Excitotoxicity may be critical in the formation of brain lesions associated with cerebral palsy. When injected into the murine neopallium at postnatal day (P) 5, ibotenate (activating NMDA and metabotropic glutamate receptors) produces neuronal death and white matter cysts. Such white matter cysts resemble those seen in periventricular leukomalacia, a lesion evident in numerous human premature newborns. The goal of this study was to assess BDNF neuroprotection against neonatal excitotoxic lesions. Cortical and white matter lesions induced by ibotenate at P5 were reduced by BDNF by up to 36 and 60%, respectively. BDNF neuroprotection involved TrkB receptors, MAPK pathway and reduced apoptosis. Although BDNF did not prevent the initial appearance of white matter lesions, it promoted secondary decrease of the lesion size. BDNF neuroprotection at P5 was maximal against lesions induced by NMDA or ibotenate but was moderate against lesions produced by an AMPA-kainate agonist. Finally, BDNF exacerbated neuronal death produced by ibotenate at P0 through increased apoptosis and p75NTR receptors, while BDNF had no detectable effect on lesions induced at P10. Altogether, these data showed that BDNF neuroprotection against neonatal excitotoxicity is dependent upon the type of activated glutamate receptors, the lesion localization and the developmental stage.
Brain-derived neurotrophic factor (BDNF) acts through p75NTR and tyrosine kinase B (TrkB) receptors (Rodriguez-Tébar et al., 1990; Klein et al., 1991). TrkB receptor is linked to the mitogen-associated protein kinase (MAPK or ERK) pathway (Han and Holtzman, 2000) and generally mediates survival signals, while p75NTR receptor is coupled to the janus kinase (JAK) pathway (Barrett, 2000) and is often associated with death signals (Rabizadeh et al., 1993).
In adult rodents, BDNF is neuroprotective in several models, including spinal cord injury (Namiki et al., 2000), focal brain ischemia (Beck et al., 1994; Schabitz et al., 2000) and excitotoxic neuronal death (Burke et al., 1994; Bemelmans et al., 1999). BDNF also promotes the regenerative sprouting of injured axons in the adult brain (Mamounas et al., 2000). In postnatal day (P) 7 rats, BDNF is neuroprotective against a hypoxic–ischemic insult (Cheng et al., 1997; Walton et al., 1999; Han et al., 2000). In vitro, BDNF can reverse N-methyl-D-aspartate (NMDA)-mediated neuronal death (Tremblay et al., 1999).
We previously established a mouse model of neonatal excitotoxic brain lesions mimicking brain damage associated with cerebral palsy (Marret et al., 1995; Tahraoui et al., 2001). Brain damage was induced by ibotenate, a glutamatergic agonist acting on NMDA and metabotropic receptors, but not alpha-3-amino-hydroxy-5-methyl-4-isoxazole propionic acid (AMPA) and kainate receptors. The pattern of excitotoxic lesion was highly dependent upon the stage of brain maturation. Neuronal death was observed in cortical layers V and VI when ibotenate was administered at P0 or P2. In contrast, ibotenate injected at P5 or P10 killed neurons in all cortical layers. Furthermore, ibotenate injected at P5 induced periventricular white matter cystic lesions, reminiscent of lesions seen in PVL, a disease affecting the white matter of a subset of human infants born between 24 and 32 weeks gestation, which in rodents roughly corresponds to the first postnatal week. The white matter lesions evident following injection of excitotoxin in P5 mice involved early microglial activation and astrocytic death with subsequent axonal breakdown and oligodendrocytic death.
In this model, vasoactive intestinal peptide (VIP) (Gressens et al., 1997) largely prevented ibotenate-induced neuronal death at P0 but had only a moderate effect at P5, pointing to the importance of stage specific aspects of neuronal maturation that confer vulnerability to excitotoxic injury as well as protection thereof by VIP. Similarly, platelet activating factor (Bac et al., 1998) protected both the cortical gray matter and the white matter against ibotenate-induced lesions in P5 mice, while melatonin (Husson et al., 2002) or nociceptin antagonists (Laudenbach et al., 2001) only protected against ibotenate-induced white matter lesions in P5 mice, suggesting distinct mechanisms for these two types of lesions.
We previously demonstrated in this model that co-injection of VIP and ibotenate in P5 mice protected the white matter against the evolution of ibotenate-induced lesions (Gressens et al., 1997). VIP did not prevent the initial appearance of white matter lesions, but rather promoted a secondary decrease of the white matter lesion size, suggesting a role in mediating axonal regrowth or sprouting (Gressens et al., 1998). We additionally identified that protein kinase C (PKC) and MAPK pathways were critical for such neuroprotection (Gressens et al., 1998). Taken together, the data we have accrued from our in vitro and in vivo studies suggested the following model of white matter neuroprotection in P5 mice: VIP activates PKC in astrocytes which release soluble factors; these released factors activate neuronal MAPK which will permit axonal sprouting.
This formulation is consistent with previous studies which have shown that VIP-treated astrocytic cultures produce and/or release trophic factors including cytokines (Brenneman et al., 1995), protease inhibitors (Festoff et al., 1996) and BDNF (Pellegri et al., 1998).
Using our mouse model of excitotoxin-induced neonatal lesions, the goals of the present study included: (i) establishing the neuroprotective profile of BDNF against white matter and cortical gray matter lesions in P5 mice; (ii) identifying the cellular and molecular mechanisms of such BDNF neuroprotection; (iii) determining the potential links between the neuroprotection afforded by VIP and BDNF; and (iv) evaluating the influence of developmental stage on BDNF neuroprotection.
Material and Methods
Swiss mice of both sexes were used in this study. Experimental protocols were approved by the institutional review committee and meet the guidelines of the ‘Institut National de la Santé et de la Recherche Médicale’.
Ibotenate (Sigma, St-Quentin Fallavier, France) and VIP (Peninsula, Belmont, CA) were diluted in phosphate-buffered saline (PBS) containing 0.01% acetic acid. NMDA (Tocris, Bristol, UK), S-bromowillardiine (a glutamatergic agonist acting against AMPA and kainate receptors) (Tocris), human recombinant BDNF (Alomone Labs, Jerusalem, Israel), neutralizing rabbit anti-BDNF antibody (Chemicon International, Temecula, CA) and neutralizing rabbit anti-p75NTR receptor antibody (Chemicon International) were diluted in PBS. K252-a (Biomol, Plymouth Meeting, PA), H89 (Biomol), PD98059 (New England Biolabs, Beverly, MA) and bisindolylmaleimide I (France-Biochem, Meudon, France) were diluted in PBS containing 10% DMSO.
Administration of Excitotoxic Drugs
Excitotoxic brain lesions were induced by injecting ibotenate, NMDA or S-bromowillardiine into developing mouse brains, as previously described (Marret et al., 1995; Gressens et al., 1997, 1998; Bac et al., 1998; Dommergues et al., 2000; Tahraoui et al., 2001; Laudenbach et al., 2001, 2002; Husson et al., 2002). Briefly, anesthetized mouse pups were kept under a warming lamp and were injected intracerebrally (into the neopallial parenchyma) on day P0, P5 or P10. Intraparenchymal injections were performed with a 25-gauge needle on a 50 μl Hamilton syringe mounted on a calibrated microdispenser. The needle was inserted 2 mm under the external surface of scalp skin in the frontoparietal area of the right hemisphere, 2 mm from the midline in the lateral-medial plane, and 2 mm (at P0) or 3 mm (at P5 and P10) anterior to bregma in the rostro-caudal plane. Histopathology confirmed that the tip of the needle always reached the periventricular white matter. Two 1 μl boluses of ibotenate, NMDA or S-bromowillardiine were injected at 20 s intervals. The needle was left in place for an additional 20 s. Then, 5 μg (at P0) or 10 μg (P5, P10) ibotenate, 5 μg (at P5) NMDA or 15 μg (at P5) S-bromowillardiine was administered to each pup.
Pups from at least two different litters were used in each experimental group, and data were obtained from two or more successive experiments.
In the first set of experiments designed to identify the glutamate receptor sub-type mediating excitotoxic injury to the white matter, and the neuro-protective action of BDNF (Fig. 1), P5 pups received an intraparenchymal injection of one of the following three drugs, co-administered with PBS (vehicle control) or BDNF: (i) ibotenate with PBS, or with 0.5, 5 or 50 ng BDNF; (ii) NMDA with PBS or with 50 ng of BDNF; or (iii) S-bromowillardiine with PBS or with 50 ng of BDNF.
In the second set of experiments designed to compare the neuro-protective effects of BDNF with VIP (Fig. 2), P5 pups received an intraparenchymal injection of ibotenate with PBS, 1 μg of VIP or 50 ng of BDNF, with additional animals receiving each of these growth factors in the presence of 1 μg neutralizing anti-BDNF antibody.
In the third series of experiments designed to identify neuro-histologic progression of the excitotoxic injury as reflected by GAP43, GFAP and Lectin staining following co-infusion of PBS or the growth factors VIP and BDNF (Figs 3 and 4), P5 pups received an intraparenchymal injection of ibotenate with PBS, 1 μg of VIP or 50 ng of BDNF, and sacrificed immediately (T0h), 4 (T4h), 8 (T8h), 24 (T24h), 48 (T48h), 72 (T72h) or 96 h (T96h) after ibotenate injection.
In the fourth series of experiments designed to identify the signal transduction mechanisms underlying the neuro-protective effects of BDNF (Fig. 5), ibotenate plus PBS, or one of the following drugs, or combination of drugs, diluted to a final volume of 2 μl was injected intraparenchymally: 50 ng BDNF; 50 ng BDNF + 2.5 μg neutralizing anti-p75NTR receptor antibody; 50 ng BDNF + 50 ng K252-A; 50 ng BDNF + 3 μg PD98059; 50 ng BDNF + 5 μg H89; or 50 ng BDNF + 0.1 μg bisindolylmaleimide I.
Some of the animals injected on P5 in experiments one, two and four above were additionally studied to quantify the extent of injury to the cortical gray matter using nissl-stained sections in animals sacrificed five days after injection (Fig. 6), as well as progression of apoptosis as identified by DNA fragmentation and caspase 3 activation in animals sacrificed 24 h after injection (Fig. 7). In addition, some of the animals injected on P5 in the second experiment were also used to quantify the effects of VIP on BDNF gene expression by real-time reverse transcription polymerase chain reaction (real time PCR) (Fig. 8).
In a fifth set of experiments, other animals were injected at P0 following the injection protocols for experiments one, two, and four above, and were studied to quantify the extent of injury to the cortical gray matter using nissl-stained sections in animals sacrificed 5 days after injection (Fig. 9), as well as progression of apoptosis as identified by DNA fragmentation and caspase 3 activation in animals sacrificed 24 h after injection (Fig. 10).
In the sixth and last experiment, animals injected at P10 following the injection protocol for experiment two above were likewise studied to quantify the extent of injury to the cortical gray matter using Nissl-stained sections following sacrifice at P15.
Determination of Lesion Size
Mouse pups were sacrificed by decapitation 4, 8, 24, 72, 96 or 120 h following the excitotoxic challenge. Brains were immediately fixed in 4% formalin for 7 days. Following embedding in paraffin, we cut 15 μm thick coronal or sagittal sections. Every third section was stained with cresyl-violet. Previous studies (Marret et al., 1995; Gressens et al., 1997; Husson et al., 2002) have shown an excellent correlation between the maximal size of the lesion in the lateral-medial and fronto-occipital axes of the excitotoxic lesions. To further confirm these observations, we cut serial sections of the entire brain in the coronal plane or in the sagittal plane. This permitted an accurate and reproducible determination of the maximal fronto-occipital (on coronal sections) or lateral-medial (on sagittal sections) diameters of the lesions (which is equal to the number of sections where the lesion was present multiplied by 15 μm). We used these linear measures as an index of the volume of the lesion.
Immunohistochemistry for White Matter Axons, Astrocytes and Activated Macrophages
To study the outcome of white matter axons, astrocytes, and activated microglia/macrophages, P5 pups were given intraparenchymal injections of 10 μg ibotenate plus 50 ng BDNF or of 10 μg ibotenate plus PBS. Animals were sacrificed at 4, 8, 24, 72, 96 and 120 h after injection. Five animals were included in each treatment group for each time point. Brains were either fixed in formalin prior to embedding in paraffin or frozen immediately at −80°C. Cryostat sections were fixed in methanol and acetone. Deparaffinized sections were labeled for glial fibrillary acidic protein (GFAP) and a growth cone-associated protein of 43 kDa (GAP-43). Frozen sections were labeled with lectin. Fifteen-micrometer-thick sections from coronal tissue containing white matter lesions and from comparable anatomic areas of the contralateral untreated hemisphere were reacted with anti-GFAP (Dako, Glostrup, Denmark), anti-GAP-43 (Roche, Meylan, France) or biotinylated Griffonea simplicifolia I isolectin B4 (Vector, Burlingame, CA). Detection of labeled antigens was performed with avidin–biotin–horseradish peroxidase kits (Vector) according to manufacturer's instructions.
To avoid regional and experimental variation in the intensity of labeling with GAP-43, ibotenate-injected hemispheres were compared with the untreated contralateral hemispheres at the same anatomic level. Several sections from each animal were assigned to different or batches for immuno reaction. Moreover, two different investigators independently performed qualitative analyses. Investigators focused on white matter around the cystic lesion to quantify GFAP and lectin-labeled cells. For each animal, cell counts were performed in a 0.03 mm2 area of the most affected section.
Cell Death Staining
Twenty-four hours following parenchymal injection of ibotenate plus 50 ng BDNF or ibotenate plus PBS at P0 or P5, pups were sacrificed and their brains fixed in formalin and embedded in paraffin. Every third section was stained with cresyl-violet. Sections adjacent to the most affected areas were used for terminal transferase-mediated dUTP nick end (Tunel) labeling, using an in situ cell death detection kit as recommended (Roche, Meylan, France). Tunel positive cells were counted in a 1 mm2 area in the cortical layers at the level of ibotenate-induced lesions. Ten nonadjacent fields from five brains were studied in each group.
Immunohistochemistry for Cleaved Caspase 3
Sections adjacent to those used for Tunel staining were reacted with anti-cleaved caspase 3 antibody (Cell Signaling, Beverly, MA). Detection of labeled antigens was performed with avidin-biotin horseradish peroxidase kits (Vector) according to manufacturer's instructions. Labeled cells were counted in a 1 mm2 area in the neocortical layers at the level of ibotenate-induced lesions. Ten nonadjacent fields from five brains were studied in each group.
Real Time PCR Analysis of the Expression of BDNF Gene Variants
To study the effects of VIP on BDNF gene expression following an excitotoxic insult, P5 pups were given intraparenchymal injections of 10 μg ibotenate plus 1 μg VIP or of 10 μg ibotenate plus PBS. Animals were sacrificed at four h after injection. Ten animals were included in each treatment group. Brains were removed and tissues immediately adjacent to the site of injection were collected from the different animals for RNA extraction. Total RNA was extracted as previously described (Lelievre et al., 2002), followed by DNase I treatment to remove any trace of genomic DNA contamination. For each sample, 600 ng were used in reverse transcription (iScript kit from BioRad, Hercules, CA, USA) according to the manufacturer's specifications. Further analysis of BDNF and β2-microglobin mRNA levels was performed by real-time PCR. We first design selective sets of primers that selectively amplify either all the BDNF variants (i.e. BDNF-5) or specific transcripts (BDNF-1, -2a, −3, or −4). Nucleotide sequences encoding for the different BDNF variants were aligned (ClustalW multialignment software available at http://clustalw.genome.ad.jp/) to determine locations and sequences that selectively identify one variant from another. Areas of overlapping for these highly-selective nucleotidic sequences were first analyzed for secondary structures using M-fold software (http://www.bioinfo.rpi.edu/applications/mfold/) to ensure absence of secondary structures. M-fold-targeted regions were imported into Oligo6 software (Molecular Biology Insights, Cascade, CO) to design highly-stringent primer sets. For BDNF variants, we chose the following oligonucleotides 5′- AGGACAGCAAAGCCACAATGT-OH and 5′-CCTTCATGCAACCGAAGTATG, 5′-AGTACTTCATCCAGTTCCACC-OH and 5′-CGTTTACTTCTTTCATGG-OH, 5′-AGCCCAGTTCCACCAG-OH and 5′-CATGCAACCGAAGTATGAAAT-OH, 5′-AGCAGCTGCCTTGATGTTTAC-OH and 5′- ATGCAACCGAAGTATGAAATA-OH, and finally 5′-AACAATGTGACTCCACTGCC-OH and 5′-ACCGAAGTAAATAA CCAT-OH, as sense and antisense primers, for BDNF-1, -2a, -3, -4 and -5, respectively. PCR amplification resulted in the generation of a single band at 92, 110, 75, 112 and 133 bp, corresponding to the regions 621–713, 191–301, 125–200, 295–407 and 277–410 of the previously published sequences (NCBI access no. AY057-908, -909, -912, -913 and -914) of mouse BDNF-1 to -5 variants, respectively. To standardize the experiments we designed, using the same approach, a primer set (5-CCGGCTTGTATGCTATC and 5-TCAGTATGTTCGGCTTC, as sense and anti-sense respectively), for the mouse β2-microglobulin gene. These primers amplified an 87 bp region encoding the nucleotides 99–185 of the published sequence (MMB2MR) of the mouse mRNA. Amplified BDNF variants and β 2-microglobulin bands were cloned into PCRII and sequenced to confirm identity. Real time PCR was set up using sybergreen-containing supermix from BioRad, for 58 cycles of a three-step procedure including a 30 s denaturation at 96°C, a 30 s annealing at 64°C, followed by a 30 s extension at 72°C. Amplification specificity was assessed by melting curve. Quantification utilized standard curves made from serial dilutions of control RNA sample or of the corresponding cDNA cloned into PCRII vector. Differences between samples were calculated as the difference between the specific ratios (BDNF variant/β2-microglobulin) calculated for each individual sample. Experiments were independently run twice. In each experiment, samples were performed in triplicates.
ELISA Determination of BDNF Content in Supernatants of Cultured Astrocytes
To study the effects of VIP on BDNF production and/or release by astrocytes, primary cultures of astrocytes were prepared from P2 mouse brains as previously described (Laudenbach et al., 2002). Astrocytes were plated at a density of 1.2 millions cells and cultured in DMEM (Invitrogen, Renfrew, UK) enriched with 10% horse serum. The cultures were maintained at 37°C in a humidified 95% air–5% CO2 atmosphere for 2–3 weeks. Fifteen hours prior to VIP exposure, culture medium was removed and cells were incubated in serum-free DMEM. Astrocytes were then treated for 1 h with 1 μM VIP or PBS. The supernatants of cultures were removed, centrifuged and frozen at −80°C. BDNF content was measured using an ELISA kit (Quantikine human BDNF, R&D, Lille, France) as directed. Samples were run in triplicate.
To study the effect of each treatment most of the data were analyzed with a Student's t-test or a one-way analysis of variance (ANOVA). In ANOVA analyses, when a main effect of treatment group was found to be significant, a Dunnett's multiple comparison tests were performed. In the subset of experiments where white matter lesion size and white matter density of GFAP-positive and lectin-positive cells were evaluated at different timepoints after ibotenate injection, results were studied using a two-way ANOVA with Treatment and Age (time elapsed after injection) as between-subject factors. When a main effect of Treatment or Age or their interaction was found to be significant, we conducted pair-wise comparisons between treatment groups at each age.
Mortality and Epileptic Manifestations
Overall mortality was low in the present study, with death seen in <5% of the animals injected at each developmental age studied (P0, P5 and P10). No significant difference was observed in a test of contingency (exact Fisher test) when the different treatment groups were compared, at each developmental age, with the animals injected with ibotenate and PBS. Epileptic manifestations including clonic or tonic seizures and apneas were observed in all excitotoxin-treated animals. However, BDNF treatment did not induce any difference in severity and frequency of seizures (frequency of seizures was quantitatively assessed during a 10 min period, once every hour during the first 6 h following excitotoxin injection; severity of seizures was qualitatively assessed according to the same schedule) when compared with controls (data non shown).
BDNF-induced Neuroprotection of the White Matter on P5
As previously described (Marret et al., 1995; Gressens et al., 1997; Tahraoui et al., 2001), although ibotenate induces cortical gray matter lesions when injected at P0, P5 or P10 (see below), it only produces significant white matter lesions when given at P5. Due to the stage-dependent enhanced sensitivity of the periventricular white matter to excitotoxic insults on P5, we focused our study of BDNF effects on periventricular white matter lesions at this developmental stage.
Co-injection of ibotenate and BDNF at P5 induced a dose-dependent reduction (up to 60% with 50 ng BDNF) of the white matter lesion size when compared with pups treated with ibotenate and PBS (Fig. 1A–F). BDNF-induced neuroprotection against ibotenate insults was replicated in pups treated with NMDA (79% reduction of lesion size when compared with pups treated with NMDA and PBS) (Fig. 1G). Furthermore, BDNF was shown to significantly protect against S-bromowillardiine-induced white matter lesions (50% reduction of lesion size when compared with pups treated with S-bromowillardiine and PBS) (Fig. 1G).
Addition of neutralizing BDNF antibody completely blocked BDNF-induced neuroprotection (Fig. 2). In addition, while co-treatment with VIP and ibotenate significantly protected the white matter, co-injection of ibotenate, VIP and neutralizing BDNF antibody significantly and partially reversed VIP-induced neuroprotection against ibotenate (Fig. 2).
Effects of BDNF on White Matter Axons, Astrocytes and Activated Microglia/Macrophages after an Excitotoxic Insult on P5
In animals injected with ibotenate and PBS, the white matter lesion increased in size during the first 24 h and remained stable thereafter (Fig. 3A). Co-treatment with ibotenate and BDNF or VIP did not modify the lesion size during the first 24 h, but induced, secondarily, a dramatic regression of the white matter lesion during the next 4 days (Fig. 3A).
Within 24 h after injection and thereafter, GAP-43 immunostaining around the white matter cyst was denser, when compared with the non-injected hemisphere, both in ibotenate-PBS and ibotenate-BDNF groups. In animals co-treated with BDNF and ibotenate, GAP-43 labeled axons emerging from these more densely labeled rims of the cyst and crossing the cyst were observed between the second and the fourth days after injection while brains treated with ibotenate and PBS did not display similar findings (Fig. 4A,B).
The number of white matter GFAP-positive cells around the site of injection significantly dropped during the first 24 h in animals co-treated with ibotenate and PBS or BDNF while VIP co-treatment largely inhibited this astrocytic disappearance (Figs 3B and 4C–E). Two days after ibotenate injection and thereafter, a progressive and intense reactive gliosis was observed around the lesion and, later on, in the whole injected hemisphere (Figs 3B and 4F). Co-treatment with VIP, but not BDNF, and ibotenate significantly reduced the extent of this reactive gliosis by 96 h after the insult (Figs 3B and 4G,H).
Griffonea simplicifolia I isolectin B4 is a marker of reactive microglia. Immunostaining with this lectin revealed a rapidly growing population of labeled cells in the ibotenate-injected hemisphere, centered on the white matter cystic lesion. Labeled cells appeared as early as 4 h after treatment, and the number of cells increased during the first 3 days (Figs 3C and 4I). Co-treatment with BDNF did not significantly modify the pattern of lectin-labeled cells in the white matter following ibotenate injection while VIP slightly reduced the intensity of microglial reaction during the first 8 hours following the excitotoxic challenge (Figs 3C and 4J).
Signal Transduction Mechanisms Mediating BDNF-induced Neuroprotection of the White Matter on P5
Co-treatment with ibotenate, BDNF and PD98059 (a MAPK kinase or Mek-1 inhibitor and thereby an inhibitor of the MAPK cascade) or K252a (an inhibitor of TrkB receptors) completely abolished BDNF-induced neuroprotection of the white matter (Fig. 5). In contrast, H89 (a PKA inhibitor), bisindolylmaleimide I (a PKC inhibitor) and neutralizing anti-p75NTR antibodies had no detectable effects on BDNF-induced neuroprotection (Fig. 5).
Neuroprotective Effects of BDNF on P5 Excitotoxic Cortical Gray Matter Lesions
Ibotenate injection on P5 produced white matter cysts (see above) and severe neuronal loss in neocortical layers II to VI, with a complete or almost complete disappearance of neuronal cell bodies in all layers along the most affected radial axis of the lesion (Fig. 1A). Co-injection of ibotenate and 50 ng BDNF induced a moderate but significant reduction of the lesion size while no detectable effect was observed with lower doses of BDNF (Figs 1B and 6A). BDNF-induced neuroprotection against ibotenate-induced cortical gray matter lesion in P5 pups was replicated in pups treated with NMDA but not in pups exposed to S-bromowillardiine (Fig. 6A).
Neuroprotective effects of BDNF were reversed by co-treatment with neutralizing anti-BDNF antibodies, PD98059 or K252a but not with neutralizing anti-p75NTR antibodies, H89 or bisindolylmaleimide I (Fig. 6B). Co-injection of ibotenate and VIP also moderately but significantly reduced the size of the cortical gray matter lesion size and this neuroprotective effect of VIP was almost completely abolished by co-administration of neutralizing anti-BDNF antibodies (Fig. 6B).
The neuroprotective effects of BDNF on ibotenate-induced cortical gray matter lesions in P5 pups were confirmed by Tunel staining and immunohistochemistry for cleaved caspase-3 (Fig. 7).
Effects of VIP on P5 BDNF Expression and on BDNF Release by Cultured Astrocytes
Real time PCR performed on extracts collected at the site of intraparenchymal injection showed a significantly increased expression of BDNF mRNA (as demonstrated by BDNF-5 primers which recognize all BDNF mRNA variants) following ibotenate + VIP injection on P5 when compared to ibotenate + PBS injection (Fig. 8A). In particular, VIP induced a significant increase of BDNF-1 and -3 mRNA variants, while BDNF-2a and -4 mRNA variants were not significantly affected (Fig. 8A).
ELISA performed on supernatants of cultured astrocytes showed a significantly increased BDNF concentration in cultures treated for one hour with VIP when compared with cultures exposed to PBS alone (Fig. 8B).
Toxic Effects of BDNF on P0 Excitotoxic Cortical Gray Matter Lesions
Ibotenate injection on P0 induced a severe neuronal loss predominating in neocortical layers V and VIa (Fig. 9A), without detectable white matter cystic lesion. Generally, a microgyric sulcus was present due to the sagging of the molecular and the upper cortical layers. Co-injection of ibotenate and lower doses (0.05–5 ng) of BDNF did not modify the size of the cortical gray matter lesion while 50 ng BDNF significantly exacerbated the severity of the lesion (Fig. 9B,C). High dose BDNF-induced toxicity was reversed by co-treatment with neutralizing anti-p75NTR antibodies but not with K252a or PD98059 (Fig. 9C). As previously described (Gressens et al., 1997), co-injection of VIP and ibotenate was neuroprotective but, surprisingly, VIP-neuroprotective effects were abolished by co-treatment with neutralizing anti-BDNF antibodies or with PD98059 (Fig. 9C).
The toxic effects of 50 ng BDNF on ibotenate-induced cortical gray matter lesions in P0 pups was confirmed by Tunel staining cleaved caspase-3 immunostaining (Fig. 10).
Lack of BDNF Effect on P10 Excitotoxic Cortical Gray Matter Lesions
Ibotenate injection on P10 induces in all pups injected a neuronal cell death in all cortical layers (data not shown) and, in some rare occasions, a white matter cyst. Due to their low incidence and small size, these white matter lesions induced by ibotenate injection on P10 were not considered in the present study. Co-injection of ibotenate and 50 ng BDNF or 1 μg VIP did not significantly modify the size of the excitotoxic cortical gray matter lesion (mean ± SEM cortical gray matter lesion size = 710 ± 63 μm for controls, 880 ± 79 μm for BDNF-treated pups and 800 ± 147 μm for VIP-treated pups).
The most salient feature of the present study is the demonstration that BDNF had a dose-dependent dramatic neuroprotective effect against excitotoxic lesions in the white matter of P5 mice, whereas only a modest neuroprotective effect against lesions of the cortical gray matter of P5 mice was observed, and seen only at the highest dose employed. In addition, the modest neuroprotective effect of BDNF on excitotoxic neuronal cell death was not evident in younger or older mice: high doses of BDNF exacerbated P0 excitotoxic neuronal death, and did not show any significant effect on P10 excitotoxic neuronal death. The fact that the neuroprotective effects of BDNF evident in P5 mice were blocked by MAPK pathway and TrkB inhibitors while the toxic effects evident in P0 mice were blocked by anti-p75NTR neutralizing antibodies implicates specific and distinct signal transduction pathways in mediating such actions, providing unique mechanistic insights and therapeutic targets for drug development to prevent brain injury in humans born prematurely.
Neuroprotection Conferred by BDNF on Excitotoxic Injury in the Developing Brain: Comparison of our Findings with Those of Others
The neuroprotective effects of BDNF against NMDA receptor-mediated neuronal cell death induced at P5 are in general agreement with previous studies showing that exogenous BDNF very significantly protect the P7 rat brain against an hypoxic-ischemic insult (Cheng et al., 1997; Han and Holtzman, 2000). The relatively moderate neuroprotection afforded by BDNF in our study when compared with the rat model might be related to (i) species differences; (ii) differences of neuronal maturation; and/or (iii) differences in mechanisms of brain injury, specifically involving potential differences in the relative contribution of glutamate receptor subtypes. The latter hypothesis is supported by the fact that, in the present study, BDNF was efficient in preventing NMDA receptor-mediated neuronal cell death but not AMPA-kainate receptor-mediated cell death.
A few previous studies have reported toxic effects of BDNF on cultured neurons (Gwag et al., 1995; Koh et al., 1995; Bamji et al., 1998). When studied, the molecular mechanisms generally involved p75NTR receptor-mediated pathway (Bamji et al., 1998; Majdan and Miller, 1999), although TrkB receptor has also been implicated (Fryer et al., 1997; Giehl et al., 2001). In the present study, the exacerbating effect of high doses BDNF on P0 cortical neurons exposed to BDNF was mediated by p75NTR receptor. Despite testing a large range of BDNF concentrations, we could not elicit any neuroprotective effect of BDNF against excitotoxic neuronal cell death at P0, suggesting that the toxic effect observed with 50 ng BDNF was real, and not the result of corresponding to a dose on the ‘right part of a bell-shape curve of neuroprotection’. BDNF-induced exacerbation of excitotoxic death of P0 cortical neurons involved, at least partially, an increase of neuronal apoptosis as shown by Tunel and caspase staining. Interestingly, although exogenous BDNF administered at P0 was inactive or toxic according to the dose administered, co-treatment with exogenous VIP and ibotenate was neuroprotective and this VIP-induced neuroprotection seemed to be mediated, at least partially, through endogenous BDNF since neutralizing anti-BDNF antibodies reversed VIP-induced neuroprotection.
Mechanistic Basis for the Neuroprotection Conferred by BDNF on Excitotoxic White Matter Lesions in P5 Mice
Glutamate Receptor-subtype Specificity
When comparing different glutamate receptors agonists, BDNF-induced neuroprotection was of the largest magnitude against a pure NMDA receptor-mediated insult (NMDA injection) when compared with NMDA + metabotropic (ibotenate) or AMPA + kainate (S-bromowillardiine) receptor-mediated in juries. These differences in neuroprotective effects of BDNF probably reflect differences in mechanisms of injury when mediated by separate glutamate receptors (Tahraoui et al., 2001).
Natural History of White Matter Lesions and Neuroprotection by BDNF
The time course of the evolution of white matter lesion size showed that BDNF did not prevent the initial lesion but rather induced a secondary decrease of white matter lesion size. Although we can not fully exclude the unlikely possibility that BDNF induces a rapid shrinkage of the cyst, data provided by GAP-43 immunostaining are compatible with the conclusion that BDNF induces axonal regrowth or sprouting, in agreement with previous studies performed in adult rats (Mamounas et al., 2000). Our working hypothesis is that spared neurons around the focal lesion or distant neurons having axons passing in the vicinity of the cyst represent potential sources for newly produced axonal processes following BDNF treatment. It should be noted that BDNF did not significantly affect the ibotenate-induced microglial activation or the ibotenate-induced changes in astroglial density, supporting the hypothesis that BDNF acts directly on neurons. Future behavioral studies in BDNF treated mice will be necessary to fully assess the functional extent of BDNF-induced neuroprotection.
Mechanistic Insights Provided via Comparison of Neuroprotection by BDNF as Compared with VIP
We previously showed (Gressens et al., 1997), and confirmed in this study, that VIP protected the white matter against ibotenate-induced lesions. The combination of the present data and previous studies (Gressens et al., 1997, 1998) permit us to propose the following formulations regarding the neuroprotective actions of VIP and BDNF in our mouse model of white matter injury: (i) VIP binds to a specific receptor on white matter astrocytes and activates a PKC pathway; (ii) PKC activation promotes astrocytic survival and astrocytic production and/or release of soluble factors including BDNF; and (iii) BDNF activates a MAPK cascade in neurons, leading to white matter neuroprotection, possibly involving axonal sprouting. Several lines of experimental evidence from our studies support this working hypothesis: (i) VIP and BDNF did not prevent the initial appearance of white matter lesions, but promoted a secondary decrease of the lesion size; (ii) VIP, but not BDNF, inhibited ibotenate-induced astrocytic death; (iii) VIP protective effects were reversed by both PKC and MPAK inhibitors while BDNF effects were only blocked by MAPK pathway inhibitors; (iv) VIP effects were partially reversed by a neutralizing anti-BDNF antibody; and (v) VIP induced an in vivo increased expression of BDNF-3 variant and an in vitro increased BDNF concentration in the supernatant of cultured astrocytes. Previous studies have shown that VIP-treated cultured astrocytes produce and/or release growth factors, including cytokines (Brenneman et al., 1995), protease inhibitors (Festoff et al., 1996) and BDNF (Pellegri et al., 1998). In the present study, cortical neurons could also contribute to the observed in vivo increased expression of BDNF-3 variant following VIP injection, as VIP can induce BDNF expression in cultured neurons (Pellegri et al., 1998).
Effects of BDNF on Excitotoxic Cortical Gray Matter Lesions
One key result of the present study is the demonstration that BDNF effects (toxicity, neuroprotection or no effect) on neonatal excitotoxic neuronal cell death strongly depends upon the stage of cortical maturation. Several factors could explain these differences: (i) the mechanism and laminar distribution of ibotenate toxicity vary according to the stage of postnatal neocortical maturation (Marret et al., 1995; Gressens et al., 1997); (ii) neuronal sensitivity to BDNF is age-dependent due to changes in the maturation of the cellular response to TrkB activation (Knusel et al., 1994; Cheng et al., 1997); (iii) diffusability or degradation of BDNF could be different in particular neocortical layers according to intrinsic differences in neuronal maturation and differentiation; and (iv) whereas TrkB receptors are expressed by neurons in all cortical layers, p75NTR expression within the developing neocortex is highly restricted to the deep cortical layers (Allendoerfer et al., 1990).
As previously reported (Han and Holtzman, 2000; Han et al., 2000), BDNF-induced neuroprotection of P5 cortical neurons was mediated by TrkB receptors and the MAPK pathway, and involved, at least partially, prevention of neuronal apoptosis, as shown by Tunel and caspase staining. Interestingly, VIP treatment induced a similar moderate but significant protection of cortical neurons against ibotenate and this VIP-mediated protective effect was completely abolished by co-treatment with neutralizing anti-BDNF antibody, suggesting that, as for the white matter (see above), VIP is working through a BDNF-mediated pathway.
In conclusion, the present study showed that the neuroprotective profile and mechanisms of BDNF against an excitotoxic insult in the newborn brain are dependent upon the type of glutamate receptors involved, the localization (gray versus white matter) of the lesion, the stage of brain development and probably the molecular context of the neural cells. These findings suggest that BDNF therapy may be particularly promising to prevent white matter injury evident in a subset of prematurely born human infants, as may be seen in conditions such as periventricular leukomalacia.
We thank Philippe Evrard for his support, Jorge Gallego for his help with statistical analysis and Latifa Ferkdadji for her help with cryostat sectioning. Supported by the INSERM, the Fonds d'Etudes et de Recherche du Corps Médical des Hôpitaux de Paris, the Société Française de Neurologie and the Fondation Grace de Monaco.
1INSERM E 9935 & Service de Neuropédiatrie, Hôpital Robert-Debré, F-75019 Paris, France, 2Mental Retardation Research Center, University of California at Los Angeles, Neuropsychiatric Institute, Los Angeles, CA 90095, USA, 3CNRS UMR 7091, Bâtiment CERVI, Hôpital de la Pitié-Salpétrière, F-75013 Paris, France and 4Laboratory of Molecular & Developmental Neuroscience, Boston, MA 02129, USA