ESR of copper and iron complexes with antitumor and cytotoxic properties.

The relatively few iron and copper metal complexes which have been examined in cells and tissues for their redox properties, radical generation properties, and antitumor activity are discussed for studies which utilized electron spin resonance spectroscopy (ESR). A common property of a number of metal complexes, which include bleomycin, adriamycin, and thiosemicarbazones described in this review, is that they are readily reduced by thiol compounds and oxidized by oxygen or reduced species of oxygen to produce radicals. Structural features of these reactions are identified by ESR spectroscopy in model systems and often in cells. Furthermore, ESR spectroscopy has been most useful to probe the environment of the complexes in cells and to measure the rate of reduction of their oxidized forms. As a result of these studies, it is anticipated that more attention will be given to the exploration of redox-active metal complexes as drugs.


Introduction
It has long been known that the interaction of radiation with cells and tissues is a complicated process beginning with the cleavage of H20 into e-, H', or OHradicals. These radical species then react indiscriminately with cellular constituents. The cytotoxic reactions are thought to involve radical attack on DNA (1)(2)(3). The extent of radiation damage to tissues can be modified by modulation of their oxygen and thiol content.
An analogous hypothesis exists for damage to tissues and cells caused by radicals which are generated by photolysis or redox chemistry which often involves a metal ion instead of high energy radiation. While other papers in this issue center on the detection of radicals, this paper focuses on metal complexes which can be catalysts for generation of the more visible radicals and radical damage. Moreover, this review is limited to a description of those metal complexes which can be detected by electron spin resonance spectroscopy (ESR). ESR only detects metal ions with unpaired electrons in the inner d andforbitals. Hence, for one electron redox reactions, it is frequently the case that the metal complex will be reduced to a diamagnetic state, i.e., Cu  is reduced to Cu (I), and only the steady-state concentration for the oxidized form ofthe metal can be detected by ESR.
There are relatively few metal complexes which have been examined in cells and tissues for their redox and radical generating properties. Whatever the reasons for this, it is not for lack of expectation that a variety of reactions will occur. Thus, for example, the area of cellular oxygen chemistry is an important domain of metallobiochemistry. Many of the proteins which interact with oxygen are metalloproteins, including, hemoglobin, cytochrome oxidase, and cytochrome P-450. There is also abundant evidence that the generation of adventious reduced species of oxygen is a common and deleterious occurrence during oxygen metabolism in an aerobic organism. While these may be initially formed as byproducts of oxygen-dependent enzymatic reaction, further reactions are likely to involve metallo-redox chemistry as described in equations (1)-(3).

MR + 02
Ma + 02 Mm + 02 7 Mm1 + 02 (1) Mm + H202 = M'b"1 + OH-+ OH' (3) Reaction (3) is often referred to as the Fenton reaction in which production of OH* occurs at a faster rate than for the direct reaction of O2 with H202, the Haber-Weiss reaction: 02' + H202 -02 + OH-+ OH' Just as reactive forms of oxygen may be generated by metal-based reactions, so too such entities as superoxide ion and hydrogen peroxide are handled by the metalloproteins superoxide dismutase and catalase which require Fe, Cu(Zn), or Mn. Thus, from many perspectives one expects a wealth of oxygen based radical chemistry to occur when exogeneous complexes of redox metals enter cells.

Copper Complexes as Antitumor and Cytotoxic Agents ESR Studies of Tridentate Cupric Complexes
More than 20 years ago French and co-workers began synthesizing metal binding ligands as potential antitumor agents with the rationale that they might extract transition metal ions from key sites in cancer cells to inactivate them (4). One class of these ligands is the a-N-heterocyclic carboxaldehyde thiosemicarbazones ( Fig. 1), some of which are active as antitumor agents in animals but are not effective in humans (5). Studies on the Cu and Fe complexes of some ofthe active ligands show that the metal complexes are much more cytotoxic than the parent compounds (6).
Detailed examination of the reaction of 2-formylpyridine thiosemicarbazonato Cu(II) with cell types has led to a scheme for intracellular reaction of the complex (Fig. 2). Upon its addition to plasma or culture media, adduct fornation between CuL+ and Lewis bases is observed in the ESR spectrum at 77°K (7). The complex or some adduct species is readily taken up by Ehrlich cells (7). Once in cells, there is little efflux of CuL+, presumably because it is bound to or reacts irreversibly with cellular structures. That cellular adducts of the copper complex do form is shown by the change in its ESR parameters as it enters cells (Fig. 3). The new spectrum taken at 770K could be modeled by the addition of CuL to excess glutathione, CuL-SG, or to cathemoglobin which has several pairs of reactive thiols, Cu-L-S-CatHb (Fig. 4) (8,9). In the absence of model studies, a good indication of the number of nitrogen, oxygen, and sulfur donor atoms bound to Cu comes from the value of g 11 and A 11 indicated in Figures 3 and 4, which can be compared to known values with the aid of Peisach-Blumberg plots (10) and from the hyperfine structure on the high field lines (g, region). The cellular adduct has ESR parameters consistent with an N2S2 coordination environment, suggesting as do the model studies that CuL forms adducts with cellular thiols.
Room temperature ESR studies of the interaction of CuL with cells show that the complex displays an immobilized spectrum indicative of its slow motion in solution (Figs. 5 and 6) (11). The intensity of the signal also decreases with time, suggesting that the complex is gradually reduced in cells. For an adduct formed from CuL+ and GSH, the room temperature spectrum should be the isotropically averaged spectrum due to rapid motion of a low molecular weight complex in a medium which is presumably about as viscous as water. Thus CuL most likely forms an adduct with a cysteine amino acid residue from a protein, for the motion for the complex is much slower.
The time course of reaction of other tridentate copper complexes, which are cytotoxic to Ehrlich cells, is similar to that of CuL. Thus, the order of rate of disappearance of the ESR of three such complexes is pyridine-2-carboxaldehyde-2'-pyridylhydrazonate copper (II) > salicylaldehydebenzoylhydrazonato copper (II) = 2-CuL. The net rate of reduction of CuL in cells as measured by ESR is faster than previously observed by following loss of the visible absorbance of the complex (11). However, both in cells and in model studies of the reaction of CuL with glutathione, there is a large increase in oxygen consumption during the total course of reaction. In addition, the thiol content of the systems decreases (7). In the model, there is a burst of production of 02and OHas measured by spin trapping reactions (7). During the reaction of CuL with GSH, a steady state is reached in which Cu(II)L concentration remains approximately constant, as the complex catalyzes the reaction of oxygen by the sulfhydryl groups of GSH. According to these results, when CuL is taken up by Ehrlich cells a redox cycle is established (Fig. 2) until oxygen is depleted or net reduction of copper is stabilized through the binding of Cu(I) to other sites in the cell. The fact that the room temperature ESR study points to the binding of CuL and the other complexes to protein thiols suggests that this redox cycle may involve the oxidation of important cysteinyl residues of enzymes.

ESR Investigations of Tetradentate Cupric Complexes
The bis(thiosemicarbazone) of 3-ethoxy-2-oxobutyraldehyde and its copper complex, Cu(II)KTS, have excellent antitumor properties in a variety of animal models. In frozen solution the ESR spectrum of CuKTS has a hyperfine structure attributable to the binding of Cu to two equivalent nitrogen donor atoms (Fig. 7). When the complex diffuses into cells this signal disappears raRidly with a pseudo first-order rate constant of  (12,13). The resultant Cu(I) species are reactive with oxygen to set in motion a redox cycle of Cu(I) and Cu(II), which catalyzes the reduction of oxygen by thiols (12). When Ehrlich cells are titrated with CuKTS, the first major site of binding of Cu(I) is metallothionein, normally a Znprotein (13). Cu(I)-metallothionein is not rapidly reactive with oxygen (14), so this structure is not involved in the catalysis of oxygen reduction. The structure of CuKTS has been studied by means of ESR because of its established antitumor properties (13,(15)(16)(17)(18). The best-resolved ESR spectrum was obtained from CuKTS doped into the nickel analog, NiKTS (18). The principal values of the ESR parameters were used to determine that the copper-ligand bonds are highly covalent and the Cu-N and Cu-S bonds may be regarded as essentially independent.
The function of the bis(thiosemicarbazone) ligand of CuKTS is both to provide a stable form of Cu(II) which can reach tumor cells and to set an appropriate redox potential for the copper center for its reaction in the reductive environment of the cell (12). Thus, 3-ethoxy-2-oxobutyraldehyde-bis(N4-dimethyl-thiosemicarbazonato) Cu(II), CuKTSM2, has a redox potential about 100 mV more negative than CuKTS (Table 1), reacts very slowly with Ehrlich cells, and at similar concentrations is not cytotoxic to tumor cells (19).
A recent reinvestigation of CuKTSM2 demonstrated that the complex destroys Ehrlich cells at higher concentrations of drug than previously used (13). The complex was known to localize in lipophilic parts of cells, presumably membrane (18). Room temperature ESR studies were undertaken to study CuKTSM2 bound in Ehrlich cells (13). The ESR signal was found to be stable and essentially immobilized in cells (Figs. 8 and 9). As the concentration of CuKTSM2 is increased, a second mobile phase was observed. Coincidently with the tran-sition from immobilized to mobile CuKTSM2, the cytotoxicity of the complex greatly increases. These ESR results and those with the tridentate copper complexes show clearly the importance of doing ambient temperature ESR measurements to probe the dynamic environment of the complexes.
Bidentate Copper Complexes of 1,10-Phenanthroline and Cytotoxic Agents It has been known for years that 1,10-phenanthroline inhibits cell proliferation (20). The proposed mechanism has been that this compound acts as a ligand for cellular zinc to mimic a condition of nutrient zinc deficiency, which is also known to inhibit growth (21). Another explanation suggests that a cuprous phenanthroline complex is the active form of the ligand (Fig. 10) (22,23). Indeed, several recent studies have inquired into the reaction of copper-phenanthroline, oxygen, a reducing agent, and DNA to degrade double-stranded DNA into acid-soluble fragments (24)(25)(26). As with other copper complexes discussed above, the phenanthroline-Cu complex cycles between Cu(II) and Cu(I) to catalyze the reduction of oxygen to reactive radical species by the reducing agent. Unpublished studies of the reaction of 1,10-phenanthroline with Ehrlich cells lends support to the possible formation of copper-phenanthroline complexes in cells. When the ligand is added to cells, copper redistribution occurs in which Cu(I) metallothionein is found (20). A reasonable intermediate in this reaction would be Cu(I)-1,10-phenanthroline. It is noted that this ligand also causes substantial losses in cellular Zn and Fe, so that its effects on copper are not clearly related to cytotoxicity.
The ESR data for copper doped into a dichloro-1,10phenanthroline zinc host indicate that the cupric ion is in a distorted tetrahedral structure (27). These parameters, particularly A. = 123 x 10-4 cm-l, are more characteristic of the type I distorted tetrahedral structure than the square planar structure (27). Bonding pa-  rameters for ternary complexes of Cu-1,10-phenanthroline and amino acids excluding histidine indicate that these adducts have a considerable amount of covalent bonding (28). Nitrogen hyperfine structure appears in the g, region for most of these compounds. More recent infrared and visible studies suggest that ternary complexes with a pyramidal square planar configuration are formed between 1,10-phenanthroline and copper-dipeptide complexes (29). ESR parameters from room temperature and frozen solutions were tabulated but no nitrogen hyperfine splittings were resolved. It is suggested'that the peptide supplies three donor atoms to the square plane and one nitrogen from 1, 1O-phenanthroline occupies the fourth planar position, while the other atom occupies the apical position. This  adduct is, in a sense, the mirror image of the adduct of CuL and protein. CuL is a tight tridentate complex, and the protein supplies the fourth donor atom to complete the square planar configuration or a fourth donor atom and an atom which binds in an axial position to complete the square pyrimidal configuration. Nitrogen atoms from 1,1O-phenanthroline (L') complete the square pyramidal configuration while the protein provides the tight tridentate binding site.
Although additional information about the speciation of coordinately unsaturated copper complexes in cells needs to be obtained, one can imagine subtle differences  Another class of copper complexes with cytotoxic activity has been discovered in the investigation of the report that cells treated with diethyldithiocarbamate (dtc) are more susceptible to bleomycin (Blm), a drug I ' I I I discussed in following sections, than control cells (30).   (5) dtc. Clusters or aggregates of Cu(dtc)2 are formed con-CuL-B-active site + protein~sistent with the low solubility of Cu(dtc)2 in aqueous 7 . .
.--1. 7_ \ ..solvents, which probably helps pull the reaction toward I -L' + :B-active-site --protein Cu-B-active site + L' "^/~~~( 6) Cu(dtC)2. Thus, dtc may enhance the activity of Blm by limiting its reaction with cellular copper. However, to the extent that Cu(dtc)2 forms in cells, a new, highly cytotoxic copper complex is generated which independupper reaction would transfer the antitumor ently augments the effects of Bhn. 24 ESR OF COPPER AND IRON COMPLEXES + g1,=2.14 For the partial spectra in the low-field region, the gain is increased twofold, the modulation amplitude was increased threefold, and the time constant was increased from 1 to 3 sec. The peak to peak height (position indicated by the solid arrow) of the center hyperfine line in the 1:2:3:2:1 pattern for the Ml = 3/2 high-field ESR line for CuKTSM2 was used to indicate the concentration of the mobile form. This line has the largest intensity for the signal in the mobile phase and overlaps at a point of low intensity for the immobile signal. The low-field line for the immobile signal (position indicated by the dashed arrow) in the g11 region does not overlap with the mobile signal and was used to indicate the relative concentration of the immobile signal. From Antholine et al. (13).
The ESR spectrum of Cu(dtc)2 has been well studied (32)(33)(34). The use of dtc to measure quantitatively inorganic cupric ions using ESR has been reviewed by Janzen (35). The source of a copper ESR signal previously observed in fatty tissues has been identified as Cu(dtc)2. This complex is formed through contact of copper-containing tissue with surgeon's gloves laced with dtc-like complexes (36). Cu(dtc)2 levels of 0.5 ,uM can be detected in tissues because the ESR lines are very sharp for this copper complex. Besides sensitivity, Cu(dtc)2 complexes are expected to form adducts in complex biological systems. Model ESR studies have already shown that mixed complexes of the type Cu(dtc)X can be identified (37).
In summary, paramagnetic copper complexes and their adducts should form in cells and be relatively stable in cells depleted of reducing equivalents. Detection and isolation of these adducts in cells is, in our opinion, the next step to better understanding for the interaction of cupric forms of these antitumor agents with cells.

ESR Studies of Cupric Bleomycin
Bleomycin is a clinically used antitumor agent (Fig. 11). It is isolated from Streptomyces verticillus as a copper complex (38). Although CuBlm is an active antineoplastic drug, it is inactive in the cleavage of DNA, a reaction thought to be the molecular basis of cytotoxicity (see FeBIm, below) (31,39,40). An inquiry into the redox reactivity of CuBIm with thiols such as glutathione and oxygen demonstrated that in contrast to the tridentate thiosemicarbazonato copper complexes, there is little redox cycling of copper and little reduction of oxygen as reductive dissociation occurs (31,(41)(42)(43).
ESR studies generally have supported the contention that the cupric site in CuBlm is square pyramidal (44)(45)(46)(47)(48)(49)(50). Data from electron spin echo spectroscopy, which is useful for determining whether imidazole is a ligand, confirms that an imidazole is involved in the binding of cupric ion (51). Nitrogen hyperfine structure is better resolved in both the 911 and g, regions at lower microwave frequencies (S-band) than at the commercial frequency (X-band) (52). Computer simulation of the Sband spectra suggests that cupric ion binds to four donor nitrogen atoms, three with AN = 10 G and one with AN = 15 G. ENDOR data confirm the assignment of at least two inequivalent nitrogens with couplings of 11 and 15 G (52). In addition, six proton couplings are well resolved in the ENDOR spectra, and the matrix ENDOR indicates that H20 is accessible to the metal site.
Ambient temperature ESR studies of CuBlm have also been carried out at several microwave frequencies to obtain rigid limit parameters for the complex in the liquid phase (53). We have argued that the cupric bleomycin structure opens up at room temperature and that the cupric ion is displaced from the square plane (53). Not only is the work important from the standpoint of the reactivity of an important antitumor complex, but these studies are fundamentally important as model studies. Thus, the comparison of complexes in frozen and mobile states can reveal dynamic changes which occur upon changing from the immobile, frozen state to the mobile, fast tumbling, liquid state. For example, analysis of the rotational correlation time suggests that copper bleomycin is cigar shaped or has segmented flexibility and rotates about a hinge (53).
It occurred to us that Zn2+ and especially Cd2+ might also form five-coordinate square pyramidal structures with Blm. If Cd2+ binding to Blm was analogous to Cu2+, the structure ofthe ligand, Blm, could be determined from C-13 NMR in the presence of diama gnetic Cd2". C-13 NMR spectra in the presence of Cu are so broadened and shifted that useful structural data has not been forthcoming. Unpublished cadmium-113 NMR results at temperatures lower than ambient temperature suggest that the secondary amine, for which lines are broadened and shifted, as well as the primary amine, the pyrimidine nitrogen, and the imidazole nitrogen are bound to Cd. The latter three donor atoms are directly implicated from the examination of 113Cd-Blm in which two-bond and three-bond 113Cd-13C spin-spin couplings are observed at carbons associated with these potential ligand sites (54). As the temperature approaches ambient temperature only the primary amine, the pyrimidine nitrogen and the imidazole nitrogen from Blm are witz, Peisach and co-workers that Fe2+ greatly stimulates the oxygen-dependent DNA strand scission activity of Blm (40,41). Since the degradation of DNA by this drug had been thought to cause cytotoxicity, it was suggested that Fe(II)Blm may be the biologically active form of the glycopeptide.
A number of elegant studies by Peisach's group (32,(56)(57)(58), together with results of Sugiura (59)(60)(61)(62)(63)(64)(65)(66) have led to the current view of oxygen activation by Fe(II)Blm shown in Eqs. bound to Cd-113. The Cd-113 isotope was also used to monitor the Cd-113 NMR spectrum of the metal itself. These data indicate that a solvent anion is bound, for example chloride ion, at ambient temperature but not at lower temperatures. Both the NMR data for CdBlm and the ESR data for CuBlm at ambient temperat/ure substantiate a change in structure as a function of temperature. We intend to use these techniques to probe how metal complexes of Blm react in media such as tumor cells, which may involve mechanisms not easily envisioned from data in the static state.
Finally, no evidence exists from low temperature ESR studies to support the formation of CuBlm-Lewis base ternary complexes such as CuBlm-pyridine adducts. However, at room temperature our first indication of ternary complex formation comes from the ESR data for CuBlm in the presence of a large excess of pyridine (11). Pyridine slows down the motion of CuBlm presumably because a ternary complex is formed; the shape of the complex also becomes spherical. Preliminary studies do suggest that CuBlm forms adducts in the presence of cells similar to the pyridine adduct (11).

Iron Complexes as Antitumor and Cytotoxic Drugs ESR Studies of Iron Bleomycin and Co(II) Blm and Ni(III) Blm
Interest in the properties of iron bleomycin (FeBlm) was enormously increased in 1978 by reports of Hor- The electronic nature of this intermediate is of interest. Its g values are contracted toward the free-electron value relative to Fe(III)Blm, presumably because of the quenching of orbital angular momentum. The g values of Fe(III)Blm are similar to those of iron prophyrins, suggesting the presence of four in-plane nigrogen donor atoms. In the iron-porphyrin system, g-value contraction only occurs when axial sulfur donor atoms replace oxygens or nitrogens (65). A similar effect is seen in Fe(III)Blm adducts (58,66). The only sulfur atom in Blm is part of the bithiazole moiety and is not thought to bind to iron in the intermediate. However, in a recent NMR study it was shown that the bithiazole group of excess Blm interacts with iron in Fe(II)Blm and NO-FeBlm, so it is plausible that bithiazole sulfur may act Spin-trapping studies have demonstrated that the redox process summarized in reactions (1)-(3) generates a large flux of hydroxyl radicals (41,63,64,68). Although OH was thought to initiate strand scission, the addition of scavengers for OHor for O2, and H202 to the reaction mixture does not affect the yield of degraded DNA (69). Thus free reduced-oxygen species are not involved in attack on the backbone of DNA. In fact, since kinetics of conversion of the product of reaction (9) to Fe(III)Blm and the rate and extent of DNA strand cleavage are identical, it is suggested that a bound form of oxygen, Fe(III)Blm-OHor Fe(III)Blm-OOH-, insensitive to radical scavengers, is the reactive species (31). This conclusion was beautifully supported by the demonstration that in the presence of H202, Fe(III)Blm was converted to the same intermediate as formed when Fe(III)Blm reacts with 02 e 2 Fe(III)Blm + H202 2 Fe(III)Blm-OHor Fe(III)Blm + Fe(III)Blm-OOH + H+ and that this intennediate degrades DNA (31).
The reductant to cleavage site stoichiometry for the DNA strand scission reaction is thought to be 4-5 to 1. Thus, at least two electron equivalents are required beyond those generated in reactions (8) and (9). Indeed, it is unlikely that the yrncipal source of reducing equivalents in cells is Fe + because of its very small free concentration. An alternative is that the other cellular reductants such as thiols may supply electrons for the overall reaction. In one study the authors demonstrated that cysteine but not glutathione stimulates the reaction of Fe(III)Blm with DNA (67). However, GSH enhances the amount of strand scission caused by Fe2" and Blm, which suggests that thiols can supply electrons HO NH2CO H2N CH3 (11) g,=2.060 to the process beyond the formation of the intermediate in reaction (9). Interestingly, although thiols are often thought to protect against radicals, in the presence of redox-active metal sites and oxygen, they may enhance radical production. The thiol dependent reduction of Fe(III)Blm has been examined by ESR spectroscopy (66). Under aerobic conditions, a redox cycle is established as shown in Eqs. (12)- (14).
Another essential feature of the mechanism of reaction of FeBlm with DNA is that the bithiazole and terminal cationic portion of Blm can intercalate and bind to DNA, thus bringing the reactive iron site into close proximity with the DNA backbone (Fig. 12) (70). Although the metal coordination and DNA binding sites are frequently portrayed as relatively independent of one another, an ESR study of Fe(II)Blm NR and its DNA adduct clearly shows that the electronic spectrum of the Fe-NO species is perturbed upon binding of the complex to DNA (Fig. 13) (71,72). Analysis of the perturbation suggests that it is occurring in the x-y plane ofthe metal coordinate site. Interestingly, NMR studies of the interaction of Fe(II)Blm-CO with poly(dA-dT) also indicate changes in the iron-binding site, particularly in proton resonances from imidazole, which is thought to be an in-plane ligand of Fe (67). Sugiura has examined the ESR properties of the Fe(III) complexes of a number of modified bleomycins (58). He concluded that the structure of the iron binding site was five-coordinate, probably similar to the structure of a biosynthetic intermediate of a Cu(II)Blm (Fig.  14). In this structure, the primary amine of Blm acts as an axial base and four nitrogen donors supply the ligands of the xy plane. This would leave the sixth coordinate site open to bind oxygen in the Fe(II) complex. When the adjacent terminal amine is hydrolyzed to yield a carboxyl group, the amine and carboxyl compete for the fifth coordination site in a pH-dependent way (58). At pH 7, the carboxyl is bound and no DNA strand scission occurs. Furthermore, the complex is now high spin with a g = 4.28 signal. Thus, like hemoglobin, its spin state changes with the nature of the axial ligand and presumably its oxygen activation properties do, as well. Studies of the pH-dependence of the ESR spectrum of Fe(III)Blm show that it undergoes a transition from low to high spin with an apparent pKa of 6.5 prior to completely dissociating at lower pH (Fig. 15) (71). Possibly, this transition involves the protonation and dissociation of the axial amine group as suggested by Sugiura's studies described above. The spin-state change can also be accomplished by the titration of Fe(II)Blm with phosphate (Fig. 16). A first hypothesis to explain the effect is that phosphate competes with the axial amine much as a carboxylate oxygen does in the depyruvamide Blm complex described above. However, the phosphate adduct is competent to carry out DNA strand cleavage. Fe(II) depyruvamide Blm is not (58). an unpublished spectrum of Co(II)Blm in the absence of DNA which is better resolved than the previously published spectra (Fig. 17). This spectrum has an apparent 1-2-3-2-1 pattern consistent with two axial nitrogen donor atoms. The ESR spectrum for Ni(III)Blm also has a five line 1-2-3-2-1 pattern on the g 1 1 feature (74,75). This again indicates the presence of two axial nitrogen donor atoms in the absence of DNA. Interestingly, the spectrum of the Ni(III) P-3A fragment has a single nitrogen donor atom consistent with the 1-1-1 three-line hyperfine pattern in the g 1 1 region consistent with a single axial nitrogen donor atom (74). Upon addition of oxygen, oxygen is bound in an axial position to Co(II)Blm and alters the ESR parameters (g = 2.098, g, = 2.007, A 1 1 Co = 20.2 G, AICo = 12.4 G) (75).
As with FeBlm NO, the addition of Co(02)Bhn to DNA changes the ESR parameters (g = 2.106, g, = 2.004, A 1 1 -= 18.9 G, A±cI = 11.5 G) (75), which argues for an interaction of the metal-binding site with DNA. A pH-dependent change in ESR parameters was attributed to a change in axial ligation from a nitrogen to an oxygen donor atom. The Co(II) complex and its dioxygen adduct of deamino-Blm which lacks the primary amine (purportedly the axial moiety) provide a model in which the axial donor atom reverts from a nitrogen donor atom to an oxygen atom (73,76). These studies with CoBlm and NiBIm probe two facets of Blm chemistry which are not as clearly described for FeBIm. Namely, Blm itself contains a multitude of potential donor atoms. A single nitrogen axial donor atom can orient the molecule into a square pyramidal configuration or two nitrogen axial donor atoms can change the Blm metal site into a rhombic octahedral configuration. Upon binding to DNA presumably through intercalation of the bithiazole group, the metallobleomycin structure reorients from rhombic to square pyramidal.

ESR Studies of Iron Adriamycin, Fe(Adr)3
The early suggestion that an iron-adriamycin complex, quelamycin, ameliorates the cardiac toxicity of adriamycin has not been substantiated in other investigations (77,78). One of the problems with this work was the use of a mixture of iron and Adr with a stoichiometry of Fe3Adr, which does not represent the actual coordination stoichiometry Fe(Adr)2 or Fe(Adr)3 (77). Recently, an iron adriamycin complex that has been difficult to characterize has been shown to bind to erythrocyte ghost membranes and is involved in lipid peroxidation (79). Other work indicates that an Fe(II)adriamycin complex bound to DNA damages DNA, presumably via a hydroxyl radical. A cycle similar to the cycle for CuL discussed previously has been proposed for the redox chemistry of iron adriamycin in cells. A schematic for this cycle taken primarily from the excellent work of Myers and co-workers is as follows (Fig.  18).
The reaction of iron adriamycin with membrane has not been observed with either CuL or FeBlm. On the other hand, the damage to DNA from hydroxyl radical appears to be quite similar to the damage caused by FeBlm.
The only ESR studies to our knowledge of FIGURE 17. X-band (A-C) and S-band (D,E) spectra for 8.6 mM CoBlm. The five hyperfine lines on a cobalt line in the g11 region indicate that two nitrogen donor atoms are in apical positions. The sample was prepared, the pH adjusted to 7.5 (or above) and the sample frozen, under a nitrogen atmosphere in a glove box. Fe3+(Adm)3 show that the ESR signal consists of two lines at g = 2.01 and g = 4.2 (80). Although the origin of the lines has not been discussed, it is presumed that the lines arise from occupancy of different Kramer's doublets of a high spin iron complex (81 maximum at about 130 min after complex formation, and disappears (80). During this reduction of iron-Adriamycin, a signal at g = 2.01 may appear which, except for the g value, appears to be a free-radical signal (82). Since the Fe(Adm)3 spectrum returns on exposure to oxygen, it is concluded that Fe(Adm)3 is slowly reduced to a ferrous Adriamycin complex and then reoxidized. Even though some of the details concerning the characterization of the iron and radical signals are not yet published, these signals are consistent with a hypothesis whereby reactive, reduced oxygen and Adriamycin radicals are generated in the presence of iron, which may have significant biological activity as described in the next section (83)(84)(85).

Radical Reaction in Systems Containing Iron Adriamycin
The anthracycline antibiotics are being used clinically against a number of neoplasms. However, the therapeutic effects of anthracycline derivatives are severely limited by their cardiotoxic effects. Active oxygen species produced during activation ofthese drugs have been suggested to be responsible for the observed cardiotoxicity. Recently, it was shown that the iron-adriamycin complex causes more extensive physiological damage (i.e., lipid peroxidation, DNA damage, etc.) than activated adriamycin alone (79,83,86). Again, formation of active oxygen species has been proposed to account for such damaging effects. In a biological milieu, active oxygen species from activated adriamycin can arise from several enzymatic and nonenzymatic reactions. These include: (i) reductive activation of adriamycin by NADPHcytochrome P450 reductase, xanthine/xanthine oxidase, NADH-quinone reductase, etc., resulting in a putative semiquinone radical which, in turn, undergoes redox cycling in the presence of oxygen to give oxy radicals; (ii) direct iron-independent reaction between the adriamycin semiquinone and hydroperoxides; (iii) thiol-dependent reduction of Fe(III)-Adr and Fe(III)-DNA-Adr complexes; (iv) lipid peroxidation induced by Fe(III)-ADP-Adr complex; (v) electron-transfer reaction between Fe(III)-Adr and Adriamycin itself. Since this chapter addresses reactions of metal-antitumor complexes, only those reactions (iii-v) involving iron and Adriamycin are considered in this section.
During thiol-dependent reduction of Fe(III)-Adr and Fe(III)-DNA-Adr complexes, oxygen consumption occurs, which was found to be sensitive to addition of catalase and superoxide dismutase. This suggests the involvement of superoxide and hydrogen peroxide. The following reactions can account for the observed oxygen consumption: Fe(III-Adr + RSH = Fe(II)-Adr + RS + H+ (15) Fe(II-Adr + 02 =Fe(III)-Adr + Q( 0-+ O2 + 2H+ --H202 + 02 RS + 02-* RSOO. (17) (18) Whereas direct ESR and ESR-spin trapping can be used to verify reactions (15 and (16), one can use oxygenuptake measurements (in the presence of spin trap) to obtain evidence for reaction (18). In fact, evidence for reaction (16) has recently been shown directly by ESR (80). Fe(III)-Adr and Fe(III)-DNA-Adr complexes also induced extensive damage to erythrocytes and DNA in the presence of thiols (87,88). Involvement of hydroxyl radicals was inferred based on absence of damage in the presence of hydroxyl radical scavengers. A Fenton reaction was proposed to generate the hydroxyl radicals.
Fe(II)-Adr + H202 = Fe(III)-Adr + -OH + -OH (19) Evidence for hydroxyl radical production during reduction of Fe(III)-Adr and Fe(III)-DNA-Adr complexes by hydrogen peroxide was recently obtained from spintrapping experiments (89). The steady-state concentrations of DMPO-OH were found to be much higher during the reduction of Fe(III)-DNA-Adr by hydrogen peroxide than of Fe(III)-Adr itself. The following reactions possibly are involved.
Fe(II)-DNA-Adr + H202 -* Fe(III)-DNA-Adr + OH + -OH (22) One can again monitor the initial rate of these reactions by using direct ESR and spin-trapping. The extent of DNA damage also was found to correlate with spintrapping results. Both the reversal of DNA damage by catalase and the detection of free hydroxyl radical would seem to rule out the involvement of perferryl ion. However, with Fe(III)-ADP-Adr complex, involvement of perferryl ion has been proposed. It was also found that the Fe(III)-ADP-Adr complex could initiate the unsaturated fatty acid decomposition which was enhanced in the presence of purified cytochrome P450 reductase and NADPH (87). The fatty acid decomposition was inhibited by both catalase and tocopherol. While it is conceivable that the existence of small amounts of preformed lipid hydroperoxides can explain the on-set of lipid peroxidation in the presence of Fe(III)-ADP-Adr complex [reaction (24)], the mechanism(s) leading to the enchanced lipid peroxidation in the enzymatic systems is not clear. A direct electron transfer reaction between Fe(III)-Adr and excess Adr also was noted (80). The ESR spectrum of Fe(III)-Adr under anerobic conditions was found to decay slowly with time presumably forming the oxidized adriamycin radical (Adi) and Fe(II)-Adr. Upon exposure to air, the ESR signal due to the Fe(III)-Adr complex reappeared.
Fe(III)-Adr + Adr = Fe(II)-Adr + Adr(oxidized radical) (25) Reoxidation of Fe(II)Adr could occur as previously shown [reaction (16)]. Although the identity of the oxidized adriamycin radical is not known, the existence of such species has been shown previously under peroxidatic and autoxidizing conditions.

Bis(ot-N-heterocyclic Carboxaldehyde Thiosemicarbazonato) Fe Complexes
An earlier section of this review described ESR studies of 2-formylpyridine thiosemicarbazonato copper which help to elucidate its behavior in Ehrlich tumor cells. Iron complexes of a-N-heterocyclic carboxaldehyde thiosemicarbazones are also cytotoxic and have antitumor properties (90). In fact, in the study of this type ofligand structure (NNS), it was thought originally that the ligand formed a ternary complex with coordinatively unsaturated iron, which was part of the active site of ribonucleotide diphosphate reductase (90 However, later it was suggested that the intact 2:1 ligand to iron complex reacts with the enzyme to inhibit the reduction of ribonucleotide to deoxyribonucleotides (6).
The ESR spectrum of bis(2-formylpyridine thiosemicarbazonato) Fe(III) at 77°K is a typical rhombic spectrum of a low spin complex with g values at 2.176, 2.135, and 1.998. In the formation of such complexes, a spinstate change usually occurs in the transition from the 1:1 to 2:1 ligand to metal structure. K1 Fe3+ + HL = FeL2+ + H+ (27) K2 FeL2+ + HL = FeL+ + H+ (28) K2 > K1 so that FeL2+ is formed in preference to FeL2+. Given this general property and the large sta-bility constants of Fe (III)L2' and Fe(II)L2 (P2 = 1026 and 1023, respectively), both complexes are likely to exist as fully coordinated 2:1 complexes. This is in contrast to high spin Fe(III)-(Adriamycin)3, which, as noted above, readily lose a ligand to become coordinately unsaturated and can fonn ternary adduct species.

General Comments
A common property of a number of the metal complexes described in this review is their facile redox chemistry, in which they are readily reduced by thiol compounds and oxidized by oxygen or reduced species of oxygen to produce reactive entities, including oxygen radicals and secondarily, organic free radicals. The interest in free radicals in the causation of tissue damage and in the mechanism of drug action runs deep as attested to by the subject matter of this volume. It is curious, therefore, that relatively little attention has been given to the exploration of redox-active metal complexes as drugs.
There are two features of metal complexes which make them attractive for study as metallodrugs. One can control and fine tune their redox/potential and their reactivity with oxygen. One can also study their paramagnetic forms in solution and in cells using ESR spectroscopy. Table 1 lists the redox properties of some Cu and Fe complexes of bis-and monothiosemicarbazones. It was shown that CuKTS and CuKTSM are reduced rapidly by cellular thiols (12); CuKTSM2 is not and reacts in a fundamentally different way with cells (13). CuL+ reacts rapidly with sulfhydryl groups (7). However, in contrast to CuKTS, Cu(I)L+ is readily reoxidized to Cu(II)L in competition with the ligand substitution reaction by which thiols (RSH) react with Cu(I)L+ to form Cu(I)SR. Presumably, this is because the tridentate CuL complex has an open in-plane coordination site for oxygen which the tetradentate CuKTS structure does not possess. Furthermore, the E1/2 values for the whole series of CuL-X complexes are much more positive than for the bis thiosemicarbazone series, so that other cellular reductants may be able to reduce these structures but not CuKTS.
The monothiosemicarbazonato iron complexes also illustrate the ease with which one can manipulate redox potential (Table 1). In addition, bis(2-formylpyridine thiosemicarbazonato) Fe(II) has some kinetic stability in air though the Fe(III) complex is thermodynamically stable. Thus one expects the redox cycle of reactions (12)- (14) to be slower for FeL2' than CuL+. These examples point out the facility with which one can alter features of reactivity of potential metallodrugs.
Paramagnetic metallodrugs have also proven to be attractive structures for studies of mechanism of action.
Usually, they have characteristic, unique ESR spectra which permit their identification in cells or model systems. For example, considering copper complexes, their geometry is usually Type II square planar or square pyramidal or blue, Type I, distorted tetrahedral.
ESR parameters g 11, A 11 , giso and Ai.0 are easily ob-tained and Peisach-Blumberg plots of g11 and A 1 1 are indicative of the donor atoms bound to cupric ion (10).
Well resolved lines in the g, region are usually the result of hyperfine structure from nitrogen donor atoms. However, interpretations from the number of well resolved lines can be ambiguous because the magnitude of A,Cu and AN are typically of the same order. More sophisticated techniques such as electron nuclear double resonance, low frequency S-band, and electron spin echo spectrosopy help determine nitrogen and proton hyperfine coupling constants, the number of equivalent nitrogen donor atoms, and the binding of imidazole, respectively. Molecular bonding parameters may also be calculated and are sensitive to the degree of covalency.
Upon addition of cupric complexes to cells new questions become apparent. Cells are well compartmentalized, complicated entities and knowledge of the environment of the copper complex becomes as important as obtaining ESR parameters. Studies of copper complexes in the liquid phase instead of the static (frozen) phase are necessary in order to optimize the use of cupric ion as a cellular probe. Until recently the problem with ESR data from cupric ion at room temperature was threefold. First, the sensitivity.is less; second, only two ESR parameters, giso and A'80, are readily extracted from fast tumbling complexes; and third, if the complex tumbles slowly, the resolution of the hyperfine structure is poorer at room temperature than in frozen solution. New instrumentation at the National Biomedical ESR Center can be used to circumvent these problems. First, the Froncisz-Hyde loop-gap resonators provide better signal to noise (91); second, a multimicrowave frequency approach allows the determination of rigid limit ESR parameters for copper from data taken at five widely varying frequencies (92); and third, although the nitrogen structure is not well resolved, the rotational correlation time, TR and the room temperature paramneters are sensitive to the dynamics of cupric ion in, for example, cells (93). Therefore, it is anticipated that uptake ofredox stable cupric complexes by cells can be used as a probe of cells much like nitroxides are used as spin labels and spin probes and that for those which react relatively slowly (t112 on the order of minutes) detailed information on the nature of the reaction may be obtained under spectral conditions identical to those under which the complex is reacting with cells.