Light-activated PACK reduces the activity of pyramidal cells in vivo
To verify the inhibitory action of the PACK silencer in awake mice, we targeted hippocampal principal cells by locally injecting AAV9.CaMKIIα.PACK-mCherry into the CA1 area of the dorsal hippocampus (Fig. 1A) and enabled illumination onto these neurons via an implanted optic fiber (Fig. 1B). To test whether applying short light pulses (10 ms) at low frequencies in vivo results in sustained inhibition of PACK-expressing CA1 neurons as previously demonstrated in vitro [11], we shined blue light at 0.05 Hz and 0.1 Hz for 1 h. The light ON phase was enclosed by a pre- and post-recording, 1 h each (Fig. 1C). Following the recording phase, histological analysis revealed that the expression of PACK-mCherry was restricted to pyramidal neurons in CA1 with labeling in cell bodies and dendrites (Fig. 1D). For LFP analysis, we only included mice, which had the optic fiber and recording electrode positioned in CA1 (Additional file 1: Fig. S1, n = 6).
First, we recorded 3-h reference LFPs in each mouse (Fig. 1E) to control for the change in LFP characteristics occurring without blue light application. Quantification of the LFP signal by determination of the line length, a measure for LFP magnitude (see the “Methods” section), revealed a decrease during the reference recordings in PACK mice (first hour set to zero; second hour −7.98 ± 1.71 mV/s, one-sample t-test: t = 11.40, n = 6, p < 0.0001, α = 0.025; third hour −10.54 ± 4.21 mV/s, one-sample t-test: t = 6.313, n = 6, p = 0.0017, α = 0.025; Fig. 1F). Next, we activated PACK with intermittent light application. During 0.05-Hz illumination, LFP magnitude was reduced directly after light pulses, followed by periods of recovery (Fig. 1G). The mean line length significantly decreased during 0.05-Hz illumination (50 min) compared to the pre-recording (−10.33 ± 1.41 mV/s, one-sample t-test: t = 7.34, n = 6, p = 0.0007, α = 0.025; Fig. 1H). Illumination with 0.1 Hz provided a stable reduction of the LFP magnitude (Fig. 1I) with a strong decrease in the line length during the light ON phase (−20.30 ± 1.73 mV/s, one-sample t-test: t = 11.75, n = 6, p < 0.0001, α = 0.025; Fig. 1J). Since there was a drop of line length already in recordings without light application, we compared the change of line length from the first to the second hour in the reference recordings to the sessions with 0.05-Hz and 0.1-Hz illumination, revealing a significant difference (reference −7.98 ± 0.70 mV/s, 0.05 Hz −10.33 ± 1.41 mV/s, 0.1 Hz −20.30 ± 1.73 mV/s, RM ANOVA: F = 24.35, n = 6, p = 0.0019). The reduction of line length was notably higher in the 0.1-Hz recordings than in the respective reference recordings (Tukey’s multiple comparison test: p = 0.003), whereas in the sessions with 0.05 Hz, it was similar to the reference recordings (Tukey’s multiple comparison test: p = 0.09).
To test the reliability of the PACK-mediated inhibition, we further analyzed the responses to light pulses applied at 0.05 Hz and 0.1 Hz. We extracted 2-s LFP snippets before and after each light pulse (Fig. 2A), plotted their overlay (Fig. 2B, D), and calculated the mean of the “before pulse” and “after pulse” line lengths for each recording session (Fig. 2C, E). We also extracted 2-s LFP snippets at corresponding time points from respective pre-recordings (“pre”) to serve as a baseline. In the 0.05-Hz session, reduction of LFP magnitude was reliable and reversible since there was a reduction after each light pulse, followed by a complete recovery during the 20-s interval between subsequent pulses (Fig. 2B). The mean “after pulse” line length was significantly smaller than “pre” and “before pulse” line length in 0.05-Hz sessions (“pre” 48.67 ± 3.88 mV/s, “before pulse” 48.38 ± 4.40 mV/s, “after pulse” 32.39 ± 2.74 mV/s, RM ANOVA: F = 28.65, n = 6, p = 0.0005, Tukey’s multiple comparison test: p < 0.01; Fig. 2C). Shining 10-ms light pulses every 10 s provided a persistent reduction in LFP amplitude (Fig. 2D). In the 0.1-Hz sessions, the line length was lower during the whole light ON period, including “after pulse” and “before pulse” snippets, suggesting a constant inhibitory effect (“pre” 49.97 ± 4.30 mV/s, “before pulse” 36.89 ± 5.16 mV/s, “after pulse” 29.31 ± 3.10 mV/s, RM ANOVA: F = 61.23, n = 6, p < 0.0001, Tukey’s multiple comparison test: “pre” vs. “before pulse” p = 0.0009, “pre” vs. “after pulse” p = 0.002, “before pulse” vs. “after pulse” p = 0.04; Fig. 2E). In summary, applying blue light at 0.1 Hz onto PACK-expressing CA1 neurons in vivo resulted in a sustained reduction of the net neuronal activity.
Next, we examined how light-induced PACK activation in principal cells alters oscillatory activity in CA1. The most dominant oscillations measured in the hippocampus of freely behaving rodents are theta (4–10 Hz), beta (12–30 Hz), and gamma (30–120 Hz) waves [25,26,27]. The theta and gamma peaks are prominent in the spectrogram snippets (Fig. 3A) and in the mean power spectral density (PSD) plot (PSD averaged across recordings; Fig. 3B) of the reference recordings acquired from PACK mice. We quantified the oscillatory power by taking the area under the curve (AUC) of the PSD plot in the respective frequency range. Similarly to the line length, the power of theta, beta, and gamma oscillations decreased significantly within the 3-h reference recordings (two-way RM ANOVA, Dunnett’s multiple comparison test, n = 6, Additional file 2: Table S1; Fig. 3E).
Applying light pulses at 0.1 Hz transiently altered the spectral power (Fig. 3C). The mean PSD during the 50-min 0.1-Hz light ON phase was visibly reduced compared to pre- and post-recordings, especially in frequencies above ~ 10 Hz (Fig. 3D). Quantification of the power change revealed a significant drop of beta and gamma power during 0.1-Hz illumination (two-way RM ANOVA, Dunnett’s multiple comparison test, n = 6, Additional file 2: Table S1, Fig. 3F). However, only the reduction of gamma power was significantly stronger during 0.1-Hz illumination than the decline in respective reference recordings (multiple paired t-test, gamma: t = 3.94, n = 6, p = 0.01, α = 0.0125; Fig. 3G). These data indicate that PACK-mediated inhibition of CA1 neurons in vivo alters mainly the power of gamma oscillations.
Light-dependent hyperactivity in bPAC-expressing mice
Activation of the PACK silencer includes cAMP production by soluble bPAC, which then opens the co-expressed SthK potassium channels in the cell membrane. The second messenger molecule, cAMP, is an important component of intracellular signaling, regulating the plasticity and excitability of neurons [28,29,30,31]. Therefore, it is crucial to investigate whether activation of bPAC alone affects network excitability.
To this end, we targeted bPAC with the AAV9.CaMKIIα viral vector to CA1 neurons and repeated the experiments like with PACK mice (Fig. 4A–D). During the reference recordings, the mean line length dropped similarly as in PACK mice (first hour set to zero, second hour −8.45 ± 0.90 mV/s, one-sample t-test: t = 9.43, n = 6, p < 0.0001, α = 0.025; third hour, −10.40 ± 4.49 mV/s, one-sample t-test: t = 8.02, n = 6, p < 0.0001, α = 0.025; Additional file 1: Fig. S3A). The changes in spectral activity included a reduction in gamma power, whereas other oscillations were not consistently altered during the reference recordings (Additional file 2: Table S2, Additional file 1: Fig. S3B-C). In the reference recordings from control mice injected with AAV9.CaMKIIα.mCherry (mCherry mice), a slight decrease in line length and a significant reduction of gamma power were also evident (Additional file 1: Fig. S2E-I). Therefore, the run-down of LFP signal in reference recordings can be attributed to habituation-reduction in arousal and exploratory behavior [32, 33].
Surprisingly, light activation of bPAC at 0.1 Hz led to sustained neuronal hyperactivity that was clearly visible in LFP snippets (Fig. 4E) as well as in corresponding spectrograms (Fig. 4G). The line length significantly increased during 0.1-Hz illumination (10.08 ± 2.46, one-sample t-test: t = 4.09, n = 6, p = 0.009, α = 0.025; Fig. 4F). In the post-recording, LFP magnitude dropped again, suggesting that the neuronal hyperactivity was reversible and related to light-induced elevation of cAMP levels (line length −8.86 ± 2.80 mV/s, one-way t-test: t = 3.16, n = 12, p = 0.0091, α = 0.025; Fig. 4F). Mainly beta and gamma oscillations were amplified by bPAC activation, although due to high variability, there were no significant differences between the pre-recording and the light ON phase (Additional file 2: Table S2, Fig. 4H, I). The elevation in neuronal activity was not induced by blue light per se since mice injected with AAV9.CaMKIIα.mCherry did not show any changes in the LFP signal during illumination (Additional file 1: Fig. S2J-N). In summary, light-induced activation of bPAC transiently elevates neuronal activity in the hippocampal CA1.
Spontaneous generalized seizures arising in PACK- and bPAC-expressing mice
In healthy control mice, which received intrahippocampal saline and recording electrodes, epileptiform activity is normally absent in the LFP [34]. Unexpectedly, in the majority of the PACK (5 out of 6) and all of the bPAC mice (n = 6), hypersynchronous activity, spreading across both hemispheres, arose at least once during LFP recordings (Fig. 5A, B, E). Most of these electrographic generalized seizures were accompanied by behavioral correlates such as freezing, nodding, forelimb clonus, or rearing according to the Racine scale [35].
The occurrence of generalized seizures in baseline recordings (“ref” and “pre”) indicates seizure initiation independent from light-induced PACK (Fig. 5C) and bPAC (Fig. 5D) activation. Illumination of bPAC-expressing neurons increased the median of average seizure count in 3 h compared to post-recordings (“pre” 0.09 [0, 0.48], “light ON” 1.13 [0.45, 3.94], “post” 0 [0, 0.86], Friedman test: Fr = 6.38, n = 6, p = 0.041, Dunn’s post hoc: p = 0.042; Fig. 5D). The median number of generalized seizures in all recordings was the highest in bPAC mice (3.50 [1.00, 9.75]; Fig. 5E). Most PACK mice experienced one generalized seizure throughout all the recordings (in total 9–28 h) (1.00 [0.75, 2.75], n = 6), whereas mCherry (n = 4) and “no virus” (n = 7) control mice had no generalized seizures (Fig. 5E). These results suggest that the dark activity of bPAC is responsible for spontaneous generalized seizures arising in mice that express PACK or bPAC in CA1 pyramidal cells.
Histological abnormalities in PACK- and bPAC-expressing CA1
An optogenetic tool suitable for long-term in vivo experiments should preserve the normal physiology and histology in the target area. For histological analysis, PACK, bPAC, and mCherry mice were perfused after the last LFP recording, 30–35 days after the intrahippocampal virus and saline injections. Coronal sections were immunolabeled with antibodies against neuronal nuclei (NeuN) and glial fibrillary acidic protein (GFAP) to investigate the histology of neurons and astrocytes, respectively.
To our surprise, we found notable widening of the pyramidal cell layer in PACK-expressing CA1 (Fig. 6A). The mean width of the pyramidal cell layer in PACK-expressing right CA1 was significantly higher than in the left side (right CA1 71.66 ± 2.01 μm, left CA1 60.68 ± 0.54 μm, paired t-test: t = 5.25, n = 8, p = 0.0012, Fig. 6B). The pyramidal cell layer was also significantly wider in bPAC-expressing CA1 compared to its contralateral counterpart (right CA1 84.73 ± 2.23 μm, left CA1 65.27 ± 0.94 μm, paired t-test: t = 8.42, n = 6, p = 0.0004; Fig. 6C). Mice, which received the same viral vector, carrying just the reporter mCherry, had similar pyramidal cell layer widths in the left and right hippocampus (right CA1 70.96 ± 3.53 μm, left CA1 69.10 ± 3.17 μm, paired t-test: t = 0.95, n = 4, p = 0.41; Fig. 6D). Cell dispersion, taken as the difference between right and left CA1 width, was the highest in bPAC-expressing CA1 (19.45 ± 2.31 μm), followed by PACK-expressing CA1 (10.98 ± 1.84 μm), and lacking in mCherry-expressing CA1 (1.86 ± 1.94 μm, one-way ANOVA: F(3,18) = 12.51, p = 0.0006, Tukey’s multiple comparison test: PACK vs. bPAC p = 0.0302, PACK vs. mCherry p = 0.040, bPAC vs. mCherry p = 0.0005; Fig. 6E). These findings led us to conclude that bPAC, and not the viral vector itself, is inducing the cell dispersion in the CA1 pyramidal cell layer.
GFAP labeling in hippocampal sections shows salient chronic astrogliosis in the PACK-expressing CA1 area (Fig. 6F, F2). The comparison of GFAP labeling in left and right hippocampi revealed strongly elevated GFAP intensity in the PACK-expressing CA1 (right CA1 8.70 ± 0.46, left CA1 3.62 ± 0.41, paired t-test: t = 13.54, n = 5, p = 0.0002; Fig. 6G). In bPAC mice, we also found notable astrogliosis in the right bPAC-expressing hippocampus (right CA1 11.62 ± 1.03, left CA1 3.73 ± 0.34, paired t-test: t = 6.01, n = 6, p = 0.0018, Fig. 6H). In mCherry mice, the GFAP intensity was slightly but significantly elevated in the right hippocampus (right CA1 3.98 ± 0.61, left CA1 3.15 ± 0.53, paired t-test: t = 4.12, n = 4, p = 0.026; Fig. 6I). Thus, it could be that either the viral vector or the presence of an electrode and an optic fiber contributed to the glial scarring in the right hippocampus. There was no significant correlation between the strength of mCherry expression and GFAP intensity in PACK, bPAC, and mCherry mice (Pearson’s correlation: r = 0.35, p = 0.2; Fig. 6J). However, the three groups formed separate clusters with (1) mCherry mice having medium mCherry expression but the lowest GFAP intensity, (2) PACK mice having the lowest mCherry expression but medium GFAP intensity, and (3) bPAC mice having the strongest mCherry expression and the highest GFAP intensity. Taken together, it seems as if bPAC expression is the main factor inducing chronic astrogliosis in PACK and bPAC mice, while the viral vector and hippocampal implantations might contribute additionally.
PACK/bPAC expression in the contralateral hippocampus prevents seizure spread in chronically epileptic mice
To find out if the PACK silencer might be useful to limit the spread of epileptiform activity, we targeted PACK to the CA1 principal cells, contralateral to ihpKA treatment. The ihpKA mouse model recapitulates the main pathological features of MTLE, including focal recurrent seizures associated with hippocampal sclerosis [21]. Most ihpKA mice have epileptiform activity that occurs in form of bursts originating in the seizure focus and propagating into the contralateral hippocampus [19, 20, 24]. These epileptiform bursts are subclinical, in other words electrographically measurable but without behavioral convulsions [24, 34]. Unexpectedly, all our PACK-injected ihpKA mice (n = 5) were free of contralateral epileptiform bursts already in baseline recordings before light activation of PACK.
We detected the epileptiform bursts with high spike load using a machine learning algorithm [36] and quantified the burst ratio, the fraction of time spent in bursts during the respective recording (Fig. 7). Although the epileptiform bursts occurred frequently in the ipsilateral hippocampus of PACK ihpKA mice (mean burst ratio during “pre” 0.19 ± 0.03, “0.1 Hz” 0.18 ± 0.01, “post” 0.20 ± 0.02, n = 5, Fig. 7E-F), the contralateral hippocampus was devoid of epileptiform bursts (Fig. 7G). Similarly, in bPAC ihpKA mice, epileptiform bursts were detected in the ipsilateral (mean burst ratio during “pre” 0.11 ± 0.05, “0.1 Hz” 0.07 ± 0.04, “post” 0.13 ± 0.07, n = 3; Fig. 7H-I) but not the contralateral hippocampus (Fig. 7J). Neither the light-induced activation of bPAC nor the whole PACK construct had any effect on the burst ratios in the ipsilateral or contralateral hippocampus (RM ANOVA, p > 0.05, Fig. 7E–J).
We compared the burst ratios in ihpKA PACK and bPAC mice to ihpKA “no virus” mice (n = 4) in a 3-h recording 33–36 days after kainate to clarify whether the lack of contralateral epileptiform bursts affects the seizure burden in the kainate-injected ipsilateral hippocampus (Fig. 7K–M). There was no significant difference in the ipsilateral burst ratios (PACK 0.19 ± 0.03, bPAC 0.11 ± 0.05, "no virus" 0.13 ± 0.03, one-way ANOVA: F = 1.53, n = 3–5, p = 0.26; Fig. 7L), whereas the mean contralateral burst ratio was evidently higher in “no virus” mice than in PACK and bPAC mice (PACK 0.00 ± 0.00, bPAC 0.00 ± 0.00, "no virus" 0.12 ± 0.04, one-way ANOVA: F = 6.26, n = 3–5, p = 0.017, Dunnett’s multiple comparison test: PACK vs. "no virus" p = 0.018, bPAC vs. "no virus" p = 0.036; Fig. 7M). These results suggest that the dark activity of bPAC in CA1 pyramidal neurons prevents the spread of epileptiform burst activity to the bPAC-expressing areas. The activity of CA1 pyramidal cells is thus critical for seizure propagation into the contralateral hippocampus. Additionally, the absence of contralateral bursts does not affect the seizure burden in the sclerotic hippocampus.