Similarities between the Abiotic Reduction of Selenite with Glutathione and the Dissimilatory Reaction Mediated by Rhodospirillum rubrum and Escherichia coli*

Various mechanisms have been proposed to explain the biological dissimilatory reduction of selenite (\batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(\mathrm{SeO}_{3}^{2-}\) \end{document}) to elemental selenium (Se°), although none is without controversy. Glutathione, the most abundant thiol in the eukaryotic cells, the cyanobacteria, and the α, β, and γ groups of the proteobacteria, has long been suspected to be involved in selenium metabolism. Experiments with the phototrophic α proteobacterium Rhodospirillum rubrum showed that the rate of selenite reduction was decreased when bacteria synthesized lower than normal levels of glutathione, and in Rhodobacter sphaeroides and Escherichia coli the reaction was reported to induce glutathione reductase. In the latter organism superoxide dismutase was also induced in cells grown in the presence of selenite, indicating that superoxide anions (O-2) were produced. These observations led us to investigate the abiotic (chemical) reduction of selenite by glutathione and to compare the features of this reaction with those of the reaction mediated by R. rubrum and E. coli. Our findings imply that selenite was first reduced to selenodiglutathione, which reached its maximum concentration within the 1st min of the reaction. Formation of selenodiglutathione was paralleled by a rapid reduction of cytochrome c, a known oxidant for superoxide anions. Cytochrome c reduction was inhibited by superoxide dismutase, indicating that O-2 was the source of electrons for the reduction. These results demonstrated that superoxide was produced in the abiotic reduction of selenite with glutathione, thus lending support to the hypothesis that glutathione may be involved in the reaction mediated by R. rubrum and E. coli. The second phase of the reaction, which led to the formation of elemental selenium (Se°), developed more slowly. Se° precipitation reached a maximum within 2 h after the beginning of the reaction. Secondary reactions leading to the degradation of the superoxide significantly decreased the yield of Se° in the abiotic reaction compared with that of the bacterially mediated selenite reduction. Abiotically formed selenium particles showed the same characteristic orange-red color, spherical structure, and size as particles produced by R. rubrum, again providing support for the hypothesis that glutathione is involved in the reduction of selenite to elemental selenium in this organism.

) were produced. These observations led us to investigate the abiotic (chemical) reduction of selenite by glutathione and to compare the features of this reaction with those of the reaction mediated by R. rubrum and E. coli. Our findings imply that selenite was first reduced to selenodiglutathione, which reached its maximum concentration within the 1st min of the reaction. Formation of selenodiglutathione was paralleled by a rapid reduction of cytochrome c, a known oxidant for superoxide anions. Cytochrome c reduction was inhibited by superoxide dismutase, indicating that O 2 ؊ was the source of electrons for the reduction. These results demonstrated that superoxide was produced in the abiotic reduction of selenite with glutathione, thus lending support to the hypothesis that glutathione may be involved in the reaction mediated by R. rubrum and E. coli. The second phase of the reaction, which led to the formation of elemental selenium (Se°), developed more slowly. Se°p recipitation reached a maximum within 2 h after the beginning of the reaction. Secondary reactions leading to the degradation of the superoxide significantly decreased the yield of Se°in the abiotic reaction compared with that of the bacterially mediated selenite reduction. Abiotically formed selenium particles showed the same characteristic orange-red color, spherical structure, and size as particles produced by R. rubrum, again providing support for the hypothesis that glutathione is involved in the reduction of selenite to elemental selenium in this organism.
Selenium is an essential trace element in the nutrition of many organisms, but it can be highly toxic depending on its concentration and speciation. High selenium concentrations may cause severe abnormalities in the development of various animals and plants (1)(2)(3). Deformation and structural modifications have been noted, especially in creatine-formed tissues (i.e. hooves, horns, hair, feather, beaks, and nails) in which appreciable quantities of selenium may accumulate. In these cases, selenium toxicity has been attributed to its ability to replace sulfur in proteins or other sulfur-containing biomolecules. In their investigations of selenite toxicity in prokaryotes, Kramer and Ames (4) did not observe any nonspecific incorporation of selenium into proteins. They demonstrated that a mutant strain of Salmonella typhimurium, which is able to overexpress oxidative stress proteins such as catalase and superoxide dismutase (SOD), 1 is significantly more resistant to selenite toxicity than the wild type. Their results suggest that free radical formation might be involved. They also considered the high reactivity of selenite with sulfhydryl groups and the formation of oxygen radicals when selenium reacted with cysteine or glutathione and concluded that selenite toxicity in bacteria might be the result of oxidative damage. Consistent with these results Bébien et al. (5) observed that two types of SOD are induced in cultures of Escherichia coli exposed to selenite, thus confirming the involvement of free radicals in selenium toxicity. In addition, glutathione reductase was induced in cultures of Rhodobacter sphaeroides (6) and E. coli (5) amended with selenite. In the bacterial domain glutathione is present in the cyanobacteria and the ␣, ␤, and ␥ groups of the proteobacteria (7). Considering these data we investigated the chemical reduction of selenite with glutathione and compared the features of this reaction with those of the dissimilatory selenite reduction in R. rubrum and E. coli.
In a chemical approach, Painter (8) observed the high reactivity of selenite with thiol groups. He was the first to demonstrate the formation of selenotrisulfides (RS-Se-SR), according to Reaction 1.
Ganther (9) studied the reaction of selenite with glutathione (GSH), the most abundant thiol found in the eukaryotic cells, the cyanobacteria, and the ␣, ␤, and ␥ groups of the proteobacteria (7). He showed that the selenotrisulfide of glutathione * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
(GS-Se-SG), which was later renamed selenodiglutathione, is a very good substrate for glutathione reductase with K m and V max values comparable with those of glutathione itself. He described this with Reaction 2.
Glutathione reductase Ganther (9) also proposed that the unstable selenopersulfide of glutathione (GS-Se Ϫ ) dismutates into elemental selenium (Se°) and reduced glutathione according to the following stoichiometry. REACTION 3 In experiments about the kinetics of selenite reduction in cultures of Rhodospirillum rubrum we observed that the rate of the reaction is decreased significantly when the organism synthesizes low levels of glutathione. 2 In the present study, we investigate the kinetics of formation of selenodiglutathione, superoxide anions, and elemental selenium during abiotic (chemical) reduction of selenite by glutathione, and we compare the features of this reaction with those of the reaction mediated by R. rubrum and E. coli. We also compare the properties of the abiotically formed Se°particles with those produced during the bacterial process.

MATERIALS AND METHODS
Selenite Reduction-Chemical reactions were performed in 50 mM Tris⅐HCl buffer (pH 7.0) at room temperature in tubes kept anoxic (Hungate, Bellco, Vineland, NJ) under a nitrogen atmosphere. The buffer was degassed with an aspirator pump for about 1 h before use. Anoxic stock solutions of selenite, glutathione, and cytochrome c were also prepared in Hungate tubes under nitrogen and with degassed buffer. The addition of reactants and sampling were performed using syringes purged with nitrogen.
Selenite-SeO 3 2Ϫ was determined using a colorimetric method based on the formation of a piazselenol complex with diaminonaphthalene (11,12).
Selenodiglutathione-GS-Se-SG was detected by monitoring its absorbance at 260 nm (9). Its appearance and disappearance were followed in a spectrophotometer (Uvikon 860, Kontron, Zü rich, Switzerland). All measurements were performed in quartz cuvettes purged with nitrogen.
Superoxide Anions-Because superoxide anions are known to reduce cytochrome c (13), relative superoxide levels can be determined by measuring the rate of cytochrome c reduction at various times during the reaction. The assays were performed in a thermostat-controlled compartment of the spectrophotometer at 25°C. Samples were prepared in glass cuvettes purged with nitrogen and closed tightly with silicon stoppers. The reaction mixture was allowed to equilibrate to the indicated temperature for 20 min before cytochrome c was added with a syringe purged with nitrogen at various reaction times. The rate of reduction was recorded at 550 nm. The cytochrome c concentration in the samples varied between 40 and 160 M, depending on the initial selenite concentration. A high concentration of 1 mg/ml or about 25 M SOD was required to inhibit cytochrome c reduction markedly under the conditions described in this work.
Oxygen-The oxygen (O 2 ) concentration was monitored using a Clark cell (oxygen membrane polarographic detector; Rank Brothers Ltd., Cambridge, England). The reaction mixture was transferred quickly to the cell after the start of the reaction and monitored for 20 -30 min. The Clark cell and syringe were purged with nitrogen before use.
Hydrogen Peroxide-Hydrogen peroxide (H 2 O 2 ) was determined using the horseradish peroxidase-coupled reaction described in Frew et al. (14). Measurements were performed 20 -30 min after the start of the reaction to avoid inhibition of the peroxidase with superoxide anions.
Elemental Selenium-Se°was determined after oxidation to selenite with concentrated nitric acid. Samples containing 20 -200 nmol of Se°w ere centrifuged in 1.5-ml Eppendorf tubes for 10 min at 15,000 ϫ g.
The supernatant was discarded, and 50 -60 l of concentrated nitric acid was added to the pellet. The tubes were closed carefully and incubated in boiling water until the orange-red color of Se°disappeared (about 2-4 min). Double-distilled water was added to the tubes for a final volume corresponding to the initial sample volume, and the samples were mixed well. The Se°transformed to selenite was determined using the method described above for selenite. The presence of Se°in the reaction mixture could also be observed qualitatively because of its turbidity visible at 400 nm, where GS-Se-SG does not absorb. Preparation and Purification of Chemically Produced Se°Particles-The Se°particles were prepared in 50 mM Tris⅐HCl buffer (pH 7.0) at room temperature under a nitrogen atmosphere. The chemically well defined detergent diheptanoyl-phosphatidylcholine (DHPC) (15) was added to the buffer for a final concentration of 5.0 mM, to avoid aggregation and stacking of the particles on the walls of the vials. The reaction was performed with 0.5 mM selenite and 2.0 mM glutathione (GSH:selenite ratio of 4:1). The reaction was allowed to develop for 2 h before the Se°particles were centrifuged for 40 min at 130,000 ϫ g at room temperature. The supernatant fluid was discarded, and the particles were washed three times with 50 mM Tris⅐HCl buffer (pH 7.0) containing 2 mM DHPC (ϭ 2 mM DHPC-Tris⅐HCl buffer). The DHPC concentration used is slightly higher than its critical micellar concentration (1.4 mM).
Isolation of Biologically Produced Se°Particles-Phototrophically grown cultures of R. rubrum amended with 0.5 mM selenite were harvested 2 days after they entered the stationary phase and were processed immediately. After centrifugation at room temperature for 5 min at 5,000 ϫ g, the cell pellet was discarded. The supernatant, containing Se°particles together with bilayer vesicles (16), was centrifuged for 40 min at 130,000 ϫ g. The resulting small pellet was resuspended in a volume of 50 mM Tris⅐HCl buffer (pH 7.0) corresponding to 1 ⁄10 of the initial volume of the culture. The contaminating bilayer vesicles were solubilized by adding the detergent DHPC to a final concentration of 15 mM (15); solubilization was achieved at room temperature. Membrane debris were removed by centrifugation at 12,000 ϫ g for 10 min immediately after solubilization. The pellet was discarded, and the supernatant containing the Se°particles was centrifuged again at 130,000 ϫ g for 40 min. The small pellet was homogenized in a volume of 2 mM DHPC-Tris⅐HCl buffer corresponding to about 1 ⁄10 of the solubilization volume. This three-stage washing procedure (centrifugation of membrane debris and subsequent centrifugation and homogenization of particles) was repeated three times.
Electron Microscopy-For transmission electron microscopy cells were fixed in 2.5% glutaraldehyde for 60 min (samples were diluted with 5% aqueous glutaraldehyde), washed with running water, and embedded in low melting point agarose. Agar blocks (ϳ1 ϫ 1 ϫ 1 mm) were fixed in 1% OsO 4 , washed in running water for 60 min, dehydrated in ethanol and acetone, and embedded in Epon-Araldit. Sections cut from the Epon-Araldit preparation were contrasted with uranyl acetate and lead citrate as described by Hess (17).
For transmission electron microscopy 10 l of Se°particles suspension with a selenium concentration of about 1 mM were transferred via pipette onto a grid of cellulose nitrate and observed without any additional treatment. For energy dispersive x-ray analysis (EDAX) three to four portions of 5-10 l of Se°particle suspension with a Se°concentration of about 10 mM were deposited on a carbon plate, and the Se°l ayer was dried at room temperature after each addition of particles. Spectra were recorded with an EDAX-DX4 microanalysis system (Philips, Eindhoven, The Netherlands).

RESULTS
Transformation of Selenite by Glutathione-Selenite was transformed rapidly in the presence of glutathione at room temperature. The yield of the reaction was much improved by increasing the GSH:selenite ratio from 2:1 to 4:1. Approximately 40 -50% and 80 -95% of the initial selenite concentrations disappeared from the reaction mixtures in samples containing GSH:selenite ratios of 2:1 and 4:1, respectively (Fig. 1,  A and B). The half-life of selenite decreased when the initial selenite and GSH concentrations were increased (Table I).
Formation of GS-Se-SG-The concentration of selenodiglutathione increased rapidly within the 1st min of the reaction. It attained a maximum within 30 -60 s and decreased over the next 10 -20 min. The higher rates of appearance and disappearance of selenodiglutathione were correlated with higher initial selenite concentrations and increased GSH:selenite ratios (Fig.  2). In samples with a starting GSH:selenite ratio higher than 2, turbidity was visible in the absorption spectra within a few minutes after initiation of the reaction (Fig. 2B). The appearance of turbidity paralleled the formation of orange-red Se°p articles. Higher turbidity correlated with higher initial selenite concentration and higher GSH:selenite ratios.
Inhibition of GS-Se-SG Degradation by SOD-The addition of SOD to the reaction mixture greatly diminished the decrease in selenodiglutathione during the first minutes of the reaction (absorption at 260 nm) (data not shown), suggesting that the superoxide anions participated to the degradation of the selenodiglutathione.
Formation of Superoxide Anions- Fig. 3 shows that maximal superoxide concentrations were reached within the 1st min of the reaction for all conditions tested. In the samples containing the lowest selenite concentration (0.125 mM) and a GSH:selenite ratio of 2:1, redox equilibrium was observed for about 10 min before the cytochrome c reduction rate began to decrease again. The superoxide anions disappeared completely about 25 min after the reaction was started. Degradation of the superoxide was faster in reaction mixtures containing the GSH: selenite ratio of 4:1. Under this condition degradation of the superoxide proceeded in two steps: a rapid step that took place within the 1st min of the reaction was followed by a significantly slower step that extended up to 25 min after initiation of the reaction (Fig. 3). Higher initial concentrations of GSH and selenite correlated with a faster degradation of the superoxide (Table I). Cytochrome c reduction, which reveals the presence of superoxide, was inhibited by SOD; the inhibition was proportional to the SOD concentration (data not shown).
Oxygen-O 2 did not accumulate in the reaction mixtures; a constant low O 2 concentration of 9.0 M (Ϯ1.5 M) was determined using the Clark cell.
Hydrogen Peroxide-Hydrogen peroxide concentrations were   Fig. 1B. b The time necessary for the disappearance of 80% of the maximal superoxide concentration was determined graphically from the curves presented in Fig. 3. c The lag time of appearance of the Se°particles was determined from the spectra of the reaction mixtures recorded at increasing times after initiation of the reaction. The values represent a mean of two experiments. The error was lower than 5% in each case. d ND, not determined. H 2 O 2 reacted rapidly with GSH. A decrease in hydrogen peroxide concentration in the presence of 0.5 mM glutathione was investigated for GSH:hydrogen peroxide ratios ranging from 4:1 to 1:1. The reaction reached equilibrium within a few seconds, with 55-70% of the H 2 O 2 disappearing from the reaction mixtures (data not shown).
H 2 O 2 also reacted with Se°. A decrease of Se°of 30% was measured after 15 min in a reaction mixture containing 0.125 mM H 2 O 2 and 0.150 mM Se°.
Formation of Se°Particles-Kinetics of the abiotic formation of Se°are presented in Fig. 4A. Turbidity measurements (400 nm) of the reaction mixtures at different times showed that precipitation of Se°particles started a few minutes after the initiation of the reaction. Shorter precipitation lag times correlated with higher initial concentrations of selenite and glutathione (Table I). According to both spectrophotometric and chemical determinations, Se°formation was slow and leveled off after 20 -90 min. In some cases, Se°concentrations decreased slowly after having reached a maximum value (Fig.  4A). The final yield of Se°was relatively low (45-80%) compared with that of the bacterially mediated reaction (see, e.g. 6,18,19) and was in most cases only slightly improved by increasing the ratio of GSH to selenite (Fig. 4B).
For the GSH:selenite ratio of 2:1, as estimated from turbidity measurements, yields of Se°were low compared with those obtained in reaction mixtures with GSH:selenite ratios of 4:1 or higher. They increased with increasing initial selenite concentrations and varied between 4.5 and 27.5% in samples contain-ing initial selenite concentrations between 0.125 and 1.00 mM (data not shown).
The selenium particles formed during chemical reduction of selenite by glutathione showed the same characteristic orangered color as the particles that are produced in bacterial cultures amended with selenite (12). The chemically produced particles, washed in buffer containing 2 mM DHPC (2 mM DHPC-Tris⅐HCl buffer), also had the same spherical morphology and diameter (35-45 nm) as the particles isolated from bacterial cultures (Fig. 5). Omission of the detergent in the reaction and/or the washing buffer led to the formation of large aggregates of both artificial and biological particles. It must be noted that the small bacterially produced selenium particles isolated from the culture medium after removing the cells by centrifugation (Fig.  5A) only represent about 3% of the total particle content of the culture. Most particles present in the culture medium had diameters varying between 250 and 350 nm (data not shown), and they sedimented during centrifugation of the cells. However, in cells of R. rubrum observed under the electron microscope the diameter of the particles varied between 35 and 45 nm, which corresponds to the size of the smallest particles isolated from the culture medium (Fig. 5C).
EDAX of both chemically and biologically produced Se°particles shows the expected emission lines for selenium at 1.37, 11.22, and 12.49 keV corresponding to the SeL␣, SeK␣, and SeK␤ transitions, respectively (Fig. 6). No additional sharp emission lines were present. A slightly larger background was observed for the biological particles (Fig. 6A) than for the abiotically produced particles (Fig. 6B). DISCUSSION In the chemical reduction of selenite with glutathione much more selenite was reduced with a GSH:selenite ratio of 4:1 compared with a ratio of 2:1. This strongly supports the process outlined in Reaction 1 in the Introduction. However, the rapid increase in the production of O 2 Ϫ within the 1st min of the reaction (as determined by cytochrome c reduction) and the concomitant formation of selenodiglutathione infer that the formation of oxygen radicals occurs during the first step of the reaction. The observed proportionality between inhibition of cytochrome c reduction and SOD concentration in the reaction mixture gives evidence that cytochrome c was reduced by the superoxide anion. These results suggest that Reaction 1 proposed by Painter (8) and Ganther (9) must be modified to take into account the formation of O 2 Ϫ . We propose to describe the first step of the abiotic reduction of selenite with glutathione by Reaction 4.
According to Bébien et al. (5), who observed a large induction of two types of SOD in E. coli grown in the presence of selenite, we propose that Reaction 4 also takes place in this organism and that it may constitute the first step of the dissimilatory reduction of selenite in all cells containing high levels of glutathione and performing intracellular selenite reduction.
In the chemical reaction the superoxide spontaneously dismutates into oxygen and hydrogen peroxide (13) according to Reaction 5.
The fate of the hydrogen peroxide formed in this reaction is described below.
In the biological reduction (Fig. 7) the highly reactive super- oxide produced during the first step of the reaction undergoes a cascade of degradation reactions catalyzed by enzymes that are induced during oxidative stress (SOD, catalase, peroxidase, and probably also cytochrome(s)). These enzymes catalyze the rapid disappearance of oxygen radicals, thus preventing oxidative damage and preserving metabolic functions and cell integrity. In a second step GS-Se-SG is converted to GS-Se Ϫ and reduced glutathione (GSH) by the glutathione reductase as described by Ganther (9) and outlined in Reaction 2 in the Introduction. In a last step, the unstable selenopersulfide converts into the reduced glutathione and elemental selenium according to Reaction 3 in the Introduction. It is not known whether this reaction is enzymatically catalyzed or not. A comprehensive view of the entire process is summarized in Fig. 7.
In cultures of many species of ␣, ␤, or ␥ proteobacteria amended with up to 1-2 mM selenite the reaction is performed to completion (see, e.g. 6,12,19,20). The threshold of selenite concentration that the cells are able to reduce may reflect, at least in part, the large requirement of oxidative stress enzymes for this process. Another challenge the cells are confronted with in the presence of selenite is the uptake of the oxyanions into the cytoplasm. Unregulated uptake would give rise to fast reactions of selenite with -SH groups of proteins, which would affect metabolic functions. Damages observed in keratine-rich tissues by higher organisms (1-3) may be the result of the absence of an efficient protective mechanism to prevent the penetration of selenite into the cell.
The abiotic degradation of selenodiglutathione (Fig. 8) may be described by Reaction 6.
GS-Se-SG ϩ GSH % GS-Se Ϫ ϩ H ϩ ϩ GSSG REACTION 6 In contrast to the biological reaction (Reaction 2), in which the reduced glutathione is continually regenerated by the action of the glutathione reductase, the abiotic reaction leads to an accumulation of oxidized glutathione (GSSG), which slows down the reaction rate. The last step of the abiotic reaction, the dismutation of GS-Se Ϫ to GSH and Se°, probably does not differ from that of the biological process.
Another important difference between the abiotic and the biotic process is the persistence of the highly reactive peroxide anion in the reaction mixture, which gives rise to a cascade of secondary reactions that are summarized in Fig. 8. Because Se°c an be reoxidized by H 2 O 2 (Reaction 9) the yield of the abiotic reduction of selenite to elemental selenium largely depends on the relative rate of formation and degradation of H 2 O 2 (Reactions 7 and 8, respectively). Because Reactions 6 and 7 proceed competitively (Fig. 8), the rate of formation of H 2 O 2 also depends on the rate of Reaction 6, i.e. the supply of GSH. Reaction 7 is supported by the observed inhibition of GS-Se-SG degradation in the presence of SOD, i.e. the removal of the superoxide anion (O 2 Ϫ ). This observation indicates that the superoxide anion participates in the transformation of selenodiglutathione to GS-Se Ϫ , according to Reaction 7.
Inhibition of the degradation of selenodiglutathione by SOD combined with the rapid degradation of the superoxide anion during the first minutes of the reaction, as shown in Fig. 3, suggests that the fast degradation of superoxide and selenodiglutathione proceeds according to Reaction 7. Spontaneous dismutation of the superoxide anion proceeds significantly slower (13). The kinetics of this last reaction possibly corresponds to the second, slower step of the superoxide degradation shown in Fig. 3.
The abiotic degradation of hydrogen peroxide takes place in two ways. 1) The fast decrease of H 2 O 2 concentrations in the presence of GSH (see "Results" and Ref. 21) indicates that hydrogen peroxide can be detoxified through reduction by GSH according to Reaction 8.
2) In agreement with the observed decrease of Se°in the presence of hydrogen peroxide (Fig. 8) we propose Reaction 9.
Considering that selenodiglutathione is a moderately stable intermediate that can be isolated from the reaction mixture (9,22) and that the oxidized glutathione accumulates in the reaction mixture, we propose that Reaction 6 develops more slowly than Reaction 7 (Fig. 8). This would promote the formation of hydrogen peroxide and the subsequent loss of Se°(Reaction 9) as shown in the abiotic reduction of selenite with glutathione ( Fig. 4).
Close correlations between the half-life of selenite, the rate of superoxide degradation, and the initiation of Se°particle formation at various initial selenite concentrations strongly support Reactions 3, 4, and 7 (Table I).
In addition to Reactions 7 and 8, which describe the fast degradation of the superoxide anion shown in Fig. 3, we have to consider the spontaneous dismutation of O 2 Ϫ (Reaction 5), which may represent the slower step of the reaction (Fig. 3). Because elemental oxygen (O 2 ) was only detected in micromolar amounts in the abiotic reaction sequence we propose that the oxygen produced by this reaction was consumed in the radical chain of reactions involving O 2 Ϫ , O 2 , and GSH, as described by Winterbourn and Metodiewa (21). Through these reactions GSH and O 2 are consumed, and O 2 Ϫ is regenerated, maintaining the consumption of reduced glutathione as long as O 2 Ϫ and O 2 are present. The entire process is summarized in Reaction 10.
The efficiency of Reaction 10 depends on pH, pO 2 , and the superoxide generation rate. The reaction is very fast at neutral pH, in air, a thiol concentration of about 0.5 mM, and superoxide production rates in the mol/min range. Estimates of the rate constants for GSH under these conditions are in the range of 10 3 M Ϫ1 s Ϫ1 (21). FIG. 7. Hypothesis for the biological reduction of selenite. Glutathione is proposed to function as electron donor in the reduction. The first intermediary product is selenodiglutathione (GS-Se-SG), which is a substrate for the glutathione reductase (GR) (9) and, according to Björnstedt et al. (24), also for the bacterial thioredoxin. The oxidized thioredoxin is regenerated by the thioredoxin reductase (TR). Reduction of selenodiglutathione by GR or TR leads to the formation of the selenopersulfide of glutathione (GS-Se Ϫ ), which dismutates into reduced glutathione (GSH) and elemental selenium (Se°). In the biological reaction degradation of the superoxide is catalyzed by enzymes that are induced under oxidative stress (5). The electron source for the regeneration of glutathione is NADPH. The numbers enclosed in circles refer to the reactions in the text. We therefore propose that the GSH disappearance according to Reaction 10 will proceed slowly in the presence of the low O 2 concentrations (9 M) and the O 2 Ϫ generation rates (10 -180 mol/ml/min) measured in our experiments.
Consumption of GSH and production of GSSG according to Reaction 10 will contribute to lower the rate and the yield of the processes described by Reactions 6 and 8 (Fig. 8). This explains the weak enhancement of the yield of Se°in the abiotic reduction of selenite with glutathione when the GSH:selenite ratio is increased from 4:1 to 8:1 (Fig. 4).
Considering 1) the high GSH levels (1-10 mM) in the cytoplasm of various ␣, ␤, and ␥ proteobacteria (7,23), 2) the high efficiency of the glutathione reductase in catalyzing the degradation of the selenodiglutathione to selenopersulfide and oxidized glutathione (9) (Reaction 2), 3) the high reaction rate observed for selenite reduction by GSH under abiotic conditions (Reaction 4), 4) the decrease of the selenite reduction rate in R. rubrum with decreased glutathione levels, 2 and 5) the production of oxygen radicals in both the abiotic reduction of selenite with glutathione and the reaction mediated by E. coli, we suggest that glutathione may be the electron donor in the dissimilatory reduction of selenite in organisms performing this reaction intracellularly and containing high GSH levels in their cytoplasm. Other biomolecules containing -SH groups may also be involved in the reduction. The reduction of selenite and selenodiglutathione by the thioredoxin-thioredoxin reductase system of E. coli was described previously (24). Involvement of this enzymatic system in the biological dissimilatory reduction of selenite is supported by the observed induction of thioredoxin and thioredoxin reductase in cultures of E. coli amended with millimolar levels of selenite (5).
In the bacterial domain only the cyanobacteria and representatives of the ␣, ␤, and ␥ groups of the proteobacteria are able to synthesize glutathione (7). Consistent with the involvement of glutathione in the dissimilatory reduction of selenite, various proteobacteria belonging to these groups have been shown to tolerate millimolar levels of selenite (6,12,19,20). In contrast, a broad survey of selenite tolerance in bacteria (63 species) demonstrated that most species (81%) are not able to grow in the presence of 0.2 mM selenite (25). Because the proposed mechanism of selenite reduction releases highly reactive oxygen species it cannot take place in strict anaerobes, which do not synthesize oxidative stress enzymes.
A different mechanism of selenite reduction may take place in Gram-positive bacteria, which are not able to synthesize glutathione (7) but can tolerate high levels of selenite. Bacillus subtilis, for example, has been shown to grow in the presence of up to 5 mM selenite (26). As for proteobacteria, however, a significant induction of thioredoxin and thioredoxin reductase has been observed in B. subtilis exposed to millimolar concentrations of selenite (27) and, generally, Gram-positive bacteria accumulate coenzyme A and other organic compounds containing disulfide groups at millimolar levels (7). Additionally, a disulfide reductase capable of reducing coenzyme A disulfide and other related disulfides is produced by bacilli (28). This suggests that the mechanism of selenite reduction in these bacteria may be similar to that proposed for the proteobacteria with a high glutathione level. It must also be mentioned that reduction of disulfides can be catalyzed by selenols, which may be present in certain cells as secondary products of the selenium metabolism (29). A prerequisite for a high tolerance toward selenite may therefore be a high cytoplasmic level of disulfide-containing molecules, catalysts for the reduction of disulfides, and a functional oxidative stress protection system.
The small sizes of the Se°particles (35-45 nm) present in cells of R. rubrum grown in the presence of selenite (Fig. 5C) suggest that the smallest particles purified from the culture medium represent the original size of biologically produced Se°p articles. The larger particles (250 -300 nm), which sediment during centrifugation of the cells, are likely produced by aggregation of the 35-45-nm particles (Fig. 5A).
EDAX provides evidence that Se°particles purified from both the bacterial cultures and from the chemical reaction mixtures contain nearly pure selenium. The larger background observed in the biological particles is possibly caused by the presence of small amounts of organic material present in the biological particles. We are presently analyzing these particles in more detail. CONCLUSION The evidence for the formation of superoxide anions during the abiotic reduction of selenite with glutathione provided in this work is consistent with the observed production of superoxide in E. coli grown in the presence of selenite (5). It lends support to the hypothesis that glutathione may be involved in the dissimilatory reduction of selenite in organisms containing high levels of glutathione. Involvement of glutathione in the dissimilatory reduction of selenite in proteobacteria is also consistent with the observed decrease of the selenite reduction rate in R. rubrum containing low levels of glutathione. 2 The most important difference between the abiotic and biotic reduction of selenite is the fate of the superoxide anion. In living cells the superoxide is transformed rapidly by oxidative stress enzymes. In the abiotic process it is involved in a cascade of secondary reactions, which leads to a significant decrease in the yield of elemental selenium.
Among the bacterial domain the proposed mechanism of selenite reduction can only take place in the cyanobacteria and the ␣, ␤, and ␥ groups of the proteobacteria, which are able to synthesize glutathione. It is incompatible with strict anaerobic organisms, which are not able to synthesize oxidative stress enzymes.
The similarities between the Se°particles obtained in the abiotic and the bacterial processes, with regard to size, color, and spherical structure, provide additional support for the role of glutathione in the dissimilatory reduction of selenite in living cells. Other reductants such as ascorbate or H 2 S did not yield particles of the same small size and with a similar spherical morphology (30). Abiotic synthesis of small Se°particles employing glutathione in the presence of DHPC might be of technological interest, e.g. for the production of selenium nanowires used in the electronic industry (10).