Ti(IV) uptake and release by human serum transferrin and recognition of Ti(IV)-transferrin by cancer cells: understanding the mechanism of action of the anticancer drug titanocene dichloride

The organometallic anticancer agent titanocene dichloride, Cp(2)TiCl(2), is now in phase II clinical trials as an anticancer drug, but its mechanism of action is poorly understood. We show here that the interactions of Cp(2)TiCl(2) with human serum transferrin (hTF) and that of Ti(2)-hTF with adenosine triphosphate (ATP) have characteristics that could allow transferrin to act as a mediator for titanium delivery to tumor cells. Such reactions may therefore be important to the anticancer activity of this new class of drugs. Cp(2)TiCl(2) reacts rapidly with human apo-transferrin under physiological conditions (100 mM NaCl, 25 mM bicarbonate, and 4 mM phosphate, pH 7.4) with carbonate as a synergistic anion. The Cp ligands are released from the drug. Two-dimensional [(1)H, (13)C] NMR studies of epsilon-[(13)C]Met-hTF show that Ti(IV) loads the C-lobe first followed by the N-lobe and binds in the specific Fe(III) sites. The protein conformational changes induced by Ti(IV) appear to be similar to those induced by Fe(III). Carbonate can act as a synergistic anion in Ti(2)-hTF but does not appear to be essential. A specific Ti(IV)-hTF adduct is formed even in the absence of bicarbonate. When the pH of Ti(2)-hTF solutions is lowered, no Ti(IV) is released at the endosomal pH of ca. 5.0-5.5, but one Ti(IV) dissociates between pH 4.5-2.0. In contrast, in the presence of 1 mM ATP, all Ti(IV) is readily released from both lobes when the pH is lowered from 7.0 to 4.5. Moreover, Fe(III) displaces Ti(IV) rapidly from the C-lobe of Ti(2)-hTF (<5 min) but only slowly (days) from the N-lobe. Thus, the species Fe(C)Ti(N)-hTF might also provide a route for Ti(IV) entry into tumor cells via the transferrin receptor. Ti(2)-hTF effectively blocked cell uptake of radiolabeled (59)Fe-hTF into BeWo cells, a human placental choriocarcinoma cell line in culture. These results imply that titanium transferrin might be recognized by the transferrin receptor and be taken up into cancer cells.     

Titanium(IV) targets phosphoesters on nucleotides: implications for the mechanism of action of the anticancer drug titanocene dichloride

Abstract Reactions between the anticancer drug titanocene dichloride (Cp2TiCl2) and various nucleotides and their constituents in aqueous solution or N,N-dimethylformamide (DMF) have been investigated by 1H and 31P NMR spectroscopy and in the solid state by IR spectroscopy. In aqueous solution over the pH* (pH meter reading in D2O) range 2.3-6.5, CMP forms one new species with Ti(IV) bound only to the phosphate group. In acidic media at pH*<4.6, three species containing titanocene bound to the phosphate group of dGMP, AMP, dTMP and UMP are formed rapidly. The bases also appear to influence titanocene binding. Only one of these Ti(IV)-bound species can be detected in the pH* range of 4.6-6.5 in each case. The order of reactivity towards Cp2TiCl2(aq) at pH* ca. 3 is GMP>TMP approximately AMP > CMP. At pH* > 7.0, hydrolysis of Cp2TiCl2 predominated and little reaction with the nucleotides was observed. Binding of deoxyribose 5'-phosphate and 4-nitrophenyl phosphate to Cp2TiCl2(aq) via their phosphate groups was detected by 31P NMR spectroscopy, but no reaction between Cp2TiCl2(aq) and deoxyguanosine, 9-ethylguanine or deoxy-D-ribose was observed in aqueous solution. The nucleoside phosphodiesters 3',5'-cyclic GMP and 2',3'-cyclic CMP did not react with Cp2TiCl2(aq) in aqueous solution; however, in the less polar solvent DMF, 3',5'-cyclic GMP coordination to [Cp2Ti]2+ via its phosphodiester group was readily observed. Binding of titanocene to the phosphodiester group of the dinucleotide GpC was also observed in DMF by 31P NMR. The nucleoside triphosphates ATP and GTP reacted more extensively with Cp2TiCl2(aq) than their monophosphates; complexes with bound phosphate groups were formed in acidic media and to a lesser extent at neutral pH. Cleavage of phosphate bonds in ATP (and GTP) by Cp2TiCl2(aq) to form inorganic phosphate, AMP (or GMP) and ADP (or GDP) was observed in aqueous solutions. In addition, titanocene binding to ATP was not inhibited by Mg(II), but the ternary complex titanocene-ATP-Mg appeared to form. These reactions contrast markedly with those of the drug cisplatin, which binds predominantly to the base nitrogen atoms of nucleotides and only weakly to the phosphate groups. The high affinity of Ti(IV) for phosphate groups may be important for its biological activity.     

Reactions of Azotobacter vinelandii nitrogenase using Ti(III) as reductant

Nitrogenase-catalyzed reactions using Ti(III) were examined under a wide variety of conditions to determine the suitability of Ti(III) to serve as a general nitrogenase reductant. Solutions prepared from H2-reduced TiCl3, aluminum-reduced TiCl3, TiCl2, evaporated TiCl3 from an HCl, solution, and TiF3 were evaluated as reductants. Three general types of reactivity were observed. The first showed that, below Ti(III) concentrations of about 0.50 mM, nitrogenase catalysis utilized Ti(III) in a first-order reaction. The second showed that, above 0.50 mM, the rate of nitrogenase catalysis was zero order in Ti(III), indicating the enzyme was saturated with this reductant. Above 2.0-5.0 mM, nitrogenase catalysis was inhibited by Ti(III) depending on the titanium source used for solution preparation. This inhibition was investigated and found to be independent of the buffer type and pH, while high salt and citrate concentrations caused moderate inhibition. [Ti(IV)] above 2.0-3.0 mM and [Ti(III)] above about 5.0 mM were inhibitory. ATP/2e values were 4-5 for [Ti(III)] at or below 1.0-2.0 mM, 2.0 from 5.0 to 7.0 mM Ti(III) where nitrogenase is not inhibited, and 2.0 above 7.0 mM Ti(III) where severe inhibition occurs. For nitrogenase-catalyzed reactions using Ti(III) as reductant, the potential of the solution changes with time as the Ti(III)/Ti(IV) ratio changes. From the change in the rate of product formation (Ti(III) disappearance) with change in solution potential, the rate of nitrogenase catalysis was determined as a function of solution potential. From such experiments, a midpoint turnover potential of -480 mV was determined for nitrogenase catalysis with an associated n = 2 value.     

Antitumor activity of bent metallocenes: electronic structure analysis using DFT computations

The antitumor activities of bent metallocenes [Cp–M–Cp]²⁺ (M = Ti, V, Nb, Mo) and complexes of them with guanine, adenine, thymine and cytosine nucleotides have been probed using electronic structure calculations. DFT/BP86 calculations have revealed that the bent metallocene–nucleotide interaction strongly depends on the stability of the hydrolyzed form of the bent metallocene dichloride [Cp₂M]²⁺ species, and in turn the stability of the [Cp₂M]²⁺ species strongly depends on the electronic structure of [Cp₂M]²⁺. Detailed electronic structure and Walsh energy analyses have been carried out for the hydrolyzed forms of four [Cp–M–Cp]²⁺ (M = Ti, V, Nb, Mo) species to find out why the bent structure is unusually stable. Energy changes that occur during the bending process in frontier molecular orbitals as well as the p(π)–d(π) overlap have been invoked to account for the anticipated antitumor activities of these species. The bonding situation and the interactions in bent metallocene–nucleotide adducts were elucidated by fragment analysis. Of the four nucleotides complexed with the four bent metallocenes, adenine and guanine show better binding abilities than the other two nucleotides. Metallocenes of second-row transition metals exhibit better binding with pyrimidine-base nucleotides. In particular, the Lewis acidic bent metallocenes interact strongly with nucleotides. The antitumor activity is directly related to the binding strength of the bent metallocene with nucleotide adducts, and the computed interaction energy values correlate very well with the experimentally observed antitumor activities.     

A (1)H NMR study of the interaction of antitumor metallocenes with glutathione

The interaction of the antitumor active metallocene dihalides Cp(2)MCl(2) (M=Ti, Nb, Mo) and 1 equiv. of glutathione was studied by (1)H NMR spectroscopy at pD 2-7 in 4 mM NaCl solutions. No interaction between glutathione and titanocene dichloride was detected at pD 2, while at pD 5-7 competitive hydrolysis of the cyclopentadienyl ligands occurred. With niobocene dichloride formation of approximately 20% of an adduct was observed at pD 2 and 5, but hydrolysis of the Cp ligands in the adduct occurred over 24 h. Molybdocene dichloride formed two stable adducts at pD 6 which were tentatively assigned as a Cp(2)Mo-glutathione chelate involving coordination of the cysteine thiol and glycine carboxylate to the metal centre, and a thiol centred 1:2 Cp(2)Mo-glutathione complex. The implications for the mechanism of antitumor action of the metallocene dihalides is discussed.     

Binding of V(IV) to human transferrin: potential relevance to anticancer activity of vanadocene dichloride

The action mechanism of vanadocene dichloride, Cp2VCl2 (Cp=eta5-C5H5), has been investigated by interaction with human serum transferrin for its promising antitumor activities. Our results have shown that Cp2VCl2 binds to transferrin and form a new complex, and the calculated apparent association constant is 1.37 x 10(5)M(-1) from the fluorescence quenching. Simultaneously, the variation of the secondary structure of transferrin occurs, most probably due to the coordination of the amino residues of protein with VIV. It was evidenced that Cp is released free in solution after VIV binding to transferrin by 1H NMR measurements. These results have shown that Cp2VCl2 forms a complex with transferrin, which may provide a possible pathway in the transport and targeted delivery of the antitumor agent.     

Structure–activity studies of Ti(IV) complexes: aqueous stability and cytotoxic properties in colon cancer HT-29 cells

As part of our research efforts in the area of titanium-based antitumor agents, we have investigated the cytotoxic activity of [Ti(4)(maltolato)(8)(mu-O)(4)], (Cp-R)(2)TiCl(2) and (Cp-R)CpTiCl(2) (R = CO(2)CH(3) and CO(2)CH(2)CH(3)), and three water-soluble titanocene-amino acid complexes-[Cp(2)Ti(aa)(2)]Cl(2) (aa = L: -cysteine, L: -methionine, and D: -penicillamine)-on the human colon adenocarcinoma cell line, HT-29. The capacity of [Ti(4)(maltolato)(8)(mu-O)(4)] to donate Ti(IV) to human apo-transferrin and its hydrolytic stability have been investigated and compared to the previously reported data on modified titanocenes with either hydrophilic ancillary ligands or the functionalized cyclopentadienyl ligands. Notably, the titanium-maltolato complex does not transfer Ti(VI) to human apo-transferrin at any time within the first seven days of its interaction, demonstrating the inert character of this species. Stability studies on these complexes have shown that titanocene complexes decompose at physiological pH while the [Ti(4)(maltolato)(8)(mu-O)(4)] complex is stable at this pH without any notable decomposition for a period of ten days. The antitumor activity of these complexes against colon cancer HT-29 cells was determined using an MTT cell viability assay at 72 and 96 h. The titanocene-amino acid and the (Cp-R)(2)TiCl(2)/(Cp-R)CpTiCl(2) (R = CO(2)CH(3)) complexes were not biologically active when human transferrin was absent; they also were inactive when human transferrin was present at dose-equivalent concentrations. (Cp-R)(2)TiCl(2) and (Cp-R)CpTiCl(2) (R = CO(2)CH(2)CH(3)) showed cytotoxic activity in HT-29 cells comparable to that which is displayed by titanocene dichloride. The titanium-maltolato complex had higher levels of cytotoxic activity than any other titanocene complex investigated. Transferrin may be important in protecting the titanium center from hydrolysis, but this may be achieved by selecting ligands that could result in hydrolytically stable, yet active, complexes.     

Genotoxicity and transmission electron microscopy studies of molybdocene dichloride

The cytotoxic effects of molybdocene dichloride (Cp(2)MoCl(2)) were investigated in V79 Chinese hamster lung cells using the micronucleus assay. Cp(2)MoCl(2) produced significant genotoxic damage whereby 0.2micronuclei/1000 binucleated cells were induced per muM of Cp(2)MoCl(2). Transmission electron microscopic analysis of thin-sectioned cells treated with Cp(2)MoCl(2) (300microM) showed distinct morphological alterations of the nuclei, condensation of chromatin, and a high incidence of polynucleated cells. Implications for the mechanism of antitumor action of molybdocene dichloride are discussed.     

Study on the amount of binding of anti-tumor metal complexes to different target sites of dGMP using trans-[en(2)Os(eta2-H(2))(CF(3)SO(3))](CF(3)SO(3)) in a competitive mode

In order to investigate the binding sites and the amount of binding of a number of anti-tumor metal complexes (cisplatin, Cp(2)TiCl(2) and (CH(3))(2)SnCl(2)) to target molecule DNA mononucleotides in aqueous solution, a 1H NMR recognition probe, trans-[en(2)Os(eta2-H(2))(CF(3)SO(3))](CF(3)SO(3)), was used in a competitive mode. The minimum percentages of binding of anti-tumor metal complexes to different sites of dGMP were also determined.     

Purification and partial characterization of camel (Camelus Dromedarius) ceruloplasmin

Adult and young camel ceruloplasmin (Cp) were isolated and purified using the single-step chromatography on amino ethyl-activated sepharose. There are no differences between the adult and the young camel protein. The molecular mass of the protein, as estimated by SDS-PAGE (denaturant conditions), was approximately 130000 Da. The electrophoretic mobility of camel Cp is slightly higher as compared to human and sheep protein suggesting that the camel Cp is homogeneous, compact and more acid. The copper content was estimated to be 5.8+/-0.3 atoms per molecule. The spectroscopic feature includes an absorption maximum at 610 nm, which could be attributed to type 1 copper. The EPR spectrum was completely devoid of any typical signal of the type 2 copper. The kinetic parameters of the adult camel Cp for the specific activity as p-phenylendiamine oxidase were determined as K(m)=0.42 mM and V(max)=0.93 microM NADH/mn/mg Cp. The optimum pH for the activity was 5.7.     

Specific inhibition by cyclodextrins of raw starch digestion by fungal glucoamylase.

alpha-, beta-, and gamma-cyclodextrins (CDs) completely inhibited raw starch digestion by glucoamylase I (GA I, MW 90,000) from Aspergillus awamori var. kawachi, and inhibited by 85% the raw starch adsorption of GA I at the CD concentrations of 1-5 mM. CDs at 1-5 mM did not inhibit gelatinized starch hydrolysis by GA I, but at the concentration of 50 mM, they inhibited such hydrolysis slightly. GA I was specifically adsorbed onto CD-Sepharose 6B, but glucoamylase I' (GA I', MW 73,000), which does not adsorb onto or digest raw starch, from the same strain was not adsorbed onto that gel. The adsorption of the glucoamylases onto raw starch and CD-Sepharose 6B was correlated to their digestion of raw starch. The hydrophobic adsorption of GA I onto CDs and raw starch occurred competitively at the Cp region, which is on the C-terminal side of Gp-I in the site for raw starch affinity of GA I, and inclusion complexes were formed.     

Inhibition of ceruloplasmin and other copper oxidases by thiomolybdate

The dietary antagonism between copper and molybdate salts prompted a study of the inhibition of copper enzymes by thiomolybdate (TM). TM strongly inhibited the oxidase activity of five copper oxidase with I50% values in the 1-5 microM range. The mechanism of the TM effect on the copper oxidase, ceruloplasmin (Cp) (E.C. 1.16.3.1), was studied in detail. In Vmax vs. E plots, TM gave parallel data suggesting irreversibility but a large number of TM molecules per Cp were required. The inhibition of Cp by TM could not be reversed by dialysis. Isolation of TM-inhibited Cp on Sephadex G-10 did not yield any active Cp molecules. Cu(II) did not restore any inhibited oxidase activity. Gel electrophoresis supported the covalent binding of Cp by TM without any extensive change in protein structure. EPR results confirmed that Cu(II) is reduced to Cu(I) after reaction with TM. However, the Mo(VI) in MoS4(2-) did not change in oxidation number. Analysis of the TM-Cp compound accounted for all six Cu atoms as found in native Cp. The data suggest the covalent binding of sulfide to Cp copper. TM also inhibited the activity of ascorbate oxidase, cytochrome oxidase, superoxide dismutase, and tyrosinase. However, no inhibition of carbonic anhydrase, a zinc enzyme, was observed at 1 mM TM.     

Characterization of the Light-induced Transient States of the Chlorophyll Proteins 668 and 743 from Atriplex rosea

The light-induced transient states of chlorophyll-protein 668 (Cp668) and its photoconverted from Cp743 were investigated using flash photolysis. Short lived transient species induced by a short flash were detected in both Cp668 and Cp743. The Cp668 transient had a half decay time of 2.0 milliseconds and showed a broad absorption band at 460 nanometers. The Cp743 transient had a half decay time of only 0.6 millisecond and had a major absorption peak at 410 nanometers in addition, to a broad absorption band around 530 nanometers. Both transient signals were quenched by oxygen. Cp668 had a temperature-dependent delayed fluorescence at room temperature with a half-life of 2.0 milliseconds, the same as the life-time of the absorption transient. This suggests that the transient species observed was a triplet state of chlorophyll.     

Lactoferrin,myeloperoxidase, and ceruloplasmin: complementary gearwheels cranking physiological and pathological processes

Copper-containing plasma protein ceruloplasmin (Cp) forms a complex with lactoferrin (Lf), an iron-binding protein, and with the heme-containing myeloperoxidase (Mpo). In case of inflammation, Lf and Mpo are secreted from neutrophil granules. Among the plasma proteins, Cp seems to be the preferential partner of Lf and Mpo. After an intraperitoneal injection of Lf to rodents, the “Cp–Lf” complex has been shown to appear in their bloodstream. Cp prevents the interaction of Lf with protoplasts of Micrococcus luteus. Upon immunoprecipitation of Cp, the blood plasma becomes depleted of Lf and in a dose-dependent manner loses the capacity to inhibit the peroxidase activity of Mpo, but not the Mpo-catalyzed oxidation of thiocyanate in the (pseudo)halogenating cycle. Antimicrobial effect against E. coli displayed by a synergistic system that includes Lf and Mpo–H₂O₂–chloride, but not thiocyanate, as the substrate for Mpo is abrogated when Cp is added. Hence, Cp can be regarded as an anti-inflammatory factor that restrains the halogenating cycle and redirects the synergistic system Mpo–H₂O₂–chloride/thiocyanate to production of hypothiocyanate, which is relatively harmless for the human organism. Structure and functions of the “2Cp–2Lf–Mpo” complex and binary complexes Cp–Lf and 2Cp–Mpo in inflammation are discussed.     

Subcellular mobility of the calpain/calpastatin network: an organelle transient

Calpain (Cp) is a calcium (Ca(2+))-dependent cysteine protease. Activation of the major isoforms of Cp, CpI and CpII, are required for a number of important cellular processes including adherence, shape change and migration. The current concept that cytoplasmic Cp locates and associates with its regulatory subunit (Rs) and substrates as well as translocates throughout the cell via random diffusion is not compatible with the spatial and temporal constraints of cellular metabolism. The novel finding that Cp and Rs function relies upon tenacious hydrophobic interactions with organelle membranes offers a unifying explanation for the paradoxical and puzzling features of Cp activation and regulation such as how nM concentrations of intracellular Ca(2+) can activate Cp molecules requiring muM to mM concentrations of Ca(2+) for in vitro activation, and how this protease can spatially and temporally locate specific substrates and translocate throughout the cell. We hypothesize that Cp and its regulatory moieties associate with organelles to facilitate the activation of this protease resulting in the cleavage of substrates and aid in its translocation throughout the cell.     

Transport of free polymannose-type oligosaccharides from the endoplasmic reticulum into the cytosol is inhibited by mannosides and requires a thapsigargin-sensitive calcium store

The transport of free polymannose-type oligosaccharides from the lumen of the endoplasmic reticulum into the cytosol has been recently demonstrated (Moore,S.E.H., et al., 1995, EMBO J., 14, 6034-6042), but at present little is known of the characteristics of this process. Here, it is shown that inhibition of the transport of endogenously synthesized metabolically radiolabeled free oligosaccharides out of the endoplasmic reticulum into the cytosol of permeabilized HepG2 cells occurs when assays are conducted in the presence of mannose (IC50, 4.9 mM), or its derivatives modified at the first carbon (C1) of the sugar ring; alpha-methyl mannoside (IC50, 2.0 mM), mannoheptulose (IC50, 1.6 mM), and alpha-benzyl mannoside (IC50, 0.8 mM), whereas other monosaccharides (50 mM), differing from mannose at position; C2 (glucose), C3 (altrose), C4 (talose), C5 (l-rhamnose), and C6 (mannoheptose), have little effect. N-Acetylglucosamine does not inhibit oligosaccharide transport and, furthermore, although mannobioses and a mannotriose inhibit free oligosaccharide transport, di-N-acetylchitobiose is without effect. It is also shown that if the transport assay buffer is either depleted of calcium ions, or supplemented with the Ca2+/Mg2+ATPase inhibitor, thapsigargin, or with calcium ionophores, free oligosaccharide transport out of the endoplasmic reticulum is inhibited. These results demonstrate that the terminal nonreducing mannosyl residues of free polymannose-type oligosaccharides and not their N-acetylglucosamine-containing reducing termini, play an important role in the interaction of the free oligosaccharide with the transport machinery, and that this transport process requires the presence of calcium sequestered in the lumen of the endoplasmic reticulum.      

Synthesis,Ti(IV) intake by apotransferrin and cytotoxic properties of functionalized titanocene dichlorides

Functionalization of cyclopentadienyl (Cp) ligands and incorporation of these into a Ti(IV) center require careful design and selection of the appropriate synthetic routes to obtain the desired product in reasonably good yields. As part of our research efforts in the area of titanocene antitumor agents, we have revisited the synthesis of Cp rings with electron-withdrawing groups and their corresponding titanocene dichlorides, (Cp-R)₂TiCl₂ and (Cp-R)CpTiCl₂, where R is CO₂CH₃ and CO₂CH₂CH₃. These complexes were characterized by elemental analysis and ¹H and ¹³C NMR and IR spectroscopies. This report presents the first detailed synthetic route for (Cp-CO₂CH₂CH₃)CpTiCl₂ and provides an alternate route for synthesis of (Cp-R)₂TiCl₂ complexes. The ability of these complexes to deliver Ti(IV) to apotransferrin was investigated to elucidate how the functionalized Cp ligands affect the titanium intake by apotransferrin. The subject complexes transfer Ti(IV) to human apotransferrin, loading both N- and C-lobes. The antitumor activity of these complexes against HT-29 cancer colon cells was determined using a 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide assay. Carboethoxy Cp functionalization results in complexes with a toxicity comparable to that of titanocene dichloride. The carbomethoxy-functionalized complexes proved to be nonactive at the time intervals studied here, regardless of their ability to donate the titanium atom to human apotransferrin.     

N-(2-hydroxypropyl)methacrylamide copolymers of a glutathione (GSH)-activated glyoxalase i inhibitor and DNA alkylating agent: synthesis, reaction kinetics with GSH, and in vitro antitumor activities

The incorporation of anticancer prodrugs into polyacrylamide conjugates has been shown to improve tumor targeting via the so-called "enhanced permeability and retention" effect. This strategy has now been expanded to include two different classes of glutathione (GSH)-activated antitumor agents prepared by radical polymerization of N-(2-hydroxypropyl)methacrylamide (HPMA) with 2-methacryloyloxy-methyl-2-cyclohexenone (7) and/or with S-(N-4-chlorophenyl-N-hydroxycarbamoyl-thioethyl)methacrylamide (8), followed by treatment with 3-chloroperoxybenzoic acid, to give the HPMA copolymers of 7 and the 8-sulfoxide, respectively. In aqueous-buffered solution at pH 6.5, GSH reacts rapidly with poly-HPMA-8-sulfoxide (k approximately 2.3 mM(-1) min(-1)) to give S-(N-4-chlorophenyl-N-hydroxycarbamoyl)glutathione (1), a tight-binding transition state analogue inhibitor of the antitumor target enzyme glyoxalase I (K(i) = 46 nM), or with poly-HPMA-7 (k approximately 0.02 mM(-1) min(-1)) to give the electrophilic antitumor agent 3-glutathio-2-methylenecyclohexenone (4). Indeed, B16 melanotic melanoma in culture is inhibited by poly-HPMA-8-sulfoxide and by poly-HPMA-7 with IC(50) values of 168 +/- 8 and 284 +/- 5 microM, respectively. These values are significantly greater than those of the unpolymerized prodrugs suggesting that the cytotoxicity of the polymer prodrugs might be limited by slow cellular uptake via pinocytosis. This prodrug strategy should be applicable to a range of different GSH-based antitumor agents.