Protected amino acids were purchased from AAPTEC or CHEMIMPEX.N¹⁰-(trifluoroacetyl)pteroic acid was purchased from Schirks. Other reagents are available from Sigma-Aldrich Chemicals, unless otherwise stated. Some of the important reagents are such as 3-acetylphenanthrene, mono-Boc-1,2-ethyldiamine. HCl, MeOCO(O)SCl, 2-mercaptopyridine, Fmoc-Lys(mtt)-Wang resin, Fmoc-Cys(4-methoxytrityl)-Wang resin, N¹⁰TFA- pteroic acid, δ-protected Fmoc-Glu(O^tBu)OH [where Fmoc = fluorenylmethyloxycarbonyl & ^tBu = tertiarybutyl], diisopropylethylamine (DIPEA), α-protected Fmoc-Glu(O^tBu)OH, TFA, benzotriazol-1-yl-oxytripyrro- lidino-phosphonium hexafluorophosphate (PyBOP), isopropanol (i-PrOH), piperidine, triisopropyl silane (TIPS), ethanedithiol (EDT), hydrazine, dichloromethane (DCM), dimethyl formamide (DMF), and diethyl ether.

¹H and ¹³C NMR spectra were recorded using a Bruker 400 MHz NMR spectrometer. LC-ESI-MS spectra were obtained using a Micromass Quattro Micro API triple quadrupole mass spectrometer with an ESI source. A Waters 1500 HPLC with a 20.0 mL loop injector was used for semi-preparative chromatography. An Agilent 1100 UV analytical HPLC chromatograph was used to identify compound purity.

KB, A-549 and MDA-MB-231 cells were obtained from American Type Culture Collection. MDA-MB-231 cells expressing high levels of FR were generated by passaging the cells for 14 weeks in folate-free cell culture medium. The cell lines were grown continuously as a monolayer in folate-free RPMI medium containing 10% fetal bovine serum and 1% penicillin/streptomycin antibiotic cocktail in a 5% CO₂: 95% air-humidified atmosphere at 37˚C.

All animal procedures were approved by the Purdue Animal Care and Use Committee in accordance with NIH guidelines. Normal rodent diets were not used, since they contain excessive amounts of folic acid which elevates serum folate levels significantly above normal physiological concentrations. Rather, all animals were maintained on a folate deficient diet (Harlan Teklad laboratories, WI) for at least 3 weeks prior to each study to lower their serum folate levels into the physiological range. Control animals were also maintained on a folate deficient diet.

Oxygen consumption rate measurements were performed at 37.0˚C ± 0.1˚C in 20 mM phosphate buffer (pH 7.4). Stock and sample solutions were prepared with double distilled deionized water and decontaminated from traces of transition metals by exposure to Chelex 100 resin using the batch method [48] . Air-saturated sample solutions were used containing micromolar amounts of quinone and millimolar amounts of ascorbate (simulating parenteral ascorbate concentrations) and 20 mM phosphate buffer (pH 7.4). Oxygen consumption rates were measured with a 5300 Oxygen Biological Monitor (Yellow Springs Instruments Co., USA) using a Clark electrode as a sensor in a YSI 5301 constant temperature bath. Rates were calculated from the initial constant slopes of [O₂] traces. Runs started in the absence of ascorbate followed by ascorbate addition, without interrupting oxygen consumption measurements.

Since 3 and 4 are not soluble in acetonitrile, half-wave reduction potentials (E_1/2) were determined in 1:3 (v/v) nitrogen purged DMSO:20 mM phosphate buffer (pH 7.4). In this solvent a single peak corresponding to a 2 electron + 2 proton reduction step is observed for quinones [49] . Solutions contained up to 500 μM quinone and 0.1 M tetra-n-butylammonium perchlorate (TBAP). A BAS CV 50W voltammetric analyzer with a glassy carbon working electrode was used in these determinations. An Ag/AgCl(sat) electrode served as the reference electrode (E' = +0.22 V vs. NHE) and a platinum wire as the counter electrode. Differential pulse voltammograms (DPV) were obtained in the potential range of −2.00 to 0.00 volts, using a 50 mV pulse amplitude and 20 mV/s of scan rate. Reduction potential values were obtained from the DPV peak potential maxima. These are very similar to the half-wave redox potentials, E_1/2, in normal polarographic measurements [50] . Since compound 2 (see Figure 1 & Table 1) represents the quinone moiety existing in 3 and 4 and 1 has been used in previous quinone + ascorbate cytotoxicity and antitumor activity studies, redox potentials and ascorbate oxidation rates were also determined for these compounds.

^aObtained after addition of 1 mM ascorbate to solutions containing 10 μM quinone in 20 mM, air-saturated, phosphate buffer. ^bMeasured in 1:4 (v/v) DMSO:20 mM phosphate buffer (pH 7.4). DPV peak potential maxima are against Ag/AgCl.).

FR⁺KB cells and FR⁻ A-549 cells were seeded into 24-well (100,000 cells/well in 500 μL) Falcon plates and allowed to form monolayers over 12 h. Spent medium was replaced with fresh medium (0.5 mL) containing increasing concentrations of 2, 3 or 4 and incubated for 2 h. The unbound test compounds 2, 3 and 4 were washed from the wells with fresh medium (3 × 0.5 ml), followed by addition of 5, 3 and 0 mM (for a control experiments without sodium ascorbate) freshly prepared sodium ascorbate solution (pH = 7.4) in medium (0.5 mL/well). Cells were then incubated for an additional 2, 12, 16 or 68 h at 37˚C. Cells were washed with fresh medium (3 × 0.5 ml) and incubated in fresh medium (0.5 mL) for 66, 56 or 0 h at 37˚C. Spent medium in each well was replaced with fresh medium (0.5 mL) containing [³H]-thymidine (1 μCi/mL), and cells were incubated for 4 h at 37˚C to allow [³H]-thymidine incorporation. Cells were then washed with medium (3 × 0.5 mL) and treated with 5% trichloroacetic acid (0.5 mL) for 10 min at room temperature. Cells were dissolved in 0.25 M NaOH (0.5 mL), transferred into individual vials containing Ecolume scintillation cocktail (3.0 mL), and counted in a scintillation counter (Packard, Packard Instrument Company). IC₅₀ values were calculated by plotting % ³H-thymidine incorporation versus log concentration of compounds 2, 3 and 4 using GraphPad Prism 4. Each % ³H-thymidine incorporation value represents the average ± standard error of the mean of at least three determinations.

Four to five week-old female nu/nu mice were inoculated subcutaneously with KB or MDA-MB-231 cells (1.0 × 10⁶/mouse in cell growth medium) on their shoulders. Growth of the tumors was measured in two perpendicular directions every 2 days using a caliper (body weights were monitored on the same schedule), and the volumes of the tumors were calculated as 0.5 × L × W² (L = longest axis and W = axis perpendicular to L in millimeters). Tumor therapy studies were performed with 3 (40 or 80 nmol/mouse) in saline (100 μL) 12 or 20 days after tumor cell implantation, when the tumors reached approximately 75 - 100 mm³ (for MDA-MB-231 tumor) and 100 - 350 mm³ (for KB-tumor) in volume.

Female nude mice bearing KB or MDA-MB-231 xenograft tumors were injected with 3 (2.0 μmol/kg for KB tumors of volume 100 - 350 mm³or 4.0 μmol/kg for MDA-MB-231 tumors of volume 75 - 100 mm³) dissolved in 100 μL of saline via lateral tail vein injection. Treatments were conducted 3× per week for two weeks. A single dose of sodium ascorbate in saline (4g/kg or 80 mg/mouse) was injected 30-minutes post 3 injection and also on days when mice were not treated with 3. Tumor volumes and body weights were measured on the same schedule. In vivo efficacy was evaluated by plotting tumor growth % versus days on therapy. Tumor growth % is defined by Equation (1),

where V and V₀ are the tumor volumes on any given day and the first day of measurement, respectively. Reported tumor growth percents are the averages ± standard errors of the mean of at least three determinations.

Upon addition of 1mM ascorbate to 10 µM quinone 1, 2, 3 or 4, an increase in the rate of ascorbate oxidation is observed (Figure 1). The reason for including compounds 2 and 1 in this study is that the former mimics the PHQ moiety in the folate conjugate while the latter serves as an appropriate comparison to quinones used by others, since most of the anticancer work with quinones + ascorbate has been performed with 1. It can be concluded from this figure that the ascorbate oxidation enhancement activity of these quinones follows the order 2 > 3 > 4 > 1, which coincides with the ranking of the quinone 2 electrons+2 protons reduction potentials (Table 1).

Quinone 2 displayed cytotoxic activity against KB cells at 3 and 5 mM ascorbate, with IC₅₀ values in the 10⁻⁸M range, depending on incubation time (Table 2). Once this behavior was detected for 2, we were confident that a similar cytotoxic behavior was to be observed for 3 and 4 and thus were prompted to synthesize the FR-targeted quinones. Incubation of quinone 3 with 3 or 5 mM ascorbate for 2 hours exhibited very low cytotoxic activity in vitro (Table 2, entries 3 - 4), with the greatest cytotoxicity seen when cells were incubated for 12 hours (IC₅₀ < 10⁻⁹ M) in the presence of either 3 or 5 mM ascorbate (Table 2, entries 6 - 7). In contrast, incubation of quinone 4 conjugate with 3 or 5 mM ascorbate for up to 68 hours was only slightly cytotoxic to KB cells (Table 2, entries 8 - 9) and no cytotoxicity was expressed when incubated for 12 hours with ascorbate. In addition, incubation of 3 in the absence of ascorbate with KB cells did not show cytotoxicity if incubated for12 hours with ascorbate (Table 2, entry 5), indicating the need for ascorbate for cytotoxicity. Compound 3 in the presence of 3 mM ascorbate did not show toxicity, while it was mildly toxic when incubated for 12 hours, in the presence of 5 mM ascorbate, with FR negative A549 cells, thus demonstrating the need for cell surface binding to FR for expression of cytotoxic activity (Table 2, entries 10 - 11). The latter was expected, since 3 is extracted from FR

negative cells upon rinsing due to its high hydrophilicity. These observations are in contrast to previous studies where KB [51] and A549 cells [21] were incubated for 1 to 2 hours with ascorbate followed by cell rinsing to remove free ascorbate and testing for cytotoxicity after incubation periods of more than one week. Ascorbate cytotoxicity to KB and A549 cells in the absence of quinones was then detected. Thus, interestingly, compound 3 is in fact accelerating the expression of this toxicity in FR+ cells.

Since the quinone conjugate 3 in combination with ascorbate demonstrated the greatest cytotoxicity in vitro, it was selected for more thorough testing against two FR+ tumor xenografts in vivo. As seen in Figure 2, using the

protocol described in Materials and methods, essentially no reduction in tumor size was detected in either xenograft model, independent of whether quinone was injected or not. This result was not anticipated, since, as mentioned above, ascorbate alone, and ascorbate + 1, have been reported to display some anti-tumor activity in murine tumor models, and because FR expression has been found to be similar in murine tumor xenografts and the cultured cancer cells from which they are derived (unpublished observations). Clearly, unknown factors dominate the behavior of ascorbate or folate-PHQ in our murine tumor models.

Although 2 electrons + 2 protons reduction potentials were measured here instead of 1 electron reduction potentials, the values still reflected the relative abilities of the quinones to undergo reduction, since previous work had shown that the rate of ascorbate oxidation in the presence of quinones increased with an increase in the one- electron redox potential of the quinone [9] . The electrochemical reaction, measured in the present work, Equation (2), corresponds to a redox potential (E_1/2), which is essentially equal to the summation of the first and second electron redox potentials, Equation (3).

Thus, since in general a quinone with a more positive E(Q/Q^•?) will also have a more positive E(Q^•?/QH₂) [52] , the trend in E_1/2 values shown here should roughly follow that in E(Q/Q^•?) values. Not surprisingly, the relative order in E_1/2 values correlates nicely with the rate of ascorbate oxidation in Table 1. The discrepancy in ascorbate oxidation rates and in reduction potentials among compounds 2, 3 and 4 is an interesting subject. First of all, at physiological pH, compound 2 should be mostly protonated at the primary amine moiety. Addition of the folate linkers to the quinone moiety not only will increase the steric hindrance upon approaching the ascorbate ion, but the amide group attached to the quinone moiety has a different pKa value than that of the amine group in compound 2. Those characteristics may be responsible for the smaller redox reactivity of 3 and 4 as compared to 2. Differences in redox reactivity between 3 and 4 may be due to differences in steric or structural conformations when approaching ascorbate, since the disulfide bond is not being reduced by ascorbate. However, the elucidation of the molecular causes for these redox reactivity differences is beyond the scope of this work, which is intended to explore the relative toxicities of releasable and non-releasable FR-targeted quinones.

Since ascorbate oxidation by quinones generates H₂O₂ and other ROS, it is expected that all PHQ derived compounds would show significant cytotoxic activity when combined with ascorbate. However, neither FA-bound 3 nor 4 showed cytotoxicity to KB cells when those compounds were incubate for only 2 hours in the presence of 3 or 5 mM ascorbate, nor compound 3, the more potent of the two folate conjugates, required 12 h of incubation with ascorbate to achieve its full potency in the subpicomolar range. Although compound 4 exhibits limited toxicity to KB cells after 68 hours of incubation with ascorbate, no cytotoxicity was expressed when incubated with ascorbate for 12 hours.

Previous research on folate-targeted therapeutic agents has suggested that folate-linked cytotoxic warheads must be released from folate to achieve maximum cytotoxicity. To promote this release, self-immolative linkers, generally containing a disulfide bond in close proximity to a labile ester bridge, have been employed to attach folate to its cytotoxic cargo [42] [53] . However, in our studies, the non-releasable FA-bound 3 displayed significantly more potent activity than the releasable FA-bound 4. Incubation of compound 4 with 5 mM ascorbate did not release the quinone even after 24 hours of incubation, as detected from HPLC sample analysis. Furthermore, a decrease in the HPLC peak of 4 was not detected either. Many studies have shown that exogenous dehydroascorbic acid is transported into cells where it is reduced to ascorbic acid by glutathione [54] . The latter is in line with the fact that the ascorbate to dehydroascorbate oxidation potential is +0.10 V, while the oxidation potential of GSH to GSSH is + 0.24 V [55] , indicating that GSH, which is located inside the cell in relatively large concentrations, should efficiently reduce the disulfide moiety in 4, as opposed to ascorbate. In fact, to our best knowledge, all the disulfide-containing folate-targeted drugs reported in the literature are known to release the drugs inside the cells, thus improving their cytotoxic activity as compared to non-releasable FR-targeted analogs [56] .

The lack of toxicity of 4 might have been caused by glucuronide conjugation of the released phenanthraquinone moiety inside the cell, as previously observed for phenanthraquinone [57] . This will decrease the amount of released PHQ available for ROS production. Another cause for the lack of toxicity of 4 could be its smaller redox potential which correlates with its smaller rate of ascorbate oxidation (Figure 1, Table 1). FA-bound 3, in contrast, would be expected to remain cell associated due to the high affinity of folate for FR [56] , forcing the ROS generator to accompany FR wherever the receptor traffics within the cell and at the cell membrane.

No tumor size reduction or decrease in tumor growth rate was detected after i.v. injection of either ascorbate, ascorbate + conjugate 3, or conjugate 3 alone. This observation contrasts with previous results seen following both i.p. administration of ascorbate alone as well as ascorbate + 1 mixture [27] . A possible explanation for the absence of anti-tumor activity in vivo may derive from the fact that neither KB nor MDA-MB-231 tumors are highly vascularized, and since ascorbate/quinone-mediated ROS production requires molecular O₂, the pressure of O₂ in the tissues could have been insufficient to enable adequate ROS generation in the tumors. Whether a more active quinone catalyst can be found that can compensate for the lower oxygen tensions that often exist in tumors will have to await further investigation.

A second possible explanation for the insensitivity of KB and MDA-MB-231 tumors to the ascorbate/quinone combination in vivo could have arisen from an inadequate concentration of the quinone moiety in the tumor. Thus, Verrax et al. reported antitumor activity of 1 + ascorbate after their i.p. injection in mice [27] . Although less ascorbate was injected in mice in their study (1 g/mouse Kg vs. 4 g/mouse Kg), 15 times more of 1 than the amount of 3 used in our work was injected. Furthermore, previous work from our lab has demonstrated that an average of ~2 μmol/Kg of FR-targeted compound will saturate KB tumors in mice, and any additional folate conjugate injected will simply be excreted [58] . This saturating concentration (i.e. ~2 μmol/Kg) is 30 times lower than the amount of 1 used by Verrax et al. Furthermore, FR-targeted compounds generally saturate all FR in KB xenografts within 5 minutes of tail injection with kinetics that are largely independent of the type of folate linker [56] . In previous work, ascorbate concentrations of up to 8 mM and H₂O₂ concentrations of up to 20 µM were detected in the extracellular fluid at 20 to 30 minutes after tail injection of 0.5 mg ascorbate/g mouse [8] . Furthermore, H₂O₂ concentrations ≥ 25 µM in vitro were found to be toxic to cancer cells [59] . In the current work we are injecting 8 times more ascorbate/g of mouse (4 mg/g). Thus, larger peak concentrations of ascorbate and even higher levels of H₂O₂ (due to the targeted PHQ) in the intracellular fluid must have been produced at the tumor tissues described in the current work. However, taken together the fact that 3 is water-soluble and 1 is hydrophobic, it is conceivable that 3 might be more rapidly excreted than 1, while 1 might partition into cells more efficiently than 3, rendering the amount of 3 available for ascorbate reduction at the tumor site insufficient to generate the cytotoxic levels of ROS induced by 1. While differences in drug metabolism, tumor size, levels of ROS detoxifying enzymes, ascorbate excretion, etc. could have also contributed to the insensitivity of our murine tumor models to the ascorbate/quinone oxidant generating system, an unambiguous description of the actual mechanism is beyond the scope of this study.

In summary, no known eukaryotic cells can survive prolonged exposure to ROS, suggesting that constitutive generators of reactive oxygen species could constitute an evasion-proof method for killing essentially all tumor cells. In an effort to design such a mutation-resistant chemotherapy, the combination of a PHQ-derived non-re- leasable FR-targeted compound with ascorbate was investigated. Our data demonstrate a significant acceleration of the ascorbate cytotoxicity against FR+ KB cells, promoted by the presence of 3, if 3 is incubated with ascorbate in vitro, but not in vivo. Since, as described above, related nontargeted therapies from other labs have shown activity against other types of implanted tumors, our observations suggest that FR-targeting of quinones to KB and MDA-MB-231 tumors may be impaired by either poor vascularization and/or a limited concentration of quinone at the tumor site as compared to nontargeted quinone-ascorbate-based therapies.