Organotypic hippocampal slices (OHCs) were prepared and cultured according to Stoppini et al. 17 . The brains of 6–7 day-old Sprague Dawley rats were removed after cold anesthesia and hippocampi were rapidly dissected under aseptic conditions in a dissection medium containing 50% HBSS, 50% Opti-MEM, penicillin 25 units/ml, streptomycin 25 μg/ml (Life technologies, Grand Island, NY, USA). Then transverse sections (400 μm) were obtained using a tissue chopper. Ten slices were placed on a 30 mm porous membrane (Millipore, Billerica MA, USA) and kept in 100 mm diameter Petri dishes filled with 5 ml of culture medium composed of 25% heat-inactivated horse serum, 25% HBSS, 50% Opti-MEM, penicillin 25 units/ml, streptomycin 25 μg/ml (Life technologies). Cultures were maintained in a humidified incubator at 35°C and 5% CO₂. One week later, cultures were transferred in defined medium composed of 25% B27 supplemented neurobasal medium, 25% HBSS, 50% Opti-MEM, penicillin 25 units/ml, streptomycin 25 μg/ml (Life technologies). All animal procedures were conducted in accordance with the European Communities Council Directive of November 24, 1986 (86/6 09/EEC), and approved by the French Ministry of Agriculture (Authorization No. 34178, ML-N).

After 2 weeks, membranes were transferred to 6-well plates, each well filled with 1 ml defined culture medium. To induce seizures, slices were treated with 25 μM kainate (Sigma-Aldrich, Saint-Louis, MO, USA) for 1 h or with 10 μM of bicuculline (Sigma-Aldrich) for 10 min. Control slices received no treatment. Slices were then transferred in a bicuculline-free or kainate free-defined culture medium during the recovery period (2, 12 and 24 h).

Since empty vessels respond to angiogenic factors, we studied the vascular remodelling after epileptic seizures using OHCs 20 25 26 . Previously, we observed the presence of angiogenesis and BBB permeability in pharmacoresistant temporal lobe epilepsies 22 . To determine if vascular remodeling was induced by seizures per se, or by cell damage or inflammation associated with severe seizures, we evaluated the vascular density in the following conditions: 1) seizures without damage induced by bicuculline; 2) inflammation induced by LPS; 3) seizures with neuronal death and inflammation induced by kainate.

To evaluate pathological angiogenesis, we measured vascular density and branching (Figure 1A). As a positive control, we also evaluated OHCs treated with VEGF. For the “point-counting” method, the number of “crossings” was counted and normalized to slice surface area in mm². The “branch point” was evaluated after magnifying the slices (Figure 1B, C). We evaluated the branching and the vascular density 24 h post-treatment. A significant increase in vascular density was observed in all conditions indicating that seizures and also inflammation are sufficient to induce a vascular remodeling, p<0.01 compared to control (Figure 1D, E). However laminin staining showed an increased vascular network in CA1 and CA3 areas after kainate treatment compared to other conditions (Figure 1D). Neuronal death also occurred in these two areas after kainate treatment 27 . We quantified the branching in CA1 and CA3 areas 24 h after kainate and bicuculline treatments. An increase of branching was observed in all conditions. Interestingly the branching was significantly higher in the case of seizures associated with neuronal death (kainate) than seizures alone (bicuculline), p<0.01 and 0.05, respectively, compared to control (Figure 1F).

We found an increase in vascular density and branching after in vitro seizures or inflammation, similar to previous results from rodent models or human tissue 21 22 . However, depending on the presence or absence of neuronal death, the vascular remodeling appeared to be different with an increase in branching in lesion areas. In the following study we compared changes in the NVU between the kainate and bicuculline models.

The NVU is mainly composed of endothelial cells characterized by limited transport due to the presence of transporters and TJs. In the brain microvasculature, cells surrounding the capillaries include astrocytes and pericytes. These cells play strategic roles in both formation and maintenance of the NVU, and also in neurovascular coupling 1 9 28 29 30 (Figure 2A). The presence of astrocyte end-feet along and surrounding blood vessels was demonstrated by GFAP staining in control organotypic cultures (Figure 2B, arrows). The presence of these cells around blood vessels is modified in pathological conditions. Indeed, we observed a reduction of astrocyte end-feet 24 h after kainate treatment. However, after seizures without lesions, end-feet appeared intact (Figure 2B, arrows).

The main characteristic of the NVU is the presence of TJ proteins which scaffold the junctions between endothelial cells. These proteins are essential for a high transendothelial electrical resistance (TEER) (Figure 2A) 29 31 . The three important TJ proteins, ZO-1, claudin-5 and occludin, are preserved in culture for several weeks 21 32 .

We studied the regulation of ZO-1, claudin-5 and occludin 24 h post-seizure induced by either kainate or bicuculline. In the kainate model, Western blot analysis revealed a significant down-regulation only for ZO-1, p<0.01. In the bicuculline model the expression of the three main TJ proteins was not affected (Figure 2C). The staining of ZO-1 and laminin revealed a regular staining of the TJ proteins along blood vessels in control slices. Similar staining was observed 24 h post-bicuculline seizures. However, at 24 h post-kainate seizures, ZO-1 staining was absent along blood vessels (Figure 2D). This experiment on OHCs showed that NVU remodeling is dependent on severity of neuronal damage induced by epileptiform stimuli.

The tissue or culture medium from OHCs can be analysed by molecular screening techniques. Due to differences in branching and the regulation of NVU between two seizure models, we can expect differences in the secretion and release of angiogenic factors or cytokines. With a protein array, we evaluated the levels of secreted cytokines 24 h after seizures induced by kainate or bicuculline. As negative and positive controls, we used the medium of non-treated slices and the medium of OHCs treated with LPS. In the medium from non-treated slices, we found only 2 cytokines: VEGF and metallopeptidase inhibitor-1 (TIMP)-1 while 24 h after LPS treatment, many additional cytokines were secreted into the culture medium. After seizure induction, we observed different patterns between kainate and bicuculline models. In kainate model, cytokines present in the culture medium indicated an inflammatory process, confirming previous results 27 33 . In contrast, 24 h after bicuculline seizures, cytokine patterns were identical to those in control slices (Figure 3A). These differences in cytokine profiles were confirmed by heatmap analysis. We focused in more detail on cytokines known to be involved in vascular remodeling by inducing angiogenesis or increased BBB permeability. VEGF, IL-1β, IL-1α, IL-6, IL-13 and also TNF-α are pro-angiogenic, while IFN-γ, IL-1β and VEGF can also participate in BBB permeability (Table 1). Most of these cytokines, including IL-1α, IL1-β but also VEGF, are present at higher levels in the kainate compared to bicuculline model (Figure 3B).

Angiogenesis and BBB permeability are hallmarks of VEGF/VEGFR-2 activation 59 . Their overexpression in the epileptic focus after experimental seizures suggests that the VEGF/VEGFR-2 system is a logical new target for refractory epilepsies 21 22 60 61 . In the kainate model, we demonstrated an upregulation and activation of VEGF/VEGFR-2 signaling, leading to the downregulation of ZO-1 21 . The lower level of VEGF in the culture medium of OHCs after bicuculline treatment suggested a different regulation of angiogenic factors that does not trigger a loss of TJ proteins.

Western blotting and immunostaining demonstrated an overexpression of VEGF only 12 h after bicuculline application, p<0.05 (Figure 4A). Increased VEGF was observed in astrocytic end-feet and surrounding blood vessels (arrows, Figure 4B). We next studied the expression and activation of VEGFR-2. An increase in VEGFR-2 expression was also detected 12 h post-seizure, p<0.05. However, the activation of VEGFR-2, measured by the phosphorylation of the receptor, was visible at 2 and 12 h post-seizure, p<0.05 for both (Figure 4C). The immunostaining revealed an overexpression of the receptor in neurons but more particularly in blood vessels (Figure 4D). After bicuculline-induced seizures, we observed a similar upregulation of the VEGF/VEGFR-2 system.

To understand differences in the regulation of tight junctions between the two seizure models, we focused on the angiopoietin system, composed of angiopoietin-1 and 2 (Ang1 and Ang2). These two proteins have opposite effects on the BBB integrity; Ang1 is involved in the maturation of blood vessels and participates in the BBB integrity, whereas Ang2 appears in early stages of the angiogenesis and disrupts the BBB 62 63 64 65 . We decided to study the regulation of these two proteins 24 h after either kainate or bicuculline seizures, a time-point where ZO-1 is downregulated in the kainate model only. By Western blotting, we observed the same level of Ang1 as in control slices but a significant downregulation of Ang2 following bicuculline seizures, p<0.01. However, after kainate seizures, Ang1 expression was significantly lower than in control slices, p<0.05, whereas Ang2 was significantly upregulated, p<0.01 (Figure 4E). These results suggest that angiopoietin system could play an important role in the regulation of TJ proteins after epileptic seizures.

Due to their easy accessibility, OHCs are excellent tools for pharmacological and biochemical assays, including drug screening for BBB protective compounds that could improve the treatments for ischemic or traumatic injuries 66 67 . We have already demonstrated the use of neutralizing antibodies in OHCs. We neutralized VEGF with an anti-VEGF antibody (bevacizumab) that prevents VEGF binding to its receptor. Despite the thickness of the culture plus insert at around 150 μm, the addition of this neutralizing antibody abolished both the down-regulation of ZO-1 protein and increased vascularisation induced by in vitro seizures 21 . Since in this study a deregulation of the Ang proteins was observed only in the model where ZO-1 was also downregulated, we tested the effect of recombinant Angiopoietin-1 (rhAng-1) applied to the culture medium. To determine if we can restore ZO-1 expression after kainate seizures, rhAng1 was added to the culture medium 4 h post-seizures. In control slices treated with rhAng1, the staining pattern of ZO-1 along blood vessels did not change compared to control slices. Treatment with rhAng1 after kainate seizures restored the presence of ZO-1 protein (Figure 5A). Western blot analysis confirmed the above results, showing a significant increase of ZO-1 expression after rhAng1 addition, p<0.05, and confirming that Ang1 plays an important role in the restoration of BBB integrity (Figure 5B).