N-Alkylaminoferrocene-Based Prodrugs Targeting Mitochondria of Cancer Cells

Intracellular concentration of reactive oxygen species (e.g., H2O2) in cancer cells is elevated over 10-fold as compared to normal cells. This feature has been used by us and several other research groups to design cancer specific prodrugs, for example, N-alkylaminoferrocene (NAAF)-based prodrugs. Further improvement of the efficacy of these prodrugs can be achieved by their targeting to intracellular organelles containing elevated reactive oxygen species (ROS) amounts. For example, we have previously demonstrated that lysosome-targeted NAAF-prodrugs exhibit higher anticancer activity in cell cultures, in primary cells and in vivo (Angew. Chem. Int. Ed. 2017, 56, 15545). Mitochondrion is an organelle, where electrons can leak from the respiratory chain. These electrons can combine with O2, generating O2−• that is followed by dismutation with the formation of H2O2. Thus, ROS can be generated in excess in mitochondria and targeting of ROS-sensitive prodrugs to these organelles could be a sensible possibility for enhancing their efficacy. We have previously reported on NAAF-prodrugs, which after their activation in cells, are accumulated in mitochondria (Angew. Chem. Int. Ed. 2018, 57, 11943). Now we prepared two hybrid NAAF-prodrugs directly accumulated in mitochondria and activated in these organelles. We studied their anticancer activity and mode of action. Based on these data, we concluded that ROS produced by mitochondria is not available in sufficient quantities for activation of the ROS-responsive prodrugs. The reason for this can be efficient scavenging of ROS by antioxidants. Our data are important for the understanding of the mechanism of action of ROS-activatable prodrugs and will facilitate their further development.


Introduction
Chemotherapy along with surgery and radiotherapy remains the most efficient and broadly applied method of cancer treatment [1]. However, despite the fact that some types of cancers are well responsive (e.g., testicular cancer to Pt(II)-based drugs [2]), chemotherapy still suffers from massive side effects and development of resistance in many other cases [3]. Therefore, the search for new side effects and development of resistance in many other cases [3]. Therefore, the search for new chemotherapeutic approaches lacking the mentioned drawbacks is warranted. One sensible possibility is to use prodrugs, which are activated under cancer specific conditions, but remain inactive and therefore nontoxic in normal cells. For example, the intracellular concentration of reactive oxygen species (ROS) (e.g., hydrogen peroxide, H2O2) in most cancer cells is elevated by more than 10-fold in comparison to normal cells [4][5][6]. This difference in the chemical composition of cellular media was already used by us [7][8][9][10][11] and other research groups [12][13][14][15][16][17] to design experimental cancer-specific prodrugs activated by H2O2. In particular, we developed Nalkylaminoferrocene-(NAAF) based prodrugs, which are activated in the presence of ROS with the formation of NAAF drugs (Scheme 1). The latter drugs amplify the ROS concentration in cancer cells, leading to their death via apoptotic and necrotic pathways. We demonstrated that NAAF-prodrugs exhibit anticancer activity towards a range of cancer cell lines [7][8][9][10][11], in vivo in several mice [9,10,18] and rat cancer models [11] as well as towards primary cancer cells [8,10,18]. Scheme 1. N-alkylaminoferrocene-(NAAF) based anticancer prodrugs targeting mitochondria due to a covalently attached carrier moiety X + . Intermediates 1, 2, 3, p-quinone methide and CO2 as well as the drug are formed from the prodrug in the medium containing high amounts of reactive oxygen species (reactive oxygen species (ROS), e.g., in cancer cells).
Since ROS are distributed heterogeneously in cells, targeting oxidatively activated prodrugs to intracellular areas with high amounts of ROS can be a feasible strategy for the enhancement of their efficacy. We have recently demonstrated the validity of this approach for the NAAF-prodrugs by redirecting them to lysosomes (LY's) [18], which are known to accumulate ROS as well as redox active transition metal ions. We have found that the LY-targeting NAAF-prodrugs are activated much more efficiently in cancer cells than their non-targeted analogs, which is reflected in their higher anticancer activity in cell cultures, primary cells and in vivo [18]. Apart from LY's, mitochondria of cancer cells are also considered as an important source of intracellular ROS due to leakage of electrons in the respiratory chain. However, in contrast to LY's lacking catalase and peroxidases, mitochondria are well protected by the ROS-neutralizing enzymes. Therefore, it is not clear whether mitochondriaderived ROS are available in cells in sufficient amounts to be explored for prodrug activation (pathway I in Scheme 1) or they are rather neutralized by the cellular antioxidative system shortly after their formation (pathway II, Scheme 1). We have previously reported on NAAF-prodrugs, which are activated by ROS in the cytoplasm of cancer cells with the generation of lipophylic cationic drugs. The latter drugs are accumulated in mitochondria [19]. In contrast, herein we report on Scheme 1. N-alkylaminoferrocene-(NAAF) based anticancer prodrugs targeting mitochondria due to a covalently attached carrier moiety X + . Intermediates 1, 2, 3, p-quinone methide and CO 2 as well as the drug are formed from the prodrug in the medium containing high amounts of reactive oxygen species (reactive oxygen species (ROS), e.g., in cancer cells).
Since ROS are distributed heterogeneously in cells, targeting oxidatively activated prodrugs to intracellular areas with high amounts of ROS can be a feasible strategy for the enhancement of their efficacy. We have recently demonstrated the validity of this approach for the NAAF-prodrugs by re-directing them to lysosomes (LY's) [18], which are known to accumulate ROS as well as redox active transition metal ions. We have found that the LY-targeting NAAF-prodrugs are activated much more efficiently in cancer cells than their non-targeted analogs, which is reflected in their higher anticancer activity in cell cultures, primary cells and in vivo [18]. Apart from LY's, mitochondria of cancer cells are also considered as an important source of intracellular ROS due to leakage of electrons in the respiratory chain. However, in contrast to LY's lacking catalase and peroxidases, mitochondria are well protected by the ROS-neutralizing enzymes. Therefore, it is not clear whether mitochondria-derived ROS are available in cells in sufficient amounts to be explored for prodrug activation (pathway I in Scheme 1) or they are rather neutralized by the cellular antioxidative system shortly after their formation (pathway II, Scheme 1). We have previously reported on NAAF-prodrugs, which are activated by ROS in the cytoplasm of cancer cells with the generation of lipophylic cationic drugs. The latter drugs are accumulated in mitochondria [19]. In contrast, herein we report on prodrugs, which are first accumulated in mitochondria and can then react with mitochondrial ROS (Scheme 1). Using these prodrugs, we clarified an important question of whether mitochondria-targeting of oxidatively activated NAAF-prodrugs favorably affects their anticancer activity. The results obtained are relevant to all other types of ROS-responsive prodrugs. They will facilitate the further development of these compounds as cancer specific drugs.
To determine active species responsible for the ROS generation, we acquired electrospray ionization (ESI) mass spectra of prodrug 5 with and without H 2 O 2 ( Figure 2, Figure S15). To determine active species responsible for the ROS generation, we acquired electrospray ionization (ESI) mass spectra of prodrug 5 with and without H2O2 (Figure 2, S15). We could detect prominent peaks corresponding to hydrolyzed prodrug 5 (5*) as well intermediates 1 and 3 (two peaks corresponding to [M] + and [M-e − + I − ] + were observed for 3) in the mass spectrum of the mixture of 5 and H2O2, whereas in the absence of H2O2 only intact (5) and hydrolyzed (5*) species were observed. These data confirm the mechanism of prodrug 5 activation outlined in Scheme 1. The ROS-generation assay ( Figure 1A) could not be accurately conducted for prodrug 7 due to its own, TMR + -derived emission in the spectral range where 2′,7′dichlorofluorescein (the product of 2′,7′-dichlorofluorescin oxidation) is detected. Fortunately, prodrug 7 activation can be directly monitored without any probe: the initial rate of fluorescence increase ((dF/dt)0, F: λex = 551 nm, λem = 576 nm) of solution of 7 is 6-fold higher in the presence of H2O2 than in its absence ( Figure 1B). The reason for that is that in the initial state (intact 7) the fluorescence of the TMR + dye is partially quenched due to the photo-induced electron transfer (PET)  To determine active species responsible for the ROS generation, we acquired electrospray ionization (ESI) mass spectra of prodrug 5 with and without H2O2 (Figure 2, S15). We could detect prominent peaks corresponding to hydrolyzed prodrug 5 (5*) as well intermediates 1 and 3 (two peaks corresponding to [M] + and [M-e − + I − ] + were observed for 3) in the mass spectrum of the mixture of 5 and H2O2, whereas in the absence of H2O2 only intact (5) and hydrolyzed (5*) species were observed. These data confirm the mechanism of prodrug 5 activation outlined in Scheme 1. The ROS-generation assay ( Figure 1A) could not be accurately conducted for prodrug 7 due to its own, TMR + -derived emission in the spectral range where 2′,7′dichlorofluorescein (the product of 2′,7′-dichlorofluorescin oxidation) is detected. Fortunately, prodrug 7 activation can be directly monitored without any probe: the initial rate of fluorescence increase ((dF/dt)0, F: λex = 551 nm, λem = 576 nm) of solution of 7 is 6-fold higher in the presence of H2O2 than in its absence ( Figure 1B). The reason for that is that in the initial state (intact 7) the fluorescence of the TMR + dye is partially quenched due to the photo-induced electron transfer (PET) We could detect prominent peaks corresponding to hydrolyzed prodrug 5 (5*) as well intermediates 1 and 3 (two peaks corresponding to [M] + and [M-e − + I − ] + were observed for 3) in the mass spectrum of the mixture of 5 and H 2 O 2 , whereas in the absence of H 2 O 2 only intact (5) and hydrolyzed (5*) species were observed. These data confirm the mechanism of prodrug 5 activation outlined in Scheme 1. The ROS-generation assay ( Figure 1A) could not be accurately conducted for prodrug 7 due to its own, TMR + -derived emission in the spectral range where 2 ,7 -dichlorofluorescein (the product of 2 ,7 -dichlorofluorescin oxidation) is detected. Fortunately, prodrug 7 activation can be directly monitored without any probe: the initial rate of fluorescence increase ((dF/dt) 0 , F: λ ex = 551 nm, λ em = 576 nm) of solution of 7 is 6-fold higher in the presence of H 2 O 2 than in its absence ( Figure 1B). The reason for that is that in the initial state (intact 7) the fluorescence of the TMR + dye is partially quenched due to the photo-induced electron transfer (PET) from the ferrocene moiety, whereas the quenching is released in the activated probe containing a ferrocenium moiety (drug in Scheme 1).
Possible equilibrium between open (fluorescent) and closed (nonfluorescent) forms of TMR should be also considered. Since it is pH dependent and all our experiments are conducted in buffered aqueous solutions at the constant pH, contribution of this equilibrium to the final fluorescence increase will be canceled out. Thus, the described above data obtained in cell free settings indicate that conjugation of positively charged moieties to the NAAF-prodrug moiety does not substantially change its ability to react with H 2 O 2 with the formation of drugs able to catalyze ROS amplification.

Study of Cellular Effects of Prodrugs and Control Compounds
To test the cellular effects of the new prodrugs, we first studied whether these compounds are uptaken by cells. For these experiments, we selected a representative cancer cell line Burkitt's lymphoma BL-2. Since natural boron concentration in these cells is negligible and prodrug 5 contains one boron atom per molecule, we could study the uptake of 5 by monitoring the amount of intracellular boron using the photometric curcumin-based assay as described in the SI. We observed that the efficiency of the uptake of 5 is 2-fold higher than that of known non-targeted prodrug 9 (p < 0.001, n = 3) and is comparable to that of LY-targeting prodrug 8 ( Figure 3A, Table S3, SI).
Molecules 2020, 25, x 6 of 15 from the ferrocene moiety, whereas the quenching is released in the activated probe containing a ferrocenium moiety (drug in Scheme 1). Possible equilibrium between open (fluorescent) and closed (nonfluorescent) forms of TMR should be also considered. Since it is pH dependent and all our experiments are conducted in buffered aqueous solutions at the constant pH, contribution of this equilibrium to the final fluorescence increase will be canceled out. Thus, the described above data obtained in cell free settings indicate that conjugation of positively charged moieties to the NAAFprodrug moiety does not substantially change its ability to react with H2O2 with the formation of drugs able to catalyze ROS amplification.

Study of Cellular Effects of Prodrugs and Control Compounds
To test the cellular effects of the new prodrugs, we first studied whether these compounds are uptaken by cells. For these experiments, we selected a representative cancer cell line Burkitt's lymphoma BL-2. Since natural boron concentration in these cells is negligible and prodrug 5 contains one boron atom per molecule, we could study the uptake of 5 by monitoring the amount of intracellular boron using the photometric curcumin-based assay as described in the SI. We observed that the efficiency of the uptake of 5 is 2-fold higher than that of known non-targeted prodrug 9 (p < 0.001, n = 3) and is comparable to that of LY-targeting prodrug 8 ( Figure 3A, Table S3, SI).  (Table S3, SI). B: Uptake of fluorescent prodrug 7 (0-25 μM, incubation time-2 h, temperature-37 °C) by BL-2 cells monitored by flow cytometry (λex = 488 nm, λem= 530 nm). C: Monitoring ROS-release in BL-2 cells (using 5(6-)chloromethyl-2′,7′-dichlorodihydrofluorescein diacetate as a fluorogenic probe, 10 μM) in the presence of prodrugs 5, 7, 9 as well as controls 11 (negative control) and a mixture of FeCl3 with 8-hydroxyquinoline (HQ) with a molar ratio of components correspondingly 1 to 2 (Fe(HQ)2, positive control); incubation time-2 h, temperature-37 °C. Student's t-test: **-p < 0.01; ***-p < 0.001.
These data indicate that the mitochondria-targeting moiety favorably affects the cellular uptake of NAAF-prodrugs. The uptake of prodrug 7 by BL-2 cells could be studied directly by making use of its intrinsic TMR + -derived fluorescence. We observed that the uptake is efficient and the loading of 7 in the cells correlates with its extracellular concentration ( Figure 3B). Furthermore, we investigated the localization of prodrug 7 in human ovarian A2780 cells by using fluorescence These data indicate that the mitochondria-targeting moiety favorably affects the cellular uptake of NAAF-prodrugs. The uptake of prodrug 7 by BL-2 cells could be studied directly by making use of its intrinsic TMR + -derived fluorescence. We observed that the uptake is efficient and the loading of 7 in the cells correlates with its extracellular concentration ( Figure 3B). Furthermore, we investigated the localization of prodrug 7 in human ovarian A2780 cells by using fluorescence microscopy. The latter cells were selected since they are adherent and, therefore, well compatible with fluorescence microscopy. We observed that prodrug 7 co-localizes with commercially available mitochondria-staining dye Mitotracker Green (MG, Figure 4). The same localization pattern was obtained for the control compound 13 lacking any cargo moiety. These data indicate that DLC's (like TMR + ) are suitable membrane carriers for the NAAF-prodrugs.
microscopy. The latter cells were selected since they are adherent and, therefore, well compatible with fluorescence microscopy. We observed that prodrug 7 co-localizes with commercially available mitochondria-staining dye Mitotracker Green (MG, Figure 4). The same localization pattern was obtained for the control compound 13 lacking any cargo moiety. These data indicate that DLC's (like TMR + ) are suitable membrane carriers for the NAAF-prodrugs.
.  Table 1). We observed that prodrug 5 exhibits overall stronger effects on the cell viability in all studied cell lines than non-targeted NAAF-prodrug 9 and unmodified ferrocene (11, negative control), but is less potent than unspecific positive control (Fe(HQ)2) (p < 0.001, n = 3 for all). For example, for BL-cells IC50 (48 h incubation) = 5 ± 1 μM for 5 versus 35 ± 2 μM for 9 (p < 0.001, n = 3). The same trend was observed for both shorter (24 h, p < 0.001, n = 3) and longer (96 h, p < 0.001, n = 3) incubation times (Table 1). However, another control compound 12, which is an empty carrier Ph3P-R + , was even more cytotoxic, e.g., IC50(48 h incubation) = 0.5 ± 0.2 μM (p < 0.001, n = 3). The toxicity was retained towards the normal cell line HDFa (IC50 < 1 μM), whereas the effect of the hybrid prodrug 5 for HDFa cells was much more moderate (IC50 = 18 ± 5 μM). The higher cancer cell selectivity of 5 versus 12 can stem either from (a) the stronger NAAF-prodrug activation due to the higher ROS amount in mitochondria than in cytoplasm of cancer cells or (b) due to the favorable modulation of the high toxicity of Ph3P-R + carrier by the NAAF-prodrug moiety. Mathematical sum of ch1 and ch2 is shown in inset C. (D-F) The same as A-C, except that MG was replaced with prodrug 7 (100 nM). (G-I) The same as A-C, except that MG was replaced with a mixture of control 13 (100 nM) and MG (1 eq). Pearson's coefficient is 0.93; (J-L) The same as A-C, except that MG was replaced with a mixture of prodrug 7 (100 nM) and MG (1 eq). Pearson's coefficient is 0.87. Next, we studied the effects of new prodrugs and controls on the viability of selected cancer cells (BL-2, A2780, human prostate cancer DU-145 cells and human immortalized T lymphocyte Jurkat cells) as well as representative normal cells (human dermal fibroblasts adult, HDFa, Table 1). We observed that prodrug 5 exhibits overall stronger effects on the cell viability in all studied cell lines than non-targeted NAAF-prodrug 9 and unmodified ferrocene (11, negative control), but is less potent than unspecific positive control (Fe(HQ) 2 ) (p < 0.001, n = 3 for all). For example, for BL-cells IC 50 (48 h incubation) = 5 ± 1 µM for 5 versus 35 ± 2 µM for 9 (p < 0.001, n = 3). The same trend was observed for both shorter (24 h, p < 0.001, n = 3) and longer (96 h, p < 0.001, n = 3) incubation times (Table 1). However, another control compound 12, which is an empty carrier Ph 3 P-R + , was even more cytotoxic, e.g., IC 50 (48 h incubation) = 0.5 ± 0.2 µM (p < 0.001, n = 3). The toxicity was retained towards the normal cell line HDFa (IC 50 < 1 µM), whereas the effect of the hybrid prodrug 5 for HDFa cells was much more moderate (IC 50 = 18 ± 5 µM). The higher cancer cell selectivity of 5 versus 12 can stem either from (a) the stronger NAAF-prodrug activation due to the higher ROS amount in mitochondria than in cytoplasm of cancer cells or (b) due to the favorable modulation of the high toxicity of Ph 3 P-R + carrier by the NAAF-prodrug moiety.
We assume that, if (a) is valid, substitution of Ph 3 P-R + in prodrug 5 for another efficient mitochondrial carrier (e.g., TMR + ) should lead to the comparable effect on cytotoxicity and cancer specificity. However, if (b) is valid, replacement of PPh 3 P-R + for a less toxic carrier (e.g., TMR + in control 13, IC 50 = 35 ± 6 µM for BL-2 cells) should lead the strong attenuation of the effect. We believe that scenario (b) is operative, since prodrug 7 (derivative of TMR + carrier) is nontoxic towards studied cancer cell lines (BL-2 and A2780). Interestingly, despite being more cytotoxic, prodrug 5 is a weaker ROS amplifier in BL-2 cells than non-targeted NAAF-prodrug 9 (p < 0.01, n = 3, Figure 3C). Since the intracellular ROS amount correlates with the efficacy of NAAF-prodrug activation [7,8,18], we conclude that the intracellular formation of the NAAF drug from prodrug 5 occurs with the lower yield than that from prodrug 9. This is in agreement with suggestion (b) discussed above.

The Mode of Action of Mitochondria-Targeting NAAF-Prodrug 5
To get a deeper insight into the mode of action of the more active prodrug 5, we first determined via which mechanism this compound induces the death of Jurkat cells. The latter cells were selected as representative cancer cells ( Figure 5A). Prodrug 5 as well as controls 9 (non-targeted NAAF-prodrug) and 12 (an empty carrier) kill the cells mostly via necrosis (red bars, Figure 5A) and apoptosis (blue bars, Figure 5A) with the former cell death phenotype being more prominent. We assume that, if (a) is valid, substitution of Ph3P-R + in prodrug 5 for another efficient mitochondrial carrier (e.g., TMR + ) should lead to the comparable effect on cytotoxicity and cancer specificity. However, if (b) is valid, replacement of PPh3P-R + for a less toxic carrier (e.g., TMR + in control 13, IC50 = 35 ± 6 μM for BL-2 cells) should lead the strong attenuation of the effect. We believe that scenario (b) is operative, since prodrug 7 (derivative of TMR + carrier) is nontoxic towards studied cancer cell lines (BL-2 and A2780). Interestingly, despite being more cytotoxic, prodrug 5 is a weaker ROS amplifier in BL-2 cells than non-targeted NAAF-prodrug 9 (p < 0.01, n = 3, Figure 3C). Since the intracellular ROS amount correlates with the efficacy of NAAF-prodrug activation [7,8,18], we conclude that the intracellular formation of the NAAF drug from prodrug 5 occurs with the lower yield than that from prodrug 9. This is in agreement with suggestion (b) discussed above.

The Mode of Action of Mitochondria-Targeting NAAF-Prodrug 5
To get a deeper insight into the mode of action of the more active prodrug 5, we first determined via which mechanism this compound induces the death of Jurkat cells. The latter cells were selected as representative cancer cells ( Figure 5A). Prodrug 5 as well as controls 9 (non-targeted NAAFprodrug) and 12 (an empty carrier) kill the cells mostly via necrosis (red bars, Figure 5A) and apoptosis (blue bars, Figure 5A) with the former cell death phenotype being more prominent.  (5)) is plotted on the Y-axis (λ ex = 638 nm; λ em = 655-695 nm). Student's t-test: **-p < 0.01; ***-p < 0.001. We could not distinguish between direct necrosis and apoptosis followed by secondary necrosis. Furthermore, at nontoxic concentrations ≤ 10 µM prodrug 5 induces the significant increase of the mitochondrial potential (green bars, Figure 5B) that has been previously observed for other targeted and non-targeted NAAF-prodrugs (e.g., 9, orange bars, Figure 5B). In contrast, the empty carrier 12 reduces the mitochondrial potential at all tested concentrations (blue bars, Figure 5B). Thus, conjugation of the NAAF-moiety to the carrier attenuates its negative effect on the mitochondria membrane potential that leads to its lower toxicity.
Finally, at the concentrations causing no cell death prodrug 5 does not affect the intracellular concentration of glutathione ( Figure 6A) and cell cycle ( Figure 6B). In contrast, typical NAAF-prodrugs induce upregulation of intracellular GSH to counteract elevated ROS and cell cycle arrest in the G0/G1 phase [18].
Molecules 2020, 25, x 9 of 15 We could not distinguish between direct necrosis and apoptosis followed by secondary necrosis. Furthermore, at nontoxic concentrations ≤ 10 μM prodrug 5 induces the significant increase of the mitochondrial potential (green bars, Figure 5B) that has been previously observed for other targeted and non-targeted NAAF-prodrugs (e.g., 9, orange bars, Figure 5B). In contrast, the empty carrier 12 reduces the mitochondrial potential at all tested concentrations (blue bars, Figure 5B). Thus, conjugation of the NAAF-moiety to the carrier attenuates its negative effect on the mitochondria membrane potential that leads to its lower toxicity.
Finally, at the concentrations causing no cell death prodrug 5 does not affect the intracellular concentration of glutathione ( Figure 6A) and cell cycle ( Figure 6B). In contrast, typical NAAFprodrugs induce upregulation of intracellular GSH to counteract elevated ROS and cell cycle arrest in the G0/G1 phase [18].

Materials and Methods
Commercially available chemicals of the best quality from Sigma-Aldrich (Schnelldorf, Germany) and Alfa-Aesar (Kandel, Germany) were obtained and used without purification. NMR spectra were acquired on a Bruker Avance 300 (Ettlingen, Germany) or Bruker Avance 400 (Ettlingen, Germany). ESI mass spectra were recorded on a Bruker ESI MicroTOF II (Bremen, Germany) or Bruker maXis 4G mass spectrometers (Bremen, Germany). C/H/N elemental analysis was performed in the microanalytical laboratory of the Department Chemistry and Pharmacy, Organic Chemistry Chairs 1 and 2 of the Friedrich-Alexander-University of Erlangen-Nuremberg. UV-visible spectra were measured on a Cary 100 UV-visible spectrophotometer (Agilent Technologies, Frankfurt am

Materials and Methods
Commercially available chemicals of the best quality from Sigma-Aldrich (Schnelldorf, Germany) and Alfa-Aesar (Kandel, Germany) were obtained and used without purification. NMR spectra were acquired on a Bruker Avance 300 (Ettlingen, Germany) or Bruker Avance 400 (Ettlingen, Germany). ESI mass spectra were recorded on a Bruker ESI MicroTOF II (Bremen, Germany) or Bruker maXis 4G mass spectrometers (Bremen, Germany). C/H/N elemental analysis was performed in the microanalytical laboratory of the Department Chemistry and Pharmacy, Organic Chemistry Chairs 1 and 2 of the Friedrich-Alexander-University of Erlangen-Nuremberg. UV-visible spectra were measured on a Cary 100 UV-visible spectrophotometer (Agilent Technologies, Frankfurt am Main, Germany) by using either quartz glass cuvettes (Hellma GmbH, Müllheim, Germany) with a sample volume of 1 mL or micro-cuvettes with a sample volume of 100 µL (BRAND GmbH, Wertheim, Germany). Fluorescence spectra were acquired on a Varian Cary Eclipse fluorescence spectrophotometer using fluorescence cuvettes (Hellma GmbH, Müllheim, Germany) with a sample volume of 1 mL. The fluorescence of live cells was quantified by using a Guava easyCyteTM 6-2L Flow cytometer from Merck Millipore. The data were processed using the inCyteTM software package (Merck Millipore, Darmstadt, Germany) and the ModFIT LTTM software (Verity Software House, Topsham, ME, USA). The fluorescence images of live cells were acquired by using a Zeiss Axio Vert.A1 microscope (Jena, Germany) using 40x/1.30 objective (oil DIC) and the following filter set: 450-490/500-550 nm for the detection of MitoTracker Green (channel 1: abbreviated ch1 in the main text of the paper); excitation/emission 538-563/570-640 nm for the detection of prodrug 7 and control 13 (channel 2: abbreviated ch2 in the main text of the paper). Statistical analysis of the data was conducted by using Student's t-test. Two data sets were considered significantly different from each other for p < 0.05.
Syntheses of new prodrugs and control compounds as well as protocols of all biological assays used in this study are described in detail in the SI to this article.

Determination of Solubility of Prodrugs 5 and 7 in Aqueous Buffers
Prodrugs 5, 7 and 9 (known control) were dissolved in DMSO at concentrations of 15, 10, 5, 2, 1.5, 1, and 0.2 mM. These solutions (10 µL) were transferred to quartz cuvettes containing 990 µL of either Dulbecco's phosphate-buffered saline (DPBS) or Roswell Park Memorial Institute (RPMI) medium containing 20% of fetal bovine serum (FBS) to obtain solutions (or suspensions) with final concentrations of prodrugs-150, 100, 50, 20, 15, 10, and 2 µM. The absorbance at 790 nm of the resulting solutions/suspensions was measured and plotted as a function of the concentration of prodrugs. Mixtures absorbing <0.04 (A 790 nm ) for FBS and 0.034 (A 790 nm ) for DPBS appeared transparent and were considered as true solutions. Using these data, we determined solubility limits for prodrugs and controls (Table S1).

Monitoring Generation of Reactive Oxygen Species (ROS) in the Presence of Either Prodrugs or Control
Compounds in Cell Free Settings 2 ,7 -Dichlorofluorescin diacetate (DCFH-DA, 4.9 mg) was dissolved in 100 µL DMF and mixed with 0.1 M NaOH (900 µL). This mixture was incubated on a shaker for 30 min at 25 • C in the dark to obtain 10 mM 2 ,7 -dichlorofluorescin (DCFH) stock solution. Next, 3-(N-morpholino)propanesulfonic acid buffer (MOPS buffer 100 mM, pH: 7.4) containing N,N,N',N'-ethylenediaminetetraacetic acid (EDTA, 10 mM), glutathione (GSH, 5 mM) and H 2 O 2 (10 mM) was prepared. Finally, DCFH was diluted in this buffer to 10 µM as the final concentration. Monitoring of the fluorescence was started (λ ex : 501 nm, λ em : 525 nm). After 5 min, the compound was added to have a final concentration of 50 µM. DMF and FeSO 4 in DMF were used as controls. The measurement was conducted for 300 min (Table S2, Figure 1A).

Monitoring Activation of Prodrug 7 in the Presence of H 2 O 2 by Fluorescence Spectroscopy
Prodrug 7 (10 µL, 50 µM in DMSO) was added into phosphate buffered saline (1000 µL, PBS: 10 mM PO 4 3− and 150 mM NaCl, pH: 7.0, final concentration 0.5 µM). After the initial scan of the fluorescence at 37 • C (λ ex : 551 nm, λ em : 576 nm), H 2 O 2 was added (10 µL, final concentration: 50 mM) into the mixture and changes of the fluorescence intensity was monitored for the next 120 min. As a control, the fluorescence change in the same mixture lacking H 2 O 2 was also monitored ( Figure 1B).

Monitoring of H 2 O 2 -Induced Activation of Prodrug 5 by Using ESI Mass Spectrometry
A freshly prepared solution of prodrug 5 (10 µL, 4 mM in DMF) was diluted with CH 3 CN (100 µL). The resulting solution was mixed with water (886 µL) followed by the addition of either H 2 O 2 (4 µL, 30% (w/w) in H 2 O) or water (4 µL, negative control). These solutions were kept at 22 • C for 5 min. Then they were diluted with CH 3 CN (1 mL) and the resulting probes were analyzed by high resolution ESI-TOF mass spectrometry in the positive mode ( Figure 2). The following reaction products could be identified ( Figure S10).

Determination of the Viability of Adherent Cells (A2780, DU-145 and HDFa)
Adherent A2780, DU-145 and NHDF cells were grown as mentioned before. Then, the cells were resuspended in the RPMI 1640 (DMEM for HDFa cell line) medium containing 5% FBS (for 24-and 48 h assays) or 10% FBS (for 96 h assays) 1% L-glutamine, and 1% penicillin/streptomycin. This suspension was spread in the wells of a 96-well microtiter plate. Every well in the plates contained 25,000 cells (exceptions: 12,500 cells for the 96h-assay and 10,000 cells for experiments with HDFa cell line) per well per 100 µL and was left at 37 • C in the chamber filled with CO 2 (5%) overnight. Stock solutions of prodrugs of different concentrations (1 µL, solvent DMSO) were added to the wells and incubated for 24-96 h. Four experiments were conducted for each concentration of the prodrug. Finally, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; 20 µL of the solution prepared by dissolving MTT (5 mg) in PBS buffer (1 mL)) was added to each well, incubated for 3 h, treated with sodium dodecyl sulfate (SDS) solution (90 µL, 10% solution in 0.01 M aqueous HCl), and incubated overnight. Then the intensity of absorbance at 590 nm was measured (MTT is converted to blue dye (λ max = 590 nm) in live cells). The absorbance at 690 nm was taken as a baseline value. These data were applied to calculate the relative number of viable cells.

Determination of the Viability of Suspension BL-2 Cells
The suspension of BL-2 cells in the RPMI 1640 medium containing 5% FBS (for 24-and 48 h assays) or 10% FBS (for 96 h assays) 1% L-glutamine, and 1% penicillin/streptomycin was spread in the wells of a 96-well microtiter plate. Every plate contained 50,000 cells (with the exception of the 96h-experiment, for which 25,000 cells were used) per well per 100 µL. Stock solutions of prodrugs of different concentrations (1 µL, solvent DMSO) were added to the wells and incubated for 24-96 h. Four experiments were conducted for each concentration of the prodrug. Finally, 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT; 20 µL of the solution prepared by dissolving MTT (5 mg) in PBS buffer (1 mL)) was added to each well, incubated for 3 h, treated with sodium dodecyl sulfate (SDS) solution (90 µL, 10% solution in 0.01 M aqueous HCl), and incubated overnight. Then the intensity of absorbance at 590 nm was measured (MTT is converted to blue dye (λ max = 590 nm) in live cells). The absorbance at 690 nm was taken as a baseline value. These data were applied to calculate the relative number of viable cells. All obtained data are summarized in Table 1.

Determination of Prodrug 5 Uptake by BL-2 Cells by Quantification of Intracellular Boron Amounts Using the Curcumin-Assay
BL-2 cells were grown as mentioned before. The suspension of the cells in the RPMI 1640 medium with 5% FBS 1% L-glutamine, and 1% penicillin/streptomycin containing 10 6 cells/mL was spread in the wells of a 6 well microtiter plates. Solutions of prodrugs (30 µL per well, solvent DMSO, each well contained 3 mL cell suspension) were added to the suspensions and incubated for 1 h. The final concentration of the prodrugs in the suspensions was 1 µM. This low concentration was used to avoid cell death, since some prodrugs in this series (e.g., prodrug 5) are very toxic. The suspensions from 12 wells treated at the same condition were combined, the combined suspensions were washed three times with ice-cold PBS (3 × 5 mL), treated with concentrated H 2 O 2 solution (200 µL, 1 M) for 30 min, and all volatiles were removed by lyophilization. Dry, lysed cells were washed with water (200 µL), and the aqueous solution obtained was acidified with HCl (400 µL, 0.1 M). Then this solution was extracted with 2-ethyl-1,3-hexanediol (160 µL, 10% in CHCl 3 , v/v), and a portion of the organic phase obtained (90 µL) was mixed with H 2 SO 4 /CH 3 CO 2 H (400 µL, 1/1, v/v). Curcumin solution in methyl isobutyl ketone (250 µL, 2 mg in 1 mL of the solvent) was added and allowed to react for 24 h. The reaction was quenched by addition of water (1 mL). The light absorbance at 550 and 780 nm of the organic phase was measured. The former value corresponds to absorbance of a curcumin-boron complex, whereas the second one is taken as a baseline. The baseline corrected absorbance at 550 nm (∆A = A(550 nm) − A(780 nm)) was proportional to the concentration of boron in the mixture (Table S3).

Determination of Prodrug 7 Uptake by BL-2 Cells by Quantification the Fluorescence Derived from 7 Using Flow Cytometry
The cells were centrifuged (5 min; 1000 rpm), washed with DPBS (2 × 10 mL) and concentrated at cell density of 10 6 cells/mL in opti-MEM. The cell suspension was split into portions as 200 µL in each. Prodrug 7 (dissolved in DMSO) was added into these suspensions (2 µL) to obtain final concentrations of prodrugs 10 and 25 µM. As a negative control pure DMSO (2 µL) was added to the cell suspension. After 120 min of incubation in the dark chamber filled with 5% CO 2 at 37 • C, the mean fluorescence of live cells (λ ex = 488 nm, λ em = 530 nm) in the suspensions was determined by using the flow cytometer ( Figure 3B).

Determination of ROS-Amplification in BL-2 Cells in the Presence of Prodrugs
The cells were centrifuged (5 min; 1000 rpm), washed with DPBS (2 × 10 mL) and the suspension at cell density of 10 6 cells/mL in DPBS was prepared. 5(6-)chloromethyl-2 ,7 -dichlorodihydrofluorescein diacetate (CM-DCFH-DA) in DMSO (8.7 µL, stock solution 10 mM) was added into 8.7 mL of cell solution to have 10 µM final concentration. The cells were incubated in the dark chamber supplemented by 5% CO 2 at 37 • C for 15 min. Then, the cells were centrifuged, washed with DPBS (1× 10 mL) and resuspended in opti-MEM medium by maintaining the cell concentration at 10 6 cells/mL. The cell suspension was split into portions as 200 µL in each. The compounds (prodrug 5, 7, 9) and the negative and positive controls (Ferrocene (11) and Fe(8HQ) 2 complex, respectively), as well as the DMSO were added from the stock solutions (2 µL). After 120 min of incubation in the dark chamber filled with 5% CO 2 at 37 • C, the mean fluorescence of live cells (λ ex = 488 nm, λ em = 530 nm) in the suspensions was determined by using the flow cytometer. All samples were applied in triplicates within the assay and the results were calculated based on three independent assays. Prodrug 7 contributes to the fluorescence signal derived from CM-DCFH-DA (λ ex = 488 nm, λ em = 530 nm). This background signal was quantified by doing the same experiment as described above, but in the absence of CM-DCFH-DA ( Figure 3B). Then we subtracted this background signal from that obtained in the presence of CM-DCFH-DA to obtain the corrected ROS-mediated fluorescence increase of CM-DCFH-DA probe ( Figure 3C).
Co-localization of prodrug 7 (or control 13) and MitoTracker Green probe in a representative cancer cell (A2780) by using fluorescence microscopy A2780 cells were seeded on a 35 mm imaging dish (µ-Dish ROS amplifier. Its mode of action relies off the mixed effect of weak ROS generation by the NAAF-part and modulation of membrane potential by the Ph 3 P-R + carrier part. Despite the unexpected mode of action, hybrid 5 exhibits favorable properties including substantially higher cytotoxicity towards cancer cells than that of non-targeting NAAF-prodrugs as well as lower cytotoxicity towards normal HDFa cells with respect to the empty carrier 12. The cancer cell specificity observed in cell-based assays may justify in vivo studies of 5 in the future. However, this experiment will become possible only after a suitable formulation of poorly water-soluble prodrug 5 is found. Since NAAF-prodrugs are only moderately stable in blood [23], intraperitoneal (i.p.) injection remains the method of choice for administration of prodrug 5 in vivo, as reported for known NAAF-prodrugs [18].
Supplementary Materials: The following are available online, Figure S1: 1 H NMR spectrum of prodrug 5 in DMSO-d 6 ; Figure S2: 13 C NMR spectrum of prodrug 5 in DMSO-d 6 ; Figure S3: 31 P NMR spectrum of prodrug 5 in DMSO-d 6 ; Figure S4: High resolution ESI-TOF mass spectrum of prodrug 5: upper plot-experimental spectrum; bottom plot-theoretical spectrum; Figure S5: 13 Figure S13: 1 H NMR spectrum of prodrug 7 in DMSO-d 6 . Impurities are marked with asterisk; Figure S14: High resolution ESI-TOF mass spectrum of prodrug 7: upper plot-theoretical spectrum; bottom plot-experimental spectrum; Figure S15: Zoomed in regions of the mass spectrum shown Figure 2B; Table S1. Solubility of prodrugs in aqueous solutions; Table S2. Initial rates of DCFH oxidation in the presence of H 2 O 2 and prodrugs/controls; Table S3. Uptake of prodrug 5 by BL-2 cells.