Localized delivery of fibroblast growth factor–2 and brain-derived neurotrophic factor reduces spontaneous seizures in an epilepsy model

Vectors.

A schematic representation of the vectors we used is shown in supporting information (SI) Fig. S1. For FGF-2, we used a previously characterized vector, TH-FGF2 (13), in which the transgene was under the control of the human cytomegalovirus (HCMV) immediate-early (IE) promoter. For BDNF, we used a vector, T0-BDNF, in which the transgene was under the control of the ICP0 promoter. The double mutant (TH-FGF2/0-BDNF) was obtained by crossing over TH-FGF2 and T0-BDNF. Promoters were chosen to promote a synergy between the NTFs, making use first of the FGF-2 property of potently inducing proliferation of hippocampal progenitors and then of the BDNF property of inducing their neuronal differentiation: the HCMV IE promoter driving FGF-2 is known to ensure robust, but transient, transgene expression, whereas the ICP0 promoter driving BDNF provides a longer-lasting expression (14).

The vector-induced transgene expression and the biological activity of the NTF proteins produced by infected cells were first studied in vitro. The data we obtained (SI Text “In vitro Characterization” and Fig. S2) indicate that: (i) vectors induce production and release of FGF-2 and BDNF at biologically active levels; and (ii) FGF-2 and BDNF supplied together, not individually, potently favor survival, proliferation, and neuronal differentiation of neural progenitors. Therefore, we elected to test the vector expressing FGF-2 and BDNF together in vivo.

The efficiency of infection and the time-course of transgene expression in vivo was examined after a single stereotaxical inoculation of the vector in the hippocampus of adult rats (SI Text “Transgene Expression in Vivo” and Fig. S3). The data show that inoculation of TH-FGF2/0-BDNF provides a bilateral, short-term (≈1 week) increase in FGF-2 expression accompanied and followed by a bilateral, slightly longer-lasting (at least 11 days) increase in BDNF expression, and that these events are not associated with significant toxicity.

Effects on SE-Induced Cell Death and Neurogenesis.

We tested TH-FGF2/0-BDNF on the pilocarpine SE–induced neurodegeneration and epileptogenesis. In this epilepsy model, an episode of SE produces intense neuronal damage and, after a latent period of 2–3 weeks, SRSs, i.e., epilepsy (15).

In the present experimental series, pilocarpine (300 mg/kg i.p.) rapidly induced a robust convulsive SE (latency, 15 ± 2 minutes) that was interrupted after 2 hours through administration of the anticonvulsant diazepam (10 mg/kg i.p.). Based on behavioral observation and EEG recordings, the severity of SE in the different animals was indistinguishable. This procedure caused damage in several brain areas: in particular (16, 17), hippocampal damage closely resembled human hippocampal sclerosis and was invariably remarkable. Typically, 3 days after pilocarpine administration, Fluoro-Jade C (FJC) staining allowed identification of several degenerating cells in CA1, CA3, and the hilus of the dentate gyrus. A pronounced edema with cell loss was also evident. A three-dimensional structural reconstruction demonstrated a 27.8 ± 3.1% reduction of hippocampal volume 3 days after pilocarpine administration, as compared with the response on control animals (P < 0.001, Mann-Whitney U test for unpaired data).

Three days after SE, these lesioned animals were randomly assigned to three groups: one group was injected in one hippocampus with the vector expressing FGF-2 and BDNF; the second group was injected with a control vector; and the third group was not treated at all (no difference was observed between these latter two groups in any of the parameters examined, and therefore they have been pooled together for statistical analysis, and collectively termed “control”). To explore the mitogenic effect of the vectors, animals were treated with bromo-deoxyuridine (BrdU, four 50-mg/kg i.p. injections, one every 2 hours) 6 days after SE (i.e., 3 days after vector inoculation, Fig. S4). This time point was chosen because it is both within the peak of SE-induced proliferation of neural stem cells (18) and of vector-induced overexpression of FGF-2 (see above). A subset of animals in each group was killed 14 days after SE (i.e., 11 days after vector inoculation) to examine possible effects on cell survival and neurogenesis, and another subset was killed 28 days after SE (25 days after vector inoculation) to examine tissue morphology. These latter animals also underwent 20 days behavioral and EEG monitoring, to examine the outcome of the treatment (Fig. S4).

Although the bulk of SE-induced damage is in the first few days and, thus, was already in place when we injected the vector, several degenerating cells were still observed 14 days after pilocarpine administration in different hippocampal subfields (CA1, CA3, hilus of the dentate gyrus; Fig. 1Aa and Ac). Administration of the vector expressing FGF-2 and BDNF did not alter this ongoing cell loss, except for a slight, nonsignificant attenuation in the injected hippocampus (Fig. 1), indicating that, in vivo, the neuroprotective effect of this vector is limited or may require more prolonged or higher-level transgene expression.

In contrast to the limited effect on cell death, the effect of the vector expressing FGF-2 and BDNF on neurogenesis was quite remarkable. In the adult hippocampus, neural progenitors are found both in the subgranular zone (SGZ) and in the caudal subventricular zone (SVZ). SE has been reported to increase proliferation of hippocampal progenitors in both areas, but SGZ progenitors differentiate into neurons that, in large part, migrate ectopically to the hilus and SVZ progenitors differentiate into glial, not neuronal, cells: that is, SE-induced neurogenesis is aberrant and may contribute to epileptogenesis (19, 20).

We used pulse-chase BrdU labeling and immunohistochemistry for markers of progenitors to determine the mitogenic and differentiative activity of the NTFs supplemented by the vectors. To explore the mitogenic effect, we counted BrdU-positive cells in the dorsal hippocampus 14 days after pilocarpine. As expected, SE per se induced a significant increase in the number of mitotically active cells (Fig. 2 and Fig. S5). In keeping with previous reports (6, 18, 20), this increase was observed in the dentate gyrus and in the pyramidal gyrus (especially CA1 strata oriens and lacunosum-molecularis, near the SVZ). Inoculation of the control vector did not alter this pattern. In contrast, inoculation of the vector expressing FGF-2 and BDNF produced a further, bilateral, highly significant increase in the number of BrdU-positive cells (Fig. 2A), indicating proliferation of progenitors. Notably, an increased number of positive cells was also observed in the ventral hippocampus (Fig. S6).

In both the SGZ and SVZ, a series of differentiation steps have been described: from early, glia-like progenitors (termed type-1 for SGZ and type-B for SVZ), to transiently amplifying cells (type-2 and type-C), to late, migrating cells (type-3 and type-A) (21, 22). Type-1 and type-B cells are positive to GFAP, nestin, and Sry-related HMG box transcription factor (Sox2); type-2 and type-B cells to nestin and Sox2, but not to GFAP; type-3 and type-A cells to doublecortin (DCX) (22, 23). Pilocarpine SE per se significantly increased the number of nestin-positive cells in the hippocampus as compared with naïve controls, but treatment with TH-FGF2/0-BDNF dramatically potentiated this effect, more markedly in the injected hippocampus (Fig. 2A). Many (≈20%) of these nestin-positive cells were also BrdU-positive, indicating that part of them were actively proliferating at the time of BrdU administration (Fig. 2A). Furthermore, a subset of these cells was also GFAP-positive (Fig. S5B), an indication of the presence of early progenitors. Finally, we observed an increased number of Sox2-positive cells in pilocarpine animals treated with TH-FGF2/0-BDNF, 7 days after SE (Fig. S5D). These data indicate that SE per se causes proliferation of hippocampal early to intermediate progenitors and that vector-mediated supplementation of FGF-2 and BDNF greatly increases this phenomenon.

In the subsequent steps of neurogenesis, progenitors begin to migrate and become DCX-positive. As stated above, this migration is partly aberrant after SE, because many SGZ progenitors do not head toward the granular layer but toward the hilus (19). Therefore, we also analyzed DCX-positive cells. Under control conditions (naïve animals), a limited number of DCX cells, with elongations projecting across the granular layer, was found in the SGZ (Fig. 2Ba) and even less so in the SVZ (essentially, just a few, weakly labeled DCX-positive cells in the CA1 stratum oriens). SE per se caused a robust increase in the number of DCX-positive cells in the dentate gyrus (Fig. 2Bb). However, these cells did not often produce elongations, tended to aggregate into clusters, and were often ectopically located (Fig. 2Bb). These observations are consistent with the aberrant aspects of SE-induced neurogenesis (19, 24). A slightly increased number of DCX-positive cells was also observed in the pyramidal gyrus (CA3 and CA1). This pattern was not modified by the administration of the control vector. In contrast, administration of TH-FGF2/0-BDNF in pilocarpine-treated animals caused a greater increase in the number of DCX-positive cells compared with pilocarpine alone or pilocarpine and control vector (Fig. 2B). Moreover, it increased the length and number of elongations, reduced cluster formation, and reduced the number of ectopic cells (Fig. 2Bc). An increased number of DCX-positive cells was also observed in the pyramidal gyrus, indicating commitment to neuronal differentiation (Fig. S5E).

Finally, we checked whether, consistent with our in vitro data, a subset of these new cells were differentiating into neurons and migrating into the appropriate location to possibly restore function. Thus, we performed double-label immunofluorescence with BrdU and MAP2abc, a marker of immature and mature neurons. SE per se did not significantly increase the number of BrdU-MAP2–positive cells; in contrast, we found several double-labeled cells in the hippocampus of pilocarpine animals treated with the double mutant (Fig. 2A).

These data suggest that injection of the vector expressing FGF-2 and BDNF in a SE-damaged hippocampus favors proliferation of early progenitors according to a pattern that, at variance with what was observed in untreated animals, appears to follow the physiological pattern more faithfully, leading to the production of cells that enter the neuronal lineage of differentiation and reducing the aberrant aspects of SE-induced neurogenesis.

Outcome: Pathology and Behavior.

To evaluate the outcome of the treatment, we compared the different experimental groups for the morphology of the hippocampus. First: compared with naïve controls, the dorsal hippocampus of pilocarpine animals remained smaller 4 weeks after SE (approximately −30%, like 3 days after SE: see above). Treatment with the double mutant produced a significant increase in hippocampal volume, with a recovery of approximately a third of the loss (Fig. S7A).

Importantly, the SE-damaged hippocampus presents a clearly reduced density of neuronal cells, most notably hilar and pyramidal cells, that was not observed in animals treated with the double mutant (Fig. S7). We measured the density of astrocytes and neurons in the hippocampus using GFAP and NeuN immunofluorescence. The density of GFAP-positive cells in the entire hippocampus was significantly increased in pilocarpine-treated animals compared with naïve controls, an indication of reactive astrocytosis (25). In the pilocarpine group of animals treated with the vector expressing FGF-2 and BDNF, the density of GFAP-positive cells decreased slightly in comparison with untreated animals, but was still greater than in naïve animals (Fig. S7B). Neuronal density was measured using NeuN. Putatively excitatory dentate gyrus granule cells were not significantly decreased by SE and treatment with the double mutant did not alter the situation (Fig. 3A). In contrast, putatively excitatory pyramidal neurons were dramatically decreased in animals that experienced SE, and the double mutant significantly increased their number (Fig. 3A). Putatively inhibitory neurons in the dentate hilus and in the stratum oriens of CA1 were even more dramatically decreased in animals that experienced SE; and, again, TH-FGF2/0-BDNF significantly improved the situation (Fig. 3B).

These data suggest that treatment with the vector expressing FGF-2 and BDNF partially heals the damage induced by SE, maintaining a good ratio between excitation and inhibition in the circuit. However, some inhibitory cells are present in the pyramidal layer and excitatory cells are present in the hilus (for example, the glutamateric mossy cells). Thus, we deepened the investigation analyzing some of the main populations of GABAergic interneurons, namely parvalbumin, somatostatin, and calbindin. Parvalbumin-containing interneurons exert perisomatic, whereas somatostatin and calbindin exert dendritic inhibition; calbindin also labels a subpopulation of principal, glutamatergic cells, both in the pyramidal and in the granular layer (26). In keeping with previous reports (27–29), we found that the number of parvalbumin-positive cells in epileptic animals was similar to that in controls (Fig. S8A), while that of somatostatin-positive cells was halved (Fig. S8B). Also, calbindin-positive principal neurons in the pyramidal and granular layer, but not calbindin-positive interneurons in the strata oriens and radiatum and in the dentate gyrus hilus, were significantly decreased in the pilocarpine group (Fig. S8C) (30, 31). Treatment with TH-FGF2/0-BDNF effectively corrected these cell losses (Fig. S8).

Avoiding the loss of somatostatin cells may imply maintenance of dendritic inhibition and, therefore, decreased susceptibility to seizures. This leads to the key question: do these beneficial effects ameliorate the outcome in terms of SRSs? We performed behavioral and EEG analyses: animals were video-EEG monitored for 20 days and the occurrence, severity, and duration of SRSs were recorded. As expected, all non vector-injected pilocarpine-treated rats exhibited SRSs beginning ≈2 weeks after SE. On average, we observed in this group four seizures per day (Fig. 4A) lasting about 1 minute each and with a mean severity score (32) of 2.5 (Fig. 4B). Administration of the control vector did not modify this pattern. In contrast, rats treated with the double mutant displayed a highly significant improvement: (i) a subset of animals (two of 11) never developed SRSs during the time frame of observation; (ii) the average number of seizures per day in this group was highly significantly reduced compared with the other two groups (Fig. 4A); (iii) the seizure severity was also significantly decreased (Fig. 4B). Thus, the FGF-2 and BDNF-expressing vector, injected in the hippocampus after the establishment of hippocampal sclerosis, provided both significant recovery from damage and significant reduction of SRSs. These results are striking, considering that this epilepsy model is accompanied by extensive damage also in extra-hippocampal regions (33), such that only a subset of SRSs may actually originate in the hippocampus (34), i.e., where we documented favorable effects by the vector. In other words, SRSs originating in the hippocampus may be essentially abolished by vector treatment.

Finally, to test whether the vector has a truly antiepileptogenic effect and is not merely dampening SRSs (i.e., blocking ictogenesis), we investigated whether the vector expressing FGF-2 and BDNF was effective in controlling seizures when injected in animals that were already experiencing SRSs. To address this question, a group of pilocarpine-treated rats was video-EEG monitored beginning 2 weeks after SE. Three weeks after the first spontaneous seizure, these animals were injected either with the control vector or with TH-FGF2/0-BDNF, and monitoring was continued for 4 more weeks. The frequency, duration, and severity of SRSs were then comparatively analyzed in the 2 weeks preceding vector inoculation and in the third and fourth weeks after inoculation. Inoculation of the FGF-2/BDNF expressing vector did not affect any of the parameters examined (Fig. S9), indicating that supplementation of FGF-2 and BDNF can interfere with epileptogenesis but not with ictogenesis.

In conclusion, the main finding of this study is that supplementation of FGF-2 and BDNF in the hippocampus lesioned by prolonged seizures effectively limits or repairs damage, maintaining a good ratio between excitation and inhibition, and significantly improves the natural history of the disease.

Understanding in depth the mechanistic basis of these effects will require further studies, but the present data provide some initial insights. FGF-2 and BDNF may provide protection against damage or gain of function of partially damaged neurons. Even if the present study did not provide support to this idea, subtle effects on specific cell populations that could not be detected with the techniques that we used cannot be excluded. Nonetheless, another hypothesis, namely, that FGF-2 and BDNF exert effects on neural stem cells and progenitors to generate new neurons, seems more plausible. The phenotype(s) of these new neurons remain uncertain, but we identified one interesting candidate in a subtype of interneurons (somatostatin). GABA-somatostatin interneurons play an important role in dendritic inhibition, and their loss may favor the development of epilepsy (28); thus, the replacement of these neurons may explain the favorable effects of FGF-2 and BDNF supplementation.

Some aspects of the present findings are worthy of note. First, virus injection was performed under conditions compatible with the clinical settings of patient observation (after the epileptogenic insult), conditions that reproduce those that may allow therapeutic intervention. Second, the disease-modifying effect is striking and can be interpreted as anti-epileptogenic, an effect that is not achieved by the currently available therapy. Further studies will be required in both respects, to establish how long the effect lasts and to challenge the hypothesis that it is truly anti-epileptogenic. Last but not least, these findings may have a heuristic value for the many other neurological diseases associated with neuronal damage.

Animals.

Male Sprague-Dawley rats (240–260 g; Harlan Italy) were used for experiments. All procedures were carried out in accordance with guidelines by the European Community and national laws and policies.

Hippocampal Volumetry.

Paraffin-embedded brains were cut in successive 10 μm sections across the entire hippocampus and one of every five sections was stained with hematoxylin and eosin. Sections were scanned and digitalized using a Nikon Coolscope. Segmentation and orientation were obtained by Amira TM 4.0 (Mercury Computer Systems), by two researchers blind to the animal groups. Three-dimensional reconstructions were then calculated and averaged using the same software.

HSV Infusion.

Under ketamine and xylazine anesthesia, a glass needle connected to a perfusion pump was implanted in the dorsal hippocampus. A total of 1.6 × 10⁶ pfu of vector were injected in a volume of 2 μl at a flow rate of 0.1 μl/min.

FJC Staining.

FJC staining was performed on 6-μm paraffin-embedded sections as previously described (35).

Immunohistochemistry and Immunofluorescence.

For BrdU immunostaining, cellular DNA was denatured with HCl. The primary antibodies used for immunohistochemistry were as follows: BrdU mouse monoclonal (1:100, Roche Molecular Biochemicals); nestin mouse monoclonal (1:100; Chemicon); MAP2abc mouse monoclonal (1:100; Chemicon); GFAP rabbit polyclonal (1:100; Sigma); parvalbumin mouse monoclonal (1:50; Swant); and calbindin rabbit polyclonal (1:50; Swant). Detection was obtained using the biotin-streptavidin system (Ultra Vision Detection System, Lab Vision Corporation). The reaction product was detected using a 3,3-diamino-benzidinetetrahydrochloride (DAB) substrate kit for peroxidase (Vector Laboratories).

For immunofluorescence, the primary antibodies were: BrdU (rat monoclonal, AbD Serotec) and nestin (mouse monoclonal, Chemicon) 1:10 and 1:10 respectively; BrdU (as above) and MAP2abc (mouse monoclonal, Immunological Sciences) 1:10 and 1:10; nestin (mouse monoclonal, Chemicon) and GFAP (rabbit polyclonal, Sigma) 1:25 and 1:100; NeuN (mouse monoclonal, Chemicon) and GFAP (as above) 1:25 and 1:100; MAP2abc (mouse monoclonal, Immunological Sciences) and GFAP (as above) 1:25 and 1:100; Sox2 (rabbit polyclonal, Immunological Sciences) and GFAP (mouse monoclonal, Immunological Sciences) 1:25 and 1:100; DCX (rabbit polyclonal, Cell Signaling) and nestin (as above) 1:20 and 1:20; DCX (goat polyclonal, Santa Cruz Biotechnology) 1:25; somatostatin-14 (rabbit polyclonal, Peninsula Laboratories) 1:25.

Telemetry.

Seizure onset, severity, and duration were assessed by video-EEG monitoring of the animals, performed by means of Phenotyper cages (Noldus Information Technology), using telemetric technology (Dataquest A.R.T. Data Acquisition, Data Sciences International).

Quantification and Statistical Analysis.

All experiments were performed strictly under double-blind conditions. The volume of the dorsal hippocampus was calculated using a stereological approach based on the principle of Cavalieri (36). The degree of cell damage was quantified using two approaches: the “neurodegeneration score” (37) and a thresholding of digital images (38). Counting of cells positive for the different markers was conducted on one of every 26 sections across the entire dorsal hippocampus (i.e., nine regularly spaced sections per animal). Quantification of GFAP-positive and NeuN-positive cells was performed in frames taken from the various hippocampal subareas. Behavioral alterations and seizure scoring were performed by means of the Observer (Noldus). Seizure detection (frequency, duration and severity), as well as single ictal events (<2 seconds) were scored for each individual rat for the whole analysis period (24 h/day for 20 consecutive days).