Deletion of KCNQ2/3 potassium channels from PV+ interneurons leads to homeostatic potentiation of excitatory transmission

KCNQ2/3 channels, ubiquitously expressed neuronal potassium channels, have emerged as indispensable regulators of brain network activity. Despite their critical role in brain homeostasis, the mechanisms by which KCNQ2/3 dysfunction lead to hypersychrony are not fully known. Here, we show that deletion of KCNQ2/3 channels changed PV+ interneurons’, but not SST+ interneurons’, firing properties. We also find that deletion of either KCNQ2/3 or KCNQ2 channels from PV+ interneurons led to elevated homeostatic potentiation of fast excitatory transmission in pyramidal neurons. Pvalb-Kcnq2 null-mice showed increased seizure susceptibility, suggesting that decreases in interneuron KCNQ2/3 activity remodels excitatory networks, providing a new function for these channels.


Introduction
Genetic studies have established that potassium channel dysfunction is responsible for multiple pediatric epilepsy disorders (Brenner and Wilcox, 2012;Geisheker et al., 2017;Niday and Tzingounis, 2018;Oyrer et al., 2018). KCNQ2 and KCNQ3 channels, in particular, have emerged as fundamental regulators of normal brain activity (Brenner and Wilcox, 2012;Geisheker et al., 2017). Pathogenic variants of these channels are strongly associated with early-onset neonatal epileptic encephalopathy and developmental disorders (Millichap et al., 2017;Mulkey et al., 2017;Oyrer et al., 2018). However, the network mechanisms by which KCNQ2/3 variants lead to hyperexcitability are not fully understood.
KCNQ2/3 channels are expressed in both pyramidal neurons and interneurons (Cooper et al., 2001). In pyramidal neurons, KCNQ2/3 channels primarily control spike frequency adaptation, a quiescence period neurons enter following a brief train of activity (Peters et al., 2005;Soh et al., 2014). However, our knowledge of KCNQ2/3 function in interneurons is limited. This gap in knowledge is partly because KCNQ2/3 function is most easily observed in neurons that undergo pronounced spike frequency adaptation, a characteristic not traditionally found in interneurons (Pelkey et al., 2017).
It is currently unknown whether loss of KCNQ2/3 function in interneurons would have effects on overall network activity, and if it does, whether it would lead to a dampening or an increase in excitability. Previous work suggests that loss of KCNQ2/3 function in interneurons would likely elevate their excitability (Lawrence et al., 2006;Pelkey et al., 2017), which in turn would increase GABAergic inhibition and decrease network excitability. However, at early times in development GABA is depolarizing due to shifted chloride equilibrium potential (Le Magueresse and Monyer, 2013). Previous work found that neonatal administration of bumetanide, which changes GABA receptor activity from depolarizating to hyperpolarizing, prevents seizures in mice with a global loss of KCNQ2 channels (Marguet et al., 2015). This may be evidence that loss of KCNQ2/3 function at this time increases network excitability through changes in interneuron GABA signaling, but the authors did not directly examine synaptic activity. Further, it is entirely unknown what effects KCNQ2/3 dysfunction in interneurons has on network excitability in juvenile and mature mice.
In this work, we developed mice lacking KCNQ2 and KCNQ3 specifically in interneurons to address their possible role at the cellular and network activity. We found that KCNQ2/3 channels regulate interneuron properties in a cell type-specific manner. Deletion of KCNQ2/3 channels primarily impacts the firing properties of PV + interneurons, but not SST + interneurons. We also find that interneuronal loss of KCNQ2/3 or KCNQ2 channels increases excitatory transmission between pyramidal neurons, perhaps as homeostatic compensation for the increased GABAergic signaling we observe.

Results and discussion
To investigate the role of KCNQ2/3 channels in interneurons, we developed mouse lines that lack KCNQ2/3 channels specifically in parvalbumin-positive (PV + ) and somatostatin-positive (SST + ) interneurons, cell types known to express KCNQ2/3 channels (Cooper et al., 2001;Lawrence et al., 2006). We crossed Kcnq2 or Kcnq3 floxed mice (Kcnq2 f/f and Kcnq3 f/f ) to Nkx2-1 cre mice (Xu et al., 2008). The Nkx2-1 cre driver is expressed starting early in development (~embryonic day 10.5) in SST + and PV + interneurons, allowing us to study the impact of KCNQ2/3 channel ablation in young and juvenile neurons. In order to identify the Cre-expressing PV + and SST + interneurons, we crossed these mice to a reporter line (Ai9) that expresses the fluorescent protein tdTomato in cells in which Cre recombinase has been active. We designated these mice as IN:Kcnq2/3 null (Nkx2-1 cre ;Kcnq2 f/f ; Kcnq3 f/f ;Ai9).
Are the synaptic changes due to loss of KCNQ2/3 from PV + , SST + , or both types of interneurons? To address this question, we tested whether ablation of KCNQ2/3 channels alters the excitability of PV + and SST + cells in mouse CA1. We distinguished these neurons based on their intrinsic excitability and firing properties ( Figure 2). We found that ablation of Kcnq2/3 from PV + -like interneurons led to an increase in the number of action potentials following suprathreshold depolarizing current pulses ( Figure 2a). In contrast, ablation of Kcnq2/3 from SST + -like interneurons did not change their firing properties ( Figure 2a). We obtained similar results for L2/3 interneurons, suggesting that KCNQ2/3 function is critical for PV + interneuron properties across multiple forebrain regions ( Figure 2a). To confirm our finding that KCNQ2/3 alters PV + interneuron excitability, we tested the effect of ablating Kcnq2 and Kcnq3 from PV + interneurons by developing Pvalb cre ;Kcnq2 f/f /Kcnq3 f/f ; Ai9 (designated as PV:Kcnq2/3 null) mice. Indeed, ablation of Kcnq2 and Kcnq3 increased the firing rate of PV + neurons in the CA1 region ( Figure 2b). The increased excitability in PV + interneurons might have been due to their increased input resistance of Kcnq2/3 null neurons ( Figure 2b; mean R IN : control 166 ± 16 MW, n = 15; PV:Kcnq2/3 null 257 ± 33 MW, n = 14; df = 27 t = À2.54 p=0.017 unpaired Student's t-test). However, as application of the pan-KCNQ blocker XE-991 did not change the input resistance in control PV + interneurons (Figure 2-figure supplement 3; R IN : control 169 ± 20 MW,+XE-991 143 ± 20 MW, n = 6; t = 1.355 df = 5 p=0.233), it's unclear whether the change in input resistance in Kcnq2/3 null neurons is due to direct loss of KCNQ2/3 activity or an indirect effect. Next, we repeated our experiments in SST cre ;Kcnq2 f/f /Kcnq3 f/f ;Ai9 neurons ( Figure 2b; refer to as SST:Kcnq2/3 null). Consistent with our earlier finding, ablation of Kcnq2/3 from SST + interneurons did not change their firing properties or input resistance (mean R IN : control 360.28 ± 34 MW, n = 6; SST:Kcnq2/3 null 327 ± 64 MW, n = 8; df = 12 t=À0.42 p=0.68 unpaired Student's t-test).
Our data suggest that ablation of KCNQ2/3 channel activity from PV + interneurons results in elevated interneuron excitability, which in turn could lead to remodeling of fast excitatory synaptic activity. As KCNQ2 pathogenic variants are much more frequent than KCNQ3 variants (Brenner and Wilcox, 2012), we also examined whether ablation of Kcnq2 alone could lead to similar effects. Indeed, we found that loss of KCNQ2 channels from PV + interneurons increased their intrinsic excitability ( Figure 3a) and the frequency of sEPSCs in pyramidal neurons (Figure 3c; Pvalb-Cre;Kcnq2 +/ + ;Ai9 1.165 ± 0.12 Hz, n = 13; Pvalb-Cre;Kcnq2 f/f ;Ai9: 2.108 ± 0.22 Hz, n = 19; p=0.0098 Mann-Whitney test). The observed changes in the sEPSC frequency were not due to changes in the intrinsic excitability of pyramidal neurons (Figure 3b). To determine whether ablation of Kcnq2 from PV +expressing interneurons affected the excitatory drive in vivo, we administered picrotoxin, a GABA A receptor antagonist. We implanted mice with bilateral frontal and parietal electrodes and validated generalized epileptiform activity coinciding with Racine scale Stage five seizures induced by intraperitoneal injection of picrotoxin (10 mg/kg) ( Figure 4a). Thereafter, we video-recorded the latency to seizure onset without implantation. All of the mice lacking Kcnq2, including Pvalb;Kcnq2 heterozygous mice, reached Stage five seizure activity within 30 min, with a significantly reduced latency to seizure onset when compared to Pvalb;Kcnq2 +/+ mice ( Figure 4b). These data confirm that loss of KCNQ2 channel activity from interneurons can lead to excitatory network hyperexcitability in vivo.
In summary, our ex vivo and in vivo work reveals that changes in interneuron excitability induced by loss of KCNQ2/3 channel activity in PV + interneurons could change the course of excitatory network development. Multiple studies have shown that early in development GABA-mediated transmission partly drives excitatory synapse maturation and formation in pyramidal neurons. This is because early in development pyramidal neurons have a depolarized GABA equilibrium potential (Le Magueresse and Monyer, 2013; Wang and Kriegstein, 2011), allowing GABA to provide a much-needed depolarizing signal to relieve magnesium block from NMDA receptors and in turn promote AMPA receptor unsilencing and synaptic input maturation. Indeed, recent work showed that pharmacologically blocking potassium channels with 4-AP to drive GABA release on newborn neurons led to robust formation of excitatory synaptic inputs, allowing for the integration of newborn cells to excitatory synaptic networks (Chancey et al., 2013). We thus speculate a similar mechanism might in play in our mice. Future work is needed to directly test this hypothesis. Importantly, our work suggests a unique mechanism compared to earlier work showing that loss of potassium channel activity critical for interneuron firing behavior leads only to depolarization block (Lau et al., 2000) and hyperexcitability through disinhibition. Therefore, our work raises the possibility that potassium channel dysfunction that elevates GABAergic transmission could also lead to remodeling of excitatory transmission in the cortex. Such changes might contribute to the severe neurodevelopmental effects seen in patients with potassium channel epileptic encephalopathy.

Slice electrophysiology
Mice (P8-P35) were anesthetized with isoflurane and rapidly decapitated. The brain was quickly removed and placed in ice-cold sucrose based cutting solution consisting of the following: 25 mM NaHCO 3 , 200 mM sucrose, 10 mM glucose, 2.5 mM KCl, 1.3 mM NaH 2 PO 4 , 0.5 mM CaCl 2 , and 7 mM MgCl 2 . Coronal or Transverse hippocampus slices including surrounding structures were cut at 300 mm using a vibrating microtome (Microm HM 650V-Thermo Fisher Scientific; or Leica VT1200S). Slices were then transferred in artificial cerebrospinal fluid (ACSF) consisting of the following (in mM): 125 NaCl, 26 NaHCO 3 , 2.5 KCl, 1 NaH 2 PO 4 , 1.3 MgCl 2 , 2.5 CaCl 2 , and 12 glucose and equilibrated at 35˚C for 30 min, and then maintained at room temperature for at least 1 hr before electrophysiological recordings. Cutting solution and ACSF were saturated with 95% O 2 and 5% CO 2 . All experiments were performed at room temperature. Whole-cell recordings were obtained using borosilicate glass electrodes having resistances of 2 to 4 MW. For current clamp whole cell recordings we used an internal consisted of the following (in mM): 130 potassium methylsulfate (or potassium gluconate), 10 KCl, 5 Tris-phosphocreatine, 10 HEPES, 4 NaCl, 4 Mg 2 ATP, and 0.4 Na 4 GTP. The pH was adjusted to 7.2 to 7.3 with KOH. CNQX (4 mM), D-AP5 (10 mM), and picrotoxin (100 mM) (Abcam) were added in all current clamp experiments (Figure 2 and Figure 3a) to block AMPAmediated, NMDA-mediated, and GABA-mediated synaptic transmission, respectively. For synaptic activity recordings in voltage-clamp configuration we used an internal solution consisting of the following (in mM): 135 Cs-MeSO 3 , 8 NaCl, 10 HEPES, 0.3 EGTA, 5 QX-314, 0.4 Na 4 GTP, and 4 Mg 2 ATP. The pH was adjusted to 7.2 with CsOH. To measure mEPSC and mIPSC, 1 mM TTX (Abcam) was present after recording spontaneous synaptic activity. We recorded s/mEPSCs in voltage-clamp mode at À70 mV, which is the reversal potential for GABA A -mediated chloride currents when using a low internal chloride concentration. For s/mIPSCs were recorded at 0 mV the reversal of AMPA-mediated currents. For mEPSCs, we performed all recordings in the presence of 1 mM TTX to block action potentials. Recordings were performed using a Multiclamp 700B amplifier (Molecular Devices), low pass-filtered at 2 kHz, sampled at 10 kHz, and analyzed offline using either Prism 7 (Graphpad), Clampfit 10 (PCLAMP; Molecular Devices), Mini analysis program (Synaptosoft), or Origin eight pro (OriginLab).

In vivo recordings
We implanted mice between the ages P30-45. For all mice we gave 48 hr post-operative recovery time before chemoconvulsive testing. Mice were implanted as previously described (Maheshwari et al., 2017). Briefly, mice were anesthetized with isoflurane (2-4% in O 2 ) anesthesia and surgically implanted with silver wire electrodes (0.005 inch diameter) inserted into the epidural space over the somatosensory cortex (1 mm posterior and 3 mm lateral to bregma) and frontal cortex (1 mm anterior and 1 mm lateral to bregma) bilaterally through cranial burr holes and attached to a microminiature connector cemented to the skull. The reference electrode was placed over the right cerebellum (1 mm posterior and 1 mm lateral to lambda) and the ground electrode was placed over the left cerebellum (1 mm posterior and 1 mm lateral to lambda).

Fluorescence In Situ Hybridization (FISH)
To prepare brain sections for FISH, mice were anesthetized with isoflurane and rapidly decapitated. Mouse brains (P19-22) were flash frozen embedded in OCT compound and coronal slices cryosectioned at a thickness of 14 mm and mounted on SuperFrost Plus Gold slides (Fisher Scientific). Sections were fixed in 4% paraformaldehyde for 15 min at 4˚C, dehydrated in 50, 70, and 100% ethanol, and air-dried at room temperature. Fluorescent RNAscope in situ hybridization (ISH) was performed using an RNAscope Fresh Frozen Multiplex Fluorescent kit according to the manufacturer's protocol to perform target probe hybridization and signal amplification (Advanced Cell Diagnostics). Probes