ] < ≈ 10 000) Ag 5 still catalyzes the oxidation of sulfur groups but only to form disulfides (S −1 ), at the same time of reaction. As far as we know, there is no catalyst that can totally oxidize sulfur groups with atmospheric oxygen, indicating the high catalytic activity of Ag 5 clusters. Moreover, the catalytic activity of Ag 5 for the partial sulfur oxidation to disulfide is (TOR ≈ 3 × 10 3 s −1 ) ≈ 2 orders of magnitude larger than the one reported previously for Au clusters, [10] as we will report elsewhere. Using intermediate concentrations of clusters, in the range ([R-SH]/[Ag 5 ] ≈ 1000-10 000), at the same reaction time, we obtain a population of different oxidation species, namely, S −1 , S +2 , S +4 , and S +6 (see XANES study of Ag 5 catalysis and Figure S11, Supporting Information).
We used density-functional theory (DFT) to determine the possible mechanism by which S-containing compounds are oxidized catalytically by Ag 5 (Figure 1c,d). First, we found that Ag 5 can adsorb oxygen molecules, activating some of them to superoxide-like species (see Theoretical Calculations, Table S2 and Figures S12-S19, Supporting Information). Using such activated oxygens, Ag 5 forms, in a fast step, a disulfide ( Figure 1c). Subsequently, as we can see in Figure 1d, the activated oxygens are able to break the S-S bond with the formation of R-(SO) 2 -Ag 5 sulfinic species (I3, S +2 ) attached to the cluster. Such sulfinic species can be further catalytically overoxidized to sulfite R-(SO 3 )-Ag 5 (I5, S +4 ) and finally, detaching the S from the rest of the organic molecule, (SO 3 )-Ag 5 (I6), to sulfate (SO 4 )-Ag 5 (P, S +6 ) species.

Other Oxidants Involved in Ag 5 Catalysis
Oxidants other than O 2 could also be involved in the catalytic activity of Ag 5 under biological conditions and understanding their impact could help elucidate whether Ag 5 has clinical applications. [19] Therefore, we investigated the catalytic activity by XANES of Ag 5 against recombinant E. coli Trx in the presence of H 2 O 2 or the hydroxyl radical (OH•). Due to the higher oxidation activity of these species, in larger nanoparticles. [11][12][13][14] Reducing cluster size from the bulk material to the subnanometric cluster scale alters the particle properties which are greatly affected by strong quantum confinement. Those cluster properties depend strongly on the number of atoms. Here we show that contrary to smaller Ag 3 clusters, which do not bind to thiols but intercalate into DNA, [14,15] Ag 5 , with only two more atoms, binds to thiols and display exceptional catalytic activities for their complete oxidation. Contingent upon oxidant concentration, Ag 5 catalyzes the oxidation of cysteine (Cys), GSH, and Trx successively to disulfide (S −1 ), sulfinic (S +2 ), sulfonic (S +4 ), and finally irreversibly detaches the sulfur atom from the rest of the molecule and oxidizing it to sulfate (S +6 ). Having shown that Ag 5 affects thiol-based antioxidants, we hypothesized that this could be effectively employed for novel pharmacological applications where redox state is important, such as in cancer. [16,17] We present here that Ag 5 , depending on the intracellular redox state, catalyzes the oxidation of redox-active Cys residues, leading to the activation of programmed cell death. We further describe driven mutations frequently found in human cancers that render cells sensitive to Ag 5 .

Ag 5 Catalyze the Oxidation of Cysteine
Ag 5 was synthesized by a modification of a previously developed electrochemical method for the synthesis of Ag 3. [14,15] As we describe in the Experimental Section, we increased the voltage with respect to the preparation of Ag 3 clusters. This allows to focus the kinetics to produce Ag 5 clusters mainly. When the voltage is further increased (see the Experimental Section) larger clusters are obtained. Interestingly such larger clusters do not display the catalytic activities here described for Ag 5 but act specifically in the lysosomes and do not cause cell mortality as it will be reported elsewhere.
Ag clusters were fully characterized by high-angle dark-field scanning transmission electron microscopy (HAADF-STEM) we carried out the experiments, at 10 min of reaction time, with concentrations of oxidants 10 times less than oxygen and a ratio [R-SH]/ [Ag 5 ] = 5000. For comparison purposes (Figure 1e), at the same conditions, when exogeneous oxygen is present with clusters we observe the formation of ≈4% of the S in an oxidation state > = −1 (1.6% S +4 and 2.4% S +6 , i.e., 16% and 24% without concentration correction, displayed in Figure 1e for a better visualization) versus 0% without clusters. We show also that H 2 O 2 without clusters oxidizes 60% (25% S +4 and 22% S +6 ) of recombinant E. coli Trx, while with clusters this proportion increases to 92% (31% S +4 and 26% S +6 ). When using OH• the oxidized proportion is 88% (34% S +4 and 29% S +6 ) while with clusters is 92% (12% S +4 and 50% S +6 ). Therefore, we can conclude that the catalytic activity of Ag 5 increases with different oxidant species in the order O 2 ≪ H 2 O 2 <OH•.

Cellular Ag 5 Uptake
To assess the ability of Ag 5 to enter cancer cells, we incubated human acute lymphoblastic leukemia MOLT-4 cells with Ag 5 at 37 °C. After 5 min incubation at 37 °C, 14% ± 1.72 SEM of the total silver offered to the cells was found inside the cell (see Figure S20a,b, Supporting Information). To elucidate the mechanism of Ag 5 transport across the plasma membrane, we compared the uptake of Ag 5 in MOLT-4 cells at 4 and 37 °C. There was no difference with temperature indicating Ag 5 transport is a passive rather than an active process (see Figure S20c, Supporting Information). Moreover, modification of the mitochondrial membrane potential did not alter the amount of silver in isolated mitochondria after Ag 5 treatment (see Figure S20d, Supporting Information). Together, these data suggest that Ag 5 can enter cells and mitochondria by passive diffusion as an uncharged species.

Ag 5 Increases Intracellular Thiol Oxidation
As Ag 5 oxidizes thiols in vitro and can diffuse into cells we sought to demonstrate that Ag 5 could oxidize intracellular thiols. We observed a significant reduction (≈20%) in the Figure 1. Ag 5 catalyze R-SH oxidation. a) Normalized S-K XANES spectra of cysteine (Cys), glutathione (GSH) and thioredoxin (Trx) before and after addition of Ag 5 . Vertical lines indicate the corresponding energy for different S-oxidation states. b) Normalized S-K XANES spectra of Cys after addition of increasing concentrations of Ag 5 . Vertical lines indicate the energy position associated to the different S-oxidation states. c) Proposed DFT mechanism for the formation of a disulfide on oxygenated Ag 5 . d) Proposed DFT mechanism for S-S bond breaking and subsequent formation of sulfinic, sulfite, and sulfate on Ag 5 . e) Population of S-oxidized species for thioredoxin (Trx) with different oxidating species, obtained from S K-XANES for [Ag 5 ]/[Trx] = 1:5000, after 10 min of reaction.

(4 of 14)
© 2022 The Authors. Advanced Functional Materials published by Wiley-VCH GmbH intracellular GSH/GSSG ratio in whole-cell lysates of A549 cells treated with Ag 5 for 1 h (see Figure S21, Supporting Information). Glutathione reductase (GR) maintains the ratio of GSH/ GSSG present in the cell by catalyzing the reduction of GSSG to GSH. [20] Western blotting of whole-cell lysates of A549 cells revealed a rapid and substantial increase in the oxidized GR form following stimulation with Ag 5 (Figure 2a). This suggests that Ag 5 may accelerate thiol oxidation by targeting sensitive thiols on redox enzymes, thereby tipping the balance toward the oxidative state. To elucidate time-dependent changes of thiol oxidation by Ag 5 we used a redox sensitive glutaredoxin-1 (Grx1)-roGFP biosensor designed to specifically measure the GSH redox potential. [21] The addition of Ag 5 to A549 cells led to a prompt oxidative response with the maximum signal observed after 7 min (Figure 2b) revealing Ag 5 rapidly mediated thiol oxidation catalysis.

Ag 5 Increases the Oxidation of Thioredoxin-Dependent Peroxiredoxin 3 (PRDX3) in A549 Lung Carcinoma Cells Mitochondria
As shown by XANES experiments, Ag 5 increases oxidation of thiol groups of E. coli recombinant Trx. Furthermore, QM/MM calculations confirmed the binding of Ag 5 in the Cys-X-X-Cys motif of the Trx (see Theoretical Calculations, Supporting Information). Additionally, in an in vitro assay, Ag 5 strongly inhibited the Trx/thioredoxin reductase (TrxR) system ( Figure 2c) (see Ag 5 catalyze Trx and TrxR and Figure S22, Supporting Information). Grx1 also possesses the Trx motif Cys-X-X-Cys, [22] so the observed activation in the Grx1-roGFP biosensor might also be due to the direct effect of Ag 5 mediated oxidation of the Cys residues in Grx1. Therefore, we explored whether Ag 5 also affected the Trx antioxidant system in cells. Trx oxidation depends on H 2 O 2 and PRDX3, an enzyme located exclusively in the mitochondrial matrix that reacts rapidly with H 2 O 2 and dimerizes via a disulfide. [23,24] Mitochondria isolated from Wistar rats' heart and liver were incubated with Ag 5 at different doses for 5 min at 37 °C. As shown in Figure 2d, PRDX3 oxidation was dependent on the Ag 5 concentration ((40-240) × 10 −9 m). In contrast, Ag + at the largest concentration of Ag 5 used did not affect the PRDX3 redox state. This finding confirms the membrane permeability of Ag 5 , as it must have crossed the inner mitochondrial membrane to oxidize PRDX3. Furthermore, 15 min after intact cells were exposed to Ag 5 , PRDX3 was oxidized (Figure 2e), confirming that Ag 5 rapidly diffuses into the cell, enter the mitochondria, and increases the oxidation of PRDX3.

Ag 5 is Cytotoxic for Human Cancer Cells
A chemical disrupting redox homeostasis would be expected to induce cell death. Treatment of A549 human lung adenocarcinoma cells and A2780 human ovarian cancer cell line with Ag 5 increased mortality (Figure 2f) (see Figure S23a, Supporting Information). Consistent with cell death being mediated through thiol oxidation by Ag 5 , this toxic effect could be abrogated by the thiol-reducing agents, dithiothreitol (DTT) and N-acetylcysteine (NAC) [25] (Figure 2f). These results were corroborated in various human cancer cell lines (see Figure S23a-e, Supporting Information).
Furthermore, PRDX3 oxidation corresponds to cytochrome c release and caspase activation, [29] consequently, we examined the Ag 5 effect on apoptotic biomarkers: PARP cleavage, Annexin-V / FITC and 7AAD staining and caspase-3/7 activation (see Figure S25a,b and Movie S1, Supporting Information). All were activated by Ag 5 , indicating Ag 5 is triggering apoptotic cell death and concordant with PRDX3 oxidation. If true, pretreatment with auranofin (AF), a TrxR inhibitor that causes the oxidation of PRDX3 [30] could be expected to potentiate Ag 5 cytotoxicity. The combination of AF and Ag 5 present greater cytotoxicity than either treatment alone (see Figure S25c,d, Supporting Information). Conversely, if PRDX3 oxidation is relevant for Ag 5 action, increasing the levels of PRDX3 should reduce Ag 5 efficacy. Transient overexpression of PRDX3 in HEK293 consistently increased the Ag 5 ED50 from 177 ± 12.9 to 193 ± 14.6 × 10 −9 m (Figure 2h).
PRDX3 plays an important role in mitochondrial redox control in cancer cells [31] thus, we studied Ag 5 cytotoxicity across a wide range of tumor cell types to see whether it correlates with PRDX3 expression. To this end, an ED50 value for Ag 5 was obtained for each cell line of a panel of 85 human tumor cell lines (see Figure S26, Supporting Information). PRDX3 mRNA levels and ED50 value are inversely correlated (Figure 2i), pinpointing PRDX3 as a primary mediator of Ag 5 action.

The Cytotoxic Activity of Ag 5 Depends on the Intracellular Redox State
Contingent upon H 2 O 2 concentration, Ag 5 catalyzes the irreversible oxidation of GSH and Trx. Thus, we reasoned that cancer cells with an increased generation or reduced disposal of H 2 O 2 will be more sensitive to Ag 5 . H 2 O 2 is generated continuously in cells that consume oxygen and it is well established that during proliferation cell respiration is increased and thus, could be used to target tumor redox vulnerability [32] (see Figure S27a, Supporting Information). Therefore, we explored whether the increased oxidative stress necessary for tumor cell division [33] is sufficient for Ag 5 to trigger cell demise. First, we confirmed that proliferation increased the A549 intracellular oxidative state, assessing the fluorescence due to 2″,7″-dichlorofluorescein (DCF) oxidation, compared to that of nonproliferating A549 cells, consistent with an increased oxidative state (see Figure S27b, Supporting Information). In good agreement, Ag 5 did not affect the viability of confluent non-proliferating cells; however, proliferating A549 cells viability decreased (Figure 3a). Likewise, serum starvation, a condition widely used to induce reversible cell  growth arrest (see Figure S27c, Supporting Information), also rendered A549 cells less sensitive to Ag 5 . We assumed that Ag 5 has a marked differential effect on proliferating and resting cells due to the displacement of the intracellular redox balance towards an oxidative state brought about by cell division. Corroborating this hypothesis, H 2 O 2 administration strongly sensitized both serum-starved and confluent nonproliferating A549 cells to the cytotoxic effect of Ag 5 (Figure 3b). Furthermore, mito-TEMPO, which catalyzes the dismutation of superoxide into H 2 O 2 specifically in the mitochondria, increased the cytotoxicity of Ag 5 compared to that with Ag 5 alone in A549 cells (Figure 3c), indicating that mitochondrial H 2 O 2 levels are important for Ag 5 activity. Conversely, Ebselen has been reported to decrease ROS levels [34] and it caused a dose-dependent reduction in the cytotoxic effect of Ag 5 on proliferating A549 cells and A2780 cells (see Figure S27d, Supporting Information). To further evaluate the role of the intracellular redox environment in the intracellular catalytic activity of Ag 5 , we treated A549 and A2780 cells with 2-deoxyglucose (2DG) that produce oxidative stress [35] and, accordingly increased Ag 5 cytotoxicity ( Figure 3e) (see Figure S27e, Supporting Information). Further confirmation of the link between Ag 5 action and the intracellular oxidative state comes from comparing the cell oxidative state, as seen by DCF fluorescence, among the A549 and A2780 cell lines. A2780 cells shown an increased fluorescence and are more sensitive to Ag 5 (see Figure S27f, Supporting Information). In summary, in the presence of an intracellular oxidative environment, Ag 5 accelerates oxidative reactions, leading to programmed cell death. However, in a reductive intracellular environment, Ag 5 has no biological effects.

Ag 5 Activate Redox Signaling Pathways
Ag 5 mediated oxidation of GR and PRDX3 should lead to oxidative stress and, consequently, activation of redox signaling pathways. Among the most well-known redox signaling pathways are ASK1, [36] NRF2-KEAP1, [37] and MTF1 [38] pathways. NRF2 and MTF1 are nucleocytoplasmic shuttling proteins that accumulate in the nucleus upon activation. In the presence of Ag 5 , we observed the nuclear accumulation of NRF2 ( Figure 3f) (see Figure S28a,b, Supporting Information) and MTF1 (Figure 3f) (see Figure S28c, Supporting Information). Confirming previous findings, we observed a significative reduction in the nuclear accumulation of MTF-1 and NRF2 in HEK293T cells treated with Ag 5 in the presence of DTT (0.5 × 10 −3 m) (see Figure S28b,c, Supporting Information). Moreover, qPCR confirmed the overexpression of genes under NRF2 transcriptional regulation, HMOX1 and GCLM, and genes transcriptional regulated by MTF-1, MT1F and MT1X in a human multiple myeloma MM.1S cell line (Table S3, Supporting Information). Furthermore, activation of the ASK1 pathway was assessed by analyzing its downstream effector c-Jun N-terminal kinase (JNK). [39] Ag 5 treatment caused a dose-and time-dependent increase in JNK phosphorylation, indicating ASK1 activation (Figure 3h,i, respectively). JNK phosphorylation activates mitochondrial death factors such as Bim and Bax, resulting in cytochrome c release and initiation of apoptosis [40] suggesting that various effectors can mediate apoptosis triggered by Ag 5 .

X-Ray Irradiation Enhances the Intracellular Catalytic Activity of Ag 5
Radiotherapy forms free radicals, notably OH•, which then damages the DNA. XANES experiments suggested that the increase of OH• in irradiated cells could sensitize them to Ag 5 . Of relevance, Ag 5 , in combination with X-ray irradiation at even at the lowest absorbed dose, 2 Gy, decreased cell viability, indicating X-ray irradiation cooperates to increase Ag 5 cytotoxicity (Figure 3j) (see Figure S29a, Supporting Information). Importantly, X-ray irradiation-induced potentiation of the effect of Ag 5 on cell death was not due to a concomitant increase in phosphorylation of the histone H2AX (γ−H2AX), a marker for DNA damage (see Figure S29b, Supporting Information), suggesting that Ag 5 acts through a different cell death mechanism, probably using OH• generated by X death-ray irradiation to accelerate irreversible Cys oxidation, as XANES experiments suggests.

Ag 5 Preferentially Cause Cell Death in RAS-Transformed Cells and Hypoxic Cells
Activated oncogenes cause metabolic changes that give rise to the accumulation of ROS. [41][42][43] We therefore reasoned that compared to non-transformed cells, transformed cells would be preferentially killed by treatment with Ag 5 . To evaluate this hypothesis, we introduced a doxycycline-inducible activated allele of HRAS (HRAS G12V ) into non-transformed immortalized mouse fibroblasts (W3T3 cells) and confirmed inducible HRAS expression by western blotting (Figure 4a). Interestingly, cells expressing HRAS had greater DCF fluorescence, indicating increased oxidation ( Figure 4b) and were more sensitive to cell death induced by Ag 5 activity than noninduced cells (Figure 4c). These effects were inhibited by DTT coadministration, confirming that HRAS-induced redox changes are responsible for the increased effects of Ag 5 (see Figure S30, Supporting Information).
Metabolic phenotypes in tumors are both heterogeneous and flexible [44,45] and thus may affect Ag 5 therapeutic efficacy. Conditions within tumoroids or multicellular tumor spheroids (MCTSs) resemble the pathophysiological conditions within human tumor tissue in many ways and are widely used for drug testing. [46] Like human tumors, large MCTSs, more than 600 µm diameter, are characterized by a heterogeneous cell subpopulation, with actively proliferating cells on the periphery and quiescent, hypoxic, and necrotic cells in the inner regions. [47] Therefore, to assess the effects of the Ag 5 on hypoxic cells, we developed A549 cell MCTSs as an ex vivo tumor model. The MCTSs were treated with Ag 5 (4.5 × 10 −6 m) four times (days 0, 2, 4, and 6, with the first day of treatment set as day 0) or vehicle (control) and examined daily. Ag 5 treatment decreased the area of the MCTSs over the seven-day period (see Figure S31, Supporting Information). Notably, a less dense region was observed in the center of the MCTSs treated with Ag 5 , but not the control MCTSs, suggesting decreased cellularity that appears to be more pronounced in hypoxic centrally located cells of the MCTSs (Figure 4d, red arrow). Comparable results were obtained in MCTSs formed from HCT116, indicating that cellular distribution rather than the cell line, was responsible for the effect (Figure 4e). The presence of these Ag 5 sensitive inner regions in the MCTSs after Ag 5 treatment could be related to the ability of the Ag 5 , given their small size and neutral charge, to penetrate the MCTSs and reach these central hypoxic regions commonly found in human tumors. [48] Then, because both hypoxia and low-glucose conditions increase intracellular oxidation, [33] we hypothesized that they would sensitize cells to Ag 5 activity. To assess this hypothesis, we first measured the degree of hypoxia in the tumoroids using a fluorescent probe to confirm the presence of hypoxic conditions and clearly observed in a 1000-cell tumoroid (Figure 4f, top row). Interestingly, exposure of the tumoroids to Ag 5 at increasing doses caused a dose-dependent reduction in hypoxic cells (Figure 4f). Nevertheless, a fraction of the cells in the tumoroid is resistant to Ag 5 , which we tentatively ascribed to nonproliferating cells with a higher intracellular reductive state; further characterization is now needed to establish whether cellular resistance is permanent or can be reversed when cells proliferate again.

Ag 5 Show Antitumoral Activity In Vivo
To determine whether the in vitro results described above would be confirmed through in vivo experiments, we evaluated the anti-tumoral activity of Ag 5 in vivo by using an orthotopic model of metastatic lung cancer [49] in which metastatic cells can be detected in the mediastinal lymph nodes 13 days after the injection of cancer cells into the lung. Therefore, to evaluate the effects of Ag 5 on an advanced cancer in which metastasis has occurred, we intravenously administered Ag 5 20 days after the implantation of cancer cells in the lung, when lymph node metastases were already evident. Tumor growth was measured over time by in vivo bioluminescence imaging, and the cancer cell burden from both lymph node metastases and the primary tumor was assessed ex vivo at the time of sacrifice. Cisplatin (CDDP), a first-line treatment for metastatic lung cancer, was used as positive control for tumor growth inhibition. As shown in Figure 4g, treatment with Ag 5 significantly reduced tumor growth compared to that in untreated animals, in which the tumors grew exponentially. Additionally, compared to control mice, Ag 5 -treated mice exhibited significantly reduced luciferase activity in both the primary tumor and mediastinal lymph nodes at the time of sacrifice (Figure 4h). Further immunohistochemical analysis of lung sections from the mice confirmed the antitumor effect of Ag 5 (Figure 4i). Additionally, no signs of toxicity, either by body weight loss or by blood cell counts, were observed in the mice after treatment with Ag 5 at the therapeutic doses used, suggesting their potential safety in humans (Table S4 and Figure S32, Supporting Information). Together, these results show the ability of Ag 5 to significantly decrease both the primary tumor and metastatic cell burden in an advanced metastatic cancer.

Conclusion
There are very few redox active therapeutic modalities. Here we have shown that the unique intracellular chemistry of Ag 5 can selectively target and disrupt cell metabolism in cells with elevated redox stress, such as cancer cells. Treating tumors based on their unique genetic profile has proved difficult, exposed the patients to severe toxicity when either monotherapy or combination therapy has been used, and acquired resistance is common. [50] As an alternative, targeting metabolic vulnerabilities shared by tumors has generated growing interest in recent years. Here we show that Ag 5 triggers tumor cell death by acting synergistically with the elevated levels of redox stress in tumor cells. Importantly, in most normal resting cells which have a low level of redox stress, so H 2 O 2 levels are low, while the oxidation of their thiols is accelerated by Ag 5 , the unstressed antioxidant systems in these cells can maintain homeostasis. However, if H 2 O 2 levels increase above a threshold level, for example in tumors, antioxidant systems are overwhelmed because they cannot counteract the increased oxidation rate caused by the presence of Ag 5 . Also, if the antioxidant systems are affected, the necessary amount of H 2 O 2 to produce cell death in the presence of Ag 5 is lower. Therefore, Ag 5 does not induce cell death in the balanced redox environment in normal cells, but when the redox balance is skewed in the presence of elevated intracellular H 2 O 2 production, Ag 5 can catalyze the irreversible oxidation of thiol groups in key molecules of both the GSH and Trx antioxidant pathways, particularly in the mitochondria that ultimately leads to oxidative stress and cell death. Of relevance, our results indicate that X-ray irradiation forming free radicals sensitized cells to Ag 5 ; the combination of Ag 5 with X-ray irradiation, at even at the lowest absorbed dose, improves the therapeutic ratio of radiation, which now deserves further studies.
Hypoxia is strongly associated with poor prognosis [51] and is a leading cause of therapy resistance, [52] the finding that Ag 5 was effective in reducing hypoxic regions of MCTS is of interest as a potential new approach for the treatment of tumor hypoxic cells. Therapeutic efficacy of Ag 5 was demonstrated in an orthotopic model of metastatic lung cancer in which Ag 5 was shown to inhibit the growth of both the primary tumor and the mediastinal lymph node metastases. These findings are of particular interest because high ROS levels have not only been described in proliferating tumor cells, but also in metastatic settings. Our findings suggest the unique intracellular redox chemistry of quantum clusters may lead to new redox-based approaches to cancer therapy.

Experimental Section
Synthesis of Ag Clusters: Unless otherwise specified, all reagents were purchased from Sigma Aldrich, Co., Spain. Thioredoxin (TRX-01 Thioredoxin1, E. coli (native)) was purchased from IMCO Corp Ltd AB. Silver sheets (99%) were purchased from Goodfellow Cambridge Ltd., Huntingdon, UK. Alumina nanoparticles (average size ≈ 50 nm) and cloth pads were purchased from Buehler, Düsseldorf, Germany.
The synthesis was carried out with a Biologic VMP300 potentiostat (Seyssinet-Pariset, France). A Methrom thermally insulated three-electrode electrochemical cell, deoxygenated just prior to the synthesis, was used with a hydrogen electrode as a reference and two Ag foils (17.5 cm 2 surface area) as counter and working electrodes. The experimental conditions were similar to those reported before, [14,15] but the voltage was increased with respect to the preparation of Ag 3 clusters (10 V) because in this way larger clusters are produced. This (besides the electrode's polishing, which determines the surface's roughness) is crucial for achieving the desired cluster's size, as it will be reported elsewhere. It has to be noted that Ag clusters with larger sizes were also synthesized (mainly consisting in Ag 7 and Ag 9 , see inset in Figure S4, Supporting Information) increasing further the voltage, to compare their biological properties with the ones reported here. Prior to the synthesis, both silver electrodes were polished with sandpaper followed by alumina (≈50 nm), washed thoroughly with MilliQ water and sonicated. An ion-selective electrode was used to verify that, after synthesis, the concentration of remaining Ag + ions are always less than 3 mg L −1 after precipitation with NaCl, and various techniques were used to characterize the synthesized Ag clusters (see also below). The samples were finally concentrated at 50 °C using a rotary evaporator (Buchi Rotavapor R-210 at a pressure of 2 mbar) (Massó Analítica S.A., Barcelona, Spain) to a final concentration of ≈30 mg L −1 as determined by flame atomic absorption spectroscopy.
Cluster's Characterization Methods: In Ag clusters' characterization methods, both UV-vis and fluorescence spectroscopy experiments were performed at room temperature using 1 cm path length Hellma quartz cuvettes (Hellma GmbH & Co. KG., Müllheim, Germany). UV-vis spectra were recorded with an Analytik Jena Specord S600 spectrometer (Analytik Jena AG, Jena, Germany) with a diode array detector, and fluorescence spectra were recorded with a Cary Eclipse Varian fluorimeter (Agilent Technologies Spain, S.L., Madrid, Spain).
Electron microscopy studies were performed on a FEI Titan Themis 60-300 Double Aberration Corrected microscope operated at 200 kV. The aberrations of the condenser lenses were corrected up to fourth order using the Zemlin tableau to obtain a sub-Angstrom electron probe. A condenser aperture of 50 µm yielding an electron probe with a convergence angle of 20 mrad was used. To avoid sample modification under the electron probe a beam current of 0.025 nA was used. STEM samples were prepared by depositing one drop of Ag clusters' solution onto holey carbon coated Cu grids. After preparation, samples were maintained under vacuum conditions. Electrospray ionization time-of flight (ESI-TOF) mass spectrometry measurements were performed using a Bruker MicroTOF massspectrometer operating in negative ionization mode. Temperature control and nitrogen drying gas (1 µL/144 min) in ESI source were employed to assist the ionization process. The ESI source conditions were as follows: source voltage −4.5 kV, heated capillary temperature 275 °C, capillary voltage −35 V and sheath gas and auxiliary gas 5 and 2 (N 2 , arbitrary units). For full-scan MS analysis, the spectra were recorded in the range of m/z 100 to 1000 with a scan speed of 1 scan/s. All spectra were acquired in reflectron mode of the TOF mass spectrometer equipped with multistep detection to obtain maximum sensitivity. The mass resolution was set at 15000 FWHM. The instrument was calibrated using a calibration solution according to the manufacturer's instructions. Due to the difficulty to charge the uncharged clusters [15] a systematic investigation was carried out to obtain the maximum sensitivity for the detection of Ag clusters. As it will be published elsewhere, concentrations of NaCl and formic acid are crucial to achieve this, approaching 0.1 ppm sensitivity for the optimal conditions. 2 µL of Ag clusters' solutions diluted 300 times to contain 25 × 10 −3 m of NaCl (which is the optimum NaCl concentration for Ag clusters' detection) was directly injected into a mobile phase of acetonitrile containing 0.1% formic acid at a flow rate of 0.2 mL min −1 . Details of the analysis of Ag clusters' peaks are given in the Supporting Information.
Ion concentration was measured using a previously calibrated pH and Ion-Meter GLP 22 (Crison Instruments S.A., Barcelona, Spain) by adding a stabilizing solution (NaNO 3 5 m) at a ratio of 2:100 to the sample at a constant temperature of 25 °C.
The total Ag content in the clusters' samples was analyzed by flame atomic absorption spectroscopy, performed with a PerkinElmer 3110 with an Ag hollow cathode lamp Lumia from PerkinElmer (Madrid, Spain) (current 10 mA).

XANES Characterization of Sulfur Oxidation States: S K-edge (2470 eV)
and L3 Ag-edge (3351 eV) X-ray absorption near edge structure (XANES) experiments were performed at de SXS beamline at the Laboratório Nacional de Luz Síncrotron (LNLS, Campinas, Brazil) which was equipped with an InSb (111) double crystal monochromator with slit aperture of 1 mm, to achieve a resolution of about 0.5 eV. Details of the experimental setup of the SXS beam line were published elsewhere. [53] X-ray absorption spectra were recorded in fluorescence mode, collecting the emitted X-ray from the S Kα 1 , 2 (at 2309.5 and 2308.4 eV, respectively) and Ag Lα 1 (at 2982.7 eV) emission lines for each measured edge. Absorption experiments were performed either in a vacuum of 10 −8 mbar at room temperature, and in a special liquid sample holder designed ad hoc for the experiment with reactive oxygen species, at room temperature and atmospheric pressure, allowing us to determinate initial reactions intermediaries as well as final compounds. The photon energy was calibrated by assigning the value 2481.5 eV to the highest maximum of Na 2 S 2 O 3 (corresponding to the so-called inner sphere), in accordance with the criteria previously reported by Vairavamurthy. [54] The final XANES spectra were obtained after background subtraction and normalization to the post edge intensity, following usual procedure described elsewhere. XANES quantification were performed with Athena software and subsequent analysis on Origin lab software. For GSH characterization, a fraction of solution was deposited by drop casting on carbon disks (Ted Pella, Inc) to have an Ag or S concentration in a detectable value. For Trx samples, they were carefully mounted in a liquid sample holder, designed ad hoc for this experiment. Phosphate buffered saline (PBS) solution was used as the solvent in all the reaction mixes with Trx, with the purpose to reproduce the same intracellular pH and ionic strength. The hydroxyl radical solutions were prepared by Fenton reaction, with H 2 O 2 and FeCl 2 . Concentration values are indicated in the Table S1, Supporting Information.
Theoretical Calculations: DFT and QM/MM calculations were performed on the systems shown in the Supporting Information, with full geometry optimization, using the B3LYP [55,56] functional, the 6-31G(d,p) [57,58] basis set for H, C, N, O, S atoms, and Lanl2dz [59] basis set for Ag. The ONIOM [60] method was used for two layers QM/MM calculations and the UFF [61] force field was used in the lower layer. The implicit water solvent was considered through the C-PCM [62] method. Transition state structures were found by using the synchronous transit guided quasi-Newton method. [63] Frequency calculations, within the harmonic oscillator approximation, were performed on the optimized geometries, to confirm that they corresponded to minimum energy structures (all real vibration frequencies) or transition states (one imaginary vibration frequency). All calculations were performed by the Gaussian09 [64] program package. Animals: All mice research in this study was done in accordance with Spanish and European regulations and approved by the USC Ethics Committee (project ID 15010/14/001). The orthotopic lung cancer model was generated by injection of luciferase-expressing A549 cells into the left lung of nude mice essentially as described. [14,49] Tumor-bearing animals were treated with cisplatin (four doses, 4 mg kg −1 ), Ag 5 (four doses, 0.25 mg kg −1 ) or left untreated (control group). Treatments were administered intravenously. In vivo tumor evolution was followed by bioluminescence in anesthetized mice using an IVIS Spectrum Image System (Perkin Elmer). Upon sacrifice ex vivo luminescence and histological analysis were performed as described previously. [14] In histological analysis, lungs were fixed in 10% neutral buffered formalin for 24 h and embedded in paraffin. Sections 4 mm thick were mounted on FLEX IHC microscope slides (Dako-Agilent, Glostrup, Denmark) and heated at 60 °C for 1 h. The immunohistochemical technique was automatically performed using an AutostainerLink 48 (Dako-Agilent). After deparaffination and epitope retrieval in EnVision FLEX target retrieval solution (high pH) for 20 min at 97 °C, the slides were allowed to cool in PT Link to 65 °C and then in Dako wash buffer for 5 min at room temperature. The immunostaining protocol included incubation at room temperature in: 1) EnVision FLEX peroxidaseblocking reagent (Dako-Agilent) for 5 min; 2) ready-to-use FLEX primary antibody (Dako-Agilent) anti-CK7 (clone OV-TL 12/30), for 20 min; 3) EnVision FLEX/HRP (dextran polymer conjugated with horseradish peroxidase and affinity-isolated goat antimouse and antirabbit immunoglobulins) for 20 min; 4) substrate working solution (mix) (3,3′-diaminobenzidine tetrahydrochloride chromogen solution) (Dako-Agilent) for 10 min; and 5) EnVision FLEX hematoxylin (Dako-Agilent) for 9 min. Sections were examined and photographed using an Olympus PROVIS AX70microscope equipped with an Olympus DP70 camera.
All rats' experiments were carried out in accordance with the UK Animals (Scientific Procedures) Act of 1986 and the University of Cambridge Animal Welfare Policy. Wistar rats (003, wildtype, female) were ordered from Charles River Laboratories UK (Margate, UK). They were housed under standard laboratory conditions with ad lib food and water and used between 10 and 12 weeks. Animals were culled by cervical dislocation with accordance to UK Home Office Schedule 1.
Mammalian Mitochondrial Preparations: Mitochondria isolation from rat liver and rat heart was prepared as described previously. [8] Bovine heart mitochondrial membranes (BHM) were prepared as described previously. [65,66] Protein concentration was determined using the bicinchoninic acid (BCA) assay with BSA as a standard.
Both, cellular and the mitochondrial pellets were digested in HNO 3 at 90 °C for 4 h. Absorbance was measured at 595 nm using a CLARIOstar microplate reader. For nonproliferating cells, proliferation was stopped using two approximations: 1) confluency: cells were allowed to grow during 96 h in medium-10% of FBS to reach the confluence and 2) serum starvation: 24 h after seeding medium was replaced with medium-0.05% of FBS for 72 h. Then, cells were treated with 1) Ag 5 for 1 h and incubated in complete medium for 24 h more or 2) with H 2 O 2 (200 × 10 −6 m) alone or in combination with Ag 5 for 1 h. Then, the medium was replaced with complete medium for 24 h more and cell death was measured by the MTT assay as described above.
Trx and TrxR Activity: Trx and TrxR activity was assessed using the Kit for assays of mammalian Trx and TrxR (IMCO) following manufacturer's instructions.
To measure Trx activity, Trx and TrxR were incubated with different doses of Ag 5 for 30 min following manufacturer's instructions. Then, absorbance was measured at 412 nm using a CLARIOstar plate reader, BMG Labtech. The effect of Ag 5 in Trx activity was determined by the inhibition of TNB signal.
TrxR activity was measured using two electron acceptors: 1) Selenocysteine (protocol adapted from Cuniff et al. [67] ). TrxR from rat liver (100 ng dissolved in PBS) was incubated with Ag 5 (1.13 × 10 −6 m) or Ag + (at the same Ag concentration) for 10 min in a 96-well plate. After that, a solution of NADPH and Selenocysteine was added to each well at a final concentration of 400 × 10 −6 m. Immediately, NADPH consumption was measured at 340 nm for 30 min using a SpectraMax platereader, Molecular Devices. 2) H 2 O 2 (protocol adapted from Zhao et al. [68] ). TrxR from rat liver (17 × 10 −9 m, dissolved in Tris-Cl pH 7. For the anaerobic conditions, insulin was incubated in TE buffer and DTT as described above in a Baker Ruskinn Invivo2 200 hypoxic chamber (0.2% O 2 and 5% CO 2 ) prior deoxygenation of samples for 30 min.
Then, insulin sample was tested in the presence of DTT (control) or passed through a Illustra MicroSpin G-25 Column, GE Healthcare containing Sephadex as manufacturer's instructions to remove DTT. After that, 5 µL of deoxygenated Ag 5 (9.6 × 10 −6 m) were added and after a 30 min incubation in hypoxic conditions, DTNB was incorporated and absorbance was measured as described above.
GSH Oxidation in Living Cells: A549 cells (2.5 × 10 4 cells) were seeded in 35 mm plate dishes (Mattek) and transduced with the Premo Cellular Redox Sensor Grx-1-roGFP following manufacturer's instructions. After that, Ag 5 (IC 50 ) were added to the dish and cells imaged during 10 min. Images were taken every 10 s using a Leica TCS SP8 confocal microscope. Fluorescence intensity emitted by each cell was measured at 405 and 488 nm and emission was collected at 500-530 nm to calculate the ratio 405/488. Images were processed using the ImageJ software. Ratio images were created using the RatioPlus Plugging and the ImageJ Look Up table "green fire blue" was used to create false-color ratio images.
GSH/GSSG Ratio: A549 cells (6 × 10 3 cells) were seeded in 96-well clear bottom white microplates and 24 h later treated with Ag 5 (IC 50 ) for 1 h. Then, oxidized, and total glutathione were measured using the GSH/GSSG-Glo Assay according to the instructions of the kit using a CLARIOstar microplate reader.
PRDX3 Transfection: HEK293 cells were transfected with pcDNA3.1-PRXD3 (GenScript, clone ID OHu14665) or the empty vector using calcium phosphate. At 24 h, the cells were counted and seeded on 96-well plates previously covered with polylysine. After one day, Ag 5 was added at the indicated amounts for 1 h, washed and cells were allowed to grow for additional 24 h. Cell viability was measured by the MTT assay. PRXD3 overexpression was checked by western blot.
Correlation Between PRDX3 mRNA Expression and Ag 5 Sensitivity: An ED50 value was obtained for each cell line by fitting a two-parametric log-logistic model with lower limit at 0 and upper limit at 1 using the drm package [69] to cell viability values relative to the negative control, as a function of drug concentration. Gene expression data were obtained from https://depmap.org/portal/download, Version 20Q2 [70] and available for 85 of our cell lines. Their correlation was assessed using Spearman's rank correlation implemented in the cor.test function of R.
Determination of ROS Using Fluorescent Dyes: H 2 DCFDA staining. A549 (3 × 10 4 cells per well) and A2780 (3.5 × 10 4 cell per well) were seeded in 24-well plate and allowed to grow for 24 or 96 h for proliferating and nonproliferating cells. W3T3 RasV12 (5 × 10 4 cells per well) were seeded in 12-well plates. Then, cells were trypsinized, washed with PBS and incubated with CM-H 2 DCFDA for 30 min at 37 °C. Stained cells were analyzed using the Guava EasyCyte flow cytometer and processed with the FlowJo program.
Immunoblotting: Unless specified, protein extracts from cultured cells were extracted using RIPA buffer with protease inhibitor cocktail (Sigma). After the electrophoresis proteins were transferred to PVDF membranes and blotted against the indicated antibodies.
PDRX3 and GR Assay: For the isolated mitochondria assays, mitochondria (0.5 mg mL −1 ) were incubated in KCl buffer supplemented with glutamate/malate 5 × 10 −3 m and different doses of Ag 5 or AgNO 3 at 37 °C for 5 min. After that, methyl methanethiosulfonate (MMTS) was added up to 80 × 10 −3 m. Then samples were centrifuged at 7500 × g for 5 min. The supernatants were removed, and the pellets suspended in loading buffer with MMTS (80 × 10 −3 m) and heated at 95 °C for 5 min. Protein was determined by western blotting with anti-PRDX3. Membrane was visualized with an Odyssey CLx.
For the cell assays, A549 cells (2 × 10 6 cells) were seeded on 100 mm cell culture dishes and 24 h later treated with Ag 5 (37.7 × 10 −6 m) in a 1:50 dilution for 15 min. Then, the treatment was removed and medium with MMTS (80 × 10 −3 m) was added to the dish and incubated for 10 min at room temperature. After that, cells were washed with cold PBS and lysed with RIPA buffer with protease inhibitor cocktail containing MMTS (80 × 10 −3 m) in nonreducing conditions. To quantify the amount of protein, an untreated control was performed without MMTS. After that, western blotting was performed by using anti-PRDX and anti-GR.
Immunofluorescence: A549 (3 × 10 4 cells per well), A2780 (8 × 10 4 cells per well), and HEK293T (8 × 10 4 cells per well) cells were grown on 24 mm coverslips and 24 h later treated with: 1) Ag 5 (IC 50 ) for 10 or 30 min, 2) DTT (0.5 × 10 −3 m) for 10 min, 3) Ag 5 (IC 50 ) and DTT (0.5 × 10 −3 m) for 10 min, and 4) NEM (100 × 10 −6 m) for 30 min. After treatment, cells were washed with PBS, fixed in formalin solution (10%) for 30 min, washed with PBS and permeabilized with 0.5% Triton X-100 in PBS for 5-10 min. Then cells were washed again and blocked in PBS containing 1% BSA for 1 h. After that, samples were incubated with the primary antibodies anti-Keap1 (overnight at 4 °C) or anti-MTF1 (2 h at room temperature). Following washing with PBS, the samples were incubated for 45 min with the corresponding secondary antibody, AlexaFluor-488 donkey anti-goat for Keap, AlexaFluor-594 goat anti-rabbit for MTF1 and Hoechst (to counterstain the nucleus). Then, for double staining NRF2/Keap1, the cells were washed and incubated with the primary antibody against NRF2 for 2 h at room temperature, washed with PBS and incubated with the AlexaFluor-594 goat anti-rabbit and Hoechst (to counterstain the nucleus) for 45 min. Coverslips with stained cells were mounted on glass slides with Fluoroshield Mounting Medium. Images were obtained using the Olympus IX51 microscope equipped with an Olympus DP72 camera and CellSens Imaging Software and processed with ImageJ or the Leica TCS SP8 confocal microscope and analyzed using LasX software.
Caspase 3-7 Assay: A549 cells (3 × 10 4 ) were seeded in 24-well plate and 24 h later cells were treated with Ag 5 (IC 50 ), in the presence of the CellEvent Caspase 3/7 reagent (8 × 10 −6 m). Time-lapse images were acquired every minute during 1 h using an inverted Leica DMI600B Wide-Field microscope equipped with an Okolab incubator controlled by an Okolab temperature controller and a DFC360FX Camera. Images were processed with the LAS AF software.
Annexin-V Assay: A549 (6 × 10 4 ) cells were seeded in 12-well plate and 24 h later treated with Ag 5 (IC 50 ) for 30 min. Then, cells were washed twice with PBS and stained with Annexin V-FITC and 7-AAD following manufacturer´s instructions. Images were obtained using Olympus IX51 microscope equipped with an Olympus DP72 camera and CellSens Imaging Software and processed with ImageJ.
Cell Irradiation: Cells were irradiated with 6 MV X-Ray beams generated by a Varian Medical Systems (Palo Alto, California) CLINAC 2100 C/D High energy linear accelerator at Radiation Physics Laboratory of University of Santiago de Compostela (http://www.usc.es/rpl). Absorbed dose to cells was determined by ionometric dosimetry corrected by Monte Carlo simulations. Irradiations were carried out by placing cells in 24-well plates surrounded by water equivalent plastic for electronic build up. A549 (3 × 10 4 cells per well) were seed in 24-well plates and allowed to grow for 96 h until they reach confluence (nonproliferative state). Then, cells were treated with Ag 5 (1.5 × 10 −6 , 2 × 10 −6 , 2.2 × 10 −6 m) for 10 min and irradiated. Target absorbed doses (between 2 and 6 Gy) were delivered by single fractions at a rate of 2 Gy min −1 . After irradiation, cells were incubated with Ag 5 until complete 1 h of treatment and then trypsinized and reseeded for colony formation. 8 days later, cultures were fixed with crystal violet and colonies (>50 cells) were counted. Survival fractions (SF) were fitted to the linear quadratic model using GraphPad Prism 5 software (GraphPad Prism Software Inc., La Jolla, CA, USA).
RNA Extraction, Reverse Transcription, and Quantitative Real-Time PCR (qRT-PCR): MM.1S cells (4 × 10 6 ) were seeded in 60 mm culture dishes and treated with Ag 5 (IC 50 ) for 30 min. Then, treatment was removed and 4 h later cells were collected, and their RNA isolated using the kit NucleoSpin ARN (Macherey-Nagel) following the manufacturer instructions. Reverse transcription was performed using the M-MLV reverse transcriptase, according to the manufacturer's instructions and analyzed by qRT-PCR using the TaqMan chemistry. Relative changes in gene expression were determined using the 2 −∆∆Ct method. [71] Multicellular Tumoroids: A549 and HCT116 multicellular tumoroids were generated by the hanging-drop method as described previously. [72] 500 cells/20 µL were dispensed into each well of a MicroWell MiniTray (Nunc). At day 5 upon the inversion of the tray, spheroids were transferred to a 96-well plate coated with agarose (1%) and treated with Ag 5 four times on alternate days. To monitor spheroids growth images of each spheroid were taken every day using an Olympus IX51 microscope. Spheroid's area was measured using ImageJ. At the end of the experiment, A549 spheroids were stained with Image-iT Green Hypoxia Reagent 5 × 10 −6 m and Hoechst (1 µg µL −1 ) for 1 h. Images of control and treated spheroids were taken on a Leica AOBS-SP5 confocal microscopy and analyzed using ImageJ software. HCT116 spheroids were stained with CellEvent Caspase 3/7 reagent (8 × 10 −6 m), Propidium iodide (1 µg µL −1 ) and Hoechst (1 µg µL −1 ) for 30 min. Then the spheroids were mixed with low-melt agarose solution (1%), the mixture was sucked into 1 mL syringes (diameter 1 mm) and allowed to gel at room temperature. After that the agarose blocks containing spheroids were imaged using Light Sheet Fluorescence Microscopy (LMSF Carl Zeiss Microscopy GmbH) and the images were processed using the Imaris Cell Imaging Software.
Statistical Analysis: All statistical analyses were performed with GraphPad Prism Version 5.0 software (GraphPad Software, Inc., La Jolla, USA). The differences were considered significant for p < 0.05. A two-way ANOVA with Bonferroni's correction was used for multiple group comparisons. Comparisons between two groups were analyzed using the 2-tailed Student's test. Mann-Whitney test was used for mice experiments.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.