Secretion of brain-derived neurotrophic factor from PC12 cells in response to oxidative stress requires autocrine dopamine signaling

Authors


Address correspondence and reprint requests to Dr David M. Katz, Department of Neurosciences, Case Western Reserve University School of Medicine, 10900 Euclid Avenue, Cleveland, OH 44106, USA.
E-mail: dmk4@po.cwru.edu

Abstract

Expression of brain-derived neurotrophic factor (BDNF) is sensitive to changes in oxygen availability, suggesting that BDNF may be involved in adaptive responses to oxidative stress. However, it is unknown whether or not oxidative stress actually increases availability of BDNF by stimulating BDNF secretion. To approach this issue we examined BDNF release from PC12 cells, a well-established model of neurosecretion, in response to hypoxic stimuli. BDNF secretion from neuronally differentiated PC12 cells was strongly stimulated by exposure to intermittent hypoxia (IH). This response was inhibited by N-acetyl-l-cysteine, a potent scavenger of reactive oxygen species (ROS) and mimicked by exogenous ROS. IH-induced BDNF release requires activation of tetrodotoxin sensitive Na+ channels and Ca2+ influx through N- and L-type channels, as well as mobilization of internal Ca2+ stores. These results demonstrate that oxidative stress can stimulate BDNF release and that underlying mechanisms are similar to those previously described for activity-dependent BDNF secretion from neurons. Surprisingly, we also found that IH-induced secretion of BDNF was blocked by dopamine D2 receptor antagonists or by inhibition of dopamine synthesis with α-methyl-p-tyrosine. These data indicate that oxidative stress can stimulate BDNF release through an autocrine or paracrine loop that requires dopamine receptor activation.

Abbreviations used
AMPT

α-methyl-p-tyrosine

2-APB

2-aminodioxydiphenylborate

BDNF

brain-derived neurotrophic factor

H2O2

hydrogen peroxide

IH

intermittent hypoxia

IP3

inositol 1,4,5-triphosphate

NAC

N-acetyl-l-cysteine

NGF

nerve growth factor

NPG

nodose-petrosal ganglia

PC12

rat pheochromocytoma cells

ROS

reactive oxygen species

SH

sustained hypoxia

TTX

tetrodotoxin

Brain-derived neurotrophic factor (BDNF) can protect neurons against oxidative damage resulting from diverse neuropathologic insults (Mattson et al. 2002). For example, treatment with exogenous BDNF can markedly attenuate the loss of dopaminergic substantia nigra neurons resulting from oxyradical damage following exposure to 6-hydroxydopamine (Altar et al. 1994) or 1-methyl-4-phenylpyridinium (Frim et al. 1994; Tsukahara et al. 1995), the active metabolite of 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine. In addition, attenuation of BDNF signaling in the intact brain by overexpression of a truncated form of the BDNF receptor, TrkB, increases susceptibility to neuronal damage after focal cerebral ischemia, suggesting that endogenous TrkB ligands are also neuroprotective (Saarelainen et al. 2000). Consistent with this hypothesis, BDNF expression increases in ischemic brain tissue (Lindvall et al. 1992; Mattson and Scheff 1994; Hughes et al. 1999; Walton et al. 1999). It is unknown, however, whether or not oxidative stress increases BDNF availability by stimulating BDNF release.

To address this issue, we defined BDNF release from transfected rat pheochromocytoma (PC12) cells, an extremely well-characterized model of neurotrophin (Heymach et al. 1996; Moller et al. 1998; Sadakata et al. 2004) and catecholamine (Greene and Rein 1977; Schubert and Klier 1977; Ritchie 1979; Pozzan et al. 1984) secretion. Following transfection of BDNF plasmids into PC12 cells, BDNF is packaged into dense core secretory vesicles (Moller et al. 1998) and can be released in response to potassium depolarization (Kruttgen et al. 1998). Thus, BDNF transfected PC12 cells have proven to be a useful model for defining mechanisms of BDNF processing and secretion (Heymach and Shooter 1995; Goodman et al. 1996; Kruttgen et al. 1998). We focused in particular on responses to intermittent hypoxia (IH), a form of oxidative stress used to model the transient fluctuations in oxygen availability that characterize obstructive sleep apnea in humans (Peng et al. 2003; Row et al. 2003; Xu et al. 2004). Our findings demonstrate that IH strongly stimulates BDNF release through mechanisms that closely resemble activity-dependent BDNF release in neurons, including mobilization of intracellular calcium through inositol 1,4,5-triphosphate (IP3)- and ryanodine-sensitive channels. Because PC12 cells synthesize and release dopamine and express D2 autoreceptors (Courtney et al. 1991; Pothos et al. 1998), we were also able to examine the possibility that IH-induced release of BDNF may be regulated by dopaminergic mechanisms. We found that blockade of D2 receptors or inhibition of dopamine synthesis largely blocked IH-induced release of BDNF, indicating that this model of oxidative stress triggers BDNF secretion in PC12 cells by stimulating an autocrine/paracrine loop that requires activation of D2 receptors by endogenous dopamine.

Materials and methods

Reagents

Tetrodotoxin (TTX), ω-Conotoxin GVIA, N-acetyl-l-cysteine (NAC), caffeine, thapsigargin, sulpiride, butaclamol, pergolide, α-methyl-p-tyrosine (AMPT) and nerve growth factor (NGF) were purchased from Sigma (St. Louis, MO, USA). Nimodipine, dantrolene and 2-aminodioxydiphenylborate (2-APB) were from Calbiochem (San Diego, CA, USA). Hydrogen peroxide (H2O2) was from Fisher (Fair Lawn, NJ, USA).

PC12 cell cultures

PC12 cells (original clone from Dr L. Green) were cultured in a humidified chamber maintained with 10% CO2 and 90% air at 37°C. Cells were grown in Dulbecco's modified Eagle's medium (Invitrogen, Carlsbad, CA, USA) supplemented with 10% horse serum, 5% fetal calf serum, 100 µg/mL streptomycin and 100 U/mL penicillin G. The medium was changed every 2 days and experiments were performed using cells maintained between passages 4 and 8. All experiments were performed in Dulbecco's modified Eagle's medium with 0.2% bovine serum albumin. To determine whether or not the capacity of PC12 cells to release BDNF is influenced by expression of neuronal properties, we compared release in native cells and cells exposed to NGF (10 ng/mL) to induce neuronal phenotypic differentiation. In the experiments involving treatment with drugs, cells were pre-incubated for 30 min with either drug or vehicle.

Primary neuronal cultures

All procedures on experimental animals were performed following national and international ethics guidelines and approved by the Institutional Animal Care and Use Committee at Case Western Reserve University. Pregnant rats (Sprague Dawley strain; Harlan, Indianapolis, IN, USA) were killed with ethyl ether at 19 days of gestation. The uterine horns were removed, placed into phosphate-buffered saline (pH 7.4) with 1% glucose. The fetuses were excised and killed by decapitation. The nodose-petrosal ganglia (NPG) were removed aseptically, digested in 0.1% trypsin (Worthington Biochemical, Lakewood, NJ, USA) with 0.01% deoxyribonuclease I (Sigma) dissolved in Ca2+- and Mg2+-free Hanks' balanced salt solution (Mediatech, Herndon, VA, USA) for 15 min at 37°C and then rinsed in 0.1% soybean trypsin inhibitor (Worthington Biomedical) dissolved in Ca2+- and Mg2+-containing Dulbecco's phosphate-buffered salt solution (Mediatech). The ganglia were further digested in 0.2% collagenase (Sigma) dissolved in Ca2+- and Mg2+-free HBSS for 30 min at 37°C and then rinsed in 2% bovine serum albumin (Sigma) dissolved in Neurobasal medium (Invitrogen), washed in culture medium and triturated through siliconized, fire-polished Pasteur pipettes. Dissociated NPG neurons were plated into 96-well flat-bottom ELISA plates (Nalge Nunc International, Naperville, IL, USA) coated with poly d-lysine (Sigma) and BDNF monoclonal antibody (from BDNF ELISA kit; Promega, Madison, Wisconsin, USA) at a density of 1.5 NPG per well. Cultures were grown for 1 day in Neurobasal medium supplemented with 2% B-27 serum-free supplement (Invitrogen), 0.5 mm l-glutamine (Invitrogen), and 1% penicillin–streptomycin–neomycin antibiotic mixture (Life Technologies, Gaithersburg, MD, USA) before the onset of hypoxic treatment.

Hypoxic exposure

For exposure of cells to IH, cell cultures were placed in a humidified Lucite chamber (length = 12 in., width = 12 in., and height =7 in.) in which gas flow (2.4 L/min) into the chamber and the duration of gas exposure were regulated by timed solenoid valves. The IH protocol consisted of alternating exposure to normoxia (21% O2, 2 min) and hypoxia (1% O2, 30 s) for 10–120 cycles at 37°C. We have previously defined the oxygen tension profile in the chamber and culture medium of cells exposed to this protocol, using an Beckman LB2 oxygen analyzer (Beckman, Fullterton, CA, USA) and an oxygen electrode (Lazar Research Laboratories Inc., Los Angeles, CA, USA), respectively (Yuan et al. 2004). Typically, the ambient pO2 of the chamber and the medium is 146 and 70 mmHg, respectively. During hypoxic challenge, a period of 45 s is required for the gas mixture in the chamber to reach 1% O2; after switching back to the normoxic gas mixture, a period of 50 s is required to return to 21% O2. During each cycle of hypoxic exposure, the pO2 of the chamber and the medium drops to 20 and 50 mmHg, respectively. During IH exposure, the pH of the culture medium remains unchanged at 7.4. Sustained hypoxia (SH) was induced for 1 h, which is equivalent to the cumulative hypoxic exposure during 120 cycles of IH (30 s/episode × 120 cycles).

Transient transfection

The cDNA for rat BDNF (generously provided by Dr EM Shooter, Stanford University) was subcloned into pBJ-5, an SRα-based expression plasmid, as described by Heymach and Shooter (1995). One day before transfection, PC12 cells were removed from the culture flask by trypsinization and equally distributed into 48-well plates coated with poly d-lysine and laminin in culture medium without antibiotics. Transient transfections were performed using Lipofectamine and Plus Reagent (Invitrogen) as indicated by the manufacturer and 0.1 µg of DNA, 1 µL Plus Reagent, and 0.5 µL of Lipofectamine were used to transfect 2 × 104 cells/well. The cells were allowed to grow for an additional 44 h post-transfection.

Secretion studies

Prior to the onset of intermittent hypoxia the cells were placed in serum-free Dulbecco's modified Eagle's medium containing 0.2% bovine serum albumin for 24 h. Exposure to IH began 20 h before the end of the experiment, and exposure of sister cultures to SH occurred during either the first or last hour. All groups were maintained in culture for the same period of time, and the culture medium was collected for BDNF analysis at the end of the culture period. Conditioned medium was cleared of cellular debris by centrifugation. In some experiments the cells were washed with cold phosphate-buffered saline and lysed in phosphate-buffered saline buffer containing 150 mm NaCl, 1% Nonidet P-40, 1% sodium deoxycholate, 1 mm EDTA (Fisher Scientific), and protease inhibitors (Sigma). The BDNF content of conditioned media and cell lysates was measured with a commercially available ELISA kit according to the manufacturer's instructions (BDNF Emax Immunoassay System; Promega). Data were collected from at least three independent experiments for each experimental condition.

Brain-derived neurotrophic factor enzyme-linked immunosorbent assay in situ

BDNF ELISA in situ was performed as previously described by our laboratory (Balkowiec and Katz 2000). Briefly, 96-well ELISA plates were UV-sterilized for 30 min, coated with poly d-lysine at room temperature for 30 min followed by anti-BDNF monoclonal antibody at 4°C for 16.5 h. Next, plates were washed and blocked, followed by two 1 h incubations with culture medium to remove any residual washing solution. NPG neurons were prepared as described above, plated in anti-BDNF-coated wells and grown for 1 day before exposure to IH. BDNF samples used to generate the standard curves were added to the same plate on the day of plating. At the end of the exposure period the plates were washed extensively to remove all cells and cell debris. ELISA was then performed according to the manufacturer's protocol.

Analysis of cell viability

PC12 cell viability following hypoxic exposure was assessed using the CellTiter 96 Assay (Promega) and expressed as the percentage change in the number of viable cells compared to normoxic controls. For NPG neurons, cell viability was assessed by cell counts. The number of neurons in each culture was estimated by counting all cells from one longitudinal strip through the center of each well (∼10% of the total area).

Dopamine analysis

Cells were isolated by centrifugation (450 g for 5 min), placed in 0.1 N HClO4, and stored at −80°C prior to analysis by HPLC with electrochemical detection. Briefly, samples were separated on a 3 µm C18 column (150 mm × 2, ESA, Inc., Chelmsford, MA, USA) using a mobile phase consisting of 75 mm NaH2PO4, 0.8 m octanesulfonic acid, 0.1 m Na2EDTA, 0.01% triethylamine (v/v) and 10% acetonitrile (v/v), pH 3.0. The mobile phase was pumped through the system at 0.3 mL/min using an ESA LC-10AD pump. Dopamine was detected and quantified using an ESA Coulochem II detector, an ESA Model 5010 conditioning cell, and an ESA Model 5014B microdialysis cell. The settings for detection were E1 =−0.1 V, E2 = +0.210 V, guard cell = +0.350 V. The limits of detection for dopamine were in the femtomole range.

Statistical analysis

Data were obtained from at least three independent experiments and expressed as means ± SEM. Statistical significance was evaluated by a Student's paired t-test or one-way anova followed by post hoc Tukey test for repeated measures. p-values less than 0.05 were considered significant.

Results

Stimulation of brain-derived neurotrophic factor release by intermittent hypoxia in PC12 cells and primary sensory neurons

To determine whether or not hypoxic stimuli evoke BDNF release, PC12 cells were transiently transfected with an expression plasmid encoding the full-length BDNF gene (Heymach and Shooter 1995) and exposed to either normoxia, sustained hypoxia (1 h) or intermittent hypoxia (120 cycles over 20 h) as described in Materials and Methods. IH of 120 cycles was chosen because this protocol has previously been demonstrated to potently increase c-fos promoter activity and induction of c-fos mRNA in PC12 cells (Yuan et al. 2004). Cells transfected with the empty pBJ5 vector were used as a negative control for the specificity of BDNF immunoreactivity. Treatment with 50 mm KCl for 1 h, a stimulus known to trigger BDNF release from PC12 cells (Heymach and Shooter 1995), was used as a positive control.

As shown in Fig. 1, no BDNF was detectable in the medium from cultures of mock transfected cells. However, BDNF was present in the medium of BDNF transfected cells grown in normoxia, indicating that some secretion occurs under control conditions (122.3 ± 7.2 pg/mL in differentiated cells and 75.7 ± 4.0 pg/mL in non-differentiated cells). Treatment of cells grown in normoxia with 50 mm KCl resulted in a greater than two-fold increase in BDNF levels, consistent with previous reports demonstrating depolarization mediated BDNF release from PC12 cells (Kruttgen et al. 1998) as well as transfected rat cortical neurons (Wu et al. 2004).

Figure 1.

Brain-derived neurotrophic factor (BDNF) release from PC12 cells in response to normoxia, sustained hypoxia (SH), intermittent hypoxia (IH) or potassium chloride (KCl) depolarization. Cells were transiently transfected with BDNF/pBJ5 expression plasmid or the empty pBJ5 vector (Mock) and then differentiated with 10 ng/mL nerve growth factor (NGF) as described in Materials and Methods. Each cycle of IH consisted of 30 s of hypoxia (pO2 of 50 mmHg) repeated every 10 min and cells were exposed to a total of 120 cycles over a period of 20 h. SH (pO2 of 50 mmHg) was delivered for 1 h, corresponding to the last hour of IH exposure. The cumulative exposure to hypoxia was the same in the IH and SH groups. KCl-treated cells received 50 mm KCl during the last hour of the experiment. Each bar represents the mean ± SEM of at least three independent experiments. *p < 0.05, **p < 0.01 vs. normoxia (anova followed by Tukey's post analysis). ##p < 0.01 (paired t-test).

Exposure of transfected cells to 120 cycles of IH increased BDNF secretion more than two-fold to levels that were comparable to those observed in KCl-treated cultures (Fig. 1; p < 0.01). On the other hand, exposure to SH for 1 h, which is equivalent to the cumulative hypoxia exposure during 120 episodes of IH, induced only a modest increase in BDNF release of approximately 50% (p < 0.05 vs. nomoxia; p < 0.01 vs. IH 120 cycles). Equivalent levels of BDNF release were observed whether cells were exposed to SH in the first or last hour of the culture period (data not shown). Although there was no difference in the magnitude of hypoxia-induced release between differentiated and non-differentiated cells, NGF treatment significantly increased the absolute level of BDNF expression (Fig. 1; p < 0.01 vs. non-differentiated cells). Therefore, and because IH was significantly more effective at inducing BDNF release than SH, the studies below focus on IH-induced BDNF secretion in NGF-differentiated cells.

To rule out the possibility that BDNF was released from cells exposed to hypoxic stimuli as a result of cell damage, cell viability was assessed by quantifying the number of living cells in normoxic and hypoxic cultures (see Materials and Methods for details). No differences were observed among control, SH (103 ± 7% of control) and IH (97 ± 3% of control) groups, indicating that cell viability was unaffected by exposure to either hypoxia protocol.

To define the time course of BDNF release in response to IH, BDNF levels in PC12 supernatants were measured after 10, 30, 60, 90, 120 cycles of IH. As shown in Fig. 2(a), release of BDNF was significantly greater than control values after 90 cycles of IH (p < 0.01).

Figure 2.

Regulation of brain-derived neurotrophic factor (BDNF) expression and release in response to intermittent hypoxia (IH). Cells were exposed to 10, 30, 60, 90, or 120 cycles of IH, respectively, and the culture medium (a) and cell lysate (b) assayed for BDNF content. The total BDNF content (medium plus lysate) is shown in (c). BDNF release as percentage of total BDNF content is shown in (d). Each bar represents the mean ± SEM of at least three independent experiments. *p < 0.05, **p < 0.01 vs. normoxia (anova followed by Tukey's post analysis).

To determine whether or not hypoxia also stimulates BDNF release from neurons, BDNF content was assayed in cultures of fetal NPG primary sensory neurons following exposure to IH. NPG neurons were chosen because they express high levels of BDNF mRNA and protein (Brady et al. 1999; Balkowiec and Katz 2000) and release BDNF in response to physiologic patterns of electrical stimulation (Balkowiec and Katz 2000). Exposure of NPG cultures to 120 cycles of IH evoked a significant increase in BDNF release compared to normoxic controls (normoxia, 37.7 ± 2.9 pg/mL vs. IH, 56.4 ± 1.6 pg/mL; p < 0.01) with no change in neuronal survival (number of neurons per well: normoxia, 1393 ± 15; IH, 1403 ± 37; n = 6).

Intermittent hypoxia increases brain-derived neurotrophic factor expression and up-regulates the release process

The increases in BDNF release that we observed in response to SH and IH could reflect the net effect of several interrelated processes, including up-regulation of the release process itself as well as increased BDNF expression (i.e. synthesis and/or decreased degradation). To distinguish among these possibilities we measured cellular BDNF content in PC12 cell lysates from cultures exposed to normoxia and different cycles of IH. Cellular content, plus the amount of BDNF released, was used as a measure of total BDNF content. As shown in Fig. 2(b), cellular BDNF content was not significantly different from controls in any of the IH groups. However, total BDNF content (lysate plus supernatant) was significantly greater than control after 120 cycles of IH (p < 0.05; Fig. 2c). Moreover, the percentage of total BDNF content that was released was significantly greater following 120 cycles of IH compared to controls. As shown in Fig. 2(d), approximately 12% of the total BDNF content is released under normoxic conditions, whereas approximately 18% is released following 120 cycles of IH, an increase of 50%.

Intermittent hypoxia-induced brain-derived neurotrophic factor release is associated with oxidative stress

Recent studies in vivo have suggested that increased generation of reactive oxygen species (ROS) contributes to IH-induced neural plasticity (Peng et al. 2003). Moreover, we have previously shown that the IH protocol used in the present study results in ROS generation in vitro (Yuan et al. 2004). Therefore, to determine whether or not oxidative stress is involved in IH-induced BDNF release, cells were treated with NAC (300 µm), a ROS scavenger. As shown in Fig. 3(a), NAC potently inhibited IH-induced BDNF release (p < 0.01).

Figure 3.

Hypoxia-induced brain-derived neurotrophic factor (BDNF) release is associated with oxidative stress. (a) Thirty minutes prior to the onset of intermittent hypoxia (IH, 120 cycles), cells were pretreated with 300 µmN-acetyl-l-cysteine (NAC), a scavenger of reactive oxygen species. (b) Dose–response of hydrogen peroxide (H2O2) on BDNF release during 10 min incubation. (c) Time course of H2O2 (10 µm) stimulated BDNF release. Each bar represents the mean ± SEM of at least three independent experiments. **p < 0.01 vs. control (paired t-test).

To determine whether or not ROS can directly stimulate BDNF release or, alternatively, are simply permissive for BDNF release in response to IH, cells were exposed to different concentrations of H2O2, a ROS generator, for 1–60 min under normoxic conditions. These experiments demonstrated that brief exposure (10 min) to 100 µm H2O2, or prolonged exposure (60 min) to 10 µm H2O2, potently stimulates BDNF release (Figs 3b and c). Cell viability was unchanged between control and H2O2-stimulated groups (10 min, 100 µm H2O2 = 96 ± 1% of control; 60 min with 10 µm H2O2 = 94 ± 2% of control).

Intermittent hypoxia-induced brain-derived neurotrophic factor release requires activation of voltage-gated sodium channels

PC12 cells respond to reduced oxygen availability with membrane depolarization, leading to Na+ and Ca2+ influx through voltage-gated channels in the plasma membrane (Seta et al. 2002). To examine the role of Na+ flux in IH-induced BDNF release, cells were exposed to normoxia or IH in the absence or presence of TTX (1 µm), a specific blocker of voltage sensitive Na+ channels. TTX completely inhibited BDNF secretion induced by IH (Fig. 4), indicating that activation of TTX sensitive sodium channels is required for release.

Figure 4.

Hypoxia-induced brain-derived neurotrophic factor (BDNF) release requires activation of voltage-gated sodium channels. Thirty minutes prior to the onset of intermittent hypoxia (IH, 120 cycles), cells were pretreated with 1 µm tetrodotoxin (TTX), a selective blocker of voltage-gated sodium channels. Each bar represents the mean ± SEM of at least three independent experiments. **p < 0.01 vs. control (paired t-test).

Intermittent hypoxia-induced brain-derived neurotrophic factor release requires Ca2+ influx and Ca2+ mobilization from internal stores

Oxidative stress can directly influence Ca2+ homeostasis by elevating intracellular Ca2+ concentration (Ermak and Davies 2002). Moreover, we previously demonstrated that neuronal secretion of BDNF requires both influx of extracellular Ca2+ and Ca2+ mobilization from internal stores (Balkowiec and Katz 2002). To determine whether IH-induced BDNF release is regulated by similar mechanisms, initial experiments examined the effect of Cd2+, a non-selective blocker of voltage-gated Ca2+ channels, on BDNF release in response to IH (Fig. 5a). Pretreatment with 200 µm Cd2+ completely blocked IH-induced BDNF release, indicating an absolute requirement for Ca2+ influx. In addition, release was significantly blocked by pre-incubation with either 1 µmω-conotoxin GIVA, a selective N-type Ca2+ channel blocker or 1 µm nimodipine, a selective L-type Ca2+ channel blocker (Figs 5b and c). The effects of N- and L-type channel blockade were approximately additive, with each treatment reducing IH-induced release by 48% and 63%, respectively.

Figure 5.

Hypoxia-induced brain-derived neurotrophic factor (BDNF) release requires Ca2+ influx. Thirty minutes prior to the onset of exposure to intermittent hypoxia (IH, 120 cycles), cells were pretreated with (a) 200 µm CdCl2 (Cadmium), a non-selective voltage-gated Ca2+ channel blocker, (b) 1 µmω-conotoxin GIVA (CgTX), a selective N-type channel blocker, or (c) 1 µm nimodipine, a selective L-type channel blocker. Each bar represents the mean ± SEM of at least three independent experiments. *p < 0.05, **p < 0.01 vs. control (paired t-test).

To define the role of Ca2+ mobilization from internal stores, cells were pretreated with either (i) dantrolene, an antagonist of ryanodine receptors, (ii) 2-APB, an antagonist of IP3 receptors, (iii) caffeine, a ryanodine receptor agonist or (iv) thapsigargin, an antagonist of the sarco/endoplasmic reticulum Ca2+-ATPase, prior to IH exposure. Pretreatment with dantrolene (50 µm) significantly inhibited IH-induced BDNF release (70%; Fig. 6a), indicating a role for Ca2+ release through ryanodine receptors. Consistent with this finding, pre-incubation of cells with caffeine (60 µm), which results in depletion of internal stores through activation of ryanodine receptors, reduced IH-evoked BDNF release to a comparable degree (50%; Fig. 6c). Pre-incubation with 2-APB (20 µm) also reduced IH-induced BDNF release (70%; Fig. 6b), indicating a role for Ca2+ release from IP3-sensitive stores. Treatment with thapsigargin (1 µm), which depletes the endoplasmic reticulum Ca2+ stores, also inhibited IH-induced BDNF release (54%; Fig. 6d). Together, these data indicate that IH-induced BDNF release requires influx of extracellular Ca2+ as well as mobilization of internal Ca2+ stores through ryanodine and IP3 receptors.

Figure 6.

Hypoxia-induced brain-derived neurotrophic factor (BDNF) release requires Ca2+ mobilization from internal stores. Thirty minutes prior to the onset of exposure to intermittent hypoxia (IH, 120 cycles), cells were pretreated with (a) 50 µm dantrolene, a ryanodine receptor antagonist or (b) 20 µm 2-aminodioxydiphenylborate (2-APB), an inositol 1,4,5-triphosphate receptor antagonist. Additional cultures were exposed to (c) 60 µm caffeine, a ryanodine receptor agonist, or (d) 1 µm thapsigargin, an inhibitor of the sarco/endoplasmic reticulum Ca2+ ATPase, for 30 min, after which the drugs were removed and replaced with fresh medium. Each bar represents the mean ± SEM of at least three independent experiments. In (a) and (b), *p < 0.05, **p < 0.01 vs. control; in (c) and (d), **p < 0.01 vs. vehicle (paired t-test).

Intermittent hypoxia-induced brain-derived neurotrophic factor release requires activation of dopaminergic receptors by endogenous dopamine

Dopamine is an abundant secretory product of PC12 cells and its release can be triggered by some of the same stimuli that trigger BDNF release, including membrane depolarization (Schubert and Klier 1977) and hypoxia (Kumar et al. 1998). Moreover, PC12 cells express the D2 class of dopamine receptors which negatively regulate dopamine release, as in brain neurons (Courtney et al. 1991; Pothos et al. 1998). We therefore asked whether or not IH-induced BDNF release might be influenced by an effect of oxidative stress on secretion of endogenous dopamine and subsequent activation of D2 receptors. To address this question we first examined whether or not inhibition of dopamine synthesis would alter IH-induced BDNF release. In these experiments, cells were treated with AMPT (1 mm), an inhibitor of tyrosine hydroxylase, the rate-limiting enzyme in dopamine synthesis, for 24 h prior to exposure to IH. AMPT pretreatment decreased cellular dopamine content by 92% (not shown) and significantly reduced IH-induced BDNF release by 58%(Fig. 7a). In a separate set of experiments, we examined whether or not pharmacologic blockade of dopamine receptors would similarly inhibit IH-induced BDNF release by treating cells with the dopamine antagonists sulpiride (1 µm) and butaclamol (1 µm) during the period of IH exposure. Because these drugs exhibit relative selectivity for the D2 subclass of dopamine receptors (D2, D3, D4) over D1 receptors (D1, D5) (De Lean et al. 1982) we will refer to them as ‘D2 antagonists’. Sulpiride and butaclamol markedly inhibited IH-induced BDNF release (sulpiride, 78%; butaclamol; 58%) compared to vehicle controls (Fig. 7b), indicating that D2 activation is required for most IH-induced BDNF release. Neither drug affected basal release under normoxic conditions. Finally, to determine whether or not dopamine receptor activation is sufficient to induce BDNF release, cells were exposed to the dopamine agonist pergolide (10 µm; relative affinity D3 > D2 > D1) (De Lean et al. 1982; Millan et al. 2002; Gerlach et al. 2003) during 20 h of normoxia, a time period equivalent to 120 cycles of IH. Treatment with pergolide resulted in a significant 60% increase in BDNF release compared to vehicle-treated controls (control, 132 ± 22 pg/mL; pergolide, 211 ± 27 pg/mL; p < 0.05).

Figure 7.

Hypoxia-induced brain-derived neurotrophic factor (BDNF) release requires activation of D2 receptors by endogenous dopamine. (a) Twenty-four hours prior to the onset of intermittent hypoxia (IH, 120 cycles), cells were pretreated with 1 mmα-methyl-p-tyrosine (AMPT), an inhibitor of tyrosine hydroxylase, the rate-limiting enzyme in dopamine synthesis, to inhibit dopamine synthesis. (b) Thirty minutes prior to the onset of IH (120 cycles), cells were pretreated with 1 µm sulpiride or 1 µm butaclamol, antagonists with relative selectivity for D2 receptors. Each bar represents the mean ± SEM of at least three independent experiments. **p < 0.01 vs. control (paired t-test).

Discussion

The present findings demonstrate that BDNF secretion from neuronally differentiated PC12 cells is markedly increased by exposure to hypoxic stimuli. Thus, our data support the hypothesis that oxidative stress can increase BDNF availability by stimulating BDNF release. Mechanisms underlying this release, including the requirement for (i) sodium influx through TTX sensitive channels, (ii) Ca2+ influx through voltage-gated channels and (iii) Ca2+ release from IP3- and ryanodine-sensitive stores, closely resemble those previously described for activity-dependent BDNF release from neurons (Balkowiec and Katz 2000, 2002). However, an unexpected finding was that most of the BDNF release induced by IH was blocked by dopamine receptor antagonists or inhibition of dopamine synthesis. These data indicate that the effect of IH on BDNF release is largely indirect and requires autocrine or paracrine signaling by endogenous dopamine.

Role of voltage-gated sodium and Ca2+ channels

The dependence of IH-induced BDNF secretion on voltage-gated sodium and Ca2+ channels is similar to that of activity-dependent neuronal BDNF release (Balkowiec and Katz 2000, 2002). These findings are consistent with the fact that acute hypoxic stimulation of PC12 cells results in membrane depolarization by inhibition of the Kv1.2 O2 sensitive potassium current (Zhu et al. 1996; Conforti et al. 2000) and with previous observations that PC12 cells express a full complement of voltage-gated Ca2+ channels (UsowicZ et al. 1990; Liu et al. 1996). Moreover, Ca2+ influx through voltage-gated Ca2+ channels has previously been shown to regulate catecholamine release from PC12 cells in response to acute hypoxic stimuli (Kumar et al. 1998; Taylor and Peers 1998; Kim et al. 2004).

Ca2+ mobilization from internal stores

In addition to influx of extracellular Ca2+, IH-induced BDNF release requires Ca2+ mobilization from internal stores through ryanodine receptors and IP3 receptors. This accords well with previous reports from our laboratory demonstrating that ryanodine receptors are required for activity-dependent BDNF release from hippocampal neurons (Balkowiec and Katz 2002). Moreover, IP3 receptors are required for hypoxia-induced catecholamine release from PC12 cells (Kim et al. 2004). The fact that regulated secretion of BDNF from PC12 cells and hippocampal neurons requires activation of both Ca2+ influx and Ca2+ mobilization from internal stores suggests a role for Ca2+-induced Ca2+ release, a mechanism by which cytoplasmic Ca2+ levels can be amplified and prolonged by Ca2+ release from internal stores (Albrecht et al. 2002). In some cell types, the effect of ROS on Ca2+ homeostasis can be attributed, in part, to Ca2+ release from internal stores (Kourie 1998). Thus, it is possible that decreased Ca2+ mobilization may contribute to the inhibition of IH-induced BDNF release that we observed following treatment of cells with NAC, a ROS scavenger.

Dopaminergic regulation of brain-derived neurotrophic factor release

Dopamine receptors, particularly the D2 subtype, have a well defined role in modulating dopamine synthesis and release in neurons (Goldstein et al. 1990) and PC12 cells (Courtney et al. 1991; Pothos et al. 1998). In addition, multiple reciprocal interactions have been found between dopamine and BDNF signaling. For example, BDNF can stimulate dopamine release from cultured mesencephalic (Blochl and Sirrenberg 1996) and striatal (Goggi et al. 2003) neurons and retinal amacrine cells (Neal et al. 2003). In addition, BDNF is required for expression of dopamine D3 receptors in the nucleus accumbens and striatum (Guillin et al. 2001; Sokoloff et al. 2002; Guillin et al. 2003) and can also up-regulate dopamine D5 receptors in cultured striatal astrocytes (Brito et al. 2004). On the other hand, BDNF expression in cultured striatal neurons (Kuppers and Beyer 2001), NT2/N cells (Fang et al. 2003) and transfected NG108-15 cells (Takeuchi et al. 2002) can be up-regulated by exogenous dopamine and D1 or D2 dopamine receptor agonists. Similarly, treatment of mice with the dopamine precursor levodopa acutely increases BDNF mRNA in mouse striatum (Okazawa et al. 1992). In NT2/N cells, BDNF up-regulation by activation of D1 receptors is associated with protection against oxygen–glucose deprivation induced cell death that is dependent on TrkB, indicating that dopamine signaling can trigger BDNF release from these cells (Fang et al. 2003). Moreover, the non-selective dopamine agonist apomorphine has been reported to stimulate BDNF secretion from embryonic ventral mesencephalic neurons (Guo et al.2002).

Our data indicate that IH is largely ineffective at stimulating BDNF release from PC12 cells when either dopamine synthesis or dopamine receptors are blocked. These data suggest that ROS generation, which is also required for IH-induced BDNF secretion, acts upstream of dopamine receptor activation. This possibility is consistent with previous observations that ROS generation can potentiate dopamine release from PC12 cells (Kim et al. 2004). Moreover, the fact that D2 antagonists, as well as AMPT, blocked IH-induced BDNF release argues against the possibility that dopamine stimulates BDNF release by acting intracellularly to promote ROS formation.

The importance of dopamine receptor activation for most IH-induced BDNF release may be related in part to the ability of dopamine to stimulate Ca2+ release from endoplasmic reticulum stores (Hernandez-Lopez et al. 2000; Takeuchi et al. 2002). Ca2+ mobilization from internal stores is a key step in regulation of neurotrophin release in general and a focus of diverse signals that can either enhance or suppress BDNF secretion (Balkowiec and Katz 2002; Canossa et al. 2002). Indeed, the present findings demonstrate that Ca2+ mobilization from the endoplasmic reticulum, as well as Ca2+ influx, are required for IH-induced BDNF release. However, increased Ca2+ release from internal stores occurs rapidly following dopamine receptor activation (Hernandez-Lopez et al. 2000; Takeuchi et al. 2002), whereas 90 or more cycles of IH are required to elicit a significant increase in BDNF release. Thus, it seems unlikely that an acute effect of dopamine receptor activation on Ca2+ mobilization alone can explain IH-induced BDNF release. Therefore, our data indicate a role for other, relatively slow, Ca2+-dependent processes mediated by dopamine. In particular, the fact that multiple cycles of IH increased the percentage of total BDNF content that is released, as well as BDNF expression, suggests that prolonged dopamine signaling up-regulates the secretory pathway used by BDNF in these cells. This could occur, for example, through transcriptional activation of genes involved in peptide packaging and/or release. This possibility is consistent with the fact that D2 agonists can activate the mitogen-associated protein kinase and protein kinase A transcriptional cascades (Yan et al. 1999; Bonci and Hopf 2005), phosphorylate the cyclic AMP response element binding protein (Yan et al. 1999) and up-regulate neuronal gene expression (Berke et al. 1998).

Nonetheless, our finding that IH-induced BDNF release was blocked by D2 antagonists seems paradoxical in view of the fact that D2 activation normally inhibits neurosecretion. However, recent studies indicate that prolonged exposure to dopamine or D2 agonists, as in our study, targets D2 receptors for internalization and degradation, whereas excitatory D1 receptors are recycled to the plasma membrane (Bartlett et al. 2005). The net result is a relative increase in dopamine mediated excitation at the expense of inhibition. Perhaps, therefore, the increases in BDNF release that we observed in response to multiple cycles of IH, or prolonged exposure to pergolide, result from a predominance of D1 mediated excitation that is normally masked by the presence of D2 receptors (Bartlett et al. 2005). Finally, the fact that BDNF can stimulate dopamine secretion (Blochl and Sirrenberg 1996; Goggi et al. 2003; Neal et al. 2003) and vice versa (present study) raises the possibility that BDNF and dopamine may mutually enhance each other's release.

Dopamine receptor blockade does not completely inhibit IH-induced BDNF release (Fig. 7b). It is possible, therefore, that the dopamine antagonists we used did not completely inhibit dopamine signaling in our cultures. Alternatively, these data may indicate that non-dopaminergic mechanisms, such as direct effects of ROS, may also be involved in IH-induced BDNF release. This possibility is supported by the fact that the BDNF release induced by short-term stimulation with H2O2 (Fig. 3) is not blocked by pretreatment of cells with AMPT (data not shown), indicating that dopamine signaling is not required for this acute effect. Perhaps therefore the small component of IH-induced BDNF release that is not blocked by dopamine antagonists results from a direct effect of ROS, possibly on Ca2+ mobilization (see above). This would be consistent with previous reports demonstrating that ROS can directly stimulate secretion of classical transmitters, such as acetylcholine, from PC12 cells (Kim et al. 2004).

Although the bulk of our studies focused on IH as a model of oxidative stress, we also found that BDNF release is weakly stimulated by exposure to 1 h of SH. Several factors may account for the fact that IH is a more potent secretagogue than SH, despite the fact that the cumulative exposure to hypoxia was the same in both paradigms. One possibility is that the repetitive reoxygenation that occurs with multiple cycles of IH, as well as the hypoxic episodes themselves, augments release by increasing the production of ROS (Yuan et al. 2004). Alternatively, or, in addition, it is possible that the onset of each hypoxic cycle, rather than the cumulative time in hypoxia, is the key stimulus for BDNF release. For example, it is well known that the peak secretory response of PC12 cells to sustained depolarization is transient, even in the presence of sustained elevations in intracellular Ca2+ (Di Virgilio et al. 1987). If this is also true for hypoxic stimuli, then the efficacy of repeated cycles of IH may in part reflect the cumulative effect of multiple peak responses at the onset of each hypoxic event.

In vivo, dopamine and BDNF are colocalized within several neuronal populations, including subsets of neurons in the substantia nigra and ventral tegmental area (Seroogy et al. 1994) and the petrosal ganglion of the glossopharyngeal nerve (Brady et al. 1999). Therefore, to the degree that our data can be extrapolated to primary dopaminergic neurons, our results suggest that oxidative stress may stimulate autocrine/paracrine signaling by dopamine, and BDNF release, in these populations. In light of evidence that BDNF can attenuate oxyradical damage (Mattson et al. 1995; Petersen et al. 2001), these findings raise the possibility that coexpression of BDNF may confer increased resistance to oxidative stress within subsets of dopaminergic neurons.

Acknowledgements

We thank Dr EM Shooter for the generous supply of BDNF expression plasmid, Dr AD Smith for performing the analyses of dopamine, and Dr MJ Zigmond for helpful suggestions on the research project and mansucript. This work was supported by US Public Health Service (NHLBI) grants to D.M.K.

Ancillary