Abstract
The subiculum represents a crucial brain pivot in regulating seizure generalization in temporal lobe epilepsy (TLE), primarily through a synergy of local GABAergic and long-projecting glutamatergic signaling. However, little is known about how subicular GABAergic interneurons are involved in a cell-type–specific way. Here, employing Ca2+ fiber photometry, retrograde monosynaptic viral tracing, and chemogenetics in epilepsy models of both male and female mice, we elucidate circuit reorganization patterns mediated by subicular cell-type–specific interneurons and delineate their functional disparities in seizure modulation in TLE. We reveal distinct functional dynamics of subicular parvalbumin+ and somatostatin+ interneurons during secondary generalized seizure. These interneuron subtypes have their biased circuit organizations in terms of both input and output patterns, which undergo distinct reorganization in chronic epileptic condition. Notably, somatostatin+ interneurons exert more effective feedforward inhibition onto pyramidal neurons compared with parvalbumin+ interneurons, which engenders consistent antiseizure effects in TLE. These findings provide an improved understanding of different subtypes of subicular interneurons in circuit reorganization in TLE and supplement compelling proofs for precise treatment of epilepsy by targeting subicular somatostatin+ interneurons.
Significance Statement
Temporal lobe epilepsy (TLE) is the most common type of refractory epilepsy and not well controlled by current medications. In this study, we reveal that subicular GABAergic interneurons are involved in seizure generalization in a cell-type–specific way. We find that subicular parvalbumin+ and somatostatin+ interneurons have distinct functional dynamics and undergo different circuit reorganizations in chronic epileptic condition. Notably, somatostatin+ interneurons exert more effective inhibition onto pyramidal neurons, engendering consistent antiseizure effects. This is therapeutically significant for precise treatment targeting the subiculum in TLE.
Introduction
Epilepsy, characterized by recurrent spontaneous seizures, is a severe neurological disease that affects millions of individuals worldwide (Thijs et al., 2019) and poorly managed by current medications and surgical treatment (Löscher et al., 2020; Sills and Rogawski, 2020). Temporal lobe epilepsy (TLE) is the most common type of refractory epilepsy (Engel, 1996; Jallon, 1997). Seizures in TLE typically originate in specific areas like the hippocampus or the amygdala and further propagate and develop into secondary generalized seizure (sGS), leading to a high risk of permanent neural damage or even sudden unexpected death in epilepsy (Sveinsson et al., 2020). However, the precise pathogenesis of sGS remains unclear. As the cause of epilepsy has been gradually considered as an imbalanced “excitation and inhibition” in neural activity in key brain regions and related circuits (Paz and Huguenard, 2015; Wang and Chen, 2019), it is therapeutically necessary to deconstruct neural circuit mechanisms underlying sGS.
The subiculum is the main output region of the hippocampus. In TLE, local degraded GABAergic signaling and hyperexcited long-projecting glutamatergic signaling synergistically contribute to seizure generation and generalization (O'Mara et al., 2001; Cohen et al., 2002; Fei et al., 2021). The mechanism of subicular glutamatergic neurons in regulating TLE has been finely studied (Cross and Cavazos, 2007; Herrington et al., 2015; C. Xu et al., 2019; Fei et al., 2022). Glutaminergic neurons in the subiculum remain relatively intact in numbers and fire synchronously during various epileptic conditions. Specifically, a subclass of subicular glutamatergic neurons, distributed in the deep layer and projecting to the anterior nucleus of the thalamus (ANT), are hyperactivated and mediate the development of sGS (Fei et al., 2022). In contrast, not much attention is paid to the role of subicular inhibitory GABAergic interneurons (INs) in TLE. Although interneurons only compose 10–15% of total subicular neurons, they crucially regulate hippocampal excitatory output through local feedforward and feedback circuits in physiological condition (Ding, 2013). Subicular interneurons are heterogeneous in molecular, morphological, and electrophysiological characteristics and experience various changes in numbers and properties in epileptic condition (Benini and Avoli, 2005; Huberfeld et al., 2007; Knopp et al., 2008). With regard to two major types of subicular INs, parvalbumin+ interneurons (PV-INs) and somatostatin+ interneurons (SST-INs) are reported to function differently in vitro or vivo epilepsy models (Wang et al., 2017; Anstötz et al., 2021). Our previous study has found that optogenetically activating PV-INs or SST-INs impose entirely different effects on sGS expression in hippocampal kindling model (Wang et al., 2017), indicating their distinct roles in TLE. However, the causal role of subicular cell-type–specific GABAergic interneurons in TLE is still not fully illustrated.
Here, combined with fiber photometry, rabies-assisted retrograde tracing, optogenetics, and chemogenetics in animal models of epilepsy, we aim to investigate the difference between subicular PV- and SST-INs, in terms of their neural excitability, input/output circuit reorganizations, and function in controlling seizures of TLE. We find these two types of interneurons could be distinguished by their response to hippocampal seizures, as well as circuit organization and reorganization patterns in chronic epilepsy, which could lead to functional difference in seizure modulation. This may be therapeutically significant for developing precise treatment of sGS in TLE by targeting subicular SST-INs.
Materials and Methods
Animals
This study used PV-Cre (stock number 008069), SST-Cre (stock number 013044), and CaKMIIα-Cre (stock number 005359) transgenic mice. All animals were genotyped according to the guidelines of The Jackson Laboratory. Both mature male and female mice (8–16 weeks) were used in fiber photometry, viral tracing, and immunochemistry studies. To avoid gender differences in epileptic seizures (Christensen et al., 2005; Scharfman and MacLusky, 2006), only mature male mice were used in behavioral tests (acute and chronic seizure modulation experiments). All animals were group-housed (4–6 per cage, single gender) before experiments with a 12 h light/dark cycle. Temperature at 23–26°C and humidity at 50–60% were maintained. All behavioral tests were performed at light cycle. All experimental procedures were approved by the ethical committee of Zhejiang Chinese Medical University and in complete compliance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals.
Viral conducts
For fiber photometry, AAV2/9-hEF1a-DIO-GCaMP6s (viral titers, 1.0 × 1013 particles/ml; Taitool Bioscience) or AAV2/9-CaMKIIα-GCaMP6s (viral titers, 5.6 × 1012 particles/ml; OBiO Technology) were injected into the subiculum of PV-Cre or SST-Cre mice. For monosynaptic retrograde tracing, AAV2/9-hEF1a-DIO-TVA-RVG-eGFP (viral titers, 1.0 × 1013 particles/ml; Taitool Bioscience) and RV-EnvA-ΔG-dsRed (viral titers, 1.0 × 1013 particles/ml; Taitool Bioscience) were injected into the subiculum of PV-Cre or SST-Cre mice; AAV2/9-hEF1a-DIO-TVA-RVG-mCherry (viral titers, 1.0 × 1013 particles/ml; Taitool Bioscience) and RV-EnvA-ΔG-eGFP (viral titers, 1.0 × 1013 particles/ml; Taitool Bioscience) were injected into the subiculum of CaKMIIα-Cre mice. For selective manipulation of subicular PV-INs or SST-INs, AAV2/9-hSyn-FLEX-ChrimsonR-tdTomato (viral titers, 1.0 × 1013 particles/ml; Taitool Bioscience), AAV2/9-EF1a-DIO-hM3D(Gq)-mCherry (viral titers, 4.46 × 1012 particles/ml; OBiO Technology), AAV2/9-hSyn-DIO-hM4D(Gi)-mCherry (viral titers, 1.0 × 1013 particles/ml; Taitool Bioscience), or AAV2/9-EF1a-DIO-mCherry (control virus, viral titers, 1.0 × 1013 particles/ml; Taitool Bioscience) was injected into the subiculum of PV-Cre or SST-Cre mice.
Stereotactic injections and surgeries
Mice were anesthetized with ∼2% isoflurane and mounted in a stereotaxic apparatus (68043, RWD Life Science) for virus injection and surgery. The body temperature of the mice was kept warm via a heating pad.
For virus injection, a glass micropipette attached to a 1 μl syringe was fixed in the injection pump (Pump 11 Elite, Harvard Apparatus). Injections were targeted into the subiculum (AP, −3.4 mm; L, −2.0 mm; DV, −1.8 mm) at 50 nl/min. In fiber photometry and behavioral experiments, the injection volume of AAV was 200 nl. After each injection, the needle was left in place for 10 min before withdrawal. The skin was then closed using surgical clips.
For electrode and optical fiber implanting, twisted-bifilar stainless electrodes (791500, 0.127 mm diameter, A-M Systems) were implanted into the right hippocampal CA3 area (AP, −2.9 mm; L, −3.1 mm; V, −3.1 mm) for both kindling stimulation and EEG monitoring. Optical cannulas (0.2 mm diameter, 0.37 NA, Inper) were separately lowered into the subiculum (AP, −3.4 mm; L, −2.0 mm; V, −1.7 mm). Three screws were then placed over the skull to fix the dental cement, among which one over the motor cortex as the ground and another over the cerebellum as reference. After surgeries, mice were group-housed 2–3 per cage for a better recovery. One week later, behavioral tests were conducted. Viral expressions and electrode locations were verified after all behavior tests. Mice with correct place of virus and optical cannula location were included in the statistics.
Calcium fiber photometry
Fiber photometry was performed in mice expressing GCaMP6s. The fluorescence signal was filtered and then collected by a photomultiplier tube. The voltage signal was converted from an amplifier to a low-pass filter (<100 Hz) and collected by fiber photometry software with a sampling rate of 100 Hz. The power of the laser intensity between the fiber tip and the mice brain regions ranged from 0.01 to 0.03 mW to avoid bleaching. Data were further analyzed by MATLAB (version R2021b, MathWorks). The values of fluorescence change were shown as ΔF/F with the equation (ΔF/F) = (F − F0) /F0 according to a MATLAB program.
During fiber photometry combined with optogenetics experiments, through a single optic cannula placed in the subiculum, 589 nm laser (20 Hz, 10 s on–off cycle, 10 cycles) and 488 nm laser were given simultaneously for selective activation of PV-INs or SST-INs and GCaMP fluorescence of pyramidal neurons recording, respectively.
Light stimulation
A 589 nm light was delivered through a 200-μm-diameter optical fiber connected to a mobile light stimulator. Laser power was adjusted to 3–5 mW. In optogenetics combined with fiber photometry experiment, a single trial lasted for 30 s (10 s pre, 10 s light delivery, and 10 s post). During light delivery period, stimulation parameters were 20 Hz and 10 ms pulse.
Hippocampal kindling model
Hippocampal kindling was conducted as in our previous study (Fei et al., 2022). EEGs were recorded by PowerLab System (PowerLab 8/35, ADInstruments) at a 1 kHz sample rate. All mice received 10 kindling stimulations daily through an external stimulator in series with a recording system. The stimulation parameters were 400 μA, 20 Hz, 1 ms monophasic square-wave pulse, and 40 pulses. Seizure severity was defined according to the modified Racine scale (Racine, 1972): (1) facial movement; (2) head nodding; (3) unilateral forelimb clonus; (4) bilateral forelimb clonus and rearing; and (5) rearing and falling. Stages 1–3 are considered as FSs, and Stages 4–5 are sGSs. The onset of sGS was defined as the timing when both forelimbs of mice were clonus and rearing. The length of latency to sGS was defined as the duration between the start of kindling and the timing of sGS onset. The length of after-discharge duration (ADD) was defined as the duration between the start of kindling and the end of paroxysmal discharge event in EEG.
KA-induced acute seizure model
In this model, seizures were typically induced by intra-amygdaloid (12 mM, 0.1 μl) or intraperitoneal (15 mg/kg) injection of KA (ab120100, Abcam; Lévesque and Avoli, 2013). EEG of the hippocampal CA3 was monitored by PowerLab System, and epileptic behavior was simultaneously video recorded in each free moving mouse in a transparent chamber. Seizure severity was defined according to our previous study (B. Chen et al., 2020). The onset of sGS was defined as the timing when both forelimbs of mice were rearing and raising. The onset of status epilepticus (SE) was defined as the timing when continuous ictal discharge (>5 min, twofold-baseline high amplitude with frequency greater than 3 Hz) appeared (Z. H. Xu et al., 2016; Zhao et al., 2020). We determined the latency to sGS and SE for each mouse in 1.5 h after KA induction.
For manipulation of acute seizure in intra-amygdaloid model, KA was injected via a cannula guided into the BLA (AP, −1.4 mm; L, −2.9 mm; DV, −4.8 mm) over 3 min with a tube attached to a 1 μl syringe. One hour before KA induction, mice with mCherry or hM3Dq were intraperitoneally injected with CNO (2 mg/kg).
KA-induced chronic epilepsy model
In the chronic model, KA (12 mM, 0.1 μl) was stereotaxically injected into the BLA (AP, −1.4 mm; L, −2.9 mm; DV, −4.9 mm) over 3 min with a glass micropipette attached to a 1 μl syringe under ∼2% isoflurane. Mice that survived from SE would undergo spontaneous recurrent seizures in the following several months (Mouri et al., 2008; Li et al., 2012).
For chronic spontaneous seizure manipulation, 2 months after KA injection, EEG synchronized with video monitoring in mice with hM3Dq was recorded 8 h/day for 3 d as Pre. During this period, mice were intraperitoneal injected with saline. A typical sGS was defined as an event that lasted >30 s synchronized abnormal EEG and behavioral tonic–clonic seizure. Mice with detectable sGSs were intraperitoneally injected with CNO (2 mg/kg) for the next 3 d (1 h before recording) and then another 3 d with saline as Post. We calculated the number and total duration of sGSs for each mouse.
RV-assisted retrograde tracing
To identify the upstream input regions to the subicular PV-INs, SST-INs, and pyramidal neurons, we used a Cre-dependent helper virus [TVA mixed with RV glycoprotein (RVG)] and RVΔG for monosynaptic retrograde tracing. For control mice, intra-amygdaloid saline injection (0.1 μl) was conducted. For chronic epilepsy mice, we injected KA (12 mM, 0.1 μl). Helper virus and RVΔG were then injected 1 month after saline or KA injection. Briefly, 60 nl helper virus was injected into the subiculum of PV-Cre, SST-Cre, or CaMKIIα-Cre mice first. After 3 weeks, 90 nl RVΔG was injected into the same location. Mice were perfused for histology 7–10 d after the second injection. Cells that expressed both fluorescence were defined as starter cells, while those that expressed only RV's fluorescence were upstream cells.
Histology and quantification
Mice were deeply anesthetized with pentobarbital and then perfused with 0.9% saline and 4% PFA in 0.1 M phosphate buffer. Brains were removed and stored in the same fixative overnight at 4°C and then dehydrated in 30% sucrose for 48–72 h. Coronal sections were cut on a freezing microtome NX70 (Thermo Fisher Scientific).
For RV cell counting, 60 μm slices were selected from bregma location approximately +2.46 to −4.96 mm with a 60 μm interval. The subiculum is located between the CA1 and the retrosplenial cortex but is considered a distinct anatomical and functional entity (Ding, 2013). Brain regions where upstream cells are located were recognized according to the brain atlas. Anteroposterior location area before −2.90 mm was defined as the anterior subiculum, while after −2.90 mm was defined as the posterior subiculum.
Slices containing each corresponding region were all included in the calculation by ImageJ 8.0 (Fuji) software. The connection strength index (CSI) was defined as the presynaptic cell number in a certain brain region/starter cell number (Sun et al., 2014; Callaway and Luo, 2015). To avoid the error from virus infection efficiency among each mouse, we normalized the data as “total inputs,” which equals to RV cell number in certain brain region/total RV cell number.
In vitro electrophysiology
PV-hM4Di or SST-hMDi mice were decapitated, and the brains were rapidly dissected and placed in ice-cold oxygenated solution containing the following solutes (in mM): 110 choline chloride, 2.5 KCl, 1.3 NaH2PO4, 25 NaHCO3, 0.5 CaCl2, 7 MgCl2, 20 D-glucose, 1.3 ascorbic acid, and 0.6 Na-pyruvate. Subsequently, 300 μm coronal slices containing the hippocampus were cut using a vibratome (VT1000, Leica) and incubated at 34.7°C for 0.5 h and then maintained at 25°C for further experiments.
Slice recording was performed as in our previous study (Fei et al., 2022). Briefly, slices were kept at 25°C in a recording chamber perfused with 3 ml/min ACSF containing the following (in mM): 120 NaCl, 11 dextrose, 2.5 KCl, 1.28 MgSO4, 2.5 CaCl2, 1 NaH2PO4, and 14.3 NaHCO3. Patch pipettes were pulled from glass capillaries and at resistances of 6–9 MΩ, which contained the following (in mM): 35 K-gluconate, 110 KCl, 10 HEPES, 2 MgCl2, 2 Na2ATP, and 10 EGTA. Patch-clamp recordings were performed by an EPC 10 amplifier (HEKA Instruments), with a low-pass filter at 3 kHz and a sample rate of 10 kHz. To record the rheobase of subicular PV- or SST-INs, a hyperpolarizing to depolarizing gradient of currents (−20 to 300 pA, 20 pA each step, 500 ms) was injected. Then slice was incubated in CNO (20 μM) for 5 min, and currents were injected again. We calculated the increase of rheobase (%) after CNO incubation for each recorded PV-IN or SST-IN.
Immunohistochemistry
For immunostaining, slices were rewarmed for 30 min and placed at a mixed diluent (PBS plus 3% BSA plus 5% normal donkey serum plus 0.3% Triton X-100) for another 2 h and then incubated with primary antibody at 4°C for 24 h. The primary antibodies used in the study were as follows: guinea pig polyclonal anti-c-Fos (1:2,000; 226004, Synaptic Systems), rabbit monoclonal anti-CaMKII (1:800; ab134041, Abcam), rabbit monoclonal anti-PV (1:1,000; pv27, Swant), rabbit monoclonal anti-SST (1:1,000; HA721460, Huaan Biotechnology). After three times of 10 min wash by PBS, slices were incubated with the corresponding secondary antibody for 2 h at room temperature. Then after another three times wash, slices were placed on a glass slide and mounted with DAPI (Vectashield Mounting Media, VectorLabs). Images were captured by confocal microscopy (SP8, Leica) or a digital slicing workstation (VS120, Olympus).
Statistics and reproducibility
Data in the manuscript were presented as means ± SEM. Experiments were repeated at least two times independently with similar results. Quantification of cells and EEG/calcium signaling analysis in this study were done by a well-trained investigator who was blinded to both mouse strain and experimental condition. In the quantification of RV cells, data collection was unblinded to experimental condition because mice injected with KA underwent remarkable cell loss and their sections were totally different from those injected with saline. Statistical analysis was performed by Prism 9 (GraphPad Software). Data with Gaussian distributions were analyzed by t test and one-way or two-way ANOVA test followed by appropriate post hoc test for multiple comparisons for statistical significance. Non-normally distributed data were analyzed by the Wilcoxon signed-rank test or Friedman with post hoc Dunn's test for multiple comparisons. Significance of results was reported in the figure legends. For all analyses, the threshold was determined at α = 0.05, and p < 0.05 was considered significant.
Results
Subicular PV- and SST-INs manifest distinct functional dynamics during secondary GSs
Firstly, to determine how PV- and SST-INs dynamically responded to seizure generalization, we recorded real-time calcium signals of PV- and SST-INs in vivo in mice that underwent intraperitoneal injection of KA-induced seizures. This was accomplished by cell-type selectively transfecting the subiculum of PV-Cre or SST-Cre mice with an AAV expressing genetically encoded calcium indicator GCaMP6s in a floxed construct (AAV-hEF1a-DIO-GCaMP6s; T. W. Chen et al., 2013; Fig. 1a–c). During a single sGS event, the GCaMP fluorescence of PV- and SST-INs robustly increased (Fig. 1d–g). Although both showed upregulated calcium activity, a difference in dynamics could be identified: For PV-INs, the calcium activity experienced a small delay before the increase; while for SST-INs, a more rapid increase was observed. Consequently, the difference values of calcium peak time from sGS onset (PV 29.39 ± 4.59 s vs SST 0.036 ± 3.42 s, p = 0.0002) were significantly larger, and the difference values to seizure termination (PV 17.35 ± 4.39 s vs SST 39.48 ± 7.90 s, p = 0.0307) were significantly shorter in PV-INs compared with SST-INs (Fig. 1h,i). Moreover, as the amplitude and duration of GCaMP fluorescence could vary among mice, we normalized the timing of fluorescence peak to the latency to sGS and whole after-discharge duration (ADD). The fluorescence peak of PV-INs appeared later than that of SST-INs, relative to the latency to sGS (relative peak time/latency: PV 2.50 ± 0.33 vs SST 1.12 ± 0.32, p = 0.0107) or ADD (relative peak time/ADD: PV 0.77 ± 0.04 vs SST 0.31 ± 0.79, p = 0.0002), indicating SST-INs responded to sGS faster than PV-INs.
Subicular PV+ and SST+ interneurons manifest distinct functional dynamics during secondary generalized seizure. a, Schematic of GCaMP6(s) injection into the subiculum and electrode implantation into the CA3 of PV- or SST-Cre mice for fiber photometry during intraperitoneal injection of KA- or kindling- induced secondary generalized seizure (sGS). b, c, Representative images of GCaMP6(s) expression and cannula placement in the subiculum of PV-Cre (b) or SST-Cre (c) mice injected with AAV-DIO-GCaMP6s, respectively (PV-GCaMP6s or SST-GCaMP6s mice). Scale bar, 200 μm. d–g, Typical EEG (top), example trace (ΔF/F, bottom), and heatmap of calcium signal from PV-GCaMP6s (d and e, n = 7) or SST-GCaMP6s (f and g, n = 7) mice during KA-induced sGS. Triangles represent seizure onset, red dashed line indicates the begin of sGS, and gray dashed line indicates the end of seizure. In heatmap, each mouse includes one trial. h, i, Quantification of GCaMP6s signal peak time from sGS onset, relative ratio to latency to sGS (h, unpaired t test; left, t(12) = 5.131, ***p = 0.0002; right, t(12) = 2.449, *p = 0.0307) and peak time to seizure termination, relative ratio to ADD (i, unpaired t test; left, t(12) = 3.017, *p = 0.0107; right, t(12) = 5.158, ***p = 0.0002) of PV-GCaMP6s and SST-GCaMP6s mice. j–m, Typical EEG (top), example trace of calcium signal (ΔF/F, bottom), and heatmap from PV-GCaMP6s (j and k, n = 5) or SST-GCaMP6s (l and m, n = 5) mice during kindling-induced sGS. Black rectangle represents the kindling stimulation period, red dashed line indicates the begin of sGS, and gray dashed line indicates the end of seizure. In heatmap, each mouse includes two trials. n, o, Quantification of GCaMP6s signal peak time from sGS onset, relative ratio to latency to sGS (n, unpaired t test; left, t(8) = 3.695, **p = 0.0061; right, t(8) = 4.365, **p = 0.0024) and peak time to seizure termination, relative ratio to ADD (o, unpaired t test; left, t(8) = 2.689, *p = 0.0275; right, t(8) = 3.017, *p = 0.0166) of PV-GCaMP6s and SST-GCaMP6s mice. Data are presented as mean ± SEM.
Since the timing of sGS initiation was unpredictable during the acute KA model, we therefore conducted hippocampal kindling in mice with GCaMP6s to strictly control the onset of individual seizure and uncovered similar results (Fig. 1j–m). During kindling-induced sGS, the fluorescence peak of PV-INs also appeared later than that of SST-INs, as evidenced by a larger difference value of calcium peak time from sGS onset (PV 8.56 ± 1.28 s vs SST −1.25 ± 2.33 s, p = 0.0061), shorter value of peak time to seizure termination (PV 9.63 ± 2.37 s vs SST 16.83 ± 1.25 s, p = 0.0275), and larger relative ratio to the latency to sGS (relative peak time/latency: PV 1.52 ± 0.08 vs SST 0.90 ± 0.11, p = 0.0024) or ADD (relative peak time/ADD: PV 0.75 ± 0.07 vs SST 0.51 ± 0.11, p = 0.0166) in PV-INs (Fig. 1n,o). Moreover, when sGS ended, the calcium activity of SST-INs went quickly back to the baseline, but PV-INs were still maintained at a high level. These results demonstrate that subicular PV- and SST-INs have distinct functional dynamics in response to hippocampal seizure.
Subicular PV- and SST-INs have distinct inputs and undergo distinct patterns of circuit reorganization in epileptic condition
Distinctions in calcium activity of subicular PV- and SST-INs could be relevant to biased innervations by upstream brain regions. To verify this hypothesis, we performed monosynaptic retrograde tracing of subicular PV- and SST-INs by rabies virus (RV; Fig. 2a). The location of starter cells within the subiculum is shown in Figure 2b. The results showed that both subicular PV- and SST-INs received multiple inputs from the hippocampus, basal forebrain, thalamus, and cortices, among which the major afferents came from the CA1 (PV-Sham 56.64 ± 2.54 vs SST-Sham 57.70 ± 3.89%, Fig. 2h). Quantificationally, subicular PV-INs received much more inputs mainly from the subiculum itself (PV-Sham 18.56 ± 1.74 vs SST-Sham 9.77 ± 0.69%, p < 0.0001, Fig. 2h), while SST-INs received much more inputs from numerous other brain regions, including the MS/DB and ipsilateral CA3 (Fig. 2d,f). These results map the major input brain regions of subicular PV- and SST-INs, indicating these interneuron subtypes have their own biased circuit organizations directly connected with multiple brain regions out of the subiculum and may play an important role in the integration of upstream signals. Notably, SST-INs are more likely to be connected by extrasubiculum neurons compared with PV-INs.
Subicular PV+ and SST+ interneurons undergo distinct patterns of circuit reorganization in chronic epileptic condition. a, Strategy of monosynaptic retrograde tracing of subicular PV- and SST-INs’ inputs in normal and epileptic conditions. b, Representative images of overlap of helper (green) and RV-infected (red) on starter cells (yellow) in the subiculum of PV- and SST-Cre mouse. The white arrows indicated the location of starter cells. Scale bar, 500 μm. c, Typical CA3 EEG of secondary generalized seizure (sGS) and ictal spike (IS) during chronic period. Triangles indicate the beginning and ending of an sGS. d–h, Representative images (d–g) and quantifications (h) of RV-labeled neurons in different brain regions of subicular PV and SST-INs in normal and epileptic conditions. Percentages are standardized using the number of RV cells versus the total number of RV cells for each mouse. MS/DB, medium septum and diagonal band; ANT, anterior nucleus of thalamus; AD, anterodorsal thalamus; AV, anteroventral thalamus; AM, anteromedial thalamus; CA3 (c-lat), contralateral CA3. Scale bar, 500 μm. N = 11, 8, 12, 7 mice. Two-way ANOVA with post hoc Sidak's multiple-comparisons test. SUB, F(1, 34) = 11.35, PV-Sham versus PV-KA, **p = 0.0011; PV-Sham versus SST-Sham, ####p < 0.0001; SST-Sham versus SST-KA, p = 0.923. MS/DB, F(1, 34) = 14.61, PV-Sham versus PV-KA, ***p = 0.0005; PV-Sham versus SST-Sham, #p = 0.0437; SST-Sham versus SST-KA, p = 0.749. ANT, F(1, 34) = 9.589, PV-Sham versus PV-KA, **p = 0.0025; PV-Sham versus SST-Sham, p = 0.981; SST-Sham versus SST-KA, p = 0.949. CA3, F(1, 34) = 12.81, PV-Sham versus PV-KA, *p = 0.0171; PV-Sham versus SST-Sham, ###p = 0.0002; SST-Sham versus SST-KA, p = 0.235. Data are presented as mean ± SEM.
In TLE with sGS, seizure activities often result from an abnormality in the neural circuit level or the “reorganization and recruitment of epileptic neural network” (Peng et al., 2013; Choy et al., 2022; Frankowski et al., 2022). Due to subiculum's pivotal role in epilepsy, we next asked whether input networks of subicular PV- and SST-INs were reorganized in chronic epileptic condition, by using a classic intra-amygdaloid KA-induced chronic TLE model (Lévesque and Avoli, 2013). Twelve millimeter KA was microinjected into the BLA of PV-Cre or SST-Cre mice. Two months after KA injection, spontaneous sGS and ictal spikes (ISs) could be steadily recorded in the hippocampal CA3 (Fig. 2c). RV-assisted monosynaptic retrograde tracing was also performed in mice with chronic spontaneous seizure. The results showed that the CA1 still offered the most abundant projections to both subicular PV- and SST-INs. However, PV-INs received more inputs from regions outside of the hippocampus, including the MS/DB, anterior nucleus of the thalamus (ANT), and the ipsilateral CA3, while its afferents from the inside subiculum significantly decreased (Fig. 2e,h). As for SST-INs, the proportion of each upstream region was unchanged in general (Fig. 2g,h).
Interestingly, we found PV-INs underwent remarkable loss (∼52.5%) in chronic epileptic condition (Fig. 3a,b), while the number of SST-INs remained relatively intact, with only 20.4% neuronal loss (Fig. 3d,e). However, compared with PV-INs, the total input cell number and connection strength index (CSI) of subicular SST-INs were remarkably reduced to around half of those under normal condition (RV cell traced from PV-INs, Sham 8,322 ± 1,194 vs KA 6,028 ± 1,204, p = 0.213; RV cell traced from SST-INs, Sham 9,097 ± 1,689 vs KA 2,795 ± 406, p = 0.0126; CSI traced from PV-INs, Sham 119.1 ± 13.83 vs KA 94.91 ± 10.75, p = 0.2044; CSI traced from SST-INs, 126.4 ± 11.12 vs KA 52.13 ± 4.85, p = 0.0001; Fig. 3c,f). As a result, SST-INs received remarkably less inputs from most regions in absolute number (Fig. 3g–l), indicating outside-driven signaling of SST-INs was weakened. These results suggest that the neural circuit reorganization pattern centered by subicular PV and SST-INs occurs with different characteristics in chronic epileptic condition.
Subicular PV+ interneurons lose in number, while SST+ interneurons lose in input connection in chronic epileptic condition. a, b, Expression (a) and quantification (b) of subicular PV-INs in mice with and without chronic seizure. N = 4, 4 mice. Mann–Whitney test, U = 0, *p = 0.0286. c, Total input RV cells and connection strength index (CSI) traced from subicular PV-INs in normal and epileptic conditions. N = 11, 8 mice. Unpaired t test; left, t(17) = 1.294, p = 0.213; right, t(17) = 1.32, p = 0.2044. d, e, Expression (d) and quantification (e) of subicular SST-INs in mice with and without chronic seizure. N = 4, 4 mice. Mann–Whitney test, U = 2, p = 0.1143. f, Total input RV cells and CSI traced from subicular SST-INs in normal and epileptic conditions. N = 12, 7 mice. Unpaired t test; left, t(17) = 2.789, *p = 0.0126; right, t(17) = 4.898, ***p = 0.0001. g–l, CSI of RV-labeled neurons in different brain regions of subicular PV- and SST-INs in normal and epileptic conditions. Two-way ANOVA with post hoc Sidak's multiple-comparisons test. CA1, F(1, 34) = 1.862, PV-Sham versus PV-KA, p = 0.109; SST-Sham versus SST-KA, ***p = 0.0007. SUB, F(1, 34) = 1.123, PV-Sham versus PV-KA, **p = 0.0064; SST-Sham versus SST-KA, p = 0.2218. MS/DB, F(1, 34) = 15.12, PV-Sham versus PV-KA, p = 0.0827; SST-Sham versus SST-KA, *p = 0.0234. ANT, F(1, 34) = 9.066, PV-Sham versus PV-KA, *p = 0.0319; SST-Sham versus SST-KA, *p = 0.3508. CA3, F(1, 34) = 11.90, PV-Sham versus PV-KA, p = 0.3696; SST-Sham versus SST-KA, *p = 0.0143. Data are presented as mean ± SEM.
Subicular pyramidal neurons lose innervation from SST-INs in epileptic condition
Except for afferents of PV- and SST-INs, we further investigated their main downstream circuits. To this end, monosynaptic retrograde tracing rabies was injected into the subiculum of CaMKIIα-Cre mice, and immunohistochemical analysis was performed to see the types of local presynaptic cells of subicular pyramidal neurons (Fig. 4a,b). For total RV-labeled upstream cells, 61.64 ± 4.22% were colabeled with CaMKII, indicating recurrent excitatory circuits within the subiculum. For presynaptic GABAergic neurons, 12.78 ± 7.19% RV cells were labeled with PV and 5.47 ± 1.34% with SST (Fig. 4c,d, left panel). Combined with the above results, this evidence suggests that PV- and SST-INs may both form local inhibitory feedforward and feedback circuits within the subiculum; PV-INs prefer to form feedback circuits as they receive a large number of local neurons, while SST-INs prefer to form feedforward circuits as they receive input from outside of subiculum.
Subicular SST+, but not PV+, interneurons lose innervation onto pyramidal neurons in epileptic condition. a, Strategy of monosynaptic retrograde tracing of subicular pyramidal neurons’ inputs in normal and epileptic conditions. b, Representative images of overlap of helper (red) and RV-infected (green) starter cells (yellow) in the subiculum of CaMKIIα-Cre mouse with and without chronic seizure. Scale bar, 100 μm. c, d, Representative images of RV-labeled neurons (green) in the anterior (c) and posterior (d) subiculum merged with CaMKII, PV, and SST immunofluorescence (magenta). Scale bar, 100 and 25 μm for enlarged images. e–g, Quantification of the proportions of CaMKII (e), PV (f), and SST (g) neurons in retrogradely labeled neurons in the anterior (A) and posterior (P) subiculum in normal and epileptic conditions. N = 10, 9 mice. Unpaired t test, CaMKII, t(18) = 2.439, *p = 0.0253, Sham-A compared with Sham-P; t(17) = 3.525, ##p = 0.0026, Sham-A compared with KA-A; t(17) = 2.619, #p = 0.0180, Sham-P compared with KA-P. PV, t(18) = 2.268, *p = 0.0359, Sham-A compared with Sham-P; t(17) = 1.754, p = 0.0975, Sham-A compared with KA-A; t(17) = 0.3787, p = 0.7096, Sham-P compared with KA-P. SST, t(18) = 0.2591, *p = 0.7985, Sham-A compared with Sham-P; t(17) = 2.779, #p = 0.0129, Sham-A compared with KA-A; t(17) = 0.6352, p = 0.5338, Sham-P compared with KA-P. Data are presented as mean ± SEM.
Moreover, as subicular pyramidal neurons were reported to be heterogeneous across anteroposterior, proximal–distal, and superficial-deep axes (O'Mara et al., 2001), we found that connections between excitatory and inhibitory neurons were unevenly distributed across an anteroposterior axis of the subiculum. Excitatory neurons tended to send more efferent to pyramidal neurons in the posterior subiculum than the anterior; For interneurons, PV-INs in the anterior made more innervation onto local pyramidal neurons, while SST-INs had no subregional difference (Fig. 4e–g).
Then we investigated the reorganization of subicular pyramidal neurons, especially changes in their afferents from local PV- and SST-INs in chronic epileptic condition. Rabies was thus injected into the subiculum of KA-treated CaMKIIα-Cre mice. The results showed that local pyramidal neurons were still connected with pyramidal neurons and interneurons within the subiculum (Fig. 4c,d, right panel). However, pyramidal neurons received more projections from local pyramidal neurons (anterior, Sham 54.42 ± 4.82 vs KA 73.54 ± 2.02%, p = 0.0026; posterior, Sham 68.78 ± 3.41 vs KA 78.81 ± 1.38%, p = 0.0180) compared with physiological condition. Less inputs from SST-INs, especially in the anterior subiculum, were also found (Sham 5.93 ± 1.22 vs KA 2.21 ± 0.38%, p = 0.0129, Fig. 4e–g). In contrast, projections from PV-INs did not change in epileptic condition (anterior, Sham 17.19 ± 2.60 vs KA 11.76 ± 1.50%, p = 0.0975; posterior, Sham 10.08 ± 1.75 vs KA 11.07 ± 1.96%, p = 0.7096). These results indicate that SST-INs’ anatomical innervation on pyramidal neurons is weakened in TLE.
Subicular SST-INs dominate in modifying the excitability of local pyramidal neurons
As distinct anatomic characteristics of subicular PV- and SST-INs on the circuit level were seen, we then tested their modulating functions on local pyramidal neurons. To achieve that, we first explored how local pyramidal neurons would activate when the two types of interneurons were inhibited. The subicular interneurons were transfected with a floxed hM4Di construct by injecting AAV-DIO-hM4Di-mCherry in PV-Cre and SST-Cre mice. The designer receptor hM4Di is a synthetic variant of the human muscarinic receptor, inhibiting neural activity through the Gi pathway with the presence of CNO (Urban and Roth, 2015). The immunochemistry results showed that subicular PV- and SST-INs were accurately and effectively transfected with virus (71.31 ± 7.39% of PV-INs expressed mCherry, and 89.09 ± 2.66% of mCherry+ neurons were PV; 76.01% of SST-INs expressed mCherry, and 88.71 ± 3.69% of mCherry+ neurons were SST; Fig. 5a–c). We first performed in vitro electrophysical recording and found that under the bath of CNO, both PV- and SST-INs increased their firing rheobase by approximately 100%, indicating chemogenetic manipulation had similar effects on inhibiting subicular PV- and SST-INs (Fig. 5d,e). For in vivo experiments, we perfused mice in 2 h after CNO injection for immunochemical analysis of c-Fos (a neural activity marker) and its colocalization with biomarkers of different neuron types. Neurons activated by per inhibited SST-IN were more than per PV-IN (c-Fos+ cell per virus, 2.91 ± 0.42 vs 1.24 ± 0.21, p = 0.0030; Fig. 5f,g). Moreover, the proportion of pyramidal neurons activated by inhibition of SST-INs was higher compared with PV-INs (c-Fos+/CaMKII+, 81.37 ± 1.80 vs 56.96 ± 5.21%, p = 0.0006), suggesting that SST-INs provided more effective inhibition onto pyramidal neurons within the subiculum. Interestingly, inhibition of PV-INs activated much more GABAergic neuronal types than that of SST-INs (c-Fos+/PV+, 4.68 ± 1.32 vs 17.01 ± 4.08%, p = 0.0122; c-Fos+/SST+, 1.70 ± 0.84 vs 15.17 ± 1.96%, p < 0.0001; Fig. 5f,g), suggesting that PV-INs might also innervate local PV- and SST-INs to form disinhibitory circuits.
Subicular SST+ interneurons dominate in modifying the excitability of local pyramidal neurons. a, Schematic of virus injection into the subiculum of PV- or SST-Cre mice (PV-hM4Di or SST-hM4Di mice) and the experiment timeline. b, c, Representative images (b) and quantification (c) of hM4Di-mCherry cells that coexpressed with PV or SST. Scale bar, 50 μm. N = 5, 5 mice. d, e, Representative traces (d) and quantification (e) of the rheobase of a PV- or SST-IN that expressed hM4Di-mCherry before and after a bath of CNO. Eight PV-INs from three mice and six SST-INs from two mice. f, Left, Representative images of subicular PV and SST-INs expressing hM4Di-mCherry (red) and merged with immunofluorescence of c-Fos (magenta). Right, Immunofluorescence of c-Fos merged with that of CaMKII, PV, or SST (green). Scale bar, 25 μm, g, Top, Quantification of the ratio of c-Fos+ cells to hM4Di-mCherry–labeled neurons in PV-hM4Di and SST-hM4Di mice after CNO injection. Bottom, Quantification of the proportions of c-Fos+ cells coexpressing CaMKII, PV, or SST in PV-hM4Di or SST-hM4Di mice. N = 8, 8 mice. Unpaired t test, t(14) = 3.578, **p = 0.0030; t(14) = 4.428, ***p = 0.0006; t(14) = 2.876, *p = 0.0122; t(14) = 6.339, ****p < 0.0001. h, Schematic of mixed virus injection into the subiculum of PV-Cre or SST-Cre mice (PV-ChrimsonCaMKIIα-GCaMP6s or SST-ChrimsonCaMKIIα-GCaMP6s mice) and capture of calcium signal. i, Experiment setup. j, k, Top, Typical case of mean fluorescence and corresponding heatmap aligned to the end of each cycle of photostimuli in PV-ChrimsonCaMKIIα-GCaMP6s (j) and SST-ChrimsonCaMKIIα-GCaMP6s (k) mice. Bottom, Single-trial (left) and quantification (right) of the change of pyramidal neurons’ mean fluorescence before and after photostimuli in PV-ChrimsonCaMKIIα-GCaMP6s (j) and SST-ChrimsonCaMKIIα-GCaMP6s (k) mice. Wilcoxon signed-rank test, PV, W = 5.0, p = 0.8457; W = −13.0, p = 0.2188, OFF versus ON. SST, W = −55.0, **p = 0.0020; W = −38.0, **p = 0.0078, OFF versus ON. Data are presented as mean ± SEM or separated dots.
In fact, in addition to the biased brain-wide inputs into subicular PV- and SST-INs, we also investigated whether subicular local inputs into PV- and SST-INs would show diversity by immunostaining RV-labeled cells with CaMKII, PV, and SST. We found that both PV- and SST-INs were mostly innervated from subicular CaMKII-positive glutamatergic neurons, while local glutamatergic projections to PV-INs were denser (Fig. 6a,b). Meanwhile, SST-INs also received intensive projections from local PV-INs, while PV-INs only received very sparse inputs from both local PV- and SST-INs (Fig. 6c–f). These results demonstrate that although both subicular PV- and SST-INs are innervated by local glutamatergic neurons directly, SST-INs might receive more local inhibitory signals than PV-INs, suggesting their complex local microcircuit within the subiculum.
Subicular PV+ and SST+ interneurons receive biased local inputs in normal condition. a, b, Representative images (a) and quantifications (b) of retrogradely labeled neurons of PV and SST-INs merging with CaMKII in the subiculum. White arrows indicate merged cells. Scale bar, 100 μm. Unpaired t test, t(11) = 4.175, **p = 0.0015. c, d, Representative images (c) and quantifications (d) of retrogradely labeled neurons of PV and SST-INs merging with PV in the subiculum. White arrows indicate merged cells. Yellow arrows indicate starter cells. Scale bar, 100 μm. Unpaired t test, t(11) = 4.853, ***p = 0.0005. e, f, Representative images (e) and quantifications (f) of retrogradely labeled neurons of PV and SST-INs merging with SST in the subiculum. White arrows indicate merged cells. Yellow arrows indicate starter cells. Scale bar, 100 μm. Unpaired t test, t(11) = 3.077, *p = 0.0105. N = 6, 7 mice. Data are presented as mean ± SEM.
Furthermore, to perform a real-time assessment of the effect of PV- and SST-INs on local pyramidal neurons, we monitored the calcium activity of subicular pyramidal neurons when simultaneously activating PV- or SST-INs by optogenetics. To accomplish that, PV- or SST-INs were transfected with a floxed ChrimsonR construct AAV-DIO-ChrimsonR-tdTomato, an engineered channelrhodopsin selectively responding to 589 nm red laser (Klapoetke et al., 2014), and pyramidal neurons were meanwhile transfected with a CaMKIIα promoter–dependent GCaMP6s AAV-CaMKIIα-GCaMP6s into the subiculum of PV-Cre and SST-Cre mice (Fig. 5h,i). Then, 589 and 488 nm lasers were given at the same time for optical activation and GCaMP fluorescence recording. When PV-INs were activated, the GCaMP fluoresce of pyramidal neurons showed diversity in change—not appearing to consistent increase or reduction; while when SST-INs were activated, the GCaMP fluoresce of pyramidal neurons decreased rapidly and consistently (Fig. 5j,k). The above data show both subicular PV- and SST-INs have a direct functional connection with local pyramidal neurons, but SST-INs modify the excitability of local pyramidal neurons more sufficiently.
Activation of subicular SST-, rather than PV-INs, alleviates epileptic seizures
Finally, we further tested their roles in seizure modulation by assessing the effects of chemogenetically activating PV- or SST-INs in KA-induced epilepsy models. First, we aimed to testify whether activation of subicular PV- or SST-INs through Cre-dependent hM3Dq and CNO could alleviate KA-induced acute seizure (Fig. 7a). The immunochemistry results showed that subicular PV- and SST-INs successfully expressed mCherry fluorescence (75.05 ± 6.34% of PV-INs expressed mCherry, and 94.07 ± 2.31% of mCherry+ neurons were PV; 68.42% of SST-INs expressed mCherry, and 87.31 ± 2.23% of mCherry+ neurons were SST; Fig. 7b). Activation of subicular PV-INs only had a tendency to retard the latency to status epilepticus (SE; PV-mCherry 29.99 ± 3.61 vs PV-hM3Dq 41.58 ± 4.13 min, p = 0.0541, Fig. 7c,e). However, activation of subicular SST-INs significantly protected against the latency to SE (SST-mCherry 38.69 ± 6.21 vs SST-hM3Dq 71.17 ± 8.13 min, p = 0.0051) and also had a tendency to retard the latency to sGS (SST-mCherry 14.88 ± 2.82 vs SST-hM3Dq 31.97 ± 7.97 min, p = 0.0501, Fig. 7d,f). These results indicate that subicular SST-INs have better antiseizure effects than PV-INs in the acute period of the KA model.
Activation of subicular SST+, rather than PV+, interneurons alleviates epileptic seizures. a, Schematic of acute seizure modulation induced by KA injection and virus injection into the subiculum of PV- or SST-Cre mice. b, Representative images and quantification of hM3Dq (red) expressing and colabeled with PV and SST (green) in the subiculum of PV-hM3Dq and SST-hM3Dq mice, respectively. Scale bar, 200 and 50 μm for enlarged images. N = 5, 5 mice. c, d, Representative CA3 EEGs of PV-Cre (c) and SST-Cre (d) mice with or without hM3Dq in the subiculum treated by CNO 1.5 h in the acute KA model. The first red row indicates end of KA injection. The second indicates onset of secondary generalized seizure (sGS). The third indicates onset of status epilepticus (SE). e, f, Effects of chemogenetic activation of subicular PV- (e) and SST-INs (f) on latency to sGS and SE. Unpaired t test, PV, to GS, t(18) = 1.154, p = 0.2638; to SE, t(18) = 2.061, p = 0.0541. SST, to GS, t(17) = 2.109, p = 0.0501; to SE, t(17) = 3.212, **p = 0.0051. g, Schematic of modulation in chronic spontaneous seizure model. i, Representative CA3 EEG of an entire sGS during chronic period and the corresponding power spectrum. j, k, Effects of chemogenetic activation of subicular PV- (j) and SST-INs (k) on the number and duration of sGSs. Fridman with post hoc Dunn's multiple-comparisons test, PV, number, Fridman statistic = 0.2222; duration, Fridman statistic = 0.3415. SST, number, Fridman statistic = 11.31, CNO versus Pre, **p = 0.0027, CNO versus Post, p = 0.090; duration, Fridman statistic = 10.3, CNO versus Pre, **p = 0.0063 CNO versus Post, p = 0.0975. Data are presented as mean ± SEM or separated dots.
Then, we studied how subicular PV- and SST-INs regulated KA-induced chronic spontaneous seizure (Fig. 7g,i). During this period, chemogenetically activating subicular PV-INs did not show any significant effects on the number and duration of seizures (seizure number/day, Pre 2.25 ± 1.25 vs CNO 1.76 ± 0.84; seizure duration/day, Pre 99.71 ± 59.85 vs CNO 81.82 ± 36.51 s; Fig. 7j). By comparison, activating subicular SST-INs via CNO not only significantly decreased the number of behavioral seizures but also shortened total duration and decreased seizure burden (seizure number/day, Pre 2.00 ± 0.42 vs CNO 0.24 ± 0.19, p = 0.0027; seizure duration/day, Pre 87.09 ± 16.54 vs CNO 6.88 ± 4.99 s, p = 0.0063; Fig. 7k). These results reveal different efficiency of PV- and SST-INs in seizure controlling, suggesting SST-INs’ fast response during seizure and more effective feedforward inhibition on pyramidal neurons could directly serve as an antiseizure role.
Discussion
Accumulating research has identified the critical role of subicular GABAergic interneurons in TLE (Palma et al., 2006; Huberfeld et al., 2007; Knopp et al., 2008; Buchin et al., 2016; Anstötz et al., 2021). However, how PV- and SST-INs, two major subtypes, are involved in seizure control is barely understood. Here, assisted by calcium fiber photometry, retrograde tracing, and chemogenetics in animal epilepsy models, we explore the difference between subicular PV- and SST-INs in circuits and functions in TLE: (1) Compared to PV-INs, SST-INs respond earlier to generalized seizure. (2) Physiologically, PV-INs receive more input projections from inside the subiculum, while SST-INs from outside. These biased circuit patterns also undergo distinct changes in epileptic condition. (3) For output, SST-INs provide a more efficient feedforward inhibition onto local glutamatergic neurons than PV-INs. In addition, PV-INs also form a local disinhibitory circuit with other interneurons. (4) The above biased circuit differences between PV- and SST-INs influence their overall network “excitation–inhibition” effects, resulting in distinct regulation of epileptic seizures and engendering consistent antiseizure effects of SST-INs. These findings provide a better understanding for circuit reorganization centered around subicular cell-type–specific interneurons.
Both subicular PV- and SST-INs show activating status during sGS, indicating compensatory protective mechanisms preventing seizure generalization would happen at the beginning period. Compared to SST-INs, PV-INs have a short delay to rise in calcium signaling in both KA and kindling-induced sGS. The delayed activation at the beginning of seizure suggests a moderate inhibition, perhaps from long-projecting signaling of distant regions, like the medial septum (Freund and Antal, 1988; Sans-Dublanc et al., 2020). Meanwhile, we find the most abundant afferents of subicular PV- and SST-INs are both from the CA1, confirming the subiculum as the major gateway of the hippocampal output (Aggleton and Christiansen, 2015). Second to the CA1 is the subiculum itself, suggesting tangled microcircuits exist in the subiculum. However, the difference is that PV-INs receive more local input and innervate relatively less pyramidal neurons, suggesting PV-INs are more likely to stand in a feedback-circuit pattern, while SST-INs connect with long-range upstream regions and provide quite a few projections to local pyramidal neurons, suggesting their feedforward role in the subicular network. Although we cannot exclude the fact that CSI does not necessarily reflect the strength of connections, since overlapping connections (one presynaptic cell drives many postsynaptic cells) are not calculated under our definition, at least it accounts for that SST-INs, which receive more input from the hippocampal CA3 (outside of the subiculum), where seizures mostly generate in the present study, activate more quickly.
Like the sculptor, interneurons chisel away part of neural excitability, maintaining the balance of “excitation–inhibition.” In chronic epilepsy, remarkable neural reorganization centered around subicular PV- and SST-INs leads to the imbalance of “excitation–inhibition,” which is in accordance with the prevailing view that epileptic seizures accompany with weakening of inhibitory systems (Houser and Esclapez, 1996; Kobayashi and Buckmaster, 2003; Maglóczky and Freund, 2005). For PV-INs, the local input density reduces after seizure, while RV distribution regarding the MS/DB, ANT, and CA3 contrarily increase, suggesting feedback inhibition is weakened in chronic epileptic condition; for SST-INs, the CSI reduces with nearly every upstream region, indicating that SST-INs lose driven factor to produce compensatory protective mechanisms during sGS. Such reorganization could be either the cause or the consequence of the generalization of epileptic neural network. For the cause, when a subtle balance of “excitation–inhibition” changes because of circuit reconnecting, newly formed inputs could be other powerful pathways to transmit signal and promote epileptic discharge synchronization throughout the brain (Lillis et al., 2015; Wenzel et al., 2019). As a consequence, when some neurons die because of excitatory toxin, other surviving ones may be recruited in new circuits to compensate for inhibition. Furthermore, we found that activating PV-INs exhibits various changes in pyramidal neurons’ calcium signal, while activating SST-INs consistently reduces the calcium signal of pyramidal neurons, which at least indicates that SST-INs are inhibitory hub within the subicular network, and their impairment onto excitatory neurons might be the cause of sGS during epileptogenesis.
Circuit reorganization of subicular PV-INs could also partly account for the difference in behavioral seizure modulation results. In acute seizures induced by KA, activating PV-INs has a tendency to delay the latency to SE; while in chronic period, activating PV-INs does not aggravate or retard seizure severity. This could be relevant to numerous PV-IN loss (Fig. 3a,b), or PV-INs are less innervated from local inputs (Fig. 2h) in chronic epileptic condition. In contrast, subicular SST-INs are continuously antiseizure in both acute and chronic periods, although they also undergo complicated circuit reorganization. Compared to our previous study that activating SST-INs is effective in abating hippocampal seizure, we find that it also alleviates seizure that initiates from the amygdala, indicating a broad-spectrum antiseizure role of subicular SST-INs, independent of seizure onset area. We should note that seizure rates in the KA model can be quite variable over time in the same animal (Bertram and Cornett, 1994; Williams et al., 2009). Some mice might display seizure clusters that occur at multiday intervals, which can confound analyses of the effects of chemogenetic manipulation. This potential variation would be reduced by applying longer seizure monitoring periods in the future. Although SST-INs do not show significant neuronal loss, they form fewer connections with local pyramidal neurons in chronic epileptic condition, further suggesting the subiculum would lose natural driven factor (SST-INs) to produce compensatory protective mechanisms during sGS. Interestingly, activating total subicular GABAergic neurons is proepileptic in the kindling model (Wang et al., 2017), suggesting subicular PV-INs in the current intra-amygdaloid KA-induced seizures are not enough to give sufficient depolarizing signaling, or there could be other subtypes of interneurons conversely modulating seizure. For instance, VIP+ interneurons (VIP-INs) activation in the ventral subiculum has recently been reported to increase the frequency of seizures induced by KA (Rahimi et al., 2023). The role of subicular VIP+ or other interneurons in chronic seizure control requires further studies. Additionally, we cannot exclude other mechanisms that contribute to different effects of subicular PV- and SST-INs on seizure modulation. For example, the subcellular locations of synapses which PV- and SST-INs make with pyramidal neurons are different: PV-INs often synapse on soma, while SST-INs synapse on the distal dendrites. Therefore, inhibition of subicular SST-INs could be more effective in a “winner-take-all” way (Miles et al., 1996; Khirug et al., 2008). Moreover, SST-INs receive intensive projections from local PV-INs, while PV-INs only receive sparse inputs from both local PV- and SST-INs, indicating a more complicated microcircuit, that is, the strong mutual interaction between the INs might also be the reason. Our results at least indicate that biased circuits between PV- and SST-INs might influence their modulation of neural excitability, thereby engendering consistent antiseizure effects of SST-INs in TLE.
Functional neural circuits involving cell-type–specific interneurons in various brain regions, such as the prefrontal cortex, and the CA1 have been studied with great interest by virtue of recent advances (English et al., 2017; Zhang et al., 2022). Our manuscript presents a thorough description of subicular PV- and SST-INs’ circuit organization in normal and chronic epileptic condition and suggests that the SST-IN subtype is a key driven factor in providing compensatory protection during sGS. Selective activation of SST-INs is sufficient to consistently play antiseizure roles in TLE. These results not only deconstruct the complex microcircuits in the subiculum for a better understanding of the role of interneurons in the subiculum in physiological conditions (e.g., spatial memory, cognition, and emotion; Cembrowski et al., 2018; Sun et al., 2019; Yan et al., 2022) but also provide a new possibility of developing a precise antiseizure medication strategy for TLE by targeting the subicular circuits centered around subtypes of interneurons.
Data Availability
The authors declare that all data used in this study are available from the corresponding authors upon reasonable request.
Footnotes
This work was supported by grants from the National Natural Science Foundation of China (Grant No. U23A20533 to Yi Wang; Grant No. 82330116 to Z.C.; Grant No. 82304460 to F.F.) and Natural Science Foundation of Zhejiang Province (Grant No. D24H310001 to Yi Wang; Grant No. LD22H310003 to Z.C.).
↵*F.F., X.W., and X.F. contributed equally to this work.
The authors declare no competing financial interests.
- Correspondence should be addressed to Yi Wang at wang-yi{at}zju.edu.cn.