Adult male Sprague-Dawley rats weighing 250–350 g were used in our experiment and were acquired from the animal facility of Capital Medical University. Animal care and protocols were approved by the Ethics Committee on Animals Care and Usage of Capital Medical University. RuBi-GABA was purchased from Abcam Biochemicals (Cambridge, UK). Isoflurane was purchased from Ruiwode (Shenzhen, China), and all other chemicals were purchased from Sigma-Aldrich (St Louis, MO, USA).

Rats were anesthetized with isoflurane (2%) and then placed on a Stoelting stereotaxic frame (Stoelting, Wood Dale, IL, USA). A 4-mm diameter craniotomy was performed over the anterior left hemisphere using a dental drill. During drilling, the skull was continuously irrigated with artificial cerebrospinal fluid (ACSF) to prevent the underlying brain from overheating. After the craniotomy was performed, we created a reservoir around the cranial window using dental cement which was then filled with RuBi-GABA solutions. A 2-mm slit was then placed in the dura which allowed the RuBi-GABA that filled the reservoir to directly diffuse into the neocortical surface.

Two screw electrodes were symmetrically placed in the skull over each hemisphere to differentially record the electroencephalogram (EEG) between each hemisphere. The reference electrode and ground electrode were implanted in the neck muscles (Figure 1A). EEGs were recorded using a Cerebus Neural Signal Processing System (Blackrock, Salt Lake City, UT, USA) at a sampling rate of 2000 Hz, and all the data were stored on a personal computer. We commenced EEG recording before RuBi-GABA pretreatment and continued to record throughout the entire experiment. Electrophysiological data were analyzed offline using the Matlab program (Math Works, Natick, MA, USA).

When we finished the surgery, we added 200 µl of ACSF or RuBi-GABA solution (0.1 or 0.2 mM in ACSF) to the reservoir, respectively. After 45 min of pretreatment, we induced focal neocortical seizures by injecting 0.5 µl of a 4-aminopyridine (4-AP) solution (25 mM in ACSF) into the neocortex at a depth of 0.5 mm, at a position 2 mm anterior to the bregma and 2.5 mm from the midline, using a commercial oocyte injection system (Drummond Scientific, Broomall, PA, USA) coupled to a glass micropipette (tip diameter approximately 100 µm). The injection was carried out over a 5-min period to minimize cortical damage, and the pipette was maintained for an additional 5 min to minimize leakage of the 4-AP.

All ictal and interictal durations of seizures in our experiment were independently and blindly analyzed offline by two experienced reviewers. We observed close agreement between the two reviewers. Because we need a few seconds to confirm the seizure onset before we start the photolysis process, so there is a lantency between seizure onset to photolysis process was started. In order to accurately show the antiepileptic effects of RuBi-GABA, so we also corrected seizure durations by subtracting the latency in the amount of time from seizures onset to the initiation of photolysis process. Owing to high time resolution, wavelet analysis was applied to depict time frequency spectrum to reflect dynamic change of seizures during uncaging.

To observe the change in EEG power and achieve more accurate results, the fast Fourier transform (FFT)-based power spectral analysis method with excellent frequency resolution was used to calculate the power of each frequency band activity in seizures. Power spectral histograms were calculated using 4000 points Hamming window with 2000 points overlap and 4000 FFT points. The EEG frequency was decomposed into several frequency activity: delta frequency band (1–4 Hz), theta frequency band (4–8 Hz), alpha frequency band (8–13 Hz), and beta frequency band (13–30 Hz). Because of the great individual variability in the power of EEG in the rats, these variations will directly affect the results of the analysis. Therefore, we normalized the power of EEG in different frequency bands for each rat. The normalized approach is the power of seizure with photolysis of RuBi-GABA divided by the mean of the power of seizures without photolysis in the rats.

To observe the change of EEG power from baseline to seizures, we randomly chose two baseline EEG segments prior to the injection of 4-AP from each rat, with a duration equal to the mean duration of seizures in the control group (CG). Then, the powers of these EEGs were calculated using the FFT and compared between baseline and CGs.

We used an implantable optical fiber (0.39NA, 200 µm diameter; Doric Lenses Inc, Quebec, Canada) terminated in 1.25-mm ceramic ferrules (5 mm length; Ruiwode, Shenzhen, China) to introduce a blue laser, produced by a high-power laser driver (473 nm; SLOC, Shanghai, China), into the neocortex (4-AP injection site) to cause the photolysis of RuBi-GABA. The optical power delivered at the fiber tip was calibrated with an optical power meter (PM121D; Thorlabs, Newtown, NJ, USA) before experiments. When seizure onset was confirmed, we manually triggered the laser driver to start the photolysis process for 30 s. To avoid the long-lasting effects of the photolysis of RuBi-GABA upon subsequent experimental results, we did not trigger the next photolysis activity until the seizure duration had completely recovered to control levels. In addition, to assure sufficient RuBi-GABA for each photolysis process and avoid the effect of GABA which released from the solution in the reservoir on the subsequent experiments, we replaced the solution in the reservoir after each photolysis activities.

Animals were divided into four different groups: a CG (n = 8), an illumination-only group (IC, n = 6), a 0.2-mM RuBi-GABA without illumination group (RC, n = 6), and a RuBi-GABA with illumination group (RI, n = 24). The RI group was further divided into four sub-groups (n = 6 for each group): a 0.1-mM RuBi-GABA with 15-mW photolysis group (RI₁), a 20-mW photolysis group (RI₂), a 0.2-mM RuBi-GABA using 10-mW photolysis (RI₃), and a 15-mW photolysis (RI₄) group. All experiments involving RuBi-GABA were carried out in a darkened room with only occasional red lamp illumination.

All rats were sacrificed 3 days after the acute experiments with an overdose of isoflurane (5%). Brain tissue was removed immediately and preserved in 10% formalin. After fixation, the tissue was embedded in paraffin. Coronal 4-µm sections were cut in a manner which included the area around the injection site and the corresponding contralateral neocortex. Sections were then examined with hematoxylin/eosin for signs of necrotic injury, TdT-mediated dUTP-biotin nick-end labeling (TUNEL) staining for evidence of apoptotic death, and glial fibrillary acidic protein (GFAP) staining to evaluate possible glial proliferation.

All data are presented as mean ± SEM. Data were analyzed using SPSS 19.0 statistics software (IBM, Armonk, NY, USA). Seizure duration and the changes of EEG power among different groups were analyzed by analysis of variance. The post hoc test used least significant difference. The Wilcoxon signed-rank test was used to compare non-normally distributed data (interictal time). A P-value of <0.05 was considered to be statistically significant.

We did not detect any seizures following the surgery and the application of ACSF containing RuBi-GABA upon the brain surface. However, within 10 ± 1.23 min of 4-AP injection, animals developed recurrent electrographic focal seizures with a mean seizure duration and interictal period of 65.92 ± 4.60 and 263.90 ± 23.00 s, respectively; these conditions remained for 1.5 h (Figures 1B and 2A,B).

To explore whether the anti-seizure effects of RuBi-GABA were related to RuBi-GABA concentration and the intensity of illumination, we measured different combinations of RuBi-GABA concentrations and light intensities. We found that the seizure durations of these groups were significantly different [F (1, 6) = 2.337, P = 0.047]. Then, we tested the combination of 0.1-mM RuBi-GABA with 15 and 20 mW of illumination (30 s). Under these conditions, seizure duration was 67.25 ± 3.57 and 75.33 ± 10.48 s, respectively. Compared with the CG, there was no statistically significant difference (P = 0.905 and P = 0.585). However, when the concentration of RuBi-GABA was increased to 0.2 with 10 and 15 mW of light intensity, the seizure duration was significantly reduced to 31.00 ± 6.05 and 20.86 ± 7.20 s, respectively. Compared with the CG, these reductions were statistically significant (P = 0.047 and P = 0.017 vs. CG, Figures 2A,B).

Because manual confirmation of seizure onset requires a few seconds, there was a latency from the onset of seizure to the initiation of photolysis. These latencies in the IC group and the RI_1–4 groups were 5.30 ± 0.96, 5.60 ± 0.45, 5.56 ± 0.41, 7.75 ± 2.21, and 4.56 ± 0.63 s, with an average latency of 5.69 ± 0.44 s.

To more accurately show the anti-seizure effect of the photolysis of RuBi-GABA, we corrected the duration of the seizures by subtracting the average latency (5.69 ± 0.44 s) from all seizure durations. There was also significant difference existed between these groups in the corrected seizure duration [F (1, 6) = 2.337, P = 0.047]. The duration of seizure after correction of the RI₁ and RI₂ groups was 61.56 ± 3.57 and 69.64 ± 10.48 s, respectively. Compared with the correct CG (60.23 ± 4.60 s), there was no statistically significant difference (P = 0.905 and P = 0.585). The corrected seizure duration of the RI₃ and RI₄ groups was 25.31 ± 6.05 and 15.13 ± 7.20 s, respectively. Compared with the corrected CG, these reductions were statistically significant (P = 0.047 and P = 0.017 vs. CG, Figures 2A,B).

To avoid the influence of long-lasting effect of RuBi-GABA photolysis to our experimental results, we also compared the corrected duration of each seizure with photolysis of RuBi-GABA and the previous seizure without photolysis process. Furthermore, we calculated the ratio of the corrected duration of seizures with and without photolysis process (normalized corrected seizure duration). Comparison of illumination or non-illumination, we did not find a significant reduction of seizure duration in IC, RI₁, and RI₂ groups, however, we found that illumination significantly shortened the duration of seizure in the RI₃ and RI₄ groups (Figure 3).

When we tested the ability of GABA photolysis to control seizures, we found a phenomenon in which RuBi-GABA uncaging can not only quickly terminate seizures but also significantly prolong the intervals between seizures. When we used 0.2 with 10 and 15 mW of illumination, the interictal interval was significantly prolonged from 263.90 ± 23.00 s in the CG to 966.00 ± 646.00 and 556.75 ± 215.83 s, respectively (P < 0.05, Figure 4). Of particular note was that five of the 12 rats in the two experimental groups were seizure free following the first illumination. However, when we applied a combination of 0.1 mM RuBi-GABA and 15 and 20 mW of illumination, the interictal interval was 273.00 ± 85.31 and 370.78 ± 164.72 s, respectively, which was not significantly longer than the CG (P > 0.05).

We used FFT to calculate the spectral power of the EEG at baseline and control seizures, and seizures that treated by RuBi-GABA uncaging. Compared with baseline, the seizure significantly increases in the spectral power of delta, theta, alpha, and beta frequency activity (Figure 5B). In particular, we found that large spectral peaks were present at approximately 4–7 Hz during the control seizures. When we tested the combination of 0.1 mM RuBi-GABA with 15 and 20 mW of illumination (30 s), only the power of delta during the seizures was significantly reduced compared with controls, but the power of other frequency bands did not change significantly (P < 0.05 vs. pre-illumination). However, in the groups receiving 0.2-mM RuBi-GABA with 15 and 20 mW of illumination (30 s), we found that the power in all frequency bands was significantly reduced compared with the CG (P < 0.05 vs. pre-illumination, Figure 5A). In particular, the large spectral peaks, which were present in the control seizure animals at approximately 4–7 Hz, disappeared (Figure 5B).

We were concerned that the products produced by uncaging of the RuBi-GABA, and the phototoxicity experienced during the photolysis process, may damage the underlying neocortex. Therefore, we performed routine hematoxylin/eosin staining, GFAP staining, and TUNEL staining 3 days after seizures to examine neuronal loss, neuronal necrosis, apoptotic death, and glial proliferation. The hematoxylin/eosin staining results showed that 4-AP-injected/RuBi-GABA-treated neocortex was completely indistinguishable from controls. We did not find any neuronal necrosis or loss. Furthermore, we did not detect TUNEL-positive cells in 4-AP-injected/RuBi-GABA-treated neocortex or the corresponding contralateral neocortex by TUNEL staining. Examination of GFAP-stained sections showed some sporadic astrocytes in the 4-AP-injected and RuBi-GABA-treated neocortex which were not detected in the control neocortex (Figure 6).