Hippocampal low-frequency stimulation prevents - eLife

Preparing to load PDF file. please wait...

0 of 0
100%
Hippocampal low-frequency stimulation prevents - eLife

Transcript Of Hippocampal low-frequency stimulation prevents - eLife

RESEARCH ARTICLE

Hippocampal low-frequency stimulation prevents seizure generation in a mouse model of mesial temporal lobe epilepsy
Enya Paschen1,2*, Claudio Elgueta3, Katharina Heining4,5, Diego M Vieira4,5, Piret Kleis1, Catarina Orcinha1, Ute Ha¨ ussler1,6, Marlene Bartos3,6, Ulrich Egert4,5, Philipp Janz1, Carola A Haas1,5,6*
1Experimental Epilepsy Research, Department of Neurosurgery, Medical Center University of Freiburg, Faculty of Medicine, Freiburg, Germany; 2Faculty of Biology, University of Freiburg, Freiburg, Germany; 3Systemic and Cellular Neurophysiology, Institute for Physiology I, Faculty of Medicine, University of Freiburg, Freiburg, Germany; 4Biomicrotechnology, Department of Microsystems Engineering – IMTEK, Faculty of Engineering, University of Freiburg, Freiburg, Germany; 5Bernstein Center Freiburg, University of Freiburg, Freiburg, Germany; 6Center for Basics in NeuroModulation, Faculty of Medicine, University of Freiburg, Freiburg, Germany

*For correspondence: [email protected] de (EP); [email protected] (CAH)
Competing interests: The authors declare that no competing interests exist.
Funding: See page 22
Received: 17 December 2019 Accepted: 13 December 2020 Published: 22 December 2020
Reviewing editor: John R Huguenard, Stanford University School of Medicine, United States
Copyright Paschen et al. This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Abstract Mesial temporal lobe epilepsy (MTLE) is the most common form of focal,
pharmacoresistant epilepsy in adults and is often associated with hippocampal sclerosis. Here, we established the efficacy of optogenetic and electrical low-frequency stimulation (LFS) in interfering with seizure generation in a mouse model of MTLE. Specifically, we applied LFS in the sclerotic hippocampus to study the effects on spontaneous subclinical and evoked generalized seizures. We found that stimulation at 1 Hz for 1 hr resulted in an almost complete suppression of spontaneous seizures in both hippocampi. This seizure-suppressive action during daily stimulation remained stable over several weeks. Furthermore, LFS for 30 min before a pro-convulsive stimulus successfully prevented seizure generalization. Finally, acute slice experiments revealed a reduced efficacy of perforant path transmission onto granule cells upon LFS. Taken together, our results suggest that hippocampal LFS constitutes a promising approach for seizure control in MTLE.
Introduction
Mesial temporal lobe epilepsy (MTLE) represents the most common form of acquired epilepsy in adults. MTLE is thought to arise from an initial precipitating insult in early childhood, such as status epilepticus (SE), complex febrile seizures, or head trauma (Engel, 2001). The most frequent histopathological hallmark of MTLE is hippocampal sclerosis, which is characterized by neuronal cell loss and gliosis (Blu¨mcke et al., 2013). In addition, it is often associated with granule cell dispersion (GCD) and mossy fiber sprouting (Thom, 2014). MTLE is of particular clinical interest since it is frequently resistant to pharmacological treatment (Engel, 2001). Hence, surgical removal of the seizure focus is an effective therapeutic intervention for many MTLE patients (Englot and Chang, 2014). However, for patients with multiple seizure foci or for those at risk of resection-related impairments, this treatment is not an option.
One alternative for these patients is electrical deep brain stimulation (DBS) which often targets either the hippocampus or the anterior thalamic nucleus (Li and Cook, 2018). Complementary to pharmacological treatment, DBS at high frequencies (HFS, 130–200 Hz at 1–5 V) (Laxpati et al., 2014; Li and Cook, 2018), either as an open-loop (Boe¨x et al., 2011; Tellez-Zenteno et al., 2006; Velasco et al., 2007) or a closed-loop stimulation, especially the Responsive Neuro Stimulation

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

1 of 26

Research article

Neuroscience
(RNS) System (Bergey et al., 2015; Geller et al., 2017; Nair et al., 2020), is currently in use to alleviate intractable seizures.
In MTLE with hippocampal sclerosis, however, the efficacy of HFS is rather variable between patients (Boe¨x et al., 2011; Velasco et al., 2007). This is in line with the hypothesis that neuronal loss and/or altered electrical resistance in sclerotic neural tissue impair the efficacy of HFS since stimulation can only be successful when targeting a sufficiently preserved network (Cue´llarHerrera et al., 2004; Velasco et al., 2007). MTLE patients with hippocampal sclerosis may therefore require specific stimulation parameters to achieve seizure control.
Interestingly, low-frequency stimulation (LFS) at 5 Hz was effective in MTLE patients in small cohort studies, including those with hippocampal sclerosis (Koubeissi et al., 2013; Lim et al., 2016). From a technical perspective, LFS would be favorable for clinical implementation due to its low duty cycle, resulting in less electric current injection and longer battery life. For a systematic assessment of seizure-suppressive effects of LFS in relation to disease parameters, studies in translational animal models are crucial.
Optogenetic stimulation offers cell- or pathway-specific modulation of neuronal activity and has been successfully applied to alleviate seizure burden in several rodent MTLE models (KrookMagnuson and Soltesz, 2015; Zhao et al., 2015). These optogenetic approaches targeting the hippocampus were either based on the inhibition of excitatory neurons or on the recruitment of inhibitory interneurons (Kim et al., 2020; Kokaia et al., 2013; Krook-Magnuson et al., 2013; Ladas et al., 2015; Ledri et al., 2014; Lu et al., 2016). In MTLE models and patients with strong hippocampal sclerosis, however, pyramidal cells and GABAergic interneurons are strongly diminished and therefore would be difficult to target (Bouilleret et al., 1999; Maglo´ czky and Freund, 2005; Marx et al., 2013; Thom, 2014).
In the present study, we applied optogenetic LFS (oLFS) to systematically investigate the efficacy of different stimulation frequencies in seizure interference. We used the intrahippocampal kainate (KA) mouse model, which replicates the major hallmarks of human MTLE pathology, comprising the emergence of spontaneous recurrent seizures and robust unilateral hippocampal sclerosis (Bouilleret et al., 1999). In this model, dentate granule cells (DGCs) with their entorhinal inputs, i.e. the perforant path (Froriep et al., 2012; Janz et al., 2017a), and CA2 pyramidal cells are preserved (Ha¨ussler et al., 2016). Using different KA concentrations, we modified the severity of hippocampal sclerosis and applied oLFS to stimulate entorhinal afferents in the diseased hippocampus in vivo while recording local field potentials (LFP). We then assessed the responses of individual DGCs to oLFS in vitro by patch-clamp recordings. In addition, we probed the translational value of LFS by applying electrical LFS (eLFS) in vivo over several weeks. We present evidence that LFS is highly effective in preventing both subclinical epileptiform activity and behavioral seizures in experimental MTLE with severe hippocampal sclerosis.

Results
Modification of the intrahippocampal KA mouse model
To achieve different degrees of hippocampal sclerosis and seizure burden as observed in human MTLE (Thom, 2014), we modified the established intrahippocampal KA mouse model by injecting three KA concentrations (10, 15, and 20 mM). To this end, we compared these KA groups at 35–40 days after KA injection with respect to GCD, cell loss in CA1 and hilus, and epileptiform activity (Figures 1A, 2 and 3).
Quantitative analysis of GCD in NeuN-stained sections revealed that the volume of the dispersed granule cell layer (GCL) was comparable between 20 and 15 mM KA but significantly smaller in the 10 mM KA group (Figure 2B, 10 mM: 0.24 ± 0.08 mm3; 15 mM: 1.56 ± 0.16 mm3; 20 mM: 1.52 ± 0.11 mm3, 10 mM vs 15 mM and vs. 20 mM p<0.001; n = 4; 6; 5 animals). Conversely, the loss of CA1 pyramidal cells, quantified as the total length of CA1 devoid of pyramidal cells, was similar in all groups (Figure 2C, 10 mM: 37.72 ± 8.84 mm; 15 mM: 49.57 ± 3.27 mm; 20 mM: 42.28 ± 6.44 mm; n = 4; 6; 5 animals).
The loss of NeuN+ hilar neurons in the sclerotic ipsilateral dorsal hippocampus (idHC) compared to the non-sclerotic contralateral dorsal hippocampus (cdHC) (Figure 2F,G) was significantly less pronounced in mice injected with 10 mM KA (Figure 2D, 10 mM: 52.78 ± 7.24% cell loss; 15 mM:

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

2 of 26

Research article

Neuroscience

Figure 1. Experimental design for in vivo LFS. Animals received intrahippocampal KA and a channelrhodopsin 2 (ChR2)-carrying virus into the entorhinal cortex to trigger epileptogenesis and the expression of ChR2-mCherry in entorhinal afferent fibers. After 16 days post-injection, recording electrodes, and (A) an optic fiber or (B) an optic fiber combined with a stimulation electrode were implanted. Following recovery from implantations, reference LFPs were recorded on 2 consecutive days for 3 hr each. (A) In the first group of experiments, the effect of oLFS on spontaneously occurring epileptiform activity was tested (week 4) in 4-hr recording sessions. A session consisted of 1 hr of ‘pre’ stimulus recording, followed by 1 hr of ‘oLFS’ pulses and 2 hr of post-stimulus recordings (‘post 1’ and ‘post 2’). Three different oLFS frequencies (1, 0.5, or 0.2 Hz) were applied on successive days in each animal (two sessions per animal). Next, generalized seizures were induced by optogenetic (10 Hz) stimulation. To test the effects of oLFS on seizures, oLFS (1 and 0.5 Hz) was applied either immediately after (overwriting) or before the pro-convulsive (10 Hz) stimulation (preconditioning) (week 5). (B) The second group of in vivo experiments assessed the effects of eLFS on epileptiform activity. First, we tested the effects of eLFS on optogenetically induced seizures (week 4, preconditioning, see above). In weeks 5 and 6, animals were stimulated daily for 1 hr (1 Hz, eLFS) following the same ‘pre’, ‘eLFS’,’ post 1’, post 2’ paradigm, as described above. In week 7, animals were stimulated twice (on 2 different days) over 3 hr continuously and twice in an on-off ‘cycle’ paradigm: after initial 30 min eLFS, eLFS stimulation was turned off for 10 min and then turned on again for 10 min. This was repeated four times, followed by another hour LFP recording (‘post 1’). (A1, B1) All animals were perfused after the last LFP recording and brain sections were processed for immunohistological procedures. (B2) Implantation scheme for eLFS. DG, dentate gyrus; FISH, fluorescent in situ hybridization; IHC, immunohistochemistry; perf., perfusion.
86.47 ± 3.90% cell loss; 20 mM: 88.54 ± 1.66% cell loss; 10 mM vs 15 mM and vs 20 mM p<0.001; n = 4; 6; 5 animals), whereas glutamic acid decarboxylase 67 (Gad67) mRNA+ interneurons were equally lost in all groups (Figure 2E, 10 mM: 78.33 ± 3.77% cell loss; 15 mM: 93.35 ± 0.71% cell loss; 20 mM: 85.20 ± 4.35% cell loss; n = 4; 3; 5 animals). Animals injected with a higher KA dosage (15 and 20 mM), which showed greater loss of NeuN+ neurons in the hilus, also had a larger volume of the dispersed GCL (Figure 2H).
Next, we investigated the characteristics of spontaneous epileptiform activity in all mice using LFP recordings from electrodes in both dorsal hippocampi (idHC and cdHC, Figure 3). We used a custom algorithm as described in the Materials and methods and illustrated in the supplements (Figure 3—figure supplement 1) for data analysis of LFP recordings. In brief, epileptiform spikes were automatically detected and analyzed for bursts, which were classified as low-, medium- and highload bursts according to their spike load using a self-organizing map (SOM, Figure 3—figure supplement 1A–C). Bursts recorded in the 10 mM KA group covered the whole range of the SOM, but a larger fraction of these bursts matched nodes representing lower spike loads compared to the 15 and 20 mM KA group (Figure 3—figure supplement 1D).
Epileptiform activity occurred in both hippocampi of all KA mice (Figure 3A,B). The idHC of 10 mM KA mice showed a smaller percentage of high-load bursts compared to the 15 or 20

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

3 of 26

Research article

Neuroscience

Figure 2. The degree of hippocampal sclerosis depends on KA concentration. (A) Representative NeuN-labeled sections of dorsal and ventral hippocampal regions treated with different KA concentrations at 37 days postinjection (PI). In each section, the electrode position is marked with a red asterisk. Epileptic hippocampi show GCD in the dentate gyrus (open arrowheads) and cell loss in CA1 (the region between filled arrowheads). Comparing KA concentration groups with respect to different markers of hippocampal sclerosis by quantification of (B) GCL volume of dispersed regions (i.e. GCD), (C) total length of cell loss in CA1, (D) % loss of NeuN+ hilar cells, and (E) loss of Gad67+ hilar interneurons in the sclerotic vs. non-sclerotic hippocampus (i.e. (G, G1) ipsilateral vs (F, F1) contralateral). One-way ANOVA; Tukey’s multiple comparison test; *p<0.05, **p<0.01 and ***p<0.001. All values are given as mean ± standard error of the mean (SEM). (H) Animals injected with higher KA concentrations (15 and 20 mM KA) display stronger hilar cell loss along with a higher degree of GCD. Scale bars 200 mm.

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

4 of 26

Research article

Neuroscience

Figure 3. Variable severity of epileptiform activity elicited by different KA concentrations. (A, B) Representative LFP traces for the 10 mM and 15 mM KA group (20 mM not shown) showing spontaneous epileptiform activity in the cdHC and idHC. (C) Automatic classification of epileptiform activity into low-load (blue), medium-load (green) and high-load bursts (orange). We used a custom algorithm as illustrated in Figure 3—figure supplement 1. In the 20 mM KA group, the percentage of high-load bursts in the cdHC is decreased, whereas for the 10 mM and 15 mM KA group the percentage of the burst classes is similar in both hippocampi. (D–G) Injections of 15 and 20 mM KA lead to an increased high-load burst ratio and a higher epileptic spike rate in the idHC but not in the cdHC. All values are given as mean ± SEM. Source data is provided in Figure 3—source data 1. (I, J) The high-load burst ratio is positively correlated with GCD and hilar NeuN+ cell loss ((I): p<0.0001, two-tailed; Pearson’s r = 0.84; (J): p<0.01, two-tailed; Pearson’s r = 0.64). The online version of this article includes the following source data and figure supplement(s) for figure 3:
Source data 1. Variable severity of epileptiform activity elicited by different KA concentrations. Figure supplement 1. Analysis of epileptiform activity and comparison of different KA concentrations.
mM KA groups. In the cdHC, however, the incidence of high-load events was similarly low for the 10 and 20 mM KA groups (Figure 3C, Figure 3—source data 1). The mean ratio of time spent in high-load bursts (mean high-load burst ratio) and epileptic spike rate were significantly lower in the idHC 10 mM KA group compared to the 15 and 20 mM KA group (Figure 3D–G,

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

5 of 26

Research article

Neuroscience
Figure 3—source data 1). In the idHC the extent of GCD and hilar cell loss was positively correlated with the high-load burst ratio (Figure 3H, p<0.0001, Pearson’s r = 0.84; and Figure 3I, p<0.01, Pearson’s r = 0.64; both n = 15 animals).
Taken together, we developed the intrahippocampal KA mouse model further, creating varying degrees of disease severity on both the anatomical and electrophysiological level. Thus, this model provides a valuable framework for the following stimulation experiments.
Application of different oLFS protocols during spontaneous epileptiform activity
Since the sclerotic hippocampus is considered as the focus of epileptiform activity (KrookMagnuson et al., 2015; Pallud et al., 2011), we targeted DGCs, the major surviving excitatory neurons, by photostimulation of entorhinal afferents. To this end, adult mice received KA into the hippocampus and a ChR2-carrying viral construct into the medial entorhinal cortex followed by LFP recordings and oLFS in the chronic epileptic phase (Figure 4A; Figure 4—figure supplement 1A). Prior to photostimulation, reference LFPs were recorded for 1 hr in ‘pre’ sessions to confirm the occurrence of spontaneous epileptiform activity in the idHC (Figure 4B) and cdHC (Figure 4—figure supplement 1B). Then, ChR2-expressing entorhinal fibers were stimulated in the sclerotic hippocampus (Figure 4C) with three frequencies (1, 0.5, and 0.2 Hz) on subsequent days. ChR2-mCherry expression and optic fiber position were verified histologically in all mice at the end of oLFS experiments (Figure 4—figure supplement 2, Figure 4—figure supplement 3).
One hour of optogenetic stimulation with pulsed light at 1 Hz or 0.5 Hz significantly decreased the high-load burst ratio and the epileptic spike rate, followed by a return to pre-stimulation levels within 2 hr independent of the KA concentration (Figure 4D–I, Figure 4—source data 1). Photostimulation at 0.2 Hz had no significantly suppressive effect on the high-load burst ratio, but on the epileptic spike rate (Figure 4J–L; Figure 4—source data 1). 1 Hz oLFS was significantly more effective than 0.2 Hz in suppressing high-load bursts and epileptic spikes, whereas we found no significant difference between 1 Hz and 0.5 Hz (Figure 4P,Q; Figure 4—source data 1). Looking at individual sessions in more detail, 1 Hz oLFS had a higher percentage of sessions with a suppression efficacy above 75% than 0.5 Hz and 0.2 Hz regarding high-load burst ratio and epileptic spike rate (1 Hz: 86.36% and 48.00%, 0.5 Hz: 56.25% and 23.08%, 0.2 Hz: 14.29% and 5.88%). As expected, oLFS (1 Hz) did not influence epileptiform activity in no-virus control mice (Figure 4M–O, Figure 4— source data 1).
To clarify whether the suppression of high-load bursts and epileptic spikes due to oLFS was locally restricted to the stimulation site (idHC), we analyzed LFPs contralaterally (cdHC), outside of the epileptic focus (Figure 4—figure supplement 1A–C). Interestingly, high-load burst ratio and epileptic spike rate were also suppressed in the cdHC (Figure 4—figure supplement 1D–O, Figure 4— figure supplement 1—source data 1). 1 Hz oLFS was significantly more effective than 0.2 Hz in reducing high-load bursts and epileptic spikes, whereas there was no significant difference between 1 Hz and 0.5 Hz (Figure 4—figure supplement 1P,Q; Figure 4—figure supplement 1—source data 1). Looking at individual sessions in detail, 1 Hz oLFS had a higher percentage of sessions with a suppression efficacy above 75% than 0.5 Hz and 0.2 Hz regarding high-load burst ratio and epileptic spike rate (1 Hz: 87.50% and 66.67%, 0.5 Hz: 50.00% and 7.14%, 0.2 Hz: 23.53% and 5.88%).
Next, we tested whether the neuronal responses to oLFS were confined to the stimulated area by analyzing the spatial and temporal occurrence of evoked responses in the idHC and cdHC. In all animals, pulsed light delivery to the idHC did not only trigger local but also delayed responses in the cdHC (Figure 4—figure supplement 4A–C). These latencies remained stable over the stimulation period of 1 hr (Figure 4—figure supplement 4D) ranging from 8 to 12 ms (Figure 4—figure supplement 4E; 8.99 ± 0.59 ms, n = 8 sessions), suggesting that photostimulation of entorhinal afferents may lead to repeated action potential generation in a subset of DGCs and subsequent propagation within the hippocampal network.
In parallel to LFP recordings and optogenetic stimulation, we assessed the animals’ motor behavior in an open-field environment. Mice frequently groomed and explored their environment during oLFS. Video tracking revealed that independent of the KA concentration and stimulation frequency, mice did not change their running behavior during oLFS compared to the ‘pre’ recording. The percentage of running time gradually declined during the recording time of 4 hr (Figure 4—figure supplement 5A–D, Figure 4—figure supplement 5—source data 1). The percentage of running time

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

6 of 26

Research article

Neuroscience

Figure 4. oLFS of entorhinal afferents interferes with spontaneous epileptiform activity in a frequency-dependent manner. (A–C) Experimental design. We targeted ChR2-mCherry expression (C, red) to excitatory neurons in the medial entorhinal cortex using viral vectors. ChR2-mCherry expression pattern for all mice included in the study are shown in Figure 4—figure supplement 2. The electrode positions are shown in Figure 4—figure supplement 3. (B, C) We locally stimulated entorhinal afferents in the sclerotic idHC for 1 hr per day, twice at each frequency applying only one frequency per session (1, 0.5, or 0.2 Hz). (D, G, J, M) Representative LFP traces (15 mM KA, idHC electrode) for the ‘pre’ and ‘oLFS’ sub-sessions (1, 0.5, 0.2 Hz and no-virus control, 1 Hz) are shown. Automatic detection of epileptiform activity is marked for low-load (blue), medium-load (green), and highload bursts (orange). (D) Photostimulation at 1 Hz effectively decreases spontaneous epileptiform activity in the idHC. (E, F) Automatic quantification of epileptiform activity shows that oLFS reduces the high-load burst ratio as well as the epileptic spike rate in all animals independently of the KA concentration (10 mM: light gray; 15 mM: dark gray; 20 mM: black) followed by a return to pre-stimulation levels within 2 hr (‘post 1’ and ‘post 2’). (G, J) oLFS with (H, I) 0.5 Hz or (K, L) 0.2 Hz has a weaker antiepileptic effect during stimulation. Single sessions (olive-green) were used to calculate the oneway ANOVA; Tukey’s multiple comparison test (all KA concentrations pooled); *p<0.05, **p<0.01, and ***p<0.001. All mice were video recorded and the running behavior was analyzed during each session as shown in Figure 4—figure supplement 5 with the source data provided in Figure 4—figure supplement 5—source data 1. (M–O) 1 Hz stimulation does not have any effect on epileptiform activity in no-virus controls (20 mM KA). All values are given as mean ± SEM. Analysis of the cdHC is shown in Figure 4—figure supplement 1 with the source data provided in Figure 4—figure supplement 1—source data 1. We noticed that local oLFS in the idHC leads to a delayed cellular responses in the cdHC as shown in Figure 4—figure supplement 4. (P, Q) Comparison of the stimulation frequencies in terms of suppression efficacy using the high-load burst ratio and epileptic spike rate (1-(‘oLFS’/‘pre’)*100)(one-way ANOVA; Dunns’s multiple comparison test (all KA concentrations pooled), mean ±95% CI; *p<0.05, **p<0.01, ***p<0.001). Source data is provided in Figure 4—source data 1.
Figure 4 continued on next page

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

7 of 26

Research article
Figure 4 continued
The online version of this article includes the following source data, source code and figure supplement(s) for figure 4:
Source data 1. oLFS effect on ipsilateral epileptiform activity. Figure supplement 1. oLFS of entorhinal afferents in the idHC interferes with spontaneous epileptiform activity in the cdHC. Figure supplement 1—source data 1. oLFS effect on contralateral epileptiform activity. Figure supplement 2. ChR2-mCherry expression pattern for all oLFS-stimulated mice. Figure supplement 3. Positions of implanted electrodes and optic fibers for all animals included in the study. Figure supplement 4. Local oLFS leads to delayed cellular responses in the cdHC. Figure supplement 4—source code 1. 1 Hz oLFS response: idHC to cdHC delay calculation. Figure supplement 5. Movement analysis of chronically epileptic mice during LFP recordings. Figure supplement 5—source data 1. oLFS effect on running behavior over time.

Neuroscience

in all ‘pre’ and ‘oLFS’ sessions over the 6 days of stimulation was stable (Figure 4—figure supplement 5E,F), indicating that hippocampal oLFS did not impair open-field running behavior of chronically epileptic mice.
So far, our findings demonstrated that 1 Hz oLFS of entorhinal afferents was highly effective in the suppression of spontaneous epileptiform activity within as well as outside of the epileptic focus. Moreover, this effect was independent of the degree of hippocampal sclerosis and seizure burden. Therefore, we pooled all KA groups in the following experiments.
Effects of oLFS and eLFS on induced behavioral seizures
In the intrahippocampal KA model, spontaneous epileptiform activity is mainly subclinical and rarely generalizes into behavioral seizures (Ha¨ussler et al., 2012; Janz et al., 2018; Klein et al., 2015). To assess the impact of oLFS on generalized seizures, we induced these seizures by 10 Hz photostimulation as described previously (Janz et al., 2018; Osawa et al., 2013). In addition to the evoked potentials during stimulation, high-amplitude epileptic spikes emerged in the LFP which gradually became rhythmic and dominant before progressing into full-blown behavioral seizures (Figure 5— figure supplement 1A). These evoked seizures displayed electrographic features highly similar to those of spontaneous generalized seizures (Figure 5—figure supplement 1B) and were accompanied by the same stereotypic myoclonic movements (e.g. rearing, falling, and convulsion).
We determined the minimum stimulus duration sufficient to reliably trigger a generalized seizure for each animal (Figure 5A, as described in the Materials and methods under Optogenetic stimulation). Interestingly, in mice with lower KA concentrations, generalized seizures were induced much faster, suggesting a higher susceptibility for seizure generalization (Figure 5B, 10 mM: 5.75 ± 0.63 s; 15 mM: 7.83 ± 0.87 s; 20 mM: 13.67 ± 1.86 s, 10 and 15 mM vs. 20 mM p<0.01; n = 3; 6; 4 animals). With ongoing seizure activity, mice exhibited behavioral symptoms equivalent to Racine stages (RS) 1 to 5 (Racine, 1972) independent of the stimulation duration (Figure 5C, n = 3; 6; 4 animals).
Next, we probed whether oLFS can interfere with generalized seizures. When 1 Hz oLFS was started directly after the pro-convulsive 10 Hz stimulus, ongoing seizures were not interrupted (Figure 5—figure supplement 2A). In contrast, pre-conditioning with 1 Hz oLFS for 30 min prior to the pro-convulsive stimulus was highly effective in lowering the probability for seizure generalization (Figure 5D,E, without (w/o) pre-oLFS: 91.42 ± 5.18%; with 1 Hz pre-oLFS: 14.87 ± 8.33%, p<0.001, n = 13 animals). 0.5 Hz oLFS was also effective in preventing evoked generalized seizures (Figure 5— figure supplement 2B,C, w/o pre-oLFS: 97.73 ± 2.27%; with 0.5 Hz pre-oLFS: 12.12 ± 8.13%, p<0.05; n = 11 animals). In trials with incomplete seizure suppression, the ensuing seizures were associated with a milder behavioral phenotype (Figure 5F, w/o oLFS: RS 2.58 ± 0.31; with 1 Hz oLFS: RS 0.29 ± 0.17, n = 13 animals; Figure 5—figure supplement 2D, w/o oLFS: RS 2.88 ± 0.32; with 0.5 Hz oLFS: RS 0.37 ± 0.25, p<0.01; n = 10 animals).
Since LFS and seizure induction were both driven by photostimulation, we performed additional experiments combining optogenetic 10 Hz stimulation with electrical 1 Hz preconditioning in the same animal. To this end, mice received intrahippocampal KA (15 mM) and the ChR2-carrying viral construct into the medial entorhinal cortex followed by a side-by-side implantation of an optic fiber and a stimulation electrode into the dentate gyrus. In addition, all animals were implanted with LFP recording electrodes as described above (Figure 1B).

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

8 of 26

Research article

Neuroscience

Figure 5. Preconditioning with LFS prevents optically evoked seizure generalization. (A, D, H) Representative LFP traces at three recording sites (cdHC, idHC, and ipsilateral ventral hippocampus (ivHC)). A schematic of the respective stimulation procedure is shown above each cdHC trace. (A, B) Local 10 Hz photostimulation of entorhinal afferents reliably induces generalized seizures in all KA groups. Evoked, generalized seizures displayed electrographic features highly similar to spontaneous generalized seizures as shown in Figure 5—figure supplement 1. The time needed to induce a generalized seizure (induction time) is longer with increasing KA concentration. One-way ANOVA; Tukey’s multiple comparison test; **p<0.01. (C) Mice exhibit behavioral symptoms equivalent to RS stage 1–5, independently of the KA concentration. (D, E) 1 Hz oLFS as well as (H, I) eLFS for 30 min before the pro-convulsive stimulus significantly decreases the seizure probability in all animals. Wilcoxon rank test, matched-pairs; ***p<0.001 (oLFS, n = 13; eLFS, n = 4 animals). Preconditioning with 0.5 Hz was also able to interfere with the generation of evoked generalized seizures as shown in Figure 5—figure supplement 2. (F, J) Trials in which seizure generalization is not prevented completely, the ensuing seizure is associated with a milder behavioral phenotype (RS). Wilcoxon rank test, matched-pairs; ***p<0.001 (oLFS, n = 13; eLFS, n = 4 animals). (G, K) Cellular response to 10 Hz stimulation quantified as mean AUC. (G) The response is reduced after 1 Hz oLFS stimulation in sessions in which seizures have been successfully suppressed. (K) No significant reduction is visible for AUC values after eLFS. Paired t-test; **p<0.01 (oLFS, n = 13; eLFS, n = 4 animals, respectively). All values are given as mean ± SEM. AUC calculation was performed in python 2.7 provided in Figure 5—source code 1. The online version of this article includes the following source data and figure supplement(s) for figure 5:
Source code 1. AUC calculation of 10 Hz oLFS evoked responses. Figure supplement 1. Comparison of spontaneous and evoked generalized seizures. Figure supplement 2. Preconditioning with 0.5 Hz prevents evoked generalized seizures.
We determined the minimum optical stimulus duration sufficient to reliably trigger a generalized seizure. Then, we probed the seizure-suppressive action of 1 Hz eLFS prior to optogenetic 10 Hz seizure induction (Figure 5H). Thirty minutes of 1 Hz eLFS was as effective as oLFS in reducing the probability of generalized seizures (Figure 5I, w/o pre-eLFS: 93.75 ± 6.25%; with 1 Hz pre-eLFS:

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

9 of 26

Research article

Neuroscience
16.67 ± 9.62%) and the associated behavior (Figure 5J, w/o eLFS: RS 2.38 ± 0.40; with 1 Hz eLFS: RS 0.33 ± 0.33, n = 4 animals).
In conclusion, preconditioning by optogenetic and electrical 1 Hz LFS was highly effective in preventing evoked generalized seizures.
Cellular responses to oLFS
In order to investigate the underlying mechanisms of the anti-epileptic effects of oLFS, we quantified the cellular responses in the sclerotic hippocampus. For this, we calculated the area under the curve (AUC) of each evoked response over the 1-hr stimulation period. We only selected sessions with high stimulation efficacy (within the 95% confidence interval (CI), compare Figure 4P).
Photostimulation (1 Hz) evoked stable response waveforms (Figure 6) which decreased slightly in amplitude over time in non-epileptic (saline-injected) mice (Figure 6A1-3, B). In chronically epileptic mice, the cellular responses declined strongly and rapidly within the first 10 min (Figure 6C1-3, D). Lower frequencies (0.5 and 0.2 Hz) altered the cellular response much less (Figure 6E–H) as evident from AUC analysis. Similarly, AUCs of evoked responses of the pro-convulsive 10 Hz pulse-train were reduced by about 40% after pre-conditioning when compared to the responses without

Figure 6. Evoked cellular responses decrease over time during continuous oLFS. (A, C, E, G) Representative examples of evoked responses in the dentate gyrus of idHC following local photostimulation of entorhinal afferents for (A) non-epileptic control (1 Hz) and (C, E, G) a chronically epileptic mouse (1, 0.5, and 0.2 Hz). (A2, C2, E2, G2) Mean evoked responses (50 ms–long light pulse) across 15 min time windows. (A3, C3, E3, G3) For each evoked response, AUCs are calculated during a [À0.1, +0.2 s] interval relative to the onset of each light pulse. AUC values that are within high-load bursts are marked in orange and are excluded for the calculation of the polynomial fit (blue line). (B, D, F, H) Polynomial fits of AUC normalized to the first AUC value (delta AUC) for all stimulation sessions (gray) and mean changes (red). Calculations were performed in python 2.7 provided in Figure 6—source code 1. The online version of this article includes the following source code for figure 6:
Source code 1. AUC calculation of 1 Hz oLFS evoked responses.

Paschen et al. eLife 2020;9:e54518. DOI: https://doi.org/10.7554/eLife.54518

10 of 26
SeizuresOlfsEpileptiform ActivityIdhcCdhc