Deferoxamine deconditioning increases neuronal vulnerability to hemoglobin

Abstract

Concomitant treatment with deferoxamine (DFO) protects neural cells from iron and heme-mediated oxidative injury, but also disrupts cell responses to iron loading that may be protective. We hypothesized that DFO treatment and withdrawal would subsequently increase neuronal vulnerability to hemoglobin. Pretreatment with DFO followed by its washout increased neuronal loss after subsequent hemoglobin exposure by 3–4-fold compared with control vehicle-pretreated cultures. This was associated with reduced ferritin induction by hemoglobin; expression of heme oxygenase-1, which catalyzes iron release from heme, was not altered. Increased neuronal loss was prevented by exogenous apoferritin or by continuing DFO or antioxidants throughout the experimental course. Cell nonheme iron levels after hemoglobin treatment were similar in DFO-pretreated and control cultures. These results indicate that DFO deconditions neurons and subsequently inscreases their vulnerability to heme-mediated injury. Its net effect after CNS hemorrhage may be highly dependent on the timing and duration of its administration. Withdrawal of DFO while heme or iron levels remain elevated may be deleterious, and may negate any benefit of prior concomitant therapy.

1. Introduction

Vessel rupture is the initial event in about 15% of strokes and has a significantly worse prognosis than stroke produced by vessel occlusion. Clinical observation and preclinical investigation suggest that blood extravasation into the brain parenchyma or subarachnoid space initiates multiple pathological cascades that lead to cytotoxicity and vascular dysfunction [1,2]. One likely mediator of this secondary injury is hemoglobin (Hb), which is present at millimolar concentrations in circulating blood and extravascular hematomas. Its toxicity is due at least in part to iron, an oxidative neurotoxin released upon breakdown of heme moieties by the heme oxygenase (HO) enzymes [3,4].

Identification of Hb as a putative target has generated interest in iron chelator therapy after intracerebral (ICH) or subarachnoid hemorrhage (SAH) [5,6]. Deferoxamine (DFO) is a chelator that has been in clinical use for decades to treat transfusional iron overload and acute iron poisoning. In primary cell cultures, it prevents the neurotoxicity of Hb or hemin [3,4]. In vivo, systemic DFO treatment has improved outcome in rodent and piglet ICH and SAH models [5–7]. However, when administered to ICH patients intravenously for five days at a dose approximating that used to treat iron overload, the trial was suspended due to an increased incidence of acute respiratory distress syndrome in the DFO group [8]. Unfortunately, when given at a much lower dose for only 3 days after hemorrhage in a Phase II trial, DFO was tolerated in carefully-selected patients but failed a futility analysis [9].

Cell iron homeostasis is a delicate process that balances metabolic needs with the toxicity produced by excess low molecular weight iron complexes [10]. Neurons may be particularly vulnerable to its disruption due to their high metabolic rates coupled with relatively weak antioxidant defenses [11]. DFO has previously been reported to be a potent disrupter of iron homeostasis [12,13]. Its direct toxicity when used to treat systemic iron overload has been attributed to iron or other metal starvation and metalloenzyme inhibition, and is managed by dose tapering and timely cessation of treatment as serum iron levels fall [14–16]. In the setting of the localized tissue iron overload produced by hemorrhagic stroke, a brief course of therapy has been utilized in animal and subsequent clinical studies in an effort to mitigate toxicity [17]. However, in the initial hours to days after ICH, most iron remains Hb-bound and sequestered within intact erythrocytes; any increase in perihematomal tissue iron levels is delayed until at least 24–72 h after hemorrhage [18,19]. Conversely, at the end of a 3–5 day course of DFO, erythrocyte breakdown is still in progress, and perihematomal tissue iron levels will be elevated for weeks thereafter [19].

Very little is known about the consequences of treating neural cells with DFO prior to or at the onset of elevation in local iron levels, and then abruptly stopping therapy while Hb levels remain elevated due to ongoing erythrolysis. DFO penetrates cell membranes slowly but it is taken up by endocytosis and accumulates selectively in lysosomes, which may provide sustained cell protection after extracellular levels have declined [20–22]. Conversely, by attenuating the protective response of cells to iron loading [12,23], an abbreviated treatment schedule could paradoxically increase subsequent iron-mediated neuronal loss. In the present series of experiments, we assessed the effect of DFO pretreatment and withdrawal on neuronal vulnerability to Hb in an established cortical primary cell culture model.

2. Experimental procedures

Primary cell cultures.

Preparation of mixed cortical cell cultures containing both neurons and glia followed a protocol approved the University of Maryland Baltimore Institutional Animal Care and Use Committee (Protocol #0518012). Experiments complied with ARRIVE guidelines and were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals (NIH Publications No. 8023, revised 1978). Neuron-glia cultures were used exclusively for these experiments because neurons survive without glia in vitro only in media containing antioxidants and iron-poor transferrin, which would confound results [24]. Cultures were prepared from fetal Swiss-Webster mice (Taconic Biosciences, Rensselaer, NY, USA) at gestational age 14–16 days following a method that has been previously described in detail [25]. After cortical tissue dissociation by trituration, the cell suspension was added to plating medium containing Dulbecco’s Modified Eagle Medium (DMEM, Gibco 11960, Life Technologies, Grand Island, NY, USA), 2 mM glutamine, 5% equine serum and 5% fetal bovine serum (both purchased from Hyclone, Logan, UT, USA). Cells were plated on a confluent mixed glial culture that had been prepared as previously described using medium containing 10% fetal bovine serum and lacking equine serum [26]. Two thirds of the medium was removed and replaced with fresh feeding medium on day 5–6, Day 10–11, and every 1–2 days thereafter. Feeding medium contained 10% equine serum, lacked fetal bovine serum, and was otherwise identical to plating medium.

Deferoxamine and hemoglobin or hydrogen peroxide exposure.

Cultures were used for experiments when greater than 11 days in vitro. Exclusion criteria were otherwise limited to bacterial or fungal contamination, cell peeling or cell death when examined under a phase contrast microscope prior to beginning the experiment. Unbiased treatment group assignments were guided by an online random number generator (www.random.org). DFO (Sigma-Aldrich, St. Louis, MO, USA) was added to cultures diluted in DMEM at 24 h before Hb or hydrogen peroxide treatment, reducing the serum content of the medium to 3.3% in the process. Control cultures were pretreated with DMEM alone. After DFO washout, exposure to human Hb A (Hemosol Inc, Etobicoke, Ontario, Canada) or hydrogen peroxide (CVS Health, Woonsocket, RI, USA) alone or with ascorbate (Spectrum Chemical Mfg Corp, New Brunswick, NJ, USA), Trolox (Sigma-Aldrich), DFO, or horse spleen apoferritin (Sigma-Aldrich) followed previously-published protocols [27,28] in Minimal Essential Medium (Gibco 11430) containing 10 mM glucose (MEM10).

In one experiment, DFO was removed in stepwise fashion from cultures prior to Hb treatment. This was accomplished by reducing its concentration by half every 17 min for five cycles, followed by medium exchange. This 17 min interval approximates the plasma half-life of iron-free deferoxamine [29,30].

Outcome Measures.

All researchers conducting these assays were blinded to treatment groups.

Neuronal death assay.

At the conclusion of all experiments, cultures were inspected and injury estimated under phase contrast microscopy. Neuronal death was then quantified by lactate dehydrogenase (LDH) release assay, which provides an accurate measure of neuronal death in this model that correlates well with more laborious and bias-prone cell counts [31]. The kinetic enzyme activity assay has previously been described in detail [32]. The low LDH activity in control cultures from the same plating was subtracted from all values to determine the activity produced by cytotoxicity. In order to combine data from cultures prepared from different platings, which vary somewhat in neuron density, all values were normalized to those in control cultures treated with N-methyl-D-aspartate (NMDA, Sigma-Aldrich) 300 μM, which produces necrosis of all neurons without injuring glia. Multiple prior experiments have demonstrated that Hb per se at micromolar concentrations kills neurons but not glial cells in this culture system, and LDH release due to Hb toxicity is of neuronal origin [3,33,34]. Control experiments demonstrated that neuronal LDH activity was not altered by pretreating cultures with DFO for 24 h.

Immunoblotting.

Medium was completely removed by aspiration and cultures were then washed with fresh MEM10. After another aspiration, cold 100 μl lysis buffer (210 mM mannitol, 70 mM sucrose, 5 mM HEPES, 1 mM EDTA, 0.1% sodium dodecyl sulfate, and 0.1% Triton X-100) was added to each well. Lysate was then collected and stored at −70 °C until used. A previously-described immunoblotting method was then used, with the following primary antibodies: antihorse spleen ferritin, Sigma-Aldrich F6136, 1:167 dilution; anti-heme oxygenase-1, Enzo Life Sciences, Farmingdale, NY, USA, ADI-SPA-896, 1:1667; anti-glutathione peroxidase-4, Abcam, Cambridge, MA, USA, EPNCIR44, 1:1000 or 1:5000; anti-actin, Sigma-Aldrich A2066, 1:800 or 1:5000. Immunoreactivity was visualized using ChemiDoc™ Touch Imaging System (Bio-Rad); band density was normalized to that of actin controls from the same culture.

Non-heme Iron Assay.

After saline wash (500 μl × 3), cells were lysed in HEPES-buffered saline containing 8.3% trichloroacetic acid (TCA) and 1.3% sodium pyrophosphate. The lysate was collected, boiled for 10 min, and centrifuged (10,000×g, 5 min). Non-heme iron in the supernatant was detected as previously detailed [35] using a solution containing 25 mM sodium ascorbate (Spectrum Chemical Mfg Corp), 1 mM ferrozine (Sigma-Aldrich) and 1.05 M sodium acetate (Sigma-Aldrich). Absorbance (562 nm) of samples and standard dilutions (High-Purity Standards, Charleston, SC, USA) was then quantified.

Statistical testing.

Data were analyzed with GraphPad Prism and GraphPad InStat software. Differences between treatment groups were assessed using one-way ANOVA and the Bonferroni multiple comparisons test when comparing three or more groups, and unpaired t-test when comparing two groups.

The experimental timeline is presented in schematic form in Fig. 1.

Fig. 1.

Timeline of culture treatment with deferoxamine (DFO), hemoglobin (Hb) and other agents and of outcome assays. All experiments were conducted at ≥ 12 days in vitro (DIV).

3. Results

Deferoxamine pretreatment increases Hb-mediated neuronal death.

In initial experiments, cultures were pretreated for 24 h with DFO at a concentration previously demonstrated to be well-tolerated in this model and protective against Hb neurotoxicity [36]. It was administered either alone or concomitant with a low concentration of Hb to simulate early erythrolysis. After medium exchange and DFO washout (> 1000x medium dilution), exposure to 3 μM Hb resulted in loss of about one-quarter of neurons in cultures pretreated with DMEM-based medium only (Figs. 2A and 3). Neuronal loss was consistently and significantly increased by DFO pretreatment; addition of Hb to the pretreatment protocol had no significant effect.

Fig. 2.

Deferoxamine pretreatment increases the neurotoxicity of hemoglobin. A) Cultures were pretreated for 24 h with deferoxamine 100 μM (DFO Pre, n = 18 cultures), DFO 100 μM plus hemoglobin 1 μM (Hb + DFO Pre, n = 8 cultures) or DMEM vehicle (DMEM Pre, n = 20 cultures). After medium exchange, cultures were then treated for 24 h with Hb 3 μM alone without DFO. Bars represent mean percentage LDH release (+S.E.M.), normalized to the mean value in sister cultures treated with N-methyl-D-aspartate 300 μM for the duration of the experiment, which releases all neuronal LDH without injuring glial cells. The LDH release in control cultures subjected to washes and medium exchange only were subtracted from all values in this and subsequent experiments to obtain the signal due to neurotoxicity. ***p < 0.001 versus value in DMEM-Pre Hb condition. B) Cultures (10/condition) were pretreated with DFO 100 μM alone as in A; after DFO washout they were then treated with Hb 3 μM alone or with 100 μM each of ascorbate (Asc), Trolox, or DFO. C) Cultures (16/ condition) were pretreated with DFO as in B; after DFO washout they were then treated with Hb 3 μM alone or with horse spleen apoferritin 2 mg/ml ***p < 0.001 versus DFO Pre-Hb condition for B and C.

Fig. 3.

Deferoxamine pretreatment increases neuronal vulnerability to hemoglobin. Photomicrographs of neuron-glia cultures under phase-contrast optics treated as follows: A) Control culture subjected to washes and medium exchanges only. Neuron soma are phase-bright and form clusters (3 clusters marked with arrows) on a monolayer of glial cells. Some clusters are spherical, so individual soma are then slightly out of focus. Multiple intact processes are apparent. B) Culture treated for 24 h with deferoxamine only; neuronal cell bodies and processes are intact. C) Culture pretreated with DMEM vehicle only for 24 h, followed by wash and then treatment with hemoglobin 3 μM alone for 24 h; processes are intact, but some soma appear swollen with poorly defined borders, consistent with a mildmoderate degree of injury. D) Culture pretreated with DFO 100 μM for 24 h, followed by washout and treatment with Hb 3 μM alone; neuronal soma and processes have degenerated to debris; glial monolayer remains intact. Scale bar = 100 μm.

To determine if gradual removal of DFO after 24-h pretreatment would alter results, its concentration was reduced by half every 17 min for five cycles. Neuronal loss after subsequent treatment with 3 μM Hb for 24 h was 82.8 ± 3.4% versus 4.2 ± 4.0% in cultures pretreated in identical fashion with DMEM only (p < 0.0001, n = 16 cultures/ condition).

Neuronal loss in DFO-pretreated cells is oxidative and iron-dependent.

Most of the cell death in cultures pretreated with DFO followed by washout and Hb exposure alone was prevented by concomitant treatment with ascorbate, the α-tocopherol analog Trolox, or DFO (Fig. 2B). Protection was also provided by horse-spleen apoferritin, which like DFO sequesters and detoxifies ferric iron (Fig. 2C).

Effect of deferoxamine on heme oxygenase-1 and ferritin expression.

Cell vulnerability to heme-mediated injury is modulated by expression of heme-oxygenase-1 (HO-1) and ferritin, both of which are rapidly induced by Hb [33,37]. Heme breakdown by HO-1 releases iron, which is potentially neurotoxic but under physiologic conditions is rapidly sequestered and detoxified by ferritin. We hypothesized that DFO would attenuate induction of HO-1 or ferritin in cultures treated with Hb. Contrary to this hypothesis, concomitant treatment with DFO had no effect on HO-1 expression after 24 h exposure to a nontoxic concentration of Hb (Fig. 4A). However, DFO did prevent ferritin upregulation (Fig. 4B). To simulate cessation of DFO therapy while hematoma breakdown continued, DFO was washed out after 24 h and cultures were treated with Hb alone for 6 h, a time point preceding neuronal death. The ferritin response to Hb was similarly attenuated by this protocol (Fig. 4C). Neither DFO nor Hb had any effect on expression of glutathione peroxidase-4 (Fig. 4D), an enzyme induced by some ferroptotic stimuli that protects neurons from heme toxicity [38].

Fig. 4.

Deferoxamine decreases ferritin expression. A) Bars represent mean band density in immunoblots (9 cultures/condition) treated for 24 h with sham media exchange, hemoglobin (Hb) 1 μM alone, deferoxamine (DFO) 100 μM alone or Hb 1 μM with DFO 100 μM, immunostained with antibody to heme oxygenase-1 (HO-1) or actin loading control. B) Cultures (11/condition) were treated as in A; immunoblots were stained with anti-ferritin. C) Mean band density of anti-ferritin stained immunoblots (15 cultures/condition) pretreated for 24 h with deferoxamine (DFO) 100 μM or vehicle (DMEM), then after washout treated with Hb 3 μM or vehicle (Sham) alone for 6 h. D) Cultures (7/condition) treated as in C, stained with anti-glutathione peroxidase-4. Bar order correlates with lane order of representative

Deferoxamine pretreatment has no effect on culture non-heme iron content.

Since DFO pretreatment markedly increased neuronal loss after Hb exposure, we tested the hypothesis that it would paradoxically increase culture iron levels. Hb exposure for 24 h produced a moderate but significant increase in non-heme iron levels. No difference was observed between DFO-pretreated and DMEM vehicle-pretreated cultures (Fig. 5).

Fig. 5.

Deferoxamine pretreatment has no effect on culture iron levels. Cultures (8/condition) were pretreated for 24 h with 100 μM deferoxamine (DFO) or vehicle (DMEM), then after washout treated with hemoglobin (Hb) 3 μM or vehicle (Sham). **p < 0.01, ***p < 0.001 versus corresponding sham condition.

Deferoxamine pretreatment protects neurons from hydrogen peroxide toxicity.

Hydrogen peroxide is toxic to neurons in this culture system by an iron-dependent mechanism that is likely mediated by lysosomal iron [28,39]. Exposure to 100–300 μM H₂O₂, concentrations previously demonstrated to selectively kill neurons [28], resulted in death of one-third and three-quarters of neurons respectively at 24 h. In contrast to its effect on Hb neurotoxicity, most neuronal death was prevented by DFO pretreatment followed by its washout prior to H₂O₂ exposure (Fig. 6).

Fig. 6.

Deferoxamine pretreatment protects neurons from hydrogen peroxide. Mean LDH release in cultures (6/condition) pretreated with deferoxamine 100 μM (DFO) or vehicle control (DMEM), and then after washout treated with indicated concentrations of hydrogen peroxide alone. **p < 0.01, ***p < 0.001 versus corresponding DMEM vehicle-pretreated condition.

4. Discussion

The study provides novel information that may be relevant to the design of DFO treatment protocols for hemorrhagic stroke. Despite the well-documented benefit of concomitant DFO against Hb or heme toxicity [3,4], DFO pretreatment alone or with Hb deconditioned neurons and increased their vulnerability to Hb when DFO was absent. This effect was associated with reduced expression of the iron-sequestering protein ferritin, and was attenuated or prevented by exogenous apoferritin, antioxidants, or continuous DFO treatment. In contrast, DFO pretreatment had the opposite effect on hydrogen peroxide neurotoxicity, which is mediated by its interaction with endogenous iron [28,39]. These results suggest that the effect of DFO on iron-mediated neuronal injury is more complicated than previously appreciated, and varies significantly with the timing of its administration and the nature of the oxidative insult. An abrupt cessation of DFO therapy prior to hematoma resolution may negate any benefit or even worsen outcome. Continuous DFO therapy or a combinatorial approach with other antioxidants or chelators may be necessary.

The vulnerability of neural and other cell populations to Hb or heme is an inverse function of their cytosolic ferritin content [23,26,32]. Mammalian ferritin is a 24-mer heteropolymer that self-assembles into a nanocage with a central mineral core that has a storage capacity of several thousand iron atoms [40]. Its expression is regulated primarily by iron regulatory protein-1 (IRP-1) and IRP-2, which bind to the iron responsive element in the 5’ untranslated region of H- and L-ferritin mRNA and inhibit translation. IRP-2 is predominant in the mouse CNS and in this primary cell culture system [32]. In the presence of iron, IRP-2 undergoes proteasomal degradation, enabling translation and rapidly increasing cell ferritin content [41]. Release of IRP-2 binding is inhibited by DFO [42]. The present results suggest that this inhibition persists for at least several hours after DFO washout, and in this model is sufficient to increase neuronal loss produced by Hb per se by three-fold or greater. Attenuation of this effect by exogenous apoferritin is consistent with a primary role of this protein in the neuronal defense against Hb.

DFO pretreatment had no significant effect on cell nonheme iron levels and did not significantly alter the iron increase in response to Hb treatment after DFO washout. The assay used in this experiment measures iron chelatable by ferrozine, but does not distinguish cellular low molecular weight iron and iron released from ferritin or other binding sites by TCA lysis and boiling. Only the former is depleted by DFO [10]. These results indicate that baseline levels of low molecular weight iron are likely very low in these primary cell cultures and are below the detection limit of the ferrozine assay. The similar increase in cell iron after Hb treatment in DFO-pretreated and control cultures suggests that Hb uptake, heme breakdown and iron export are not altered by DFO pretreatment.

In contrast to the deleterious effect of DFO pretreatment on Hb neurotoxicity, it robustly protected neurons from hydrogen peroxide toxicity. Like Hb toxicity, hydrogen peroxide toxicity is iron-dependent and attenuated by concomitant treatment with DFO [39]. The toxic interaction of exogenous hydrogen peroxide and endogenous iron occurs predominantly in lysosomes, generating highly reactive hydroxyl radicals via the Fenton reaction [43,44]. Incubation of cells with DFO results in its accumulation and storage selectively in lysosomes [20,22,45]. The protective effect of DFO pretreatment on hydrogen peroxide toxicity is likely a consequence of lysosomal accumulation of both compounds. The opposite effect of DFO pretreatment on Hb neurotoxicity suggests that its iron participates in deleterious free radical reactions in a different cell compartment that does not accumulate and store DFO.

This cell culture model quantifies the neurotoxicity of extracellular Hb, and its relevance to ICH is dependent on the presence of extracellular Hb in the hematoma and adjacent tissue. Extracellular Hb has been detected by magnetic resonance imaging in clinical intracerebral hematomas as early as 24 h after hemorrhage, and persists for weeks thereafter [46,47]. However, direct quantification of the concentrations of extracellular hemoglobin and its breakdown products in the days after hemorrhage is lacking in both experimental models and clinical samples. After intraventricular hemorrhage in preterm rabbit pups, low micromolar concentrations of oxyhemoglobin and methemoglobin were measured in CSF at 24 h and increased over the following 48 h [48], which is similar to in vitro observations using this culture system [49]. Quantification of the time course of hemolysis and hemoglobin content in an intracerebral hematoma and surrounding tissue seems a worthy topic for future investigation.

DFO preconditioning has been protective in ischemic stroke models, due at least in part to prolyl hydroxylase inhibition and induction of HIF-1 regulated genes [50]. The present results indicate an opposite deconditioning effect on heme or iron-mediated neuronal injury. Cellular deconditioning by DFO may negate any benefit of a short course of therapy after hemorrhagic stroke, as cells with attenuated endogenous defenses are subsequently exposed to toxic Hb concentrations without DFO. In animal studies, DFO was usually continued until completion of the experiment or for seven days, a time point at which hematoma resorption is nearing completion in rodent models [5,6,51]. Okauchi et al. reported that continuing treatment for 7–14 days was necessary to reduce long term atrophy and neurological deficits in aged rats, although DFO toxicity was evident with a 14 day course [17]. The larger hematoma volumes common to clinical ICH may require an even longer duration of therapy, which is unlikely to be tolerated with systemic administration. Strategies that selectively target DFO to the CNS may be preferable. Intranasal DFO, administered either as an aqueous solution or encapsulated within mucoadhesive microparticles, produces micromolar brain and CSF concentrations with minimal systemic absorption [30,52]; the former method improved outcome in a rat middle cerebral artery occlusion model. In a mouse blood injection SAH model, DFO was more protective when administered via intracerebroventricular versus intraperitoneal injection [6]. Direct administration of DFO into a hematoma is likely feasible in the setting of a modern neurocritical care unit and could be combined with minimally invasive surgery to reduce clot burden. Further investigation of brain-targeted, long course DFO therapy in hemorrhagic stroke models is warranted.