当前位置: 首页 > 期刊 > 《毒物学科学杂志》 > 2005年第1期 > 正文
编号:11294889
Lipids versus Proteins as Major Targets of Pro-Oxidant, Direct-Acting Hemolytic Agents
http://www.100md.com 《毒物学科学杂志》
     Department of Cell and Molecular Pharmacology, Medical University of South Carolina, Charleston, South Carolina 29425

    Department of Medicine and Pharmacology, Vanderbilt University, Nashville, Tennessee 37232

    ABSTRACT

    Lipid peroxidation and the accompanying translocation of phosphatidylserine (PS) from the inner to the outer leaflet of the lipid bilayer have recently been identified as key components of a signaling pathway for phagocytosis of apoptotic cells by macrophages. Drug-induced hemolytic anemia has long been known to be caused by an accelerated uptake of damaged (but intact) erythrocytes by macrophages in the spleen, and this process has been associated with enhanced formation of reactive oxygen species (ROS). However, the role of lipid peroxidation in hemolytic injury has remained unclear, and the effect of hemolytic agents on the distribution of PS in the erythrocyte membrane is unknown. The present studies were undertaken to determine whether lipid peroxidation and PS translocation could be detected in rat and human erythrocytes by three types of direct-acting hemolytic agents—dapsone hydroxylamine, divicine hydroquinone, and phenylhydrazine. 2',7'-Dichlorodihydrofluorescein diacetate was employed as a probe for intracellular ROS formation; lipid peroxidation was assessed by GC/MS analysis of F2-isoprostanes; and PS externalization was measured by annexin V labeling and the prothrombinase assay. The data confirmed that all three hemolytic agents generate ROS within erythrocytes under hemolytic conditions; however, no evidence for lipid peroxidation or PS translocation was detected. Instead, ROS production by these hemolytic agents was associated with extensive binding of oxidized and denatured hemoglobin to the membrane cytoskeleton. The data suggest that the transmembrane signal for macrophage recognition of hemolytic injury may be derived from oxidative alterations to erythrocyte proteins rather than to membrane lipids.

    Key Words: erythrocyte; hemoglobin; hemolytic anemia; lipid peroxidation; phosphatidylserine translocation; reactive oxygen species.

    INTRODUCTION

    The hemolytic anemia that can accompany therapy with arylamine drugs, such as dapsone, and certain natural products, such as the fava bean pyrimidine aglycone, divicine, has been well documented (Beutler, 1969). However, the molecular target(s) within erythrocytes that underlies this hemotoxicity and how the damaged cells are recognized for uptake by splenic macrophages are not well understood. Early studies established that the hemolytic response was associated with an intracellular oxidative stress (Cohen and Hochstein, 1964), as evidenced by the formation of reactive oxygen species (ROS), oxidation of hemoglobin to methemoglobin, and depletion of erythrocytic GSH. Additional support for this concept is given by the fact that individuals who are genetically deficient in erythrocytic glucose-6-phosphate dehydrogenase activity, and hence have a diminished capacity to regenerate NADPH, are significantly more susceptible to hemolytic anemia (Beutler, 1978).

    While there is broad acceptance of the role of ROS in initiating hemolytic injury, there has been longstanding uncertainty about the molecular targets of attack, whether lipid or protein. It has also been unclear how the internal oxidative damage is transmitted across the cell membrane to activate (or inhibit) signaling mechanism(s) for macrophage recognition. Indeed, there is little agreement as to what constitutes normal erythrocyte senescence or what the normal external signals for macrophage uptake might be. Evidence exists for all three structural components of the cell—lipid, carbohydrate, and protein—as internal targets and/or as external signals in a variety of normal and pathological situations (for review, see Bratosin et al., 1998).

    Lipid peroxidation and the resultant perturbation of the structural integrity of the plasma membrane have long been considered to be capable of initiating the hemolytic response (Comporti, 1993; Hochstein, 1988), though how generalized destruction of membrane lipids could stimulate a selective macrophage response was not clear. The more recent reports that lipid peroxidation in nucleated cells correlates with accumulation of phosphatidylserine (PS) in the outer leaflet of the lipid bilayer (Tyurina et al., 2000), and that externalization of this phospholipid provokes a phagocytic response by stimulating PS receptors on macrophages (Tyurina et al., 2004; Zwaal and Schroit, 1997), have provided a plausible mechanism. Alternatively, studies have shown that hemoglobin oxidation products accumulate within the aging erythrocyte (Clark, 1988), and that association of denatured hemoglobin with the skeletal protein may be an initiating event in erythrocyte senescence (Rettig et al., 1999). We have observed that hemolytic agents enhance oxidative damage to cytoskeletal proteins (Grossman et al., 1992; McMillan et al., 1995, 2001), principally in the form of disulfide-linked adduction to denatured hemoglobin monomers, and that the extent of this adduction correlates highly with the hemolytic response (Jollow and McMillan, 2001).

    The role of lipid peroxidation as an initiating event has long been controversial. On the one hand, it was reported in the early 1970s that incubation of redox-active hemolytic agents, such as phenylhydrazine (PHZ), with erythrocytes induced the formation of superoxide anion radical and hydrogen peroxide (Jain and Hochstein, 1979), and that these ROS could initiate lipid peroxidation via transition metal–catalyzed formation of hydroxyl radicals (Ciccoli et al., 1994; Goldstein et al., 1980). Rice-Evans and Hochstein (1981) reported that exposure of erythrocyte ghosts to PHZ caused lipid peroxidation, as judged by the widely utilized thiobarbituric acid-reactive substances (TBARS) assay. Conversely, other investigators have reported an absence of TBARS formation in erythrocytes exposed to PHZ when correction was made for an interfering absorption (Winterbourn and Carrell, 1972; Vilsen and Nielsen, 1984). Using TBARS and the degradation of the fluorescent fatty acid probe, cis-paranaric acid, we reported that hemolytic concentrations of PHZ, but not dapsone hydroxylamine (DDS-NOH), induced lipid peroxidation in rat and human erythrocytes (McMillan et al., 1998).

    The present studies were undertaken to reevaluate the postulate that lipid peroxidation plays a key role in initiating the hemolytic response. To test this hypothesis, we examined the effect of three different types of hemolytic agents—PHZ as the classical direct-acting hemotoxicant (Azen and Schilling, 1963), DDS-NOH as an arylhydroxylamine-nitroso redox pair (Glader and Conrad, 1973), and divicine as a hydroquinone-quinone redox pair (Arese et al., 1989)—on membrane lipids and cytoskeletal proteins in rat and human erythrocytes. To ensure that comparisons were made under equivalent experimental conditions, hemolytic activity was determined using the rat 51Cr-labeled erythrocyte in vitro exposure/in vivo survival assay. ROS generation was measured with the use of the intracellular probe, 2',7'-dichlorodihydrofluorescein diacetate (DCFDA) (Wrona et al., 2005). Lipid peroxidation was assessed by direct GC/MS measurement of F2-isoprostanes, and loss of PS asymmetry was measured by annexin V labeling of intact erythrocytes and by activation of prothrombin. Lastly, oxidative damage to cytoskeletal proteins was assessed by immunostaining with anti-hemoglobin antibodies. We conclude from the present data that oxidative attack on cytoskeletal proteins, rather than on membrane lipids, underlies the process of splenic uptake of erythrocytes damaged during the course of pro-oxidant, drug-induced hemolytic anemia.

    MATERIALS AND METHODS

    Chemicals and materials.

    DDS-NOH (4-amino-4'-hydroxylaminodiphenylsulfone) and the hydroquinone form of divicine (2,6-diamino-5-hydroxy-4[3H]-pyrimidinone) were synthesized as described previously (Grossman and Jollow, 1988; McMillan et al., 1993). PHZ, cumene hydroperoxide (CuOOH), calcium ionophore A23187 [GenBank] , N-ethylmaleimide (NEM), and rabbit anti-rat and anti-human hemoglobin antibodies were purchased from Sigma-Aldrich (St. Louis, MO). Hank's balanced salt solution (HBSS, without calcium chloride, magnesium sulfate, phenol red, and sodium bicarbonate) was obtained from Invitrogen (Life Tech Corp., Grand Island, NY). Alexafluor 647-conjugated annexin V and DCFDA were purchased from Molecular Probes (Eugene, OR). in sterile saline (1mCi/ml, pH 8) was obtained from New England Nuclear (Billerica, MA). Bovine factors V and Xa were purchased from Enzyme Research Laboratories (South Bend, IN). Chromogenic substrate S-2238 was purchased from Chromogenix (DiaPharma Group, Inc., Westchester, OH). All other reagents were of the best commercially available grade.

    Erythrocyte incubation conditions.

    Male Sprague-Dawley rats (75–100 g) were purchased from Harlan Laboratories (Indianapolis, IN) and maintained on food and water ad libitum. All procedures were performed with the approval of the Medical University of South Carolina University Committee on Animal Use and Care. Animals were acclimated for 1 week to a 12-h light-dark cycle before their use. Erythrocytes were collected from anesthetized rats, and from healthy human volunteers following informed consent, into heparinized tubes and washed three times with isotonic phosphate-buffered saline supplemented with 10 mM D-glucose (PBSG, pH 7.4) to remove the plasma and buffy coat. The erythrocytes were then resuspended in PBSG and used the same day they were collected.

    Erythrocytes were incubated with the test compounds at 37°C for up to 2 h at a 40% hematocrit (unless otherwise noted) in PBSG. The erythrocyte suspensions were incubated with a range of hemolytic concentrations of each test compound. DDS-NOH was dissolved in acetone, and divicine and PHZ were dissolved in PBSG. In each experiment, vehicle controls were performed and compared with untreated erythrocyte suspensions. Because no differences were observed in the vehicle-treated versus untreated erythrocytes, only the vehicle controls are shown in the figures. After the incubation, the erythrocytes were washed once with PBSG and then processed for analysis of F2-isoprostane formation, PS translocation, and skeletal protein alterations as described below.

    Measurement of hemolytic activity.

    The in vivo survival of rat 51Cr-labeled erythrocytes was determined after in vitro incubation with DDS-NOH, divicine or PHZ for 2 h at 37°C as described previously (Harrison and Jollow, 1986). After the incubation the erythrocytes were washed once, resuspended in PBSG (40% hematocrit), and aliquots (0.5 ml) were administered intravenously to isologous rats. T0 blood samples were taken from the orbital sinus 30 min after administration of the labeled red cells. Additional blood samples were taken every 48 h for 7 days. At the end of the experiment, the samples were counted in a well-type gamma counter, and the data were expressed as percentage of the T0 sample.

    Measurement of ROS formation.

    Rat and human erythrocytes were suspended in PBSG to a 10% hematocrit. DCFDA, dissolved in DMSO to a final concentration of 0.6 mM, was added to the erythrocyte suspensions and allowed to preincubate for 15 min at 37°C. Immediately after the addition of various concentrations of the test compounds, fluorescence was measured at 5-min intervals for 20 min (excitation 488 nm, emission 529 nm) on a Molecular Devices SprectraMAX Gemeni XS fluorescence microplate reader (Molecular Devices, Sunnyvale, CA).

    Preparation of erythrocyte membrane ghosts.

    Red cell ghosts were prepared from vehicle- and hemolytic agent-treated erythrocytes as described previously with modification (Grossman et al., 1992). Briefly, washed erythrocytes were centrifuged, and the packed cells were lysed in 30 ml of anaerobic, ice-cold phosphate buffer (5 mM, pH 8.0). The membrane ghosts were pelleted by centrifugation at 20,000 x g for 10 min. The supernatant was removed by aspiration, and the ghosts were repeatedly washed with phosphate buffer until the control cells yielded white ghosts, typically 3–4 washes.

    Measurement of lipid peroxidation.

    Lipid peroxidation was assessed by measuring the content of F2-isoprostanes in red cell ghosts prepared as described above from erythrocytes treated with the vehicle and the test compounds. The work-up procedure for quantifying F2-isoprostanes by GC/MS is described in detail elsewhere (Morrow and Roberts, 1999). Briefly, lipids were extracted from the ghost suspensions (250 μl) with chloroform/methanol (2:1, v/v). The extracted lipids were then subjected to alkaline hydrolysis to release the esterified F2-isoprostanes. Treatment of erythrocyte suspensions for 1 h at 37°C with the lipid-soluble peroxide, CuOOH, was used as a positive control for lipid peroxidation (van den Berg et al., 1992).

    Measurement of PS translocation.

    Loss of PS asymmetry in response to treatment with the hemolytic agents was determined using Alexafluor-conjugated annexin V labeling as described previously (de Jong et al., 1997). Following incubation, aliquots (50 μl) of the suspensions were removed and washed twice with annexin binding buffer (1 ml). After the last wash, aliquots of the packed red cells (2 μl) were combined with annexin binding buffer (997 μl), and AlexaFluor 647-conjugated annexin-V solution (1 μl) was added to the erythrocyte suspensions to bring the total volume to 1 ml. After 30 min incubation in the dark at room temperature, the samples were washed and resuspended in annexin binding buffer (1 ml), and analyzed on a Becton Dickinson FACSCalibur analytical flow cytometer (BD Biosciences, San Jose, CA). The combination of NEM (10 mM) plus calcium (1.2 mM) and ionophore A23187 [GenBank] (4 μM) was used as a positive control for PS translocation (Kuypers et al., 1996).

    PS translocation was also assessed using the prothrombinase assay as described previously (Kuypers et al., 1996). Briefly, after incubation, erythrocyte suspensions were washed once with HBSS, and 1 μl of packed cells was added to 1 ml of Tris buffer (pH 7.4) at 37°C. Bovine factor V (0.33 U/ml) and bovine factor Xa (0.33 U/ml) were added, followed by 0.13 mg of prothrombinase. After 4 min, the reaction was quenched with 15 mM EDTA. The erythrocytes were pelleted by centrifugation, and 75 μl of the supernatant was added to 1 ml of chromogenic substrate S-2238 working solution. The increase in absorbance at 405 nm was determined over 1 min at 37°C in a Shimadzu UV-160 Spectrophotometer (Shimadzu Corporation, Kyoto, Japan) equipped with software for enzyme kinetics. The amount of thrombin formed per unit time was determined by comparison with thrombin standards.

    Electrophoretic analysis of membrane cytoskeletal proteins.

    Red cell ghosts were washed exhaustively to remove unbound hemoglobin (four washes). The ghost protein was then solubilized in NuPAGE LDS Sample Buffer (Invitrogen, Carlsbad, CA) and heated at 70°C for 10 min. Solubilized proteins (15 μg) were loaded and resolved on 4–12% NuPAGE Bis-Tris gels with MOPS Running Buffer (Invitrogen) at 200 V (constant) for 50 min. Resolved proteins were transferred to PVDF membranes for immunoblot analysis according to the Invitrogen protocol. Blotted proteins were blocked in TBST (pH 7.5) containing 5% nonfat dry milk and incubated in TBST containing 1% BSA and primary antibody (rabbit anti-rat and anti-human hemoglobin, 1:10,000 vol/vol). After washing and incubation with the peroxidase-conjugated secondary antibody (anti-rat IgG), the immunoblots were developed using ECL detection (Amersham Biosciences, Piscataway, NJ).

    Statistical analysis.

    Data obtained from individual experiments are expressed as means ± SD and were compared by one-way ANOVA. Differences were considered significant at p < 0.05.

    RESULTS

    Effect of Hemolytic Agents on Rat Erythrocytes

    Hemolytic activity and ROS generation.

    Because the objective of these studies was to examine the relationship between ROS formation and membrane lipid alterations under hemolytic conditions, studies were undertaken to confirm that the test compounds provoked hemolytic responses that were comparable to those of previous studies. Experimentally, rat 51Cr-labeled erythrocytes (40% suspensions) were incubated for 2 h at 37°C with concentrations of the test compounds previously shown to induce a half-maximal hemolytic response (TC50): DDS-NOH (150 μM), divicine (1.5 mM), and PHZ (800 μM). After incubation, the erythrocytes were washed and returned to isologous rats. In agreement with previous studies (McMillan et al., 1998; McMillan and Jollow, 1999), all three compounds induced hemolytic responses as evidenced by increased rates of removal of radioactivity from the circulation as compared to the control (Fig. 1). Also in agreement with the earlier studies, frank lysis of the treated erythrocytes during the 2 h incubation was less than 1% and was not significantly different from the control.

    To assess ROS formation in response to these hemolytic conditions, rat erythrocytes were preincubated with DCFDA for 15 min before addition of the test compounds. Three hemolytic concentrations of each compound were to be examined in the rat erythrocytes: the TC50, one concentration above the TC50, and one below it. However, preliminary experiments indicated that the concentration of erythrocytes in the suspensions had to be reduced from the physiological hematocrit (40%) to a 10% hematocrit to avoid significant quenching of the fluorescence signal by the cells. Concentrations of the hemotoxicants were reduced accordingly (four-fold) to maintain the same compound-to-erythrocyte ratio that was used in the subsequent experiments.

    After addition of the test compounds, fluorescence intensity was measured at 5-min intervals for 20 min. As shown in Figure 2, incubation of erythrocytes with DCFDA in the absence of the hemolytic agents resulted in a slight but measurable increase in fluorescence over the 20-min time period, consistent with the known low-level, steady production of ROS in normal red cells (Jandl et al., 1960). ROS formation in DDS-NOH-treated rat erythrocytes (Fig. 2A) showed what appeared to be a short lag phase of about 5 min before increasing linearly with time for 20 min. ROS formation in divicine- and PHZ-treated rat erythrocytes (Figs. 2B and 2C) was linear with time throughout the incubation period. All three hemotoxicants provoked concentration-dependent increases in ROS formation. Control incubations confirmed that the observed increases in fluorescence intensity were not due to interference from the compounds, buffer, or cellular artifacts (data not shown). Notably, at roughly equivalent hemolytic concentrations the extent of ROS formation in PHZ-treated erythrocytes was considerably greater than that in erythrocytes treated with divicine or DDS-NOH.

    Membrane lipid peroxidation.

    To determine whether lipid peroxidation was associated with hemolytic activity and ROS generation, rat erythrocytes (40% suspension) were exposed to a range of hemolytic concentrations of the test compounds for 2 h at 37°C. After the incubation period, the intact erythrocytes were washed with PBSG and lysed under anaerobic conditions to prepare membrane ghosts. Esterified isoprostanes were extracted from the ghosts and hydrolyzed, and the total F2-isoprostane content in the samples was quantified by GC/MS (Morrow and Roberts, 1999). As shown in Figure 3, none of the test compounds produced elevations in F2-isoprostane formation that were significantly different from the vehicle control.

    CuOOH was used as a positive control in this assay because of its established capacity to induce lipid peroxidation (van den Berg et al., 1992). However, in the present studies, its use for this purpose was associated with extensive frank hemolysis of the red cells in vitro. To avoid artifactual formation of peroxides due to enhanced ROS attack on broken cell membrane fragments, the experimental protocol involved sedimentation of the intact erythrocytes under strict anaerobic conditions prior to cell lysis for ghost preparation. When applied to CuOOH, this protocol resulted in a marked decrease in the recovery of the red cells as the concentration of CuOOH was increased. Thus, although lipid peroxidation was detected in the CuOOH-treated cells in a concentration-dependent manner (Fig. 3), insufficient material remained in the samples at the highest concentration tested (2 mM) to allow determination of statistical significance.

    Membrane PS asymmetry.

    To determine if translocation of PS from the inner to the outer leaflet of the lipid bilayer of erythrocytes could be detected after exposure to these hemolytic agents, rat erythrocyte suspensions (40%) were incubated with the test compounds as described above. After incubation, the red cells were treated with Alexafluor-647-labeled annexin for 30 min, and the extent of annexin V labeling was assessed by flow cytometry. As shown in Figure 4A, binding of annexin V to control rat erythrocytes occurred in <2% of the total erythrocyte population. In contrast, more than 75% of erythrocytes incubated with the positive control, NEM plus calcium and ionophore, were annexin-positive. Annexin V binding in red cells treated with hemolytic concentrations of DDS-NOH or PHZ was not significantly different from the control values. In divicine-treated cells, a small but statistically significant increase in annexin-positive erythrocytes was observed, but only at the highest concentration tested (3 mM). Nevertheless, the percentage of annexin-positive erythrocytes detected under these extreme conditions remained below 10% of the total cells.

    Figure 4B shows the results of the prothrombinase assay in rat erythrocytes, which was used to confirm the results of the annexin labeling assay. Similar to the annexin binding assay results, TC50 concentrations of the test compounds failed to cause a statistically significant increase in the PS-dependent conversion of prothrombin to thrombin. Collectively, these data indicate that the response of rat erythrocytes to pro-oxidant, direct-acting hemolytic agents is not associated with membrane lipid peroxidation or loss of PS asymmetry.

    Effect of Hemolytic Agents on Human Erythrocytes

    To determine whether the responses observed in rat erythrocytes were applicable to humans, we examined ROS formation, lipid peroxidation, and PS translocation in erythrocytes collected from human volunteers. For the human erythrocyte suspensions, the TC50s of the test compounds were estimated based on the ability of each agent to oxidize GSH as previously reported (McMillan et al., 1995). Using GSH oxidation as a surrogate for the hemolytic response, the TC50s for DDS-NOH, PHZ, and divicine in human erythrocytes are estimated to be 300 μM, 800 μM, and 3.0 mM, respectively.

    As shown in Figure 5, these hemolytic agents also enhanced the formation of ROS in human erythrocyte suspensions in a concentration-dependent manner. The rate of ROS generation in human erythrocytes was roughly comparable to the rat red cells (Fig. 2), and as in the rat cells, PHZ was the most active ROS producer of the three hemolytic agents.

    In regard to the effect of hemolytic agent-induced ROS generation on membrane lipids, neither divicine nor PHZ caused statistically significant increases in F2-isoprostane formation as compared to the control (Fig. 6). DDS-NOH caused a modest but statistically significant increase at the lowest concentration tested, but the response decreased as the concentration of DDS-NOH was increased. This inverse concentration dependence, though not statistically significant, was also observed in the rat erythrocytes. It seems likely that statistical significance for DDS-NOH in human cells was achieved not because production of F2-isoprostanes was higher in human cells than in rat cells, but because the background level of F2-isoprostanes was much lower in the human control.

    Of note, human red cells were more resistant to frank lysis induced by treatment with CuOOH. After treatment with 2 mM CuOOH, a sufficient number of red cells were recovered to observe unequivocal evidence of the peroxidative nature of this compound in the GC/MS assay under these experimental conditions (Fig. 6).

    Examination of the ability of the test compounds to induce PS translocation (Fig. 7) also indicated that human erythrocytes do not show this response under these conditions, with the possible exception of divicine. As with the rat cells, this hemolytic agent showed a modest but statistically significant annexin binding response (Fig. 7A) at the highest concentration tested (6 mM). Since the concentration dependence for the hemolytic activity of divicine in humans is unknown, it is not clear whether this increase supports a possible role for PS translocation, or if this response is an artifact due to the very high concentration of divicine in the suspension.

    Effect of Hemolytic Agents on Membrane-Bound Hemoglobin

    Given that robust formation of ROS could be detected within erythrocytes (Figs. 2 and 5), but that the integrity of the lipid bilayer appeared to be unaffected (Figs. 3, 4, 6, 7), it became important to frame the lack of responsiveness of the membrane lipids against the backdrop of another oxidative response that was occurring under these experimental conditions. To this end, we have shown previously that ROS generated by hemolytic agents in erythrocytes could convert hemoglobin free SH-groups to reactive thiyl radical species (Bradshaw et al., 1995), and that this reaction is associated with the formation of disulfide-linked hemoglobin-skeletal protein adducts and lipid-bound hemoglobin monomer and polymers that can readily be observed on immunoblots stained with hemoglobin antibodies (Jollow and McMillan, 2001).

    As shown in Figure 8, all three hemolytic agents significantly enhanced the appearance of membrane-bound hemoglobin in both rat and human erythrocytes. All three compounds induced the formation of the lipid-bound hemoglobin monomer (16 kDa band) as well as the selective binding of hemoglobin to numerous skeletal proteins in both species. The patterns of selective hemoglobin binding produced by DDS-NOH and divicine were remarkably similar in both rats and humans, though more staining in the spectrin region (175–200 kDa) was observed in the human erythrocytes (Fig. 8B). Interestingly, the pattern of hemoglobin binding with PHZ was notably different from the two other compounds, in that much less selective binding to skeletal proteins was observed, and there were new bands that appeared above and below the hemoglobin monomer and dimer (32 kDa band). The human and rat responses to PHZ were remarkably similar (though staining was more intense in the human). Although the functional significance of this hemoglobin-skeletal protein adduct formation is not known and there are numerous potential targets to consider, the occurrence of this phenomenon across multiple hemolytic agents in two species supports its inclusion as a key component of the hemolytic sequelae.

    DISCUSSION

    In previous studies comparing the ability of DDS-NOH and PHZ to induce lipid peroxidation under equivalent hemolytic conditions (McMillan et al., 1998), we used the TBARS and cis-paranaric acid assays as indicators of lipid peroxidation. The results of those studies indicated that PHZ, but not DDS-NOH, caused lipid peroxidation in rat and human erythrocytes in a concentration-dependent manner. This conclusion was weakened by subsequent reports which showed that hydrogen peroxide (or anything that generates it) can create fluorescent pigments of hemoglobin that interfere with photometric-based assays, i.e., give false positives (Nagababu and Rifkind, 2000). Given the ability of PHZ, but not DDS-NOH, to produce auto-fluorescence in erythrocytes (unpublished observation), it was conceivable that our data on PHZ-induced lipid peroxidation were artifactual.

    For this reason we reexamined the role of lipid peroxidation in drug-induced hemolytic anemia using the more reliable GC/MS-based assay of F2-isoprostanes, which are by-products of the peroxidation of arachidonic acid. In contrast to other assays for lipid peroxidation, GC/MS analysis of F2-isoprostanes is sensitive and molecularly specific and, hence, is less vulnerable to the problems associated with UV absorbance- and fluorescence-based assays (Halliwell and Whiteman, 2004). In addition, we determined whether we could detect a related response, i.e., PS externalization, and one that has been shown recently to be a signal for recognition of apoptotic cells by macrophages.

    The hemolytic agents examined in these studies were chosen because they represent three chemically distinct types of redox pairs. DDS-NOH, the hemolytic metabolite of dapsone, is an example of an arylhydroxylamine-nitroso redox pair, which undergoes a coupled oxidation-reduction reaction with oxyhemoglobin and molecular oxygen (Kiese, 1974), yielding methemoglobin and ROS (ferryl heme and hydroxyl radical), respectively (Bradshaw et al., 1997). Divicine, is the aglycone metabolite of the fava bean pyrimidine -glucoside, vicine. Divicine is thought to undergo auto-oxidation within erythrocytes, generating the quinone form of divicine and hydrogen peroxide by a superoxide-dependent chain mechanism (Winterbourn et al., 1989). PHZ undergoes oxidation to yield a compound-centered (i.e., phenyldiazene) free radical (Goldberg and Stern, 1977).

    In regard to human erythrocytes, hemolytic activity could not be determined directly, i.e., by returning the damaged 51Cr-tagged red cells to the donors analogous to the rat red cell assay. Thus, while all three compounds are known to provoke hemolytic responses in humans, their hemolytic concentration ranges for the in vitro incubation experiments could only be estimated. The concentrations used in the human red cell experiments were derived from previous studies in which we utilized GSH-protein mixed disulfide formation as an index of relative susceptibility of rat versus human erythrocytes to pro-oxidant hemolytic compounds (McMillan et al., 1995).

    Collectively, the data indicate that the responses of rat and human erythrocytes to these pro-oxidant, direct-acting hemolytic agents are not associated with membrane lipid peroxidation or loss of PS asymmetry. Parallel studies in rat and human erythrocytes confirmed that there was ample formation of ROS in the incubates under these experimental conditions for all three types of hemotoxicants (Figs. 2 and 5), and that these treatments were hemotoxic in rat erythrocytes (Fig. 1). It should be noted that false positive results still occurred with the F2-isoprostanes assay when the samples were prepared without strict care to exclude oxygen during the work-up of the samples. When treated erythrocytes were lysed and the ghosts prepared under aerobic conditions, the concentration-dependent formation F2-isoprostanes could be demonstrated (unpublished observations). This response, however, was eliminated entirely if the ghosts were washed under anaerobic conditions and was likely due to (redox-active) membrane-bound hemoglobin that becomes associated with the treated erythrocytes in a concentration-dependent manner.

    As shown in Figures 3 and 6, DDS-NOH produced somewhat curious behavior in what appeared to be a concentration-dependent decrease in production of F2-isoprostanes. In rat erythrocytes, the response was elevated but not statistically significant; in human erythrocytes, F2-isoprostane formation was elevated significantly at all concentrations tested, but statistical significance was achieved by the presence of the lower background values in human cells; the absolute values in the treated erythrocytes were comparable in the two species.

    That hemoglobin is a target for ROS is well established. Early studies demonstrated that the Heinz bodies that appear transiently in erythrocytes isolated from patients undergoing a hemolytic crisis consist of oxidized and denatured hemoglobin. Subsequent studies have shown that hemoglobin binding to skeletal proteins may play a role in the removal of senescent erythrocytes from the circulation (Waugh and Low, 1985). We have shown previously that hemoglobin thiyl radicals can be detected in rat erythrocytes under hemolytic conditions, and that addition of cysteamine to DDS-NOH-treated erythrocytes can convert a hydroxyl radical EPR signal to a cysteamine thiyl radical signal (Bradshaw et al., 1997). The presence of membrane-bound hemoglobin in all the treatment groups (Fig. 8) indicates that conversion of oxyhemoglobin to a "reactive intermediate" may be a general characteristic of all pro-oxidant hemolytic agents. The differences in the patterns of selective hemoglobin binding among these compounds presumably reflects differences in their rates of redox cycling and ROS generation, though the quantitative relationship between ROS formation and hemolytic activity is unclear and warrants further investigation. The reason for the differences in the responses between the two species is likewise unclear, but may be explained in part by the greater number of free SH-groups per hemoglobin tetramer in the rat (6) versus the human (2) and/or by the presence of fast-reacting SH-groups in rat -chains (Rossi et al., 1998).

    That lipid peroxidation does not contribute to the mechanism underlying hemolytic injury was somewhat surprising. As noted above, ROS formation occurs in erythrocytes under hemolytic conditions, and this response is associated with the appearance of membrane-bound hemoglobin (Fig. 8). Since unsaturated lipids are known to be targets for ROS-induced lipid peroxidation, the absence of this response deserves an explanation. One possibility, illustrated in Figure 9, is that, since the initial ROS formation occurs within the heme pocket, it is likely that these species are generated predominately within the hemoglobin "liquid matrix." In view of their short lifespan, it is unlikely that significant ROS would reach the lipid bilayer. In contrast, the close proximity of ROS generation to hemoglobin SH-groups would favor accessible sulfur atoms as preferred targets. Hemoglobin thiyl radicals have sufficient stability to reach and react with cytoskeletal proteins, but they have insufficient lipophilicity to penetrate into the bilayer to initiate lipid peroxidation. In contrast, when ROS are generated outside of erythrocytes by exogenous peroxides, such as CuOOH (Figs. 3 and 6), the lipid bilayer can be penetrated and subjected to peroxidation. It is noteworthy that CuOOH provoked significant direct lysis of erythrocytes, to the extent that this response became the limiting factor in its use as a positive control for F2-isoprostane formation. The direct lytic activity associated with lipid peroxidation in our controls further supports the notion that lipid peroxidation plays little or no role in pro-oxidant drug-induced hemolytic anemia.

    It is of interest that the duration of linear formation of ROS in these incubates (Figs. 2 and 5) was significantly longer than that predicted by the stability of DDS-NOH and divicine in erythrocyte suspensions (McMillan et al., 1993; unpublished observations) suggesting that some meta-stable radical species may be an intermediary between the initial formation of ROS and attack on the skeletal protein. Since complete depletion of erythrocytic GSH exacerbates hemolytic activity (Bolchoz et al., 2002; Bowman et al., 2004), this intermediate species is unlikely to be GSH itself. Hemoglobin thiyl radicals appear to be reasonable candidates (Fig. 9), but future studies will be needed to clarify the pathway(s) linking ROS generation and the hemoglobin-skeletal protein adduct formation.

    It should be noted that the present studies deal only with pro-oxidant hemolytic agents and do not exclude a role for lipid peroxidation and/or loss of PS asymmetry in other types of erythrocyte pathology. Although erythrocytes lack mitochondria and nuclei and hence do not have a complete apoptosis apparatus, they do contain many of the components of the apoptotic cascade, such as caspase 3, Bcl-X(L), and Bak (Mandal et al., 2002; Walsh et al., 2002). The presence of these mediators suggests that a truncated sequence of apoptosis may occur in erythrocytes, and hence could have played a causative role in hemolytic injury.

    In summary, we have examined the effects of three chemically distinct types of redox-active compounds on membrane lipid peroxidation and loss of PS asymmetry in a model system where effects can be directly related to ROS formation and commitment to hemolytic removal by the spleen. When adequate care is taken to exclude oxygen during ghost preparation and preassay handling, there is no evidence that lipid peroxidation plays a role in the hemolytic response. The absence of appearance of PS in the outer leaflet of the membrane under hemolytic conditions suggests that PS does not serve as a primary factor in the uptake of intact but damaged erythrocytes into splenic macrophages. Similarly, since skeletal protein damage had occurred, as evidenced by hemoglobin adduction with various proteins of this meshwork, without a PS response, it seems unlikely that spectrin damage invokes a PS-dependent uptake mechanism under these experimental conditions.

    NOTES

    The authors certify that all research involving human subjects was done under full compliance with all government policies and the Helsinki Declaration.

    ACKNOWLEDGMENTS

    This study was supported by National Institutes of Health grants AI46424, DK48831, CA77839, and GM15431. The authors also acknowledge the MUSC GCRC (RR01070) and the BSB/CRI Animal Facility (C06 RR015455) that are supported by the Extramural Research Facilities Program of the National Center for Research Resources. The authors would like to thank Jennifer Schulte and Cynthia Reich for their technical assistance in the preparation of this manuscript. The authors also gratefully acknowledge the Medical University of South Carolina's Flow Cytometry Facility and Dr. Kumar Sambamurti for use of the fluorescence microplate reader.

    REFERENCES

    Arese, P., Mannuzzu, L., and Turrini, F. (1989). Pathophysiology of favism. Folia Haematol. 116, 745–752.

    Azen, E. A., and Schilling, R. F. (1963). Role of the spleen in acetylphenylhydrazine (APH) anemia in rats. J. Lab. Clin. Med. 62, 59–71.

    Beutler, E. (1969). Drug-induced hemolytic anemia. Pharmacol. Rev. 21, 73–103.

    Beutler, E. (1978). Hemolytic anemia in disorders of red cell metabolism. In Topics in Hematology (M. M. Wintrobe, Ed.), pp. 75–92. Plenum Medical, New York.

    Bolchoz, L. J., Morrow, J. D., McMillan, D. C., and Jollow, D. J. (2002). Primaquine-induced hemolytic anemia: Effect of 6-methoxy-8-hydroxylaminoquinoline on rat erythrocyte sulfhydryl status, membrane lipids, cytoskeletal proteins and morphology. J. Pharmacol. Exp. Ther. 303, 141–148.

    Bowman, Z. S., Oatis, J. E., Whelan, J. L., Jollow, D. J., and McMillan, D. C. (2004). Primaquine-induced hemolytic anemia: Susceptibility of normal vs. glutathione-depleted rat erythrocytes to 5-hydroxyprimaquine. J. Pharmacol. Exp. Ther. 309, 79–85.

    Bradshaw, T. P., McMillan, D. C., Crouch, R. K., and Jollow, D. J. (1995). Identification of free radicals produced in rat erythrocytes exposed to hemolytic concentrations of phenylhydroxylamine. Free Radic. Biol. Med. 18, 279–285.

    Bradshaw, T. P., McMillan, D. C., Crouch, R. K., and Jollow, D. J. (1997). Formation of free radicals and protein mixed disulfides in rat red cells exposed to dapsone hydroxylamine. Free Radic. Biol. Med. 22, 1183–1193.

    Bratosin, D., Mazurier, J., Tissier, J. P., Estaquier, J., Huart, J. J., Ameisen, J. C., Aminoff, D., and Montreuil, J. (1998). Cellular and molecular mechanisms of senescent erythrocyte phagocytosis by macrophages. A review. Biochimie 80, 173–195.

    Ciccoli, L., Signorini, C., Alessandrini, C., Ferrali, M., and Comporti, M. (1994). Iron release, lipid peroxidation, and morphological alterations of erythrocytes exposed to acrolein and phenylhydrazine. Exp. Mol. Pathol. 60, 108–118.

    Clark, M. R. (1988). Senescence of red blood cells: Progress and problems. Physiol. Rev. 68, 503–554.

    Cohen, G., and Hochstein, P. (1964). Generation of hydrogen peroxide in erythrocytes by hemolytic agents. Biochemistry 3, 895–900.

    Comporti, M. (1993). Lipid peroxidation. Biopathological significance. Mol. Aspects Med. 14, 199–207.

    de Jong, K., Geldwerth, D., and Kuypers, F. A. (1997). Oxidative damage does not alter membrane phospholipid asymmetry in human erythrocytes. Biochemistry 36, 6768–6776.

    Glader, B. E., and Conrad, M. E. (1973). Hemolysis by diphenylsulfones: Comparative effects of DDS and hydroxylamine-DDS. J. Lab. Clin. Med. 81, 267–272.

    Goldberg, B., and Stern, A. (1977). The mechanism of oxidative hemolysis produced by phenylhydrazine. Mol. Pharmacol. 13, 832–839.

    Goldstein, B. D., Rozen, M. G., and Kunis, R. L. (1980). Role of red cell membrane lipid peroxidation in hemolysis due to phenylhydrazine. Biochem. Pharmacol. 29, 1355–1359.

    Grossman, S. J., and Jollow, D. J. (1988). Role of dapsone hydroxylamine in dapsone-induced hemolytic anemia. J. Pharmacol. Exp. Ther. 244, 118–125.

    Grossman, S. J., Simson, J., and Jollow, D. J. (1992). Dapsone-induced hemolytic anemia: Effect of N-hydroxy dapsone on the sulfhydryl status and membrane proteins of rat erythrocytes. Toxicol. Appl. Pharmacol. 117, 208–217.

    Halliwell, B., and Whiteman, M. (2004). Measuring reactive species and oxidative damage in vivo and in cell culture: How should you do it and what do the results mean Br. J. Pharmacol. 142, 231–255.

    Harrison, J. H., Jr., and Jollow, D. J. (1986). Role of aniline metabolites in aniline-induced hemolytic anemia. J. Pharmacol. Exp. Ther. 238, 1045–1054.

    Hochstein, P. (1988). Perspectives on hydrogen peroxide and drug-induced hemolytic anemia in glucose-6-phosphate dehydrogenase deficiency. Free Radicic. Biol. Med. 5, 387–392.

    Jain, S. K., and Hochstein, P. (1979). Generation of superoxide radicals by hydrazine. Its role in phenylhydrazine induced hemolytic anemia. Biochim. Biophys. Acta 586, 128–134.

    Jandl, J. H., Engle, L. K., and Allen, D. W. (1960). Oxidative hemolysis and precipitation of hemoglobin I. Heinz body anemias as an acceleration of red cell aging. J. Clin. Invest. 39, 1818.

    Jollow, D. J., and McMillan, D. C. (2001). Oxidative stress, glucose-6-phosphate dehydrogenase and the red cell. Adv. Exp. Med. Biol. 500, 595–605.

    Kiese, M. (1974). Methemoglobinemia: A Comprehensive Treatise. CRC Press, Cleveland, OH.

    Kuypers, F. A., Lewis, R. A., Hua, M., Schott, M. A., Discher, D., Ernst, J. D., and Lubin, B. H. (1996). Detection of altered membrane phospholipid asymmetry in subpopulations of human red blood cells using fluorescently labeled annexin V. Blood 87, 1179–1187.

    Mandal, D., Moitra, P. K., Saha, S., and Basu, J. (2002). Caspase 3 regulates phosphatidylserine externalization and phagocytosis of oxidatively stressed erythrocytes. FEBS Lett. 513, 184–188.

    McMillan, D. C., Bolchoz, L. J., and Jollow, D. J. (2001). Favism: Effect of divicine on rat erythrocyte sulfhydryl status, hexose monophosphate shunt activity, morphology, and membrane skeletal proteins. Toxicol. Sci. 62, 353–359.

    McMillan, D. C., Jensen, C. B., and Jollow, D. J. (1998). Role of lipid peroxidation in dapsone-induced hemolytic anemia. J. Pharmacol. Exp. Ther. 287, 868–876.

    McMillan, D. C., and Jollow, D. J. (1999). Favism: Divicine hemotoxicity in the rat. Toxicol. Sci. 51, 310–316.

    McMillan, D. C., Schey, K. L., Meier, G. P., and Jollow, D. J. (1993). Chemical analysis and hemolytic activity of the fava bean aglycon divicine. Chem. Res. Toxicol. 6, 439–444.

    McMillan, D. C., Simson, J. V., Budinsky, R. A., and Jollow, D. J. (1995). Dapsone-induced hemolytic anemia: Effect of dapsone hydroxylamine on sulfhydryl status, membrane skeletal proteins and morphology of human and rat erythrocytes. J. Pharmacol. Exp. Ther. 274, 540–547.

    Morrow, J. D., and Roberts, L. J., 2nd (1999). Mass spectrometric quantification of F2-isoprostanes in biological fluids and tissues as measure of oxidant stress. Methods Enzymol. 300, 3–12.

    Nagababu, E., and Rifkind, J. M. (2000). The origin of red cell fluorescence caused by hydrogen peroxide treatment. Free Radic. Biol. Med. 29, 659–663.

    Rettig, M. P., Low, P. S., Gimm, J. A., Mohandas, N., Wang, J., and Christian, J. A. (1999). Evaluation of biochemical changes during in vivo erythrocyte senescence in the dog. Blood 93, 376–384.

    Rice-Evans, C., and Hochstein, P. (1981). Alterations in erythrocyte membrane fluidity by phenylhydrazine-induced peroxidation of lipids. Biochem. Biophys. Res. Commun. 100, 1537–1542.

    Rossi, R., Barra, D., Bellelli, A., Boumis, G., Canofeni, S., Di Simplicio, P., Lusini, L., Pascarella, S., and Amiconi, G. (1998). Fast-reacting thiols in rat hemoglobins can intercept damaging species in erythrocytes more efficiently than glutathione. J. Biol. Chem. 273, 19198–19206.

    Tyurina, Y. Y., Shvedova, A. A., Kanai, K., Tyurin, V. A., Kommineni, C., Quinn, P. J., Schor, N. F., Fabisiak, J. P., and Kagan, V. E. (2000). Phospholipid signaling in apoptosis: Peroxidation and externalization of phosphatidylserine. Toxicology 148, 93–101.

    Tyurina, Y. Y., Tyurin, V. A., Zhao, Q., Djukic, M., Quinn, P. J., Pitt, B. R., and Kagan, V. E. (2004). Oxidation of phosphatidylserine: A mechanism for plasma membrane phospholipid scrambling during apoptosis Biochem. Biophys. Res. Commun. 324, 1059–1064.

    van den Berg, J. J. M., Op den Kamp, J. A. F., Lubin, B. H., Roelofsen, B., and Kuypers, F. A. (1992). Kinetics and site specificity of hydroperoxide-induced oxidative damage in red blood cells. Free Radic. Biol. Med. 12, 487–498.

    Vilsen, B., and Nielsen, H. (1984). Reaction of phenylhydrazine with erythrocytes. Cross-linking of spectrin by disulfide exchange with oxidized hemoglobin. Biochem. Pharmacol. 33, 2739–2748.

    Walsh, M., Lutz, R. J., Cotter, T. G., and O'Connor, R. (2002). Erythrocyte survival is promoted by plasma and suppressed by a Bak-derived peptide that interacts with membrane-associated Bcl-X(L). Blood 99, 3439–3448.

    Waugh, S. M., and Low, P. S. (1985). Hemichrome binding to band 3: Nucleation of Heinz bodies on the erythrocyte membrane. Biochemistry 24, 34–39.

    Winterbourn, C. C., and Carrell, R. W. (1972). The absence of lipid peroxidation in human red cells exposed to acetylphenylhydrazine. Br. J. Haematol. 23, 499–505.

    Winterbourn, C. C., Cowden, W. B., and Sutton, H. C. (1989). Auto-oxidation of dialuric acid, divicine and isouramil. Superoxide dependent and independent mechanisms. Biochem. Pharmacol. 38, 611–618.

    Wrona, M., Patel, K., and Wardman, P. (2005). Reactivity of 2‘,7’-dichlorodihydrofluorescein and dihydrorhodamine 123 and their oxidized forms toward carbonate, nitrogen dioxide, and hydroxyl radicals. Free Radic. Biol. Med. 38, 262–270.

    Zwaal, R. F. A., and Schroit, A. J. (1997). Pathophysiological implications of membrane phospholipid asymmetry in blood cells. Blood 89, 1121–1132.(David C. McMillan, Christ)