Inhibition of p75 Neurotrophin Receptor Attenuates Isoflurane-mediated Neuronal Apoptosis in the Neonatal Central Nervous System

All studies performed on animals were approved by the Veteran Affairs San Diego Institutional Animal Care and Use Committee and conform to relevant National Institutes of Health guidelines.

Neonatal mouse neurons (The Jackson Laboratory, Bar Harbor, ME) were isolated using a papain dissociation kit (Worthington Biochemical, Lakewood, NJ) as previously described.19Mix cortical and hippocampal neurons were isolated from 1-day-old pups (PND1) and grown in culture for 5 to 21 days in vitro  (DIV). Neurons were cultured in Neuobasal A media supplemented with B27 (2%), 250 mm GLUTMax1, and penicillin/streptomycin (1%). Cells were cultured on poly-d-lysine/laminin (2 μg/cm²)–coated plates or coverslips at 37°C in 5% CO²for 5, 14, or 21 days before experiments. Cleaved-caspase 3 (Cell Signaling, Danvers, MA) and/or caspase-activated DNase (CAD)–positive cells (Santa Cruz Biotech, Santa Cruz, CA) were used to determine apoptosis via  immunoblot and/or immunofluorescence and deconvolution microscopy. Antibodies to phospho-JNK and GAPDH were obtained from Cell Signaling and Imgenex (San Diego, CA), respectively. Cell death was normalized to total caspase-3 or the neuronal marker doublecortin (Abcam, Cambridge, MA) or NeuN (Chemicon/Millipore, Billerica, MA) of cells imaged. TAT-Pep5 was purchased from CalBiochem (Gibbstown, NJ).

Neurons were placed in a plexiglass chamber within an incubator and exposed to 1.4% isoflurane delivered from a calibrated vaporizer in a gas mixture of 5% CO², 21% O², balance nitrogen gas at a flow rate of 2 l/min. The concentration of isoflurane was continuously monitored by a Datex Capnomac (DRE Medical, Inc., Louisville, KY).

Animals were exposed to isoflurane (1.4%; air flow served as the carrier) at a flow rate of 2 l/min for 4 h. Body temperature was monitored with an 8 mm, 22-gauge temperature probe (Mon-a-therm 6510; Mallinckrodt Medical, Inc., St. Louis, MO) that was affixed to the dorsal neck. The temperature in the incubator was maintained at 37°C; this results in core body temperature in mouse pups in the range of 36.5–37.5°C (data not shown).

Proteins in cell lysates were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using 10% acrylamide gels (Invitrogen, Carlsbad, CA) and transferred to polyvinylidene difluoride membranes (Millipore, Billerica, MA) by electroelution. Membranes were blocked in 20 mm phosphate-buffered saline (PBS) Tween (1%) containing 4% bovine serum albumin (BSA) and incubated with primary antibody overnight at 4°C as previously described.20Primary antibodies were visualized using secondary antibodies conjugated to horseradish peroxidase (Santa Cruz Biotech) and ECL reagent (Amersham Pharmacia Biotech, Piscataway, NJ). All displayed bands are expected to migrate to the appropriate size and were determined by comparison to molecular weight standards (Santa Cruz Biotech). The amount of protein per fraction was determined using a dye-binding protein assay (Bio-Rad, Hercules, CA). Image J was used for densitometric analysis of immunoblots.∥

Neurons were exposed to isoflurane for 4 h, and media from both control and isofluorane-treated neurons were frozen at –80°C before enzyme-linked immunosorbent assay (ELISA) (Innovative Research, Novi, MI). A 96-well plate was precoated with biotinylated plasminogen activator inhibitor-1 (PAI-1) for 30 min before three washes with wash buffer (provided by the supplier). tPA standard (0.05–10 ng/ml) and unknown samples (i.e. , media from control and isofluorane-exposed neurons) were added to the plate for 30 min, washed three times, and incubated with anti-tPA primary antibody for 30 min, followed by anti-rabbit horseradish peroxidase–conjugated secondary antibody for an additional 30 min. After three additional washes, tetramethylbenzidine substrate was added to the wells for 10 min, and the reaction was quenched with 1 M H²SO⁴and read at 450 nm on a spectrophotometer (TECAN Infinite M200, San Jose, CA).

Neurons were prepared for immunofluorescence microscopy as previously described.21The following antibodies were used for immunofluorescence: cleaved caspase 3, drebrin, doublecortin (Abcam), and NeuN (Chemicon/Millipore). Tissue and cells were fixed with 4% paraformaldehyde in PBS for 10 min at room temperature, incubated with 100 mm glycine (pH. 7.4) for 10 min to quench aldehyde groups, permeabilized in buffered Triton X-100 (0.1%) for 10 min, blocked with 1% BSA/PBS/Tween (0.05%) for 20 min, and then incubated with primary antibodies (1:100) in 1% BSA/PBS/Tween (0.05%) for 24–48 h at 4°C. Excess antibody was removed by incubation with PBS/Tween (0.1%) for 15 min, and the samples were incubated with fluorescein isothiocyanate (FITC) or Alexa-conjugated secondary antibody (1:250) for 1 h. To remove excess secondary antibody, cells were washed six times at 5-min intervals with PBS/Tween (0.1%) and incubated for 20 min with the nuclear stain Dapi (1:5000) diluted in PBS. Cells were then washed for 10 min with PBS and mounted in gelvatol for microscopic imaging. Deconvolution images were obtained as described21and captured with a DeltaVision deconvolution microscope system (Applied Precision, Inc., Issaquah, WA). The system includes a Photometrics CCD (Tuscon, AZ) mounted on a Nikon TE-200 (Melville, NY) inverted epi-fluorescence microscope. Between 30 and 80 optical sections spaced by approximately 0.1–0.3 μm were taken. Exposure times were set such that the camera response was in the linear range for each fluorophore. Lenses included 100× (NA 1.4), 60× (NA 1.4), and 40× (NA 1.3). The data sets were deconvolved and analyzed using SoftWorx software (Applied Precision, Inc.) on a Silicon Graphics Octane workstation. Image analysis was performed with Data Inspector program in SoftWorx. Maximal projection volume views or single optical sections were visualized. Colocalization of pixels was assessed quantitatively by CoLocalizer Pro 1.0 software (Tokyo, Japan). Overlap coefficient according to Manders was used to determine the degree of colocalization on whole cells or membrane regions of interest after subtracting background through normalized threshold values.22The values were defined as 0 to 1, with 1 implying that 100% of both components overlap with the other part of the image. Statistical analysis was performed using Prism 4 (GraphPad Software, Inc., La Jolla, CA).

The drebrin pixels (green) were normalized to doublecortin pixels (red). Because dendritic filopodial spines are nearly completely devoid of intermediate filaments and microtubules (as indicated by doublecortin staining)23and only contain components of the actin cytoskeleton (as indicated by drebrin staining), a neuronal microtubule marker (doublecortin) was used to normalize any changes in spinal actin cytoskeletal expression. Drebrin is a filamentous (F-)actin binding protein that stabilizes the actin cytoskeleton within dendritic filopodial spines.23,24A reduction in dendritic filopodial spines is indicated by decreased drebrin protein expression. Approximately 15–20 neurons were counted per experiment (4–6 experiments total).

Neurons were incubated with small interfering ribonucleic acid (siRNA) (Ambion, Inc., Foster City, CA; 200 pmol; sense: CAGCUGCAAGCAGAACAAGtt; antisense: CUUGUUCUGCUUGCAGCUGtt) using Lipofectamine 2000 transfection reagent (Invitrogen) for 72 h and then exposed to isoflurane. Scrambled siRNA served as control.

Under deep pentobarbital anesthesia, a midline thoracotomy was performed and the descending thoracic aorta was occluded. A 20-gauge needle was inserted into the left ventricle, and the animal was perfused transcardially with 20 ml of heparinized saline followed by 20 ml of 4% buffered formalin. The right atrium was incised to permit free flow of perfusion fluid. The brain was removed and postfixed for 24–48 h in fixative. The brain was sectioned into 4-mm blocks, placed in tissue holders, and dehydrated in graded concentrations of ethanol in a tissue processor (Autotechnicon; Technicon Instruments, Tarrytown, NY). The tissue holders were then transferred into a paraffin bath in an oven at 37°C under vacuum suction. The paraffin blocks were separated from tissue holders and mounted in a Histostat 820S microtome (Reichert Scientific Instruments, Depew, NY); 5- to 10-μm sections were then cut, placed on glass slides, and incubated overnight at 37°C. The tissue sections were then prepared for immunohistochemistry.

Brains were transcardially perfusion fixed with standard Karnovsky's fix, 4% paraformaldehyde, 1% gluteraldehyde, 0.1 M cacodylate buffer with 5 mm CaCl². PND5-7 animals were fixed with 2% paraformaldehyde, 2.5% glutaraldehyde, 0.1 M cacodylate buffer, and 5 mm CaCl²to prevent tissue artifacts. Hippocampi were dissected from whole brains after 24 h, and 400-μm vibratome slices were prepared and refixed for an additional 24 h. Brains were blocked (i.e. , dissected) to include hippocampal areas: one hemisphere for sagittal orientation, and one hemisphere for coronal. Blocks were refixed for an additional 24 h before postfixation with 1% OsO⁴in 0.1 M cacodylate buffer, en bloc  stained with uranyl acetate, and embedded with flat orientation to locate appropriate hippocampal regions of interest. Each block was thick sectioned, stained with toludine blue, and retrimmed to isolate hippocampal areas before preparation of grids. Grids (70-nm sections) were stained with uranyl acetate and lead nitrate for contrast and observed on the JEOL 1200 EX-II electron microscope (JEOL, Tokyo, Japan) equipped with a digital camera system. Twenty-five random low-magnification micrographs of the stratum radiatum were obtained from each specimen. Micrographs were analyzed for the quantity of synapses and for synapse abnormalities (reduction or changes in synapse and dendritic filopidal spine morphology, i.e. , degradation of cytoskeletal architecture). The dendritic profiles were characterized by abundant organelles such as mitochondria and endoplasmic reticulum and frequent contacts from vesicle-filled axon terminals. Spine synapses were identified by an electron-dense region presynaptically associated with vesicles and that lacked cellular organelles or contained a spine aparatus (as indicated by cytoskeletal architecture) postsynpatically as described previously.25–27Dendritic filopodial spines were also identified as dendritic extensions rich in F-actin and possessing postsynaptic densities (PSD) at their surfaces as previously described.28,29Approximately 25 electron micrographs (3350 μm²) per animal were analyzed in a blinded fashion for total synapse number per area (synapse #/3350 μm²).

All parametric data were analyzed by unpaired t  tests or ANOVA Bonferroni multiple comparison as appropriate; post hoc  comparisons were made by Student-Neuman-Keuls tests. Significance was set at P < 0.05. Statistical analysis was performed using Prism 4 (GraphPad Software, Inc.).

Cortical and hippocampal neurons were isolated on PND1 and were grown in culture up to 5, 14, and 21 days (days in vitro - DIV5, DIV14, and DIV21). Isoflurane (1.4%, 4 h) exposure increased cleaved caspase-3 (Cl-Csp3) relative to total caspase-3 (T-Csp3) in DIV5 neurons (fig. 1A), but not in DIV14 (n = 3, P = 0.22), or DIV21 (n = 3, P = 0.68) (fig. 1B) neurons. These data are consistent with the age-dependent neurotoxicity of isoflurane that has been demonstrated previously.1,9,30 

tPA is a critical modulator of BDNF.31tPA release is activity-dependent,18and anesthetics suppress neuronal activity; we therefore hypothesized that isoflurane would lower tPA release. DIV5 primary neurons were exposed to isoflurane (1.4%, 4 h), and tPA levels in the culture medium were measured by ELISA. Neurons exposed to isoflurane had less tPA (n = 6, P = 0.0013) in the media compared to control (fig. 2A). Addition of 0.03–3 μg/ml tPA to the medium dose-dependently attenuated isoflurane-mediated increase in Cl-Csp3 at 2 h (fig. 2B) after exposure. Exogenous application of matrix metalloproteinase-7, which can potentially cleave proBDNF to mBDNF, and matrix metalloproteinase-9, which cleaves proNGF to mNGF, did not reduce isoflurane-mediated neuronal apoptosis (fig. 2C), whereas 3 μg/ml tPA reduced (n = 3, P = 0.02) isoflurane-mediated (n = 3, P = 0.02) apoptosis. In addition, tPA application increased levels of phosphorylated Akt (n = 3, P = 0.002) (fig. 2D), a downstream signaling component of TrkB activation. Application of plasmin also attenuated isoflurane-mediated neuronal apoptosis (n = 3, P = 0.004); the neuroprotective effects of both plasmin and tPA were blocked by the serine protease inhibitor, aprotinin (fig. 2E).

Neonatal primary neurons (DIV5) were exposed to isoflurane (1.4%, 4 h) and stained for dendritic filopodia (i.e. , immature filopodial spines) using a neuronal F-actin marker, drebrin, and the dendritic shaft/neuronal microtubule marker, doublecortin. Basally, drebrin staining revealed thin protrusions (termed dendritic filopodia) common to immature neurons24(fig. 3A). Exposure of DIV5 neurons to isoflurane significantly reduced (n = 6, P = 0.01) the number of dendritic filopodial spines. Treatment with the proteases, plasmin (n = 6, P = 0.02) or tPA (n = 5, P = 0.04), significantly attenuated the isoflurane-induced loss of neuronal dendritic filopodia (fig. 3B). Quantitation is expressed as drebrin pixels (green) normalized to doublecortin immunofluorescence (red pixels) along the dendrites only (soma immunofluorescence was not included) (fig. 3C). In addition, Western blot images demonstrate that addition of exogenous tPA significantly mitigated (n = 3, P = 0.006) the isoflurane-mediated (n = 3, P = 0.005) reduction in drebrin protein expression (fig. 3D).

If isoflurane increases proBDNF signaling via  the p75^NTR, then knockdown or inhibition of this receptor should reduce toxicity related to this pathway. Primary neurons (DIV5) were treated with the p75^NTRinhibitors (Fc-p75^NTR, a chimera expressing only the extracellular domain of the p75^NTR, or TAT-Pep5, an intracellular inhibitor of p75^NTR) or siRNA for p75^NTRbefore isoflurane exposure (1.4%, 4 h). Preincubation with Fc-p75^NTR(1.35 μg/ml) (fig. 4A, n = 3, P = 0.001 vs.  basal, P = 0.003 vs.  isoflurane) or TAT-conjugated Pep5 (10 μm) (fig. 4B, n = 3, P = 0.007 vs.  basal, P = 0.006 vs.  isoflurane) significantly attenuated isoflurane-mediated neuronal apoptosis. p75^NTR-siRNA resulted in significant knockdown (approximately 62%; n = 3, P = 0.005) of the neurotrophin receptor after 72 h and attenuated isoflurane-mediated apoptosis (n = 3, P = 0.001 vs.  basal-scrambled control [SC], P = 0.001 vs.  isoflurane-scrambled control) compared to SC neurons (fig. 4C).

We next tested whether TAT-Pep5 was efficacious in reducing anesthetic-induced neurotoxicity and loss of synapses in PND5 mice in vivo . Neonatal C57BL/6 mice (PND5-7) were given TAT-Pep5 (intraperitoneal injection; 10 μm) 15 min before isoflurane exposure (1.4%, 4 h). Isoflurane significantly enhanced cleaved caspase 3 (Cl-Csp3) in the CA1 (n = 6, P = 0.003) and CA3 (n = 7, P = 0.007) of the hippocampus in PND5 mice (fig. 4D) 2 h after exposure. TAT-Pep5 significantly attenuated isoflurane-mediated increase in Cl-Csp3 in both the CA1 (n = 7, P = 0.007) and CA3 (n = 7, P = 0.007) regions. In contrast, isoflurane exposure did not enhance Cl-Csp3 in PND21 mice (fig. 4E), indicating that the brains of mice at this age are less vulnerable to isoflurane-induced neurotoxicity.

Synapse number in the hippocampus (stratum radiatum; mossy fiber to CA3 and Shaeffer collaterals from CA3 to CA1) of neonatal mouse pups (PND5-7) 2 h after isoflurane exposure (1.4%, 4 h) was quantitated by electron microscopy. Synapses with postsynaptic densities and presynaptic vesicles are clearly visible in representative control images (fig. 5Aa). By contrast, there was a significant reduction in total synapses in PND5-7 pups exposed to isoflurane (fig. 5Ab; n = 6, P = 0.01). This reduction in synapses was significantly attenuated by pretreatment with TAT-Pep5 (fig. 5Ac; n = 5, P = 0.01). Higher magnification (6600×) electron microscopy images for control (fig. 5Ba), isoflurane (fig. 5Bb), and TAT-Pep5 (fig. 5Bc) are shown in figure 5B. Quantitation of data is represented by the graph in figure 5C. These findings extend our results from in vitro  studies demonstrating that (1) PND5-7 animals are vulnerable to anesthetic-induced neurotoxicity and (2) these neurotoxic effects from isoflurane can be mitigated by TAT-Pep5 in vivo .

The results of the current study are consistent with the neurotoxicity of isoflurane that has been demonstrated previously.^1–3Primary neurons (DIV5) and mouse pups (PND5-7) were vulnerable to injury during the period of synaptogenesis, whereas the toxicity was not apparent at the age of 14 and 21 days DIV and in vivo . In vitro , isoflurane exposure reduced tPA release, and the administration of either tPA or plasmin reduced neurotoxicity. Reduction of p75^NTRsignaling, either with siRNA-mediated knockdown of p75^NTRor with pharmacologic inhibition with Fc-p75^NTRor TAT-Pep5, abolished isoflurane injury. In vivo , isofluorane-mediated injury to the CA1 and CA3 sectors of the hippocampus was apparent. Administration of TAT-Pep5 before isoflurane exposure abolished injury. In aggregate, the results suggest that isoflurane neurotoxicity is mediated at least in part by a reduction of tPA release and preferential signaling of proBDNF via  the p75^NTR.

There is a convincing body of evidence to indicate that the brain is vulnerable to anesthetic neurotoxicity during synaptogenesis.10In the developing brain, synaptic connections with appropriate target neurons are essential for neuronal survival because neurons are dependent on trophic support from their targets.32–34Loss of synaptic connection leads to apoptosis. An important element contributing to synaptogenesis and to consolidation and maturation of synapses is the neurotrophin BDNF.11Neurotrophins are synthesized as proneurotrophins (proNTs), which subsequently undergo proteolytic cleavage to yield the biologically active mature neurotrophins.35mBDNF preferentially bind to the TrkB receptor to promote neuronal survival and synaptogenesis.13proBDNF serves as a ligand for the p75^NTRand initiates neuronal apoptosis.12Up to 60% of the secreted neurotrophins are in the pro-form.15Several proteases, including plasmin and matrix metalloproteinases, proteolytically cleave proNTs; however, the most significant protease is plasmin.12,13Plasmin exists as plasminogen, and the key enzyme that cleaves plasminogen is tPA, which is secreted from presynaptic terminals, and its secretion is activity-dependent.16–18 

Our data show that isoflurane reduces the amount of tPA present in culture media of 5-day-old primary neu-rons. tPA is released from neurons in an activity-dependent manner,16–18and this reduction in tPA may be due to neuronal suppression by isoflurane. By contrast, proBDNF undergoes both activity-dependent and constitutive release.15Therefore, by suppressing neuronal activity during the key period of synaptogenesis, anesthetics might reduce tPA release and subsequent plasmin activation, thereby enhancing proBDNF signaling via  p75^NTR. Increase in the activation of JNK, a downstream consequence of p75^NTRactivity,36with isoflurane exposure supports this premise. The consequence of this is neuronal death, either due to a failure of synaptic stabilization and consolidation (with the subsequent failure of affected neurons to obtain target derived trophic factor support) or p75^NTR-mediated apoptosis. Previous application of tPA led to the activation of Akt (P-Akt), a downstream kinase of TrkB signaling critical to neuronal survival,37attenuated anesthetic-induced neurotoxicity as well as the loss of synaptic filopodial spines in developing neurons. The present studies used both cellular and animal models to demonstrate the novel involvement of tPA-plasmin-proBDNF-p75^NTRsignaling in anesthetic neurotoxicity.

In addition to its proteolytic function, tPA has a number of nonproteolytic activities. tPA binds to the low-density lipoprotein receptor (LRP-1) and triggers signaling cascades that lead to phosphorylation of ERK1/238and Akt39; this effect of tPA is independent of its proteolytic activity. Both ERK1/240and Akt41have antiapoptotic activity. The LRP-1 receptor is expressed in neurons and glia.39tPA can also bind to and directly activate the urokinase-like plasmingoen activator receptor and trigger intracellular signaling.42The possibility that the reduction in anesthetic neurotoxicity might be a function of the nonproteolytic effects of tPA (activation of survival signaling) exists. Aprotinin is a serine protease inhibitor of the proteolytic activity of tPA.12,43Data showing that aprotinin abrogated the protective efficacy of tPA suggests, in our model systems, that the salutary effect of tPA is mediated by its proteolytic activity and not by a receptor-mediated effect. Additional unpublished findings from our laboratory indicate that pretreatment of neurons (DIV5) with 4-aminopyridine, a potassium channel blocker that serves to enhance vesicular release from presynaptic terminals,44reduces isoflurane-mediated cell death in a dose-dependent manner, suggesting that enhanced neuronal activity can alleviate the neurotoxic effects from isoflurane. Furthermore, plasmin administration reduced neurotoxicity, indicating that the proteolytic activity (presumably cleavage of proBDNF to mBDNF) is required for prevention of anesthetic neurotoxicity.

Although plasmin is the major protease that cleaves proNTs, other enzymes such as matrix metalloproteinases can also cleave proNTs.12,13Matrix metalloproteinases-9 cleaves proNGF to yield mNGF, and matrix metalloproteinases-7 can similarly generate mBDNF. To determine whether these enzymes play a role in the prevention of isoflurane neurotoxicity, we added matrix metalloproteinases-7 or matrix metalloproteinases-9 to neurons exposed to isoflurane. The failure of either of these matrix metalloproteinases to reduce isoflurane toxicity indicates that these enzymes do not cleave proBDNF to mBDNF in our model systems, thus supporting the premise that it is tPA and plasmin that have neuroprotective efficacy.

In addition to reducing mBDNF-TrkB signaling (i.e. , decreased Akt phosphorylation), isoflurane increased p75^NTRsignaling. p75^NTRmodulates synaptic transmission, axonal elongation, and growth cone collapse, and it initiates neuronal apoptosis.14,45–47In p75^NTR−/−mice, dendritic complexity, which is increased substantially, can be reduced by p75^NTRoverexpression,48indicating a negative modulatory effect of p75^NTRon dendritic spine development. Fc-p75^NTRis a chimera that expresses extracellular domain (amino acids 1–250).49,50This domain serves to bind and scavenge BDNF and therefore potently reduces p75^NTRsignaling. Our data, which show that inhibition of p75^NTRwith TAT-Pep5, Fc-p75^NTR, or targeted siRNA reduces anesthetic-mediated neuronal apoptosis, indicate that anesthesia results in p75^NTR-dependent neuronal apoptosis in the neonatal brain, perhaps by proBDNF agonism at p75^NTR. Furthermore, TAT-Pep5 attenuation of anesthesia-induced neurotoxicity suggests that the neurodegenerative effects from anesthetics may be in part due to alterations to the actin cytoskeleton. The question that remains is whether isoflurane neurotoxicity is due to the loss of mBDNF-TrkB signaling, increased proBDNF-p75^NTRsignaling or a combination of both.

In the development of therapeutic options to prevent anesthetic neurotoxicity, a central concern is whether these interventions are directly toxic themselves. In our studies, we investigated the potential of the protective effect of p75^NTRinhibition against anesthetic neurotoxicity. Reduction of p75^NTRactivity by extracellular inhibition with Fc-p75^NTR, by intracellular inhibition with TAT-Pep5, or by knockdown of p75^NTRwith siRNA was clearly reduced by isofluorane-mediated injury. Importantly, though, the interventions did not directly injure neurons, indicated by the fact that Cl-Csp3 immunoblot (fig. 4, A–C) was similar in the nonanesthetized control and p75^NTRinhibition groups. A second concern is that protective agents might inhibit developmental neuroapoptosis. This is a normal physiologic process in the developing brain that is essential for the removal of excess neurons; inhibition of apoptosis therefore might produce the undesirable effect of preventing this normal clearance of unneeded neurons. Our data clearly show that, although p75^NTRinhibition was effective in reducing isoflurane neurotoxicity, the basal level of apoptosis (Cl-Csp3 immunoblot) was not different in the nonanesthetized control and p75^NTRinhibition groups (fig. 4, A–C). These data indicate that short-term inhibition of p75^NTRactivity does not have direct toxicity and does not inhibit normal developmental neuroapoptosis. It should be noted that injury was evaluated only in the hippocampus and not in other structures that are injured by anesthesia, such as the cortex and thalamus. Whether p75^NTRinhibition adversely affects the cortex and thalamus is not currently clear. In addition, the effect of long-term inhibition of p75^NTRduring synaptogenesis needs further investigation.

In the majority of anesthetic neurotoxicity studies focus on neuroapoptosis. However, the current investigation indicates that the injury is not restricted to neuronal death; it also includes the loss of dendritic filopodial spines and synapses. The quantity of dendritic filopodial spines was dramatically reduced after isoflurane exposure both in vitro  and in vivo . Most of the excitatory synapses are located on dendritic filopodial spines.51The loss of filopodial spines would therefore significantly reduce excitatory transmission and should lead to loss of synapses. Indeed, electron microscopy of the stratum radiatum in the hippocampus (mossy fiber-CA3 and CA3-CA1 connections) demonstrated a reduction in the number of synapses as well as ablation of existing synapses compared to nonanesthetized control state (fig. 5). However, a limitation to the latter EM findings is the absence of measurements of presynaptic profile diameter, active zone length, or number of synaptic vesicles across conditions in addition to alterations in organelle and dendritic shaft volume. More thorough EM analysis in future studies may yield a comprehensive morphologic profile of changes that occur with isoflurane and other anesthetics during synaptogenesis. The current results are consistent with those reported by Nikizad et al.  52; a general anesthetic dose of midazolam led to a significant, albeit transient, reduction in the quantity of proteins associated with synapses (e.g. , synaptophysin, synaptobrevin, SNAP-25). Of interest is the observation that the reduction in synapse number in the mossy fiber to CA3 connections (approximately 50%) was considerably greater than the extent of apoptosis in either the CA3 sector or the dentate gyrus. The relationship between synapse loss and neuronal apoptosis will require further investigation.

A recent finding suggests that proBDNF undergoes proteolysis intracellularly and is released in the mature form.53These data, which are in contradiction to those reported previously by Lu et al. ,31raise the question of the relevance of proBDNF to isoflurane neurotoxicity, at least within the framework that has been proposed herein. In support of our central premise, our data (not shown) indicate that proBDNF, when measured with an antibody that is specific for the proBDNF form, is present in the neuronal lysates. When tPA or plasmin are administered, the level of proBDNF is reduced. Moreover, isoflurane exposure does not affect proBDNF levels. Collectively, these data indicate that neurons exposed to isoflurane are also exposed to proBDNF and that tPA and plasmin, which mitigate isoflurane toxicity, reduce proBDNF.

Our study has limitations. We were unable to measure proBDNF levels in the culture media of control and isofluorane-exposed neurons. The most sensitive ELISA kit currently available does not have the requisite sensitivity to detect the very small quantities of proBDNF that might be present in the culture medium. In the absence of proBDNF measurements, the argument for a central role of proBDNF in isofluorane-mediated neuroapoptosis is weakened. However, a key piece of evidence that supports our hypothesis is the effect of Fc-p75^NTR, a peptide that binds proBDNF49,50and reduces p75^NTRsignaling. This peptide strongly reduced isoflurane toxicity, and its effect was comparable to tPA, plasmin, TAT-Pep5, and siRNA against p75^NTR. In aggregate, these data strongly suggest that proBDNF-p75^NTRsignaling plays an important role in isoflurane toxicity. A second limitation is that the means by which isoflurane reduced tPA release are not clear. tPA release is activity-dependent, and isoflurane potently reduces neuronal activity. While there are convincing data to demonstrate the neuronal suppression effect of isoflurane in vivo , such assurance cannot be provided for in vitro  neuronal cultures due to the lack of measurable electrical activity data in cultured neurons. Nonetheless, data showing a decrease in tPA in the culture media after isoflurane exposure does suggest  a reduction in neuronal activity.16,17Finally, the sample sizes in most of the in vitro  studies are small. This increases the possibility of an α error. However, the consistency of the data between the in vitro  neuronal cultures and the in vivo  studies (where sample sizes were larger) argues against this possibility.

In summary, our findings support our hypothesis that anesthetic-induced neurotoxicity in the neonatal mouse central nervous system is a function of decreased synaptic tPA release, reduced mBDNF signaling via  TrkB, and enhanced proBDNF signaling via  p75^NTR(fig. 6). The results strongly indicate that TAT-Pep5 or other interventions that limit p75^NTRsignaling or enhance TrkB signaling, may serve as potential therapeutic approaches for the prevention of anesthetic-induced neurotoxicity in the developing central nervous system.

We are grateful for the assistance from the University of California San Diego Cancer Center Digital Imaging Shared Resource, La Jolla, California: in particular James Feramisco, Ph.D. (Professor of Medicine, University of California, San Diego, La Jolla, California), Kersi Pestonjamasp, Ph.D. (Junior Faculty, University of California, San Diego), Steve McMullen, B.A., J.D. (Technician, University of California, San Diego). We are also grateful for the technical support from Yue Hu, B.S., Michael Kidd, B.S., Blake Chin-Lee, B.S., and Stephanie Cipta, B.S. (Technicians, University of California, San Diego).