Generation of superoxide via the interaction of nitrofurantoin with oxyhemoglobin.

Nitrofurantoin was found to interact with HbO2 to cause the concomitant formation of methemoglobin and superoxide. The rate of formation of methemoglobin and superoxide was linearly dependent upon the concentration of nitrofurantoin and could be inhibited by superoxide dismutase, catalase, or the prior conversion of HbO2 to ethylioscyanoferrohemoglobin. The ability of nitrofurantoin to interact with HbO2 and cause superoxide formation may represent one mechanism by which it produces red cell toxicity and suggests that generation of superoxide in erythrocytes may occur via a different mechanism than that which occurs in microsomes.

phate dehydrogenase are susceptible to red cell hemolysis when administered NF (6). Since this enzyme is required to maintain an adequate level of NADPH in the red cell and because the primary function of NADPH in erythrocytes is to serve as a source of reducing equivalents for the reduction of GSSG to GSH, it would thus seem plausible that the mechanism for NF-mediated red cell toxicity would require (a) an increase in NADPH utilization and (6) simultaneous production of an intermediate dependent on GSH for its detoxification. For example, conversion of NF to a free radical in the presence of molecular oxygen would result in an increase in superoxide and hydrogen peroxide levels in the red cell. Since GSH can act as a scavenger of free radicals as well as serve as a source of reducing equivalents for the metabolism of hydrogen peroxide to water via glutathione peroxidase, GSH would be converted to GSSG at an increased rate, and the rate of NADPH oxidation would increase to convert the CSSG to GSH via glutathione reductase. These effects of NF on the generation of activated oxygen species, coupled with the fact that NF inhibits glutathione reductase (7), might well provide the basis for NF-mediated red cell toxicity.
Recent experiments in our laboratory on the mechanism of NF-induced red cell toxicity have revealed that NF causes a marked depletion of red cell GSH levels (8) with a concomitant increase in hydrogen peroxide and MetHb formation (3). Depletion of red cell GSH, however, could be prevented by ethyl isocyanide (9), a high affinity ligand capable of binding to ferrohemoglobin and displacing oxygen (10). This observation suggests that Hb may participate in NF-induced GSH depletion, perhaps via the reduction of NF to a free radical or subsequent reduction products, or by the NF-mediated release of superoxide from HbOn.
Since the formation of nitroaromatic anion free radicals in microsomes appears to be achieved solely by the participation of a flavoprotein, the mechanism of NF-induced red cell toxicity may differ from those tissues which contain microsomal enzymes. We have thus undertaken a study of the interaction between NF and purified human HbOp to investigate directly NF-induced formation of superoxide from HbOz. from Nitrofurantoin-HbOz Interaction horse heart cytochrome c were obtained from Sigma. Ethyl isocyanide was synthesized according to the method of Reisberg and Olson (10). NF was the gift of Norwich-Eaton Pharmaceuticals. Bovine serum albumin, type I1 recrystallized, was purchased from Sigma Chemical Company. Methods Autoxidation of Hb0"The rate of autoxidation of HbOz was determined using a method previously described by Mieyal and Blumer (13). Solutions of 1.0 PM HbOZ and of 0-840 p~ NF, both 20 mM in potassium phosphate buffer, pH 7.5, were added to opposite chambers of Yankeloff mixing cuvettes, allowed to reach thermal equilibrium, and a flat baseline was recorded with the Cary 219 spectrophotometer operating in the autoslit mode at 37 "C. The sample cuvette was then inverted, and the difference spectrum was recorded repetitively from 400-600 nm. In experiments using superoxide dismutase or catalase, the enzyme was placed in the cuvette chamber containing the HbOz. A plot of the change in the absorbance of the Soret peak uersus time described the autoxidation of Hb07, which remained linear for a period of at least 1%-2 h. The slope of this line was taken as the initial rate. To convert this absorbance change into the number of moles of HbOp autoxidized, a value of €415 = 514 mM" for the HbO, tetramer was used (11).
Superoxide Formation-Acetylated cytochrome c is capable of being reduced by superoxide anion with the acetylation resulting in a greater than 95% decrease in its ability to be reduced by flavoprotein reductases (12). Thus, the inclusion of acetylated cytochrome c in a solution of HbO, containing NF permits the measurement of superoxide anion generated. A solution of 40 p~ HbO, (final concentration) which was 100 mM in potassium phosphate buffer, pH 7.5, and contained 100 p~ Na2EDTA, was placed in 1-ml cuvettes to which was added acetylated cytochrome c to achieve a final concentration of 17.5 p~. The sample and reference cuvettes were allowed to equilibrate at 37 "C for several minutes. A fixed volume of 5 p1 of a solution of dimethylformamide, which contained varying concentrations of NF (0-336 mM), was added to the sample cuvette. The reduction of acetylated cytochrome c was monitored at 550 nm, and c5s0 = 19.5 mM" was used as the value for the reduced minus oxidized extinction coefficients (14). When 5 pl of dimethylformamide was added alone to the sample cuvette it produced a finite, but small, rate of change in A5:d as compared to that measured in the presence of NF. This rate, however, was in all cases subtracted from the rate measured in the presence of NF. Furthermore, dimethylformamide was found to be 15-fold less potent than NF in facilitating autoxidation of HbOs. In experiments using superoxide dismutase or ethyl isocyanide the enzyme or ligand was added prior to the addition of NF in dimethylformamide and in such a manner so as to achieve the final concentration required.

RESULTS
A typical difference spectrum obtained when solutions of HbOa and NF are mixed is presented in Fig. 1. The Soret absorbance of HbOa in the absence of NF occurs at 415 nm with LY and / 3 bands at 577 and 542 nm, respectively. The addition of NF to HbOz produces a hypsochromic shift in the Soret absorbance and decreases in the absorbances of the a and p bands of HbOa, consistent with the difference spectrum obtained in the conversion of HbOz to MetHb. These alterations in the Soret and the (Y and , 8 band absorbance wavelengths result in the initial difference spectrum recorded in Fig. 1 which is indicative of MetHb formation. The magnitude of the absorbance changes in the difference spectrum is also time and NF concentration dependent, thus providing spectral evidence for the interaction of NF with HbOz which results in an accelerated rate of autoxidation of HbOs to MetHb. The concentration dependence of this NF-mediated oxidation process is shown graphically in Fig. 2. In the absence of NF, HbOz is autoxidized to MetHb at a finite rate, resulting in a nonzero intercept at the ordinate. The concentration dependence as shown in Fig. 2 suggests that this autoxidation process may be saturable, although the very limited solubility of NF2 The solubility of NF in the absence of protein is 840 p~. Thus, when the mixing cell technique is employed, the maximum concentration of NF that could be attained is 420 p~.
in this system precluded achieving a concentration of NF in excess of 420 FM. At the NF concentration of 420 ELM, however, a 15-fold increase in the rate of conversion of HbOz to MetHb was observed.
The inset to Fig. 2 is the double reciprocal plot of the rate of autoxidation uersus NF concentration corrected for the basal rate of HbOz autoxidation which occurs in the absence of NF. This treatment of data yields the interaction constant for NF with HbOz which is approximately 500 PM and may be thought of as that concentrat.ion of drug producing the halfmaximal stimulation of HbOa autoxidation.
A plausible mechanism of NF-mediated autoxidation of HbOe to MetHb is the release of superoxide anion. Thus, a study of the effects of superoxide dismutase and catalase on NF-stimulated Hb02 autoxidation was undertaken, and the results are presented in Table I. The presence of either superoxide dismutase or catalase, in concentrations that produced the maximum levels of inhibition attained, resulted in the partial inhibition of HbO, autoxidation. Furthermore, there appeared to be a slight increase in the degree of inhibition when superoxide dismutase and catalase were present in combination as compared to either being present singularly. T o take into consideration the possibility that the inhibit,ory effects of superoxide dismutase or catalase result from nonspecifk protein effects, similar experiments were performed using bovine serum albumin in a protein concentration (mg/ m l ) comparable to that present with superoxide dismutase. When such an experiment was performed, bovine serum albumin (200 pg/ml) was found to inhibit NF-induced autoxidation by approximately 15% (Table I). Such an inhibitory effect may occur either as a result of binding of NF to bovine serum albumin, thereby lowering the effective N F concentration, or by bovine serum albumin stabilizing HbOn against NF-induced autoxidation. In contrast, however, superoxide dismutase (Table I) at a n equivalent protein concentration of 200 pg/ml, was able to inhibit HbO, autoxidation by 48%, some 2.5-fold greater than that produced by bovine serum albumin, The inhibition of HbO, autoxidation by catalase (41%) or superoxide dismutase (48%) ( Table I) suggests that as Hb02 is converted to MetHb, oxygen is converted to superoxide. Superoxide thus formed may dismute to yield hydrogen peroxide, or considering the inhibitory effects of superoxide dismutase, cause further oxidation of HbOn to MetHb. Furthermore, the inhibitory effects of catalase alone, or in combination with superoxide dismutase, imply that hydrogen peroxide formed as a result of the dismutation or reduction of superoxide may also be able to convert HbOl to MetHb in this system.
To further examine the hypothesis that superoxide may be

Effect of inhibitors on NF-mediated superoxide generation
The fmal concentrations used were: NF, 1.68 m t d ; HbOz, 40 pM; ethyl isocyanide, 67 mM. Each experiment represents the mean k S.E. for at least two determinations. Bovine serum albumin at 1 mg/ ml , a protein concentration equivalent to that of 2600 units/ml of superoxide dismutase, had no effect on the rate of acetylated cytochrome c reduction.  . 42 versus 840) was not large enough to achieve saturation, and the rate of superoxide generation remained linear over the NF concentration range employed. Table I1 shows that the reduction of acetylated cytochrome c could be inhibited by superoxide dismutase or by complexing the heme iron of HbO, with ethyl isocyanide. The inhibition of acetylated cytochrome c reduction was dependent on the concentration of superoxide dismutase with a maximum of 83% inhibition of superoxide measured at 2600 units of superoxide dismutase/ml, while the complete conversion of Hb02 to ethylisocyanoferrohemoglobin decreased superoxide formation by 94%. A relatively large concentration of superoxide dismutase was required to inhibit acetylated cytochrome c reduction. For example, as shown in Table 11, 520 units of superoxide dismutase/ml resulted in only 18% inhibition, while 1300 and 2600 units/ml resulted in 53% and 83% inhibition, respectively. In order to preclude a generalized protein effect from being responsible for this inhibition, experiments were performed in which bovine serum albumin was present at an equivalent protein concentration as superoxide dismutase at 2600 units/ml ( i e . 1 mg/ml). Bovine serum albumin, at 1 mg/ml, had no measurable effect on superoxide generation as evidenced by acetylated cytochrome c reduction. The relatively slow rate of superoxide generation, however, may require a larger ratio of superoxide dismutase to Hb02 in order t o trap the superoxide formed before it can either participate in the further oxidation of Hb02 to MetHb or undergo dismutation to hydrogen peroxide (vide infra).

The Interaction of NF with
HbOp-The interaction of NF and HbOz results in the production of MetHb and superoxide anion, and several plausible mechanisms may exist to account for this observation. Nitrofurantoin may interact with HbOz to facilitate the release of bound oxygen from the ferrous heme iron atom as superoxide anion resulting in the formation of MetHb.
HbO, + NF + Hb3' + NF + 02" In this mechanism NF is not altered but in interacting with HbOz serves to decrease the affinity of Hb for Oz. Such an alteration in oxygen affnity may be produced either by a direct interaction of NF with or near the heme iron atom or by an allosteric mechanism whereby NF in interacting with Hb affects the quaternary structure of the apoprotein or by a combination of both effects. We have obtained preliminary evidence in support of an interaction of NF in proximity to the paramagnetic heme iron atom in ferrihemoglobin (15). In these studies, using TI relaxation rate measurements, NF was found to interact with MetHb at a site distinct from, but in proximity to, the heme iron atom, and the interaction, as evidenced by the paramagnetic effect, continued to occur in the presence of directly coordinating ligands such as fluoride and cyanide (z.e. fluoride enhanced the paramagnetic effect while cyanide diminished but did not completely remove the paramagnetic effect). Since HbOY is not paramagnetic, however, a similar study could not be performed with NF and HbOz. Further support for such an interaction may be derived by analogy from studies on aniline-catalyzed Hb02 autoxidation by Mieyal and Blumer and from TI relaxation rate studies on the interaction of aniline, 2,6-dimethylaniline, benzene, and toluene with ferrihemoglobin (16)(17)(18)(19). Mieyal and Blumer have proposed that aniline interacts with Hb02 at a site distinct from the heme iron atom and alters the affinity of Hb for oxygen thus stimulating autoxidation (13). We have found that, at equivalent concentrations, aniline and NF produce comparable rates of HbOz autoxidation. Novak and Mieyal have presented NMR evidence in support of an interaction of aniline with MetHb at a site near the heme iron atom (16). These NMR studies demonstrated that aniline continued to interact with hemoglobin in proximity to the Paramagnetic heme iron atom in the presence of ligands such as cyanide or fluoride (i.e. fluoride enhanced the paramagnetic effect while cyanide diminished but did not completely remove the paramagnetic effect). Furthermore, these results are also interesting in light of the fact that aniline binds cooperatively to MetHb (20). Thus either a direct interaction of NF with HbOZ near the heme iron atom or an NF-induced conformational change in the apoprotein or a combination of both effects may produce the NF-mediated increase in the rate of autoxidation of HbOz resulting in the generation of superoxide anion.
In an alternative mechanism, NF may facilitate the release of molecular oxygen from HbOz. NF may then be reduced to the nitro anion free radical via Hb thus yielding superoxide, from the subsequent transfer of one electron from the free radical species to molecular oxygen, and MetHb. NF + HbO:! -+ NF" + Hb3+ + Oz NF" + 0 2 + NF + 0 2 " Some evidence may also be presented in support of this mechanism (9). When erythrocytes were incubated with NF under anaerobic condltions, cellular GSH levels decreased sigrufcantly, as compared to controls, presumably as a result of GSH acting as a scavenger of the NF free radical. This NFproduced depletion of GSH could be partially inhibited (-50%) by ethyl isocyanide, thus implicating Hb in the reduction of NF. In general, however, the above mechanism is not thermodynamically favorable. The standard 1-electron reduction potentials of NF (-264 mV) (21) and Hb (+150 mV) (22) do not favor the direct reduction of NF by ferroheme. The interaction of NF with Hb, however, especially at the concen-tration ratios employed, may substantially alter the standard reduction potential of the Hb3+/Hbz' couple thereby making this reaction thermodynamically favorable. To date, no information is available on the 1-electron reduction potential of the Hb3'/HbZ' couple in the presence of NF.
An equally plausible explanation for the apparent protective effect of ethyl isocyanide in erythrocytes under anaerobic conditions may be obtained from the work of Harada and Omura (5). The reduction of NF to the nitro anion free radical may occur via a flavoprotein reductase, and further reduction of the free radical species may be catalyzed by Hb. The subsequent reduction product(s) may then be responsible for GSH depletion via conjugation. Thus, ethyl isocyanide may effectively inhibit further reduction of the NF nitroaromatic anion free radical by Hb and thereby prevent GSH depletion.
Of the two mechanisms suggested for NF-stimulated autoxidation of HbOZ and generation of superoxide, that which proposes the formation of superoxide via interaction of NF with HbOz appears to be the more thermodynamically favorable. The interaction of NF with HbOz generates a steady state of superoxide anion as a function of NF concentration. Hence the rate of autoxidation of HbOz is linear with time (1%-2 h), and the initial rate is a function of NF concentration ( Fig. 2). A certain amount of superoxide thus generated may serve to further oxidize HbO2 to MetHb or dismute to form hydrogen peroxide which is also capable of oxidizing HbOz to MetHb (vide infra). The effect of either superoxide dismutase or catalase, therefore, is to inhibit HbOz autoxidation via conversion of superoxide to hydrogen peroxide or hydrogen peroxide to water, respectively. When superoxide dismutase and catalase are present simultaneously, a slightly greater inhibitory effect on HbOz autoxidation was observed (Table  I). Although the systems are not directly comparable since substantially larger concentrations of NF and HbOz were employed, the same effect is also observed for NF-mediated superoxide generation where addition of superoxide dismutase was able to markedy inhibit the reduction of acetylated cytochrome c (Table 111). Thus, in the absence of a 1-electron reduction potential for the Hb3+/Hb2+ couple in the presence of NF, the NF-stimulated release of superoxide from HbO2 appears to be a viable mechanism.
Relationship to Other Results Involving Hb-Ligand Interactions-Hirota and Itano have studied the affiity of a series of substituted nitrosobenzenes for ferrohemoglobin (23). The nitroso species is an intermediate in the reduction pathway converting nitro-containing compounds to amines (eg. nitrobenzene to aniline). The series of nitrosobenzenes studied was found to have relatively large affinities for ferrohemoglobin and in all cases bound cooperatively (Hill coefficients of 1.8-3.8). These investigators proposed a model of nitrosoareneferrohemoglobm interaction in which the iron of the ferroheme is bonded to the nitrogen atom of the nitroso group. Formation of the I-electron reduced nitrosobenzene radical as the product in the spontaneous conversion of ferrohemoglobm to MetHb by nitrobenzene was suggested (23).
Menadione may also generate superoxide anion via an analogous interaction with Hb. Goldberg and Stern have proposed that Hb catalyzes the 1-electron reduction of menadione to its free radical semiquinone (24). This semiquinone then reduces oxygen to superoxide.
Many other classes of compounds have been shown to oxidize HbOz to MetHb via mechanisms that do not appear to be applicable to NF. Wallace and Caughey (25) and Kawanishi and Caughey (26) have presented evidence for a mechanism of MetHb formation produced by agents capable of acting as 1-electron reductants. In this model, agents such as phenols and ferrocyanides donate a single electron to dioxygen bound as Hb02. A second electron is donated from the ferrous iron of HbOf thus yielding MetHb and hydrogen peroxide as the products. Castro et al. have proposed a similar mechanism to explain the oxidation of HbOz by various organic reductants (27). In their model, HbOz reacts with the organic reductant via the abstraction of a proton yielding a free radical as well as peroxide and MetHb. It appears unlikely that NF acts via an analogous mechanism. First, such a mechanism does not explain the generation of superoxide as has been observed with NF. Second, a greater inhibitory effect of catalase on the rate of Hb02 autoxidation would be expected than we have observed. Third, if NF were to act as a reducing agent via the donation of a proton or an electron, it is difficult to imagine a stable product resulting from such an oxidation of NF.
Participation of Superoxide in NF-mediated HbO2 Autoxidation-Superoxide anion is generally thought of as a reducing agent, donating a single electron and being oxidized to molecular oxygen. Our data showing that superoxide dismutase diminishes the rate of NF-mediated HbOz autoxidation implies that superoxide anion generated during the autoxidation reaction may actually oxidize HbO2 to MetHb. Other investigators have also reported such an effect of superoxide. Winterbourn et al. have shown that generated superoxide can both convert HbOz to MetHb and reduce MetHb to HbOz (28). The oxidation reaction is favored when the absolute concentration of MetHb is low, while if the MetHb concentration is large relative to HbOn, MetHb reduction by superoxide wiU OCCLW. Lynch et al. also found that superoxide could convert Hb02 to MetHb (29).
The rate of HbOz autoxidation cannot be directly compared to the rate of superoxide generated since the two systems differ significantly in HbO2 concentration. The experimental limitation imposed by present instrumentation4 precludes a thorough study of the concentration dependence of HbO2, at a fmed concentration of NF, on the rate of autoxidation. Similarly, an analogous series of experiments using 40 ~L M Hb02 and NF in order to compare rates of autoxidation and superoxide generation could not be performed. In this regard no definitive conclusion may be drawn from these experiments Present instrumentation theoretically requires an absolute absorbance of less than 4.0 absorbance units at any one wavelength in order to allow the recording of a base-line for observation of difference spectra. A maximum Soret absorbance of less than 4.0 units corresponds to a total HbOa concentration of less than 8 PM. NF, which has a broad visible absorbance maximum at 377 nm, also has significant absorbance in the region of the HbO, Soret. Thus, only a limited HbO, concentration effect on the rate of NF-induced autoxidation could be studied, and it would be invalid to extrapolate these results to the system of 40 pM HbO2 used in the studies on superoxide generation. Furthermore, decreasing the HbOn concentration in studies on superoxide generation would severely limit the detection of superoxide because of the inherent insensitivity of the assay (e.g. a decrease in the Hb02 concentration from 40 to 10 PM would decrease the rate of acetylated cytochrome c reduction (Fig. 3) by a factor of  4).
regarding the ability of superoxide anion to participate in further reduction or oxidation of Hb. These studies have, however, demonstrated that NF interacts with Hb02 and is capable of producing MetHb and superoxide. Such an interaction in the erythrocyte would cause an increase in NADPH oxidation and may represent one mechanism by which NF causes red cell toxicity.