Astrocytes Protect against Isoflurane Neurotoxicity by Buffering pro-brain–derived Neurotrophic Factor

ANESTHETIC neurotoxicity, characterized in animal models by widespread induction of neuronal cell death, disruption in synapse formation and stabilization, and impairment of neurocognitive development, appears to occur primarily during the period of active synaptogenesis.¹  A primary mechanism contributing to anesthetic neurotoxicity during this period has been attributed to the modulation of brain-derived neurotrophic factor (BDNF),^2,3  a central regulator of neurogenesis,⁴  synaptogenesis,⁵  and neurotransmission.⁶  In addition to playing a critical role in normal brain development,⁷  BDNF signaling is instrumental to learning and long-term memory consolidation^8,9  and repair of the brain and spinal cord after injury in the adult.^10,11  Whether BDNF induces prodeath or prosurvival signaling in postsynaptic neurons is determined by the relative balance of the proneurotrophin form of BDNF (proBDNF) and the proteolytically cleaved mature form. Postsynaptic signaling of mature BDNF promotes neurite formation and stabilization of existing synapses via activation of tropomyosin receptor kinase B, whereas proBDNF signaling induces postsynaptic neuronal death via activation of p75 neurotrophin receptor (p75^NTR, or low-affinity nerve growth factor receptor). Anesthetic neurotoxicity occurs via decreased proteolytic cleavage of proBDNF,²  leading to augmentation of postsynaptic proBDNF/p75^NTR binding, resulting in disruption of synaptogenesis and induction of postsynaptic neuronal death. This observation has generated concern about potential effects of volatile anesthetic exposure on brain development in neonates and the possibility of neurocognitive sequelae but may also theoretically impact synaptogenesis and neurogenesis in the normal and/or injured adult brain.

Growth, maintenance, and repair of the neuronal network, in both neonates and adults, are coordinated by resident astrocytes,^12,13  specialized glia that are the most abundant cells in the human brain. In addition to their role in neuronal housekeeping and protection, astrocytes have been shown to play a significant role in neurotransmission, such that the association between astrocyte processes and neuronal synapses has been coined the “tripartite synapse.”¹³  Astrocytes modulate synaptic transmission via glutamate uptake¹⁴  and are central to synapse formation and stabilization.^15,16  An individual astrocyte may contact up to 100,000 neurons¹⁷  serving to integrate signals within the neuronal network. Moreover, astrocytes themselves communicate with adjacent astrocytes via intercellular gap junctions to function as a coordinated syncytium,¹⁸  providing an additional astrocyte-dependent layer of neuronal regulation. More recently, astrocytes have also been shown to express p75^NTR, which can bind and internalize proBDNF, removing it from the extracellular space.¹⁹  However, the role of astrocytes in BDNF-mediated neuronal signaling has not been defined, and whether astrocytes have the capacity to functionally buffer increases in pro-BDNF and protect adjacent neurons from proBDNF/p75^NTR-mediated cell death has not been investigated. In the current study, we used primary neuronal, astrocyte, and neuronal-astrocyte co-cultures to test the hypothesis that astrocytes mitigate increases in proBDNF via astrocyte-dependent p75^NTR binding, resulting in a reduction in BDNF/p75^NTR-mediated neuronal cell death.

All experiments were performed according to the protocols approved by the Stanford University Animal Care and Use Committee (Stanford, California) and followed the National Institutes of Health guidelines for animal welfare.

Relatively pure astrocyte cultures were prepared from postnatal day 1 to 3 Swiss Webster mice.²⁰  In brief, after euthanasia, brains were removed in a sterile field and cortices were freed of meninges in ice-cold Eagle’s minimal essential medium (Gibco, USA) and incubated in 0.05% trypsin/EDTA (Life Technologies, USA) for 30 min at 37°C followed by mechanical dissociation. Neocortical cells were plated at a density of two hemispheres per 10 ml Dulbecco’s modified Eagle medium (Gibco) with 10% equine serum (ES, HyClone; GE Healthcare Life Sciences, USA), 10% fetal bovine serum (FBS, Hyclone), and 10 ng/ml epidermal growth factor (Sigma Chemicals, USA).

Relatively pure neuronal cultures were prepared from cortices of embryonic day 15 or 16 mice.²⁰  Cortices were collected in ice-cold Eagle’s minimal essential medium, digested with 0.05% trypsin/EDTA for 15 min at 37°C, mechanically dissociated, and then plated in Dulbecco’s modified Eagle’s medium containing 26 mM NaHCO₃ (Sigma Chemicals), 24 mM glucose (Sigma Chemicals), 5% FBS, and 5% ES. On day in vitro 2 (DIV 2), medium was exchanged with glial-conditioned medium, which has same composition as neuronal plating medium but lacking FBS, incubated with DIV 14 primary glial cultures for 7 days, filtered, and stored at −20°C until use. Before use, conditioned medium was supplemented with B-27 (2%; Gibco) and cytosine arabinoside (3 μM; Sigma-Aldrich, USA), the later added to inhibit glial proliferation.

For co-culture experiments, two culture conditions were used. Method 1: Neurons from a single dissection were plated on near-confluent astrocyte cultures (DIV 14).²¹  Method 2: To assess whether age-matched astrocytes also protect neurons from isoflurane toxicity, co-cultures were prepared from a single neocortical dissection in a similar manner to neuronal cultures but increasing the serum to 7.5% each of FBS and ES, adding 5 ng/ml epidermal growth factor and omitting B-27 and cytosine arabinoside. Neuronal, astrocyte, and neuronal-astrocyte co-cultures were maintained in a 37°C humidified incubator with 5% CO₂ in room air atmosphere.

Cultures were characterized by quantitating the cell types immunohistochemically using neuronal, astrocyte, and microglial markers (fig. 1, A and B). Microglial content in our astrocyte cultures ranged from 0 to 4.6% (average 1.5 ± 1.3%) and no detectable neurons. Our neuronal culture microglial content ranged from 0.4 to 1.5% (average 1.1 ± 0.7%) and less than 1% astrocytes, whereas neuronal-astrocyte co-cultures ranged from 0.5 to 3.1% (average 1.5 ± 1.0%) microglial content. The relative maturity of our astrocytes was assessed using markers for immature (vimentin) and mature (glial fibrillary acidic protein [GFAP]) astrocytes and showed a decline in the ratio of vimentin to GFAP with DIV (fig. 1, C and D).

Primary neuronal and mixed neuronal-astrocyte co-cultures at neuronal DIV 7 were exposed for 1 h to either 2% isoflurane (assessed with a Datex 245 Airway Gas Monitor; Datex Corp., USA) in premixed carrier gas (5% CO₂, 21% O₂, and balance nitrogen) or to carrier gas alone. Gas flow was maintained at a rate of 2 l/min in a sealed, humidified incubator at 37°C. Neurons were assessed for cell death 24 h after isoflurane exposure. In additional experiments, isoflurane-induced neuronal cell death was assessed in primary neuronal cultures after application of conditioned medium from neuronal-astrocyte co-cultures, in co-cultures after astrocyte-targeted knockdown of p75^NTR, and in co-cultures with and without the astrocyte–glutamate transporter inhibitor dihydrokainic acid (Tocris Bioscience, United Kingdom). Finally, neuronal cell death was assessed in neuronal, astrocyte, and neuronal-astrocyte co-cultures 24 h after application of recombinant proBDNF (1 to 100 pg/ml; R+D Systems, USA) plus protease inhibitor (diluted to 1X; G-Biosciences, USA), with and without the p75^NTR inhibitor transactivating transcriptional activator peptide 5 (TAT-Pep5, 10 μM; EMD Millipore, USA).

Neuronal viability in primary neuronal cultures was assessed after staining with Hoechst 33342 (5 μM; Sigma Chemicals) and propidium iodide (PI, 5 μM; Sigma Chemicals). PI stains dead cells, whereas Hoechst is a cell-permeant nucleic acid stain that labels nuclei of both live and dead cells. PI-positive cells were manually counted by a blinded investigator, whereas numbers of Hoechst-positive cells were calculated using an automated macro (Image J, v1.49b; National Institutes of Health, USA). PI-positive and Hoechst-positive cells were counted in three microscopic fields per well at ×200 magnification. The number of PI-positive cells was expressed as a percent of the total number of cells. In co-cultures, neurons were differentiated from astrocytes using fluorescence immunocytochemistry with antibody specific for neuronal nuclei (NeuN) (described in further detail below in “Fluorescence Immunocytochemistry”). Dead neurons in co-cultures were identified as double positive for PI and NeuN and expressed as a percent of the total number of NeuN-positive cells. Immunofluorescence was visualized with an epifluorescence microscope (Zeiss Axiovert 200M; Carl Zeiss AG, Germany) at ×200 magnification.²² 

p75^NTR messenger RNA (mRNA) and protein expression were assessed in primary astrocyte cultures by reverse transcription quantitative polymerase chain reaction (PCR) and fluorescence immunocytochemistry (below), respectively, 24 h after transfection with small-interfering RNA (siRNA) against p75^NTR (30 pmol per well, Silencer Select cat. no. 4390771; Life Technologies) or negative mismatch control (cat. no. AM4615; Life Technologies). Transfection was achieved using Lipofectamine-2000 (Life Technologies) according to the manufacturer’s instructions. Transfection with Lipofectamine-2000 reagents has been demonstrated²³  to preferentially target astrocytes in neuronal-astrocyte co-cultures. In the current study, we verified astrocyte-specific targeting in co-cultures using a reporter plasmid (30 pmol per well pDS-Red2-N1; Clontech Laboratories, USA), assessed with fluorescence immunocytochemistry 24 h after transfection. p75^NTR protein expression was then assessed in neuronal-astrocyte co-cultures 24 h after transfection with p75^NTR siRNA (30 pmol per well).

Fluorescence immunocytochemistry was performed on cell cultures in 24-well plates.²²  In brief, cultures were fixed in 4% paraformaldehyde for 30 min at room temperature and nonspecific binding was blocked with 5% normal goat serum and 0.3% Triton X-100 in PBS for 1 h. Cells were then incubated with primary antibody to the neuronal marker NeuN (1:500, cat. no. mab377; EMD Millipore), the astrocyte markers GFAP (1:500, cat. no. ab7260; Abcam, USA) and vimentin (1:500 cat. no. ab8978; Abcam), and/or the microglial marker Iba1 (1:500, cat. no. ab178680; Abcam) overnight at 4°C. Cells were washed and subsequently incubated with Alexa Fluor 488-nm-conjugated or 594-nm-conjugated secondary antibody (1:500; Invitrogen, USA) for 1 h. Cells were counterstained with 4′,6′-diamidino-2-phenylindole (0.5 μg/ml; Sigma-Aldrich), a cell-permeant nuclear dye, for total cell count.

Total RNA was isolated with TRIzol^® (Life Technologies) from cultures 24 h after transfection. Reverse transcription was performed using the TaqMan^® MicroRNA Reverse Transcription Kit for total RNA (Life Technologies).²⁴  Predesigned primer/probes for PCR were obtained from Life Technologies for mouse p75^NTR (#Mm01309638) and glyceraldehyde 3-phosphate dehydrogenase (#Mm99999915) mRNA. Real time-quantitative PCR was conducted using the TaqMan^® Assay Kit (Life Technologies).²⁴  Measurements for p75^NTR mRNA were normalized to within-sample glyceraldehyde 3-phosphate dehydrogenase (ΔCt), and comparisons were calculated as the inverse log of the ΔΔCt from the mean of the control treatment group.²⁵ 

Concentrations of proBDNF in culture media were assessed before and immediately after isoflurane exposure, with and without transfection with p75^NTR siRNA 24 h before, by enzyme-linked immunosorbent assay (ELISA) according to the manufacturer’s directions (cat. no. BEK-2217-2P; Biosensis, Australia). Absorbance at 450 nm was measured via microplate reader (Versamax; Molecular Devices, USA).

All cultures were inspected microscopically before randomly assigned to experimental treatment, and data collection and analysis were performed blinded to treatment group. No data were excluded from inclusion in the final analysis. Sample sizes were determined from previous experience using these cell cultures.²¹  All data reported are representative of at least three independent experiments with n = 6 to 8 cultures per treatment, using two to three images per culture averaged. proBDNF ELISA data are representative of at least three independent experiments, with n = 12 cultures per treatment. All data are reported as means ± SD. For comparisons between two groups, statistical difference was determined by unpaired, two-tailed Student t test using Prism 6.0d software (GraphPad, USA). For comparisons between multiple groups, a one-way ANOVA with Bonferroni-corrected post hoc comparison of all means to the control group (for proBDNF and dihydrokainate dose–response curves) or post hoc comparison of all means to all other means (for proBDNF ELISA measurements) using Prism 6.0d software. In all analyses, a P value of less than 0.05 was considered significant.

A 1-h exposure to 2% isoflurane induced a significant increase in cell death (PI-positive cells) in primary neuronal cultures (fig. 2, A and B). However, DIV 7 neurons growing in the presence of either DIV 21 astrocytes (fig. 2, C and D) or age-matched astrocytes (fig. 2, E and F) were protected from the effect of the same isoflurane exposure. This finding demonstrates that the observed protection from isoflurane neurotoxicity afforded by astrocytes occurred independently of the relative maturity of the astrocytes from DIV 7 to 21.

Next, to test whether this effect was due to a secreted factor specific to neuronal-astrocyte co-cultures, the normal glia-conditioned media used in neuronal cultures were replaced with freshly harvested co-cultures conditioned media 1 h before isoflurane exposure. Media replacement did not result in any significant protection of neurons from isoflurane toxicity (fig. 3, A and B), suggesting that the protective effect of astrocytes was dependent on the physical presence of astrocytes. No differences in the total numbers of neurons were observed between treatment groups (data not shown).

p75^NTR protein and mRNA expression in primary astrocyte cultures was significantly reduced 24 h after transfection with p75^NTR siRNA relative to mismatch control sequence using the transfection agent Lipofectamine-2000 (fig. 4, A–C). Co-cultures transfected with reporter plasmid using Lipofectamine-2000 resulted in preferential targeting of astrocytes, with less than 2% transfected cells identified as neurons. Co-cultures transfected with p75^NTR siRNA using this method also showed a significant decrease in p75^NTR protein expression (fig. 4, D and E), however, to a smaller degree relative to transfection in primary astrocyte culture alone (fig 4B), consistent with preferential knockdown in astrocytes. Transfection with p75^NTR siRNA 24 h before isoflurane exposure resulted in a significant increase in neuronal cell death (fig. 5, A and B) relative to control transfection (fig. 5, C and D), suggesting that astrocyte p75^NTR expression contributed to neuronal protection. Treatment with either p75^NTR siRNA or isoflurane alone resulted in a modest increase in proBDNF levels in co-culture media (fig. 5E). However, combined treatment with both p75^NTR siRNA and isoflurane resulted in a further increase in proBDNF levels versus either treatment alone (fig. 5E bar furthest right). Together with the observation that only combined treatment resulted in increased neuronal cell death (fig. 5, A and B), this finding suggests that the reduction in astrocyte p75^NTR with siRNA critically reduced the buffering capacity of astrocytes to isoflurane-induced increases in proBDNF. No differences in the total numbers of neurons were observed between treatment groups (data not shown).

Application of increasing levels of proBDNF (coadministered with protease inhibitor to prevent conversion to mature BDNF), beginning with physiologic levels measured in media, resulted in a progressive increase in neuronal cell death in both primary neuronal (fig. 6A) and neuronal-astrocyte co-cultures (fig. 6B). In co-cultures, pretreatment with p75^NTR siRNA augmented neuronal cell death from 100 pg/ml proBDNF, whereas pretreatment with the intracellular p75^NTR inhibitor TAT-Pep5 blocked the neurotoxic effect from the same proBDNF dose (fig. 6B). Thus, similar to the observations in neuronal cultures,^2,3  proBDNF-induced cell death of neurons in co-culture is mediated by p75^NTR signaling and mitigated by p75^NTR expression on astrocytes. Interestingly, addition of the same concentrations of proBDNF to primary astrocyte cultures did not induce cell death (fig. 6C), providing evidence for a functional difference between astrocytes and neurons in response to binding of proBDNF to p75^NTR. Because glutamatergic excitotoxicity has been proposed as a potential mechanism contributing to anesthetic neurotoxicity,²⁶  we investigated whether astrocyte-mediated protection from isoflurane neurotoxicity was mediated by astrocytic glutamate uptake. Addition of the astrocyte–glutamate transporter inhibitor dihydrokainic acid did not result in any effect on isoflurane-mediated cell death (fig. 6D), suggesting that glutamate buffering does not contribute to astrocyte-mediated protection from isoflurane neurotoxicity in these co-cultures. No differences in the total numbers of neurons or astrocytes were observed between treatment groups (data not shown).

This study is the first to demonstrate that astrocytes protect neurons from isoflurane toxicity and that this protection occurs via reduced availability of proBDNF to bind to neuronal p75^NTR (fig. 7). Previous finding²  suggested that isoflurane neurotoxicity in primary neuronal cultures is limited to the period of active synaptogenesis (DIV 4 to 7) and decreases after approximately at DIV 10. Relatively few studies have investigated the role astrocytes play in the development of brain injury from isoflurane exposure. Lunardi et al.²⁷  demonstrated that prolonged (24 h) exposure to high-dose (3%) isoflurane impaired growth of DIV 4 but not DIV 15 astrocytes. Culley et al.²⁸  observed that isoflurane exposure to astrocytes at DIV 16 did not affect astrocyte survival or the ability to support synaptogenesis. Most recently, Ryu et al.²⁹  demonstrated that isoflurane preexposure of astrocytes at DIV 10 impaired axonal growth in developing co-cultures but did not contribute to astrocyte cell death. In the current study, DIV 7 neurons were protected from isoflurane exposure whether in the presence of age-matched (DIV 7) or DIV 21 astrocytes. These observations demonstrate that astrocyte-mediated protection is present in relatively immature astrocytes and is maintained in more mature astrocytes.

Significantly more baseline neuronal injury was observed in neurons cultured alone than in neurons cultured with age-matched astrocytes, and the least injury was seen in neurons cultured on mature astrocytes. Neither conditioned media from astrocyte cultures nor conditioned media from co-cultures provided protection from isoflurane neurotoxicity. This suggests that (1) concentrations of additional secreted factors resulting from neuron–astrocyte cross-talk were either too low or otherwise did not contribute to the neuroprotective effect observed in this study; and (2) the physical presence of astrocytes is necessary for neuroprotection from isoflurane toxicity.

Astrocyte-mediated neuroprotection that requires astrocytes in close proximity is shown in the current model (fig. 7) of synaptic BDNF signaling. proBDNF is released from the presynaptic terminal into the synapse and likely has a limited period of activity before conversion to mature BDNF by plasmin. Debate exists as to the physiologic relevance of proBDNF to in vivo signaling.³⁰  Bergami et al.¹⁹  previously demonstrated that binding of proBDNF to p75^NTR results in endocytosis and removal of proBDNF from the synaptic cleft. In this study, we observed a significant increase in proBDNF levels in media from co-cultures with astrocyte-targeted knockdown of p75^NTR after isoflurane exposure. This suggests that extrasynaptic levels of proBDNF increased secondary to reduced uptake mediated by astrocyte p75^NTR.

Ryu et al. recently reported reduced levels of mature BDNF in media from co-cultures of neurons with astrocytes subjected to isoflurane preexposure, suggesting an alteration in BDNF signaling; however, whether this was due to alterations in proBDNF levels or to changes in plasmin activity was not determined. In the current study, application of increasing levels of recombinant proBDNF resulted in a significant corresponding increase in neuronal cell death in both neuronal cultures and neuronal-astrocyte co-cultures. Interestingly, the lowest toxic dose (2.5 pg/ml) in both culture models approximated the increase in proBDNF levels we observed in media from co-cultures subjected to astrocyte-targeted p75^NTR knockdown plus isoflurane exposure, a treatment that increased the neuronal cell death.

At a dose of 100 pg/ml proBDNF, neuronal cell death in co-cultures increased substantially, likely representing saturation of astrocytic proBDNF uptake. This was exacerbated by p75^NTR knockdown and reversed by inhibition of p75^NTR signaling, verifying a proBDNF/p75^NTR mechanism of neuronal cell death, possibly via rat sarcoma virus homolog family member A (RhoA)–mediated induction of proapoptotic pathways.³  Notably, proBDNF did not kill astrocytes indicating that proBDNF binding to p75^NTR has different effects on cell survival between astrocytes and neurons under these conditions. This is supported by previous observations²⁷  that although prolonged exposure of immature astrocytes to high concentrations of isoflurane impaired astrocyte maturation and morphological development, it paradoxically reduced the activation of rat sarcoma virus homolog family member A. Moreover, activation of p75^NTR by the lower-affinity ligand nerve growth factor failed to induce cell death of cultured hippocampal astrocytes.³¹  Future investigations delineating p75^NTR cell signaling pathways in astrocytes may yield further insight into their potential to coordinate synaptic transmission.

Astrocytes regulate synaptogenesis, synaptic function, and synapse elimination through modulation of glutamatergic signaling and via contact-dependent signals.^15,16  Astrocytes have been shown to protect neurons from glutamate-mediated excitotoxicity during pathophysiologic stresses such as stroke,³²  traumatic brain injury,³³  and spinal cord injury.³⁴  The results from the current study demonstrate that astrocytic glutamate uptake did not contribute to astrocyte-mediated protection from isoflurane, instead suggesting that modulation of postsynaptic proBDNF/p75^NTR signaling was responsible for astrocyte-mediated neuroprotection. Functional modulation of BDNF signaling by astrocytes may be relevant in the adult hippocampus where BDNF signaling is central to normal learning and memory formation.^8,9  The results from the current study suggest that astrocytes may provide a novel cellular mechanism to manipulate neuronal BDNF signaling.

Although BDNF is a known regulator of neonatal brain development coordinating prodeath and prosurvival signaling, several lines of evidence also support a role for BDNF signaling in recovery from injury in the adult brain after stroke,³⁵  traumatic brain injury,³⁶  and spinal cord injury.³⁷  However, therapies targeting manipulation of BDNF signaling are limited by the short plasma half-life of BDNF³⁸  and relatively poor blood–brain barrier penetrance.³⁹  Future studies using in vivo models of central nervous system injury in adult animals are needed to investigate the effects of targeted overexpression of astrocyte p75^NTR and inhibition of neuronal p75^NTR as potential targets for the development of new therapies to encourage recovery.

Caution should be exercised in extrapolating these in vitro results to the clinical setting. For example, despite evidence from the current study demonstrating protection of DIV 7 neurons from isoflurane toxicity by DIV 7 astrocytes, previous study¹  has demonstrated isoflurane-induced neuronal cell death in vivo at postnatal day 7. DIV-7-cultured brain cells are unlikely to fully recapitulate the phenotypes of cells developing in vivo, and in particular, the ratio of astrocytes to neurons in culture is likely higher than seen in vivo at this age and may contribute to the difference between these observations.

Differences between in vivo and in vitro observations may also result from loss of the anatomic organization of the intact brain and loss of physiologic contributions from other cell types that are absent in culture. Effects of isoflurane on brain perfusion may contribute to in vivo effects and are also absent in culture. Although microglia are present throughout the brain, in our isolated cell cultures, they are present at low levels (on average < 2%), suggesting that in these in vitro observations they are unlikely to contribute measurably to the protection seen, even though they may express p75^NTR.

Finally, it is interesting to note that astrocytes in the human brain are larger, more complex, and greater in number relative to neurons than in the rodent brain.⁴⁰  This suggests that human brains may inherently posses a greater capacity for neuroprotection from anesthetic toxicity versus rodent brains and that bolstering astrocyte defense may be a potential therapeutic strategy.

The authors thank Stephen Milburn, B.A. (Department of Anesthesiology, Perioperative, and Pain Medicine, Stanford University, Stanford, California), for his assistance with article preparation, and Bryce Small, B.S. (Department of Anesthesiology, Perioperative, and Pain Medicine, Stanford University), for his contribution to figure 7.

Support was received from the National Institutes of Health (Bethesda, Maryland; grant no. T32 GM089626) and the American Heart Association (Dallas, Texas; grant no. 14FTF-19970029) to Dr. Stary, and from the National Institutes of Health (grant no. R01 NS084396) to Dr. Giffard.

The authors declare no competing interests.