The hyperpolarization‐activated cyclic nucleotide‐gated 4 channel as a potential anti‐seizure drug target

Abstract

Background and Purpose

Hyperpolarization‐activated cyclic nucleotide‐gated (HCN) channels are encoded by four genes (HCN1–4) with distinct biophysical properties and functions within the brain. HCN4 channels activate slowly at robust hyperpolarizing potentials, making them more likely to be engaged during hyperexcitable neuronal network activity seen during seizures. HCN4 channels are also highly expressed in thalamic nuclei, a brain region implicated in seizure generalization. Here, we assessed the utility of targeting the HCN4 channel as an anti‐seizure strategy using pharmacological and genetic approaches.

Experimental Approach

The impact of reducing HCN4 channel function on seizure susceptibility and neuronal network excitability was studied using an HCN4 channel preferring blocker (EC18) and a conditional brain specific HCN4 knockout mouse model.

Key Results

EC18 (10 mg·kg⁻¹) and brain‐specific HCN4 channel knockout reduced seizure susceptibility and proconvulsant‐mediated cortical spiking recorded using electrocorticography, with minimal effects on other mouse behaviours. EC18 (10 μM) decreased neuronal network bursting in mouse cortical cultures. Importantly, EC18 was not protective against proconvulsant‐mediated seizures in the conditional HCN4 channel knockout mouse and did not reduce bursting behaviour in AAV‐HCN4 shRNA infected mouse cortical cultures.

Conclusions and Implications

These data suggest the HCN4 channel as a potential pharmacologically relevant target for anti‐seizure drugs that is likely to have a low side‐effect liability in the CNS.

Abbreviations

cHCN4KO

conditional HCN4 knockout

ECoG

electrocorticography

eYFP

enhanced yellow fluorescent protein

FFT

fast Fourier transform

HCN

hyperpolarization‐activated cyclic nucleotide‐gated

Ih

hyperpolarization‐activated current

IR‐DIC

IR differential interference contrast

MEA

multi‐electrode array

MOI

multiplicity of infection

QC

quality control

PTZ

pentylenetetrazole

SLICK‐H

((SLICK‐ H is the H variant of the SLICK series of transgenic mice as defined in (Heimer‐McGinn & Young, 2011)

What is already known

• HCN channels generate pacemaker activity critical for the hyper‐synchronous neuronal network activity underlying seizures.

• HCN4 channels are highly expressed in thalamic nuclei, a region linked to seizure generalization.

What this study adds

• Reducing HCN4 channel function decreases seizure susceptibility and neuronal network excitability in adult mice.

• HCN4 channel block has no major impact on other measured behaviours.

Clinical significance

• HCN4 channels are potential pharmacologically relevant targets for anti‐seizure drugs with minimal adverse effects.

1. INTRODUCTION

Hyperpolarization‐activated cyclic nucleotide‐gated (HCN) channels are widely expressed in the brain (Biel, Wahl‐Schott, Michalakis, & Zong, 2009; Pape, 1996; Robinson & Siegelbaum, 2003). Four separate genes (HCN1–4) encode HCN channels in the CNS (Moosmang, Biel, Hofmann, & Ludwig, 1999; Santoro et al., 2000). These channels carry a non‐selective cation conductance, I_h, that controls fundamental neuronal functions including defining resting membrane potential, modulating the integration of dendritic synaptic input and the control of synaptic transmission and controlling neuronal firing properties (Bender & Baram, 2008; Biel et al., 2009; Nava, Dalle et al., 2014). There is a strong link between changes in HCN channel function and hyperexcitability occurring in epilepsy (Benarroch, 2013; DiFrancesco & DiFrancesco, 2015; Reid, Phillips, & Petrou, 2012) including transcriptional changes in HCN channels in both acquired and genetic rodent models of epilepsy (e.g. Powell et al., 2008; Strauss et al., 2004). There is also an association between neuronal hyperexcitability and genetic change in HCN1, HCN2 and HCN4 (Becker et al., 2017; Bonzanni et al., 2018; Campostrini et al., 2018; Dibbens et al., 2010; Marini et al., 2018; Nava et al., 2014). This relationship is complex, with evidence for both increased and decreased I_h associating with heightened neuronal excitability. For example, HCN1 and HCN2 knockout mice both have increased seizure susceptibility (Ludwig et al., 2003; Santoro et al., 2010), while a persistent increase in I_h is observed following induced febrile seizures (Chen et al., 2001). At a genetic level, both gain‐ and loss‐of‐function HCN1 variants, as measured in heterologous expression assays, associate with epilepsy (Marini et al., 2018; Nava et al., 2014). Pharmacologically, the broad‐spectrum I_h blockers ivabradine, caesium and ZD7288 all have anticonvulsant activity in seizure models (Kitayama et al., 2003; Luszczki, Prystupa, Andres‐Mach, Marzeda, & Florek‐Luszczki, 2013; Matsuda, Saito, Yamamoto, Niitsu, & Kogure, 2008). Together, these data suggest that HCN channels, as a class, are important modulators of neuronal excitability in epilepsy.

Here, we explore the role of HCN4 channels in defining neuronal network excitability in the context of epilepsy. HCN4 channels are highly expressed in the thalamus (Oyrer et al., 2019; Santoro et al., 2000), a brain region implicated in seizure generalization (Blumenfeld, 2005). HCN4 channels activate more slowly and at the most hyperpolarized potentials when compared to HCN1 and HCN2 channels (Sartiani, Mannaioni, Masi, Novella Romanelli, & Cerbai, 2017). This means that they are only likely to be engaged during network activity that causes robust and extended hyperpolarizing episodes. Physiologically, HCN4 channels are implicated in modulating thalamic and cortical oscillations (Zobeiri et al., 2019). The thalamic expression and biophysical properties of HCN4 channels position them as potential modulators of seizure activity. Consistent with this, the widely used anti‐epileptic drug, gabapentin, reduced the function of HCN4 channels through a left‐shift in the voltage of activation (Tae et al., 2017). Increased HCN4 channel function may also be part of the mechanism underlying hyperexcitability in some forms of epilepsy with evidence including increased HCN4 mRNA levels in the pilocarpine rodent model of temporal lobe epilepsy that correlates with increased I_h in dentate granule cells (Surges et al., 2012). Other evidence includes a switch from HCN2 to HCN4 channel expression in thalamocortical neurons in a cortical stroke model characterized by seizure development (Paz et al., 2013). Here, we show through pharmacological and molecular strategies that reducing HCN4 channel function decreases both network excitability and seizure susceptibility in adult mice, emphasizing these channels as potential targets for anti‐seizure drugs.

2. METHODS

2.1. Validity of animal species or model selection

Wild‐type P21 C57BL/6J mice (IMSR Cat# JAX:000664, RRID:IMSR_JAX:000664) are routinely used for testing anti‐seizure drugs in proconvulsant assays (Loscher, 2011). To explore the role of HCN4 channels in defining neuronal network excitability, we developed an in vivo molecular approach. Global HCN4 channel knockout is embryonic lethal in mice due to its critical role in cardiogenesis (Stieber et al., 2003). To overcome this limitation, Herrmann, Stieber, Stockl, Hofmann, and Ludwig (2007) engineered an HCN4‐floxed mouse that is viable and provides the opportunity to knock out HCN4 in a spatial‐ and/or temporal‐specific manner. In this study, we developed a double transgenic mouse model that crosses the HCN4‐floxed mouse with the Tg (Thy1‐Cre/ERT2,‐eYFP)HGfng/PyngJ mouse line (also known as SLICK‐H (SLICK‐ H is the H variant of the SLICK series of transgenic mice as defined in (Heimer‐McGinn & Young, 2011), IMSR Cat# JAX:012708, RRID:IMSR_JAX:012708) (Heimer‐McGinn & Young, 2011) to create an inducible brain‐specific knockout.

2.2. Ethical statement

All experiments were performed in accordance with the Prevention of Cruelty to Animals Act, 1986, under the guidelines of the National Health and Medical Research Council (NHMRC) Code of Practice for the Care and Use of Animals for Experimental Purposes in Australia and were approved by the Animal Ethics Committees at the Florey Institute of Neuroscience and Mental Health and the Monash Institute of Pharmaceutical Sciences. Animal studies are reported in compliance with the ARRIVE guidelines (Kilkenny, Browne, Cuthill, Emerson, & Altman, 2010) and with the recommendations made by the British Journal of Pharmacology.

2.3. Animal welfare

Anaesthesia and analgesia were used where appropriate, as described below. Mice were monitored daily and were rapidly killed by cervical dislocation, an ANZCCART‐approved method.

2.4. Housing and husbandry

Animals were housed in standard 15 × 30 × 12 cm cages, maintained under 12‐h dark and light cycles, and had access to dry pellet food and tap water ad libitum. Post‐weaning P21 male C57BL/6J mice (IMSR Cat# JAX:000664, RRID:IMSR_JAX:000664) were ordered from the Animal Resources Centre (WA, Australia) or the Monash Animal Research Platform (VIC, Australia). To generate the conditional brain HCN4 knockout (cHCN4KO) mouse model, homozygous HCN4‐floxed mice (Herrmann et al., 2007) were crossed with heterozygous HCN4‐floxed × hemizygous Tg (Thy1‐Cre/ERT2,‐eYFP)HGfng/PyngJ mice (also known as SLICK‐H mice; Heimer‐McGinn & Young, 2011, Jackson Laboratory [Stock No: 012708], IMSR Cat# JAX:012708, RRID:IMSR_JAX:012708). Mice were genotyped by TransnetYX (TN, USA) using PCR of DNA sourced from tails. Cre‐mediated recombination was induced based on a standard Jackson laboratory protocol (Heffner, 2011). Tamoxifen (75 mg·kg⁻¹, i.p.) was injected for five consecutive days (P42–P46) and mice left for at least 10 days before experimentation commenced at P56 (Figure 1a). Tamoxifen‐treated SLICK‐H × homozygous HCN4‐floxed mice constitute our conditional HCN4 knockout (cHCN4KO) model with tamoxifen‐treated SLICK‐H mice forming our negative control group. For behavioural experiments involving the SLICK‐H and cHCN4KO mouse models, the mix of male to female mice was approximately 50:50.

FIGURE 1.

(A) Cartoon to illustrate: the cre construct in the SLICK‐H mouse line: the position of the lox p sites and the deletion of exon 4 in the HCN4 floxed mouse: and the experimental timeline. (b) Brain section showing typical Thy1 promoter eYFP brain expression in SLICK‐H mice (repeated in n = 3 mice, Scale bar = 600 μm). (c) Body weight of SLICK‐H and cHCN4KO mice 10 days after tamoxifen treatment. Data shown are individual values and mean ± SEM (SLICK‐H n = 23, cHCN4KO n = 25). (d) HCN mRNA expression in the cHCN4KO brain (n = 5) relative to the mean of the SLICK‐H control (n = 5). GAPDH was used as a reference gene for normalization and each run was analysed separately. HCN4 mRNA levels were reduced in the cHCN4KO mouse relative to SLICK‐H controls. The expression of HCN1 and HCN2 mRNA in thalamus and cortex tissue was unchanged. (e) Typical examples of protein expression on Western blots of HCN isoforms (100 mg of total protein per lane) in cHCN4KO (−) and SLICK‐H (+) mouse tissue. Depending on the brain region and the HCN isoform being analysed, the exposure time ranged between 5 and 400 s. HCN1 in thalamus was at background levels. (f) HCN protein expression in the cHCN4KO brain relative to the mean of the SLICK‐H control. Blots were analysed separately by ImageJ. (HCN1 protein expression levels of thalamus‐enriched tissue were too low to analyse). (Thalamus n = 6, Cortex n = 7). Exploratory Western blots of other isoforms showed no differences compared to SLICK‐H controls (Figure S1). F test analysis showed no significant differences in variance. No reduction in HCN4 expression was seen in one mouse. As this was outside confidence levels, it was excluded from analysis. Data expressed as individual values and mean ± SEM. ^* P < 0.05 versus SLICK‐H mean via Wilcoxon matched pairs signed‐rank test

2.5. Molecular characterization of the cHCN4KO mouse model

2.5.1. Imaging

Adult P56 SLICK‐H mice (n = 3) were anaesthetized via inhalation of 1%–3% isoflurane and killed by cervical dislocation. Brains were removed from the skulls and coronal slices of 150 μm thickness were prepared using a Vibratome. Slices were mounted in ProLong Diamond antifade reagent (Thermo Fisher, cat. no. P36962) and enclosed with no. 1.5H coverslips. Confocal image stacks were obtained using a Zeiss LSM780 system with a 20X (0.8 NA) air objective and viewed in ImageJ (ImageJ, RRID: SCR_003070; Schneider, Rasband, & Eliceiri, 2012).

2.5.2. Quantitative RT‐PCR

Brains were isolated from anaesthetized mice and dissected to give hippocampus, cortex and thalamus‐enriched tissue. Tissues were snap frozen in liquid nitrogen then stored at −80°C prior to being homogenized using the TRIzol method, purified using the RNeasy mini kit (Qiagen, Hilden, Germany) and assayed for quality and quantity as previously described (Phillips, Kim, Vargas, Petrou, & Reid, 2014). cDNA was prepared using random primers and the transcriptor high‐fidelity cDNA synthesis kit (Roche Holding AG, Basel, Switzerland). Quantification was performed using a Rotor‐Gene 6000 real‐time PCR machine (Corbett Research, NSW, Australia). Wells were loaded with cDNA (20 ng for hippocampus and 12.5 ng for cortex and thalamus‐enriched tissue), TaqMan Gene expression master mix (Applied Biosystems, CA, USA) and mouse specific HCN probes: HCN1 (Mm00468832_m1), HCN2 (Mm00468538_m1) and HCN4 (Mm01176086_m1). The housekeeping gene GAPDH was used for normalization (Mm99999915_g1). All probes were labelled with a fluorescent reporter dye at the 5′ end and a non‐fluorescent quencher (NFQ) at the 3′ end (Thermo Fisher Scientific, MA, USA). The amplification protocol consisted of two denaturing steps at 60°C (2 min) and 95°C (10 min) and 40 cycles alternating at 95°C (10s) and 60°C (60s). All RT‐qPCR reactions were performed in triplicate to ensure the reliability of single values and data analysed using Rotor‐Gene 6000 series software 1.7 (Rotor‐Gene 6000 series software, RRID:SCR_017552, Corbett Research, NSW, Australia). Data were analysed as previously described (Pfaffl, 2001). Results of HCN mRNA expression levels in the cHCN4KO brain (n = 5 per run) are plotted relative to the mean of the SLICK‐H controls (n = 5 per run). This normalization reduces sources of variation that would be present in both sample groups.

2.5.3. Western blotting

The immuno‐related procedures used comply with the recommendations made by the British Journal of Pharmacology (Alexander et al., 2018) except for loading controls. Region‐specific brain tissue samples obtained as previously described in Section 2.5.2 were homogenized in RIPA buffer (10‐mM Tris pH 8.0, 1% Triton X‐100, 0.1% sodium deoxycholate, 1% SDS, 140‐mM NaCl). The homogenates were then incubated with rotation for 30 min at 4°C, centrifuged for 20 min at 12,000 g and supernatants retained. Sample aliquots were diluted one in 10 for quantitation by the Bradford method (Bio‐Rad, CA, USA). Samples were then diluted to a final protein concentration of 5 μg·μl⁻¹ of protein, 2‐M urea and 1× SDS reducing loading buffer (Laemmli, 1970) and heated for 1 h at 37°C. Equal protein concentrations of control and experimental samples were separated by electrophoresis (7% and 8% poly‐acrylamide SDS‐PAGE) along with pre‐stained Precision Plus Dual Colour Standards (Bio‐Rad, CA, USA), transferred to Cellulose Nitrate, Grade SO45A3330R (Advantec, Tokyo, Japan) and satisfactory transfer of protein by Western blotting monitored by Ponceau S (Sigma Aldrich, cat. no. P3504). Linear standard curves of control protein (25, 50, 75 and 100 μg) were used to determine suitable protein concentrations for quantitative assessment. Filters were blocked in 0.5% skim milk, 1× TBS 1.0% and IGEPAL CA‐630 (Sigma Aldrich, cat. no. I8896), ‘blocker’, for 1 h before overnight incubation at 4°C with primary antibodies: 1:500 mouse anti‐HCN4 (UC Davis/NIH NeuroMab Facility Cat# 73‐150, RRID:AB_10673158), 1:500 rabbit anti‐HCN2 APC‐030 (Alomone Labs Cat# APC‐030, RRID:AB_2313726) or 1:500 rabbit anti‐HCN1 19405 (Abcam Cat# ab19405, RRID:AB_444898). The filters were washed and incubated with secondary antibodies; either goat anti‐mouse polyclonal 32430 (Thermo Fisher Scientific Cat#32430, RRID:AB_1185566) 1:250, or goat anti‐rabbit Poly HRP 32260 (Thermo Fisher Scientific Cat# 32260, RRID:AB_1965959) 1:15,000 for 1 h at 21°C. All antibody dilutions were in ‘blocker’ and diluted antibodies were stored at 4°C and used a maximum of twice within a week of dilution. Western blots were analysed individually as follows: the protein signal was visualized with Clarity Western ECL Substrate (Bio‐Rad, CA, USA) and the signal captured by a Bio‐Rad ChemiDoc™ MP imaging system (Image Lab Software, RRID:SCR_014210) and quantified using ImageJ (ImageJ, RRID: SCR_003070; Schneider et al., 2012). Exposure times varied greatly (5–400 s) depending on the brain region and the HCN isoform being analysed; this was due to the differing expression levels of HCN isoforms in different regions of the brain. For each blot, the data for SLICK‐H control mice (n = 3 or 4) were averaged and individual values for cHCN4KO on that blot compared to this average. In order to use data from multiple blots and reduce unwanted sources of variation, these values were then plotted against their respective controls (each control average set at a value of one). Data were not averaged across multiple blots. In a subset of Western blot experiments, the number of animals was less than five and these have been described as exploratory in Figure S1.

2.6. Seizure and behavioural testing

Prior to all behavioural experiments, mice were acclimatized for 1 h in a dimly lit behavioural room. Time to maximal hindlimb extension seizures was measured and used as the most robust seizure endpoint for all seizure assays (Chiu et al., 2008).

2.6.1. Subcutaneous (s.c.) pentylenetetrazole (PTZ) proconvulsant assay

Mice were injected subcutaneously with 100 mg·kg⁻¹ of PTZ and monitored for a maximum of 40 min for maximal hindlimb extension seizures. PTZ was injected s.c. at either 10 or 30 min after the injection of EC18 (10 mg·kg⁻¹, i.p.) or saline control.

2.6.2. Kainic acid proconvulsant assay

Mice were injected intraperitoneally with 30 mg·kg⁻¹ of kainic acid 10 min after the injection of EC18 (10 mg·kg⁻¹, i.p.) and monitored for a maximum of 60 min. Time to maximal hindlimb extension seizures was measured and defined as the seizure endpoint. Mice that did not have any seizures were excluded from analysis as this is likely due to poor injection of kainic acid (SLICK‐H n = 2, cHCN4KO n = 1).

2.6.3. Thermogenic seizure assay

Mice were placed into an enclosed chamber heated to 42 ± 1°C to induce heat‐mediated clonic–tonic seizures that model febrile seizures (Reid et al., 2013). Thirty minutes prior to being placed in the thermally controlled chamber, mice were either injected with EC18 (10 mg·kg⁻¹, i.p.) or saline control. Mice were culled by cervical dislocation immediately after the first observed seizure, or 20 min after being placed in the chamber if they remained seizure‐free in line with ethics requirements.

2.6.4. Open field exploratory locomotion assay

Mice were individually placed in a square mouse open field 27.3 × 27.3 × 20.3 cm arena (Med Associates Inc., St. Albans, VT) and allowed to move freely for 1 h. Recorded data indicating distance travelled and resting duration were then compiled using locomotion analysis software (MED Associates Activity Monitor software, RRID:SCR_014296, Med Associates Inc., Vermont, USA).

2.6.5. Light/dark transition test

A square mouse open field arena was divided into two equal 27.3 × 13.7 cm light and dark zones using a black acrylic box insert with a small opening. The light zone was illuminated at 750lx. Mice were individually placed inside the dark zone at the start of the test and allowed to freely explore both zones for 10 min. Recorded data indicating the number of entries and duration spent in each zone were compiled using locomotion analysis software (MED Associates Activity Monitor software, RRID:SCR_014296, Med Associates Inc., Vermont, USA).

2.6.6. Elevated plus maze

The elevated plus maze was raised 40 cm from the floor and consisted of two open arms and two enclosed arms with 55‐cm high walls extending from the centre. Mice were individually placed in one of the enclosed arms and allowed to freely roam the maze for 10 min while being recorded by a ceiling‐mounted video camera. Entries into each arm along with their respective durations were measured using tracking software (EthoVision XT, RRID:SCR_000441, Noldus, Wageningen, the Netherlands).

2.6.7. Grip strength

Mouse limb neuromuscular strength was assessed using a grip strength isometric force transducer (Bioseb, Chaville, France) connected to a 10 × 10cm wire grid. Mice were placed on the wire grid and gently pulled backwards by their tails until they released. The peak force applied by the mouse's limbs was recorded for three separate trials to ensure the reliability of each trial. The maximal force applied by each mouse over the three trials was used for statistical analysis.

2.7. Pharmacokinetic studies

2.7.1. Intraperitoneal administration of EC18 to C57BL/6J mice

On the day of experiments, EC18 was freshly dissolved in saline and a 200‐μl solution equivalent to 10 mg·kg⁻¹ was administered into the lower abdominal quadrant of male C57BL/6J mice (IMSR Cat# JAX:000664, RRID:IMSR_JAX:000664). Blood and brain samples were collected at 10, 20, 30 and 60 min following the initial dosing. Concentrations of EC18 in plasma and brain homogenate were measured using LC MS/MS assay. To more accurately determine the brain concentration of EC18, the concentration of EC18 remaining within the brain microvasculature was subtracted from the brain homogenate concentration using the mouse brain microvascular volume of 0.017 ml·g⁻¹ (Nicolazzo et al., 2010).

2.7.2. Preparation of calibration standards and mouse samples

A stock of EC18 (10 mg·ml⁻¹) was first prepared in MilliQ water. For plasma samples, working standard solutions with concentrations of 0.5, 1, 2, 5 and 10 μg·ml⁻¹ were prepared by serial dilution of the stock solution in MilliQ water. The low‐ and high‐quality control (QC) solutions of 0.5 and 10 μg·ml⁻¹ were prepared in the same manner using an independently prepared stock solution. Calibration and QC samples were prepared by spiking 10 μl of the working solutions into 90 μl of blank plasma obtained from C57BL/6J mice (IMSR Cat# JAX:000664, RRID:IMSR_JAX:000664). The mixture was then vortexed for 5 s. Plasma samples were prepared in the same manner, except that 10 μl of MilliQ water was added into 90 μl of plasma sample instead of the working solution. To each mixture, 0.5 ml of ethyl acetate was added, followed by vortex mixing for 20 min at room temperature (RT) and centrifugation for 5 min at 1,000 g. The supernatant was then collected and dried with nitrogen gas. The residue was reconstituted in 100 μl of MilliQ water and analysed by the LC MS/MS method described in the following section. To measure the concentration of EC18 in brain samples, working standard solutions with concentrations of 0.1, 0.5, 1, 5 and 10 μg·ml⁻¹ were prepared by serial dilution of stock solution (10 mg·ml⁻¹) in MilliQ water. The low‐ and high+QC solutions of 0.1 and 10 μg·ml⁻¹ were prepared in the same manner using an independently prepared stock solution. Calibration and QC samples were prepared by spiking 30 μl of the working solutions into 270 μl of blank brain homogenate to achieve concentrations of 30, 150, 300, 600, 1,500 and 3,000 ng·g⁻¹. The mixture was then vortexed for 5 s. Brain homogenate samples were prepared in the same manner, except that 30 μl of MilliQ water was added into 270 μl of brain homogenate sample instead of the working solution. To each mixture, 1.5 ml of ethyl acetate was added, followed by vortex mixing for 20 min at RT and centrifugation for 5 min at 1,000 g. The supernatant was then collected and dried with nitrogen gas. The residue was reconstituted in 100 μl of MilliQ water and analysed by the LC MS/MS method described in the following section. Each of the four QC samples at each concentration was measured and compared to determine the intraday assay precision and accuracy. Precision was expressed as relative SD (%) and accuracy was calculated as the difference between the measured and nominal concentration expressed as percentage (Table S1).

2.7.3. LC MS/MS analysis

An Ascentis® Express C₁₈ column (2.7‐μm particle size, 2.1 × 50mm internal diameter) with a Phenomenex Security Guard™ C₁₈ guard column (2.0 × 4.0 mm) was used for the measurement of EC18. Samples of 10 μl were injected into a Shimadzu HPLC system consisting of two LC‐30AD pumps: an SIL‐30AC autoinjector and a DGU‐20A₅ degasser (Shimadzu, Kyoto, Japan). Mobile phases A and B were 0.1% (v/v) formic acid in MilliQ water and methanol respectively. The gradient profile developed to determine the concentration of EC18 was 0–1 min, 90% A; 1–1.5 min, 75% A; 1.5–2 min, 75% A; 2–2.5 min, 50% A; 2.5–3.5 min, 50% A; and 3.5–4 min, 90% A. A Shimadzu LC MS/MS‐8050 triple quadrupole mass spectrometer (Shimadzu, Kyoto, Japan) was used to perform the MS in the electrospray ionization positive mode by multiple reaction monitoring (m/z 533.10 to 258.25). Nitrogen was used as the nebulizing gas and drying gas and the temperatures of the desolvation line and heat block were set at 250°C and 400°C respectively. The dwell time was set at 100 ms.

2.8. Electrocorticography monitoring

2.8.1. Electrocorticography (ECoG) electrode implantation

Electrocorticography (ECoG) electrode implantation was performed at P26–P30 on male C57BL/6J mice (n = 13) (IMSR Cat# JAX:000664, RRID:IMSR_JAX:000664) and at P52–60 on SLICK‐H (n = 8) and cHCN4KO (n = 10) mice. Mice were weighed then anaesthetized via inhalation of 1%–3% isoflurane and placed on a stereotaxic frame. Three holes, each 1 mm in diameter, were drilled to secure stainless‐steel skull screws that act as epidural ECoG electrodes. Two screws were placed bilaterally over the primary somatosensory (S1) cortex and used as the active channel electrodes (±3.0 mm lateral to the midline, −1.0 mm caudal to bregma). The third screw was placed immediately caudal to the lambdoid suture 0.5 mm lateral from the midline towards the right side of the skull and used as the reference channel electrode. A ground electrode made of silver wire was affixed to the skull immediately caudal to the lambdoid suture 0.5 mm lateral from the midline towards the left side of the skull. The reference, ground and two active channel electrodes were connected to a mouse EEG head mount (8201‐EEG Pinnacle technology Inc., KS, USA) via silver leads (Cat No. 785500, A‐M Systems Inc., WA, USA). Self‐curing acrylic resin (Vertex‐Dental B.V., Soesterberg, the Netherlands) was used to hold the head mount and skull screws in place. Lidocaine hydrochloride (2%, s.c., Ilium, NSW, Australia) was injected subcutaneously prior to surgical incision to provide local anaesthesia to the scalp. Meloxicam (1 mg·kg⁻¹, i.p., Ilium, NSW, Australia) was given as an analgesic and mice were left to recover for 72 h before experimentation.

2.8.2. Electrocorticography (ECoG) recording of low‐dose PTZ‐induced spiking

Mice were independently housed during recordings in a clear 30 × 15 × 15 cm plexiglass container. Prior to recording, the head mount was linked to a mouse preamplifier (8406‐SE, Pinnacle Technology Inc.) and connected to a 4‐channel data conditioning/acquisition system (8200‐K1‐SE3, Pinnacle Technology Inc.). Data were acquired at 2,000 Hz and filtered (40 Hz low‐pass and 0.5 Hz high‐pass) using Sirenia Acquisition 1.7.5 software (Sirenia Acquisition, RRID:SCR_016183, Pinnacle Technology Inc.).

For C57BL/6J mice (IMSR Cat# JAX:000664, RRID:IMSR_JAX:000664), each recording consisted of a 30‐min baseline measurement followed by an injection of EC18 (10 mg·kg⁻¹, i.p.). Ten minutes post administration of EC18, male C57BL/6J mice were injected with PTZ (60 mg·kg⁻¹, s.c.) to induce spiking and recorded for an additional 30 min. For SLICK‐H and cHCN4KO mouse recordings, PTZ was injected immediately after the 30‐min baseline recording to induce spiking and recorded for an additional 30 min. Mice were then culled by cervical dislocation within 30 min of administration of PTZ.

2.8.3. Electrocorticography (ECoG) signal analysis

Recordings were converted to the European data format then imported into LabChart Reader software 8.1 (LabChart Reader software, RRID:SCR_017551, ADInstruments, NSW, Australia). Five‐minute power spectrograms were produced by running a fast Fourier transform (FFT) algorithm with a cosine‐bell data window. The window size was 1,024 data points with an overlap of 87.5%. Results are plotted as heatmaps with power expressed as μV².

2.8.4. Quantification of PTZ‐induced spikes

Spikes were detected and quantified using automated simple threshold cyclic measurements. Spike detection thresholds ranging between 200 and 300 μV were set manually per mouse based on relative baseline and spiking amplitudes. The rate of PTZ‐induced spikes detected per minute is calculated for a maximum of the first 10 min post PTZ administration or to the time that a seizure occurred.

2.9. Primary neuronal culture

2.9.1. Culture preparation

Cells were dissociated from cortices of P0–P2 C57BL/6J mice (IMSR Cat# JAX:000664, RRID:IMSR_JAX:000664) (n = 8–10), pooled to give an evenly mixed population and cultured on polyethyleneimine/laminin‐coated 24‐well multielectrode array plates (Multichannel Systems, Reutlingen, Germany) or for cDNA analysis in coated 6‐well plates, as previously described (Gazina et al., 2018; McSweeney et al., 2016). Cells were plated at a density of 2 × 10⁶ per well in the six‐well plates and 375,000 per well in the 24‐well multi‐electrode array (MEA) plate in culture medium as described (Gazina et al., 2018). At Day 3 in vitro, cytosine arabinoside (5 μM, Merck KGaA, Darmstadt, Germany) was added to the medium to inhibit glial proliferation and removed 48 h later. Medium was replaced every 2 days subsequently.

2.9.2. shRNA design and virus production

HCN4 shRNA vectors were designed by Vector Biolabs (Malvern, PA, USA). U6 promotor‐driven shRNA screening was completed on five clones. shRNA sequence CCGGCTCCAAACTGCCGTCTAATTTCTCGAGAAATTAGACGGCAGTTTGGAGTTTTTG (MISSION® TRC shRNA TRCN0000252673) was shown to reduce mouse HCN4 transcript by ~83% and was used for viral production. AAV(1/2)‐GFP‐U6‐mHCN4‐shRNA (AAV‐HCN4 shRNA, 3.4 × 10¹³ GC·ml⁻¹) was engineered and included the shRNA sequence driven under a U6 promotor and GFP under a CMV promotor. A standard AAV(1/2)‐GFP‐U6‐scrmb‐shRNA (AAV‐control, 7.1 × 10¹³ GC·ml⁻¹) was used as a negative control.

2.9.3. Viral infection and validation

The cortical cultures were left to mature for 12 days before addition of either AAV‐HCN4 shRNA or AAV‐control at a multiplicity of infection (MOI) of 15,000 per plated cell. Cells were monitored for GFP fluorescence from 6 days post transfection. Expression was visible in both cultures at 7 days post transfection.

2.9.4. Immunocytochemistry

Cells were cultured as previously described in Section 2.9.1 and grown for 10 days post viral infection in an 8‐well Nunc™ Lab‐Tek™ Chamber Slide System (Thermo Fisher, cat. no. 177402PK) for immunocytochemistry. Cells were fixed in 4% paraformaldehyde in 0.1‐M phosphate buffer pH 7.4 (PB) for 10 min at RT, permeabilized in 0.3% Triton X‐100 in PB for 6 min and then blocked in 1% FBS/22.52 mg·ml⁻¹ of glycine in PB containing 0.1% Tween20 for 30 min at RT. Cells were washed three times with PB between each procedure. Blocked cells were incubated at 4°C for 16 h with primary antibodies: 1:500 Guinea Pig anti‐NeuN (Sigma Aldrich, Millipore Cat# ABN90P, RRID:AB_2341095) and 1:500 Chicken anti‐GFP (Abcam Cat# ab13910, RRID:AB_300798). Cells were washed in PB and incubated with secondary antibodies: 1:1,000 CF®594 Donkey anti‐Guinea Pig (Biotium Cat# 20170–1, RRID:AB_10854394) and 1:10,000 Goat anti‐Chicken Alexa Fluor® 488 (Molecular Probes Cat# A‐11039, RRID:AB_142924) for 1 h at RT. After washing with PB, the wells were removed from the chamber base and the slide mounted in ProLong Diamond antifade reagent (Thermo Fisher, cat. no. P36962) and enclosed with no. 1.5H coverslips. Confocal images were obtained using a Zeiss LSM780 system with a 20X (0.8 NA) air objective and viewed in ImageJ (ImageJ, RRID:SCR_003070; Schneider et al., 2012).

2.9.5. Analysis of viral knock‐down of HCN expression

Cells were cultured for 14 days post transfection before harvesting. Three independent experiments were conducted, each involving pooled cells from eight to 10 neonates. Cells were harvested by scraping from the six‐well plates and pelleted by centrifugation. Total RNA was prepared by the TRIzol method, purified using the RNeasy mini kit (Qiagen, Hilden, Germany) and assayed for quality and quantity as previously described (Phillips et al., 2014). cDNA was prepared using random primers and the high‐fidelity cDNA Synthesis Kit (Roche Holding AG, Basel, Switzerland). qPCR was carried out on 10‐μg cDNA samples, in triplicate to ensure the reliability of single values, using hydrolysis probes HCN1 (Mm00468832_m1), HCN2 (Mm00468538_m1), HCN4 (Mm01176086_m1) and GAPDH (Mm99999915_g1). Results were analysed as described previously (Phillips et al., 2014).

2.9.6. Microelectrode Array (MEA) data acquisition

Ten days post viral transfection, two 24‐well MEA plates, each containing 12 wells of AAV‐HCN4 shRNA or AAV‐control transfected cortical cells, were analysed on a Multiwell‐MEA headstage (Multichannel Systems, Reutlingen, Germany). The plates were placed in a temperature and CO₂ controlled enclosed recording system and allowed to equilibrate for 5 min prior to baseline data collection for 6 min at a rate of 20,000 Hz. Data acquisition was carried out using the Multiwell Screen software (Multichannel Systems) and signals were filtered as previously described (Gazina et al., 2018). The plates were then removed from the headstage and 250 μl of medium (50% of the total medium) was replaced with 250 μl of pre‐warmed medium containing 5 μl of 1‐mM EC18 or water (vehicle). The plates were returned to the incubator for 2 h prior to re‐analysis as per the method described above.

2.9.7. MEA feature extraction and analysis

Voltage signals were high‐pass filtered at 300 Hz and spikes detected with custom MATLAB scripts that used a criterion of peak amplitude greater than six times the SD of noise. Noise levels were calculated in 20s blocks. Bursts in single channels (single‐channel bursts) and network bursts were detected using an adaptive algorithm based on firing rates as described previously (Mendis, Morrisroe, Petrou, & Halgamuge, 2016). Bursts on single channels were defined as rapid successions of three or more spikes while network bursts were defined as events in which more than 20% of single‐channel bursts overlapped in time. For EC18 experiments, means ± SEM (n = 11–12) are presented as the percentage change relative to baseline, to reduce unwanted variation between wells as previously described (Mendis et al., 2019).

2.10. Materials

Tamoxifen (Sigma Aldrich, cat. no. T5648) was freshly dissolved in corn oil (Sigma Aldrich, cat. no. C8267) by shaking at 60 rpm overnight at 37°C. Pentylenetetrazole (Sigma Aldrich, cat. no. P6500) and kainic acid monohydrate (Sigma Aldrich, cat. no. K0250) were dissolved in normal saline to prepare an injectable solution at concentrations of 200 and 3 mg·ml⁻¹ respectively. The HCN4 channel preferring blocker EC18 (Prof Romanelli, Florence, Italy) was dissolved in MilliQ water to make a 10 mg·ml⁻¹ stock solution and further diluted in normal saline to 1 mg·ml⁻¹ for injection.

2.11. Group size

We set the group size as the number of independent values in each experiment. For MEA experiments, this was the number of independent wells used. The number of animals and independent wells used in each experiment was predetermined based on analyses of similar published works (Chiu et al., 2008; Mendis et al., 2016; Reid et al., 2013). Statistical analysis was undertaken only for experiments where each group size was at least five (n ≥ 5).

2.12. Randomization

For experiments investigating the impact of brain‐specific HCN4 knockout on PTZ‐induced seizure susceptibility equal numbers of the SLICK‐H and cHCN4KO mice were chosen based on genotype. The boxes were then deidentified and assigned a random number. Mice were then selected for injection randomly from a maximum of three mice in any given box. Mice numbers were decoded by an independent researcher after experimentation. For other experiments, no formal method was used to generate a randomisation sequence for animals used. For EC18 experiments, male C57BL/6J mice were pulled from the combined cohort sequentially with alternate mice receiving either drug or vehicle control.

2.13. Blinding

Different researchers were involved while conducting seizure and behavioural experiments with the operators being blinded to the treatment.

2.14. Data and analysis

For all in vivo experiments involving EC18, only male mice were used to minimize possible variation between sexes. All other experiments included both male and female mice to minimize the number of litters used in line with ethical considerations. Statistical significance of Kaplan–Meier curves was determined using a log‐rank (Mantel–Cox) test. Analyses where comparisons were relative to wild type or control, including quantitative RT‐PCR, Western blot and MEA data used the Wilcoxon signed‐rank test. The Shapiro–Wilk test was used to establish if data had a normal distribution prior to other statistical analysis. F tests were used routinely to establish if differences in variance existed between groups. If F tests showed a significant difference in variance between groups, a nonparametric two‐tailed Mann–Whitney U‐test was used. A two‐factor ANOVA test was used to determine sex differences in behavioural assays. All other statistical comparisons were based on Student's two‐tailed unpaired t test. GraphPad Prism 7 was used as the statistical analysis software (GraphPad Prism, RRID:SCR_002798, CA, USA). Unless stated otherwise, all data are presented as mean and error bars indicate the SEM. Statistical significance was set at P < 0.05. The data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology (Curtis et al., 2018).

2.15. Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Harding et al., 2018) and are permanently archived in the Concise Guide to PHARMACOLOGY 2019/20 (Alexander et al., 2019).

3. RESULTS

3.1. Validation of HCN4 channel knockout in an adult mouse model

We generated a double transgenic mouse model by crossing the HCN4‐floxed mouse with the Tg (Thy1‐Cre/ERT2,‐eYFP)HGfng/PyngJ mouse line, also known as SLICK‐H (Heimer‐McGinn & Young, 2011) (Figure 1a). This conditional line has a tamoxifen‐inducible Cre‐mediated recombination system driven by the Thy1 promoter (Heimer‐McGinn & Young, 2011). We have chosen this mouse model as the Thy1 promoter expresses Cre recombinase widely in the CNS, but expression is absent in cardiac tissue. The inducible nature of this mouse line allows us to test the impact of HCN4 knockout without developmental confounds. eYFP is driven independently as a marker of cells in which recombination should occur in the SLICK‐H mouse (Figure 1a,b). In the HCN4 floxed mouse, exon 4 of HCN4 is flanked by loxP sites and deleted using the Cre/loxP system (Figure 1a). Tamoxifen (75 mg·kg⁻¹, i.p.) was injected for five consecutive days (P42–P46) and mice left for at least 10 days before experimentation commenced (Figure 1a). Tamoxifen‐treated hemizygous SLICK‐H × homozygous HCN4‐floxed mice constitute our conditional HCN4 knockout (cHCN4KO) model with tamoxifen‐treated hemizygous SLICK‐H mice forming our negative control group.

The conditional knockout of brain HCN4 channels had no significant effect on the weight of the mice (Figure 1c). We used quantitative RT‐qPCR and Western blot analysis to explore the extent of reduction in HCN4 channel expression in thalamus, cortex and hippocampus tissue of cHCN4KO mice. HCN4 mRNA levels were reduced significantly in all tissues of the cHCN4KO mouse compared to the SLICK‐H controls (Figures 1d and S2). The expression of HCN1 and HCN2 mRNA in cortex and thalamus‐enriched tissue was unchanged (Figure 1d). Western blot analysis confirmed a significant reduction in HCN4 channel protein in cortex (n = 7) and thalamus‐enriched tissue (n = 6) of cHCN4KO mice (Figure 1e,f). A similar reduction of HCN4 protein expression was seen in an exploratory Western blot on hippocampal tissue (Figure S1). Exploratory Western blots of HCN1 and HCN2 isoforms showed similar levels in cHCN4KO and SLICK‐H in hippocampal tissues (Figure S1). SLICK‐H HCN1 protein levels in thalamus‐enriched tissue were too low for ImageJ density scans, likely due to the low expression of this isoform in this brain region (Figure 1f) (Santoro et al., 2000). Typical uncropped Western blots are shown in Figures S3–S7.

3.2. Neuronal excitability and seizure susceptibility is reduced in the conditional HCN4 knockout (cHCN4KO) mouse model

We next investigated the seizure susceptibility of the cHCN4KO model using two standard proconvulsant tests: susceptibility to subcutaneous pentylenetetrazole and response to kainic acid. cHCN4KO mice were significantly less susceptible to PTZ (100 mg·kg⁻¹, s.c.)with an increase in latency to hindlimb extension relative to the SLICK‐H control (Figure 2a, Table S2). Also, cHCN4KO mice were significantly less susceptible to kainic acid (30 mg·kg⁻¹, i.p.) with an increase in latency to hindlimb extension relative to the SLICK‐H control (Figure 2b, Table S2).

FIGURE 2.

Brain‐specific HCN4 knockout reduces seizure susceptibility and ECoG activity. (a) pentylenetetrazole (PTZ; 100 mg·kg⁻¹s.c) increased hindlimb extension seizure latency. P < 0.05 versus SLICK‐H via log‐rank (Mantel–Cox) test (SLICK‐H n = 7, conditional HCN4 knockout (cHCN4KO) n = 7). (b) Kainic acid (30 mg·kg⁻¹) increased hindlimb extension seizure latency. P < 0.05 versus SLICK‐H via log‐rank (Mantel–Cox) test. (SLICK‐H n = 6, cHCN4KO n = 8). Mice that did not have any seizures were excluded from analysis (SLICK‐H n = 2, cHCN4KO n = 1). (c) Electrocorticography (ECoG) recording showing the effect of brain‐specific HCN4 knockout on low‐dose pentylenetetrazole‐induced spiking over a period of 60s. (d) ECoG power spectra showing the effect of brain‐specific HCN4 knockout after the administration of low‐dose PTZ (60 mg·kg⁻¹) over a 5‐min period. (e) Rate of PTZ‐induced spiking over a 10‐min period after low‐dose PTZ administration. F test analysis showed significant differences in variance (SLICK‐H n = 8, cHCN4KO n = 10). Data expressed as individual values and mean ± SEM. ^* P < 0.05 versus SLICK‐H via two‐tailed Mann–Whitney U‐test

To directly test the impact of HCN4 knockout on brain excitability, ECoG recordings were made following the injection of low‐dose PTZ (60 mg·kg⁻¹, s.c.) (Figure 2c). Sustained spiking on ECoG was evident within minutes of PTZ injection in both the cHCN4KO and SLICK‐H control mice (Figure 2c,d). However, cHCN4KO mice displayed significantly less spiking than SLICK‐H control mice consistent with a reduction in pro‐convulsant‐induced excitability (Figure 2c–e). These data strongly support the premise that HCN4 channels are important arbiters of neuronal network excitability and modulate seizure susceptibility.

3.3. Subtle behavioural phenotype in the cHCN4KO mouse model

The cHCN4KO mice performed at close to SLICK‐H control mice levels in a set of behavioural tests. cHCN4KO mice showed a reduction in distance travelled and an increase in resting duration early in the locomotion test when compared to the SLICK‐H control mice (Figure 3a,c). This normalized with time and there was no significant difference in the overall distance travelled or resting duration (Figure 3b,d). The grip strength of cHCN4KO mice was significantly greater than the SLICK‐H mice (Figure 3e), but there was no difference on the elevated plus maze (Figure 3f,g) or in performance on the rotarod (Figure 3h). cHCN4KO mice did show a significant increase in the number of light zone entries in the light/dark transition test, although the duration spent in the light zone was not different relative to the SLICK‐H control mice (Figure 3i,j). Additional parameters extracted from the behavioural tests are presented in Table S3. These data suggest that a reduction in neuronal HCN4 channel expression can protect mice from proconvulsant seizures with no major impact on other measured behaviours.

FIGURE 3.

Behavioural characterization of the SLICK‐H and conditional HCN4 knockout (cHCN4KO) mouse models. (a,b) Distance travelled (m) in the open field locomotor cell over a 60‐min period (SLICK‐H n = 22, cHCN4KO n = 25). (c,d) Resting duration (s) in the open field locomotor cell over a 60‐min period (SLICK‐H n = 22, cHCN4KO n = 25). (e) Maximum force (kg·10⁻³) measured during the grip strength test (SLICK‐H n = 23, cHCN4KO n = 25). (f,g) Elevated plus maze (EPM) open arm duration (f) and entries (g) relative to the closed arm (SLICK‐H n = 18, cHCN4KO n = 19). (h) Maximum latency to fall (s) measured during the rotarod performance test (SLICK‐H n = 23, cHCN4KO n = 25). (i,j) Light zone duration (s) (i) and entries (j) during the light/dark transition test (SLICK‐H n = 19, cHCN4KO n = 22). F test analysis showed no significant differences in variance. Data expressed as individual values and mean ± SEM. ^* P < 0.05 versus SLICK‐H via Student's two‐tailed unpaired t‐test. A two‐factor ANOVA test found that there were no sex differences in any behavioural measures

3.4. The HCN4 preferring compound EC18 reduces seizure susceptibility

EC18, a structural derivative of the bradycardic drug zatebradine, blocks recombinant HCN4 channels with approximately sixfold increased sensitivity over HCN1 and HCN2 channels (Del Lungo et al., 2012; Romanelli et al., 2016, 2019). We first established the pharmacokinetic profile of EC18 in the C57BL/6J mouse strain. An LC MS/MS was developed, allowing the measurement of EC18 levels in plasma and brain samples. There was a rapid increase in EC18 plasma levels following an injection of EC18 (10 mg·kg⁻¹, i.p.), peaking at 10 min and reducing to low levels within 1 h (Figure 4a). Brain homogenate levels were lower and peaked at 20 min before rapidly declining (Figure 4a). These data provided a time profile on which to design our testing of EC18 in proconvulsant seizure tests. Wild‐type male P40 mice were injected with EC18 (10 mg·kg⁻¹, i.p.) and after 10 min, their seizure susceptibility was tested using the PTZ (s.c.) proconvulsant assay (Figure 4b). Mice injected with EC18 showed significantly longer latency to hindlimb extension than those injected with saline (Figure 4b, Table S4). A separate cohort of male mice was left for 30 min following injection, a time point where EC18 levels in the blood and brain have dropped significantly (Figure 4c). Mice injected with EC18 at this later time point also showed a significantly longer latency to hindlimb extension (Figure 4c, Table S4). Although it is difficult to make direct comparisons due to cohort‐to‐cohort differences in controls, the seizure protection of EC18 at this later time point appeared to be decreased, consistent with the reduced brain and plasma levels. In another experiment, EC18 (10 mg·kg⁻¹, i.p.) was injected 30 min prior to placing wild‐type P21 male mice in a heated chamber that models febrile seizure syndromes (Reid et al., 2013). EC18 delayed the development of heat‐mediated clonic–tonic seizures (Table S4). Importantly, EC18 (10 mg·kg⁻¹, i.p.) was not effective in reducing seizure susceptibility in the cHCN4KO mouse (Figure 4d, Table S5) but was effective in the SLICK‐H control (Table S5). We also tested the impact of EC18 on the rate of low‐dose PTZ‐induced cortical spiking recorded with ECoG. EC18 (10 mg·kg⁻¹, i.p.) significantly reduced ECoG cortical spiking in wild‐type mice (Figure 4e–g), consistent with a dampening of cortical excitability.

FIGURE 4.

Effect of the HCN4 channel blocker EC18 on seizure susceptibility and ECoG activity. (a) Plasma and brain concentrations of EC18 (10 mg·kg⁻¹) following an i.p. injection. Data expressed as individual values and mean ± SD on a Log₁₀ scale. (b) pentylenetetrazol (PTZ) s.c. induced hindlimb extension seizure latency 10 min post EC18 (10 mg·kg⁻¹, i.p.) injection in C57BL/6J mice. P < 0.05 versus saline via log‐rank (Mantel–Cox) test (saline n = 5, EC18 n = 5). (c) PTZ s.c. induced hindlimb extension seizure latency 30 min post EC18 (10 mg·kg⁻¹, i.p.) injection in C57BL/6J mice. P < 0.05 versus saline via log‐rank (Mantel–Cox) test (saline n = 17, EC18 n = 10). (d) Effect of EC18 (10 mg·kg⁻¹, i.p.) on PTZ s.c. induced hindlimb extension seizure latency of the conditional HCN4 knockout 9cHCN4KO mouse 10 min post injection. P = 0.670 versus cHCN4KO + saline via log‐rank (Mantel–Cox) test (cHCN4KO) + saline n = 8, cHCN4KO + EC18 n = 7). (e) Electrocorticography (ECoG) recording showing the effect of EC18 on low‐dose 60 mg·kg⁻¹ of PTZ‐induced spiking over a period of 60 s. (f) ECoG power spectra showing the effect of EC18 after the administration of low‐dose PTZ (60 mg·kg⁻¹) over a 5‐min period. (g) Rate of low‐dose 60 mg·kg⁻¹ of PTZ‐induced spiking over a 10‐min period after PTZ administration in C57BL/6J mice. F test analysis showed no significant differences in variance (saline n = 7, EC18 n = 6). Data expressed as individual values and mean ± SEM. ^* P < 0.05 versus saline via Student's two‐tailed unpaired t‐test

3.5. The HCN4 channel preferring blocker EC18 has minimal impact on locomotion

Behaviourally, EC18 (10 mg·kg⁻¹, i.p.) injected at 30 min prior to testing caused an initial reduction in locomotion (Figure 5a), but no difference was observed in overall distance travelled (Figure 5b), nor was there a change in resting duration (Figure 5c,d).

FIGURE 5.

Effect of EC18 (10 mg·kg⁻¹, i.p.) on locomotor activity. (a,b) Distance travelled (m) 30 min post injection of EC18 in the open field locomotor cell over a 15‐min period (^* P < 0.05, saline n = 6, EC18 n = 6). (c,d) Resting duration (s) 30 min post injection of EC18 in the open field locomotor cell over a 15‐min period (saline n = 6, EC18 n = 6). F test analysis showed no significant differences in variance. Data expressed as individual values (b,d) and mean ± SEM. ^* P < 0.05 versus saline via Student's two‐tailed unpaired t‐test

3.6. EC18 reduces firing and network burst activity through HCN4 channels in cortical cultures

Having established that EC18 shows seizure protective properties in mice, we next tested its impact on neuronal network behaviour. We developed a culture‐based assay in which HCN4 expression was reduced using a shRNA knock‐down approach. Cortical neuronal cultures were infected using AAV1/2‐HCN4 shRNA (AAV‐HCN4 shRNA) or AAV1/2‐scrambled shRNA that acted as a control (AAV‐control). GFP expression in the cortical cultures was robust for both viruses and colocalized with the neuronal marker, NeuN (Figure 6a–f). A wide field view of these cortical cultures is shown in the supporting information (Figure S8). Quantitative RT‐qPCR confirmed that HCN4 mRNA is reduced by ~70% in AAV‐HCN4 shRNA relative to AAV‐control treated cells (Figure 6g). However, HCN1 mRNA levels were significantly increased, as were HCN2 mRNA levels although to a lesser extent (Figure 6g). Neuronal network activity was measured in 24‐well MEA plates. Raster plots of firing patterns were generated to allow quantification of network activity (Figure S9). Mean firing rate and network burst rate are measures that are known to be reduced by the commonly used anti‐epileptic drugs carbamazepine and sodium valproate (Colombi, Mahajani, Frega, Gasparini, & Chiappalone, 2013). In AAV‐control cultures, EC18 reduced the mean firing rate and reduced network burst rate (Figure 6h,i) consistent with a reduction in network excitability (Colombi et al., 2013). Importantly, in AAV‐HCN4 shRNA infected cultures EC18 had minimal impact on mean firing rate or network burst rate (Figure 6h,i). Furthermore, EC18 also increased the burst duration in AAV‐control infected cultures (1.19 ± 0.04), an effect that was not seen in AAV‐HCN4 shRNA infected cultures (1.00 ± 0.1). It is important to note that the interpretation of how molecular HCN4 channel knock‐down directly modulates neuronal networks in this culture assay is confounded by compensatory changes in HCN1 and HCN2 mRNA levels. Although we cannot exclude that the lack of effect of EC18 is due to a change in the ‘state’ of the network, our data are consistent with the idea that EC18 is acting through the HCN4 channel to modulate neuronal network excitability.

FIGURE 6.

EC18 reduces cultured neuronal network excitability in an HCN4 channel dependent manner. (a–f) GFP and NeuN expression in neurons infected with (a–c) AAV‐control or (d–f) AAV‐HCN4 shRNA (scale bar = 50 μm). (g) HCN mRNA expression of AAV‐HCN4 shRNA infected culture relative to AAV‐control in three independent experiments involving cells pooled from eight to 10 neonatal mice. (h) Relative change in network burst (NB) rate (Hz) and (i) mean firing rate (MFR) of AAV‐control and AAV‐HCN4 shRNA infected cortical pooled cells from eight to 12 neonatal mice after 10‐μM EC18 treatment. Data expressed as mean ± SEM. ^* P < 0.05 versus AAV‐control via Wilcoxon matched pairs signed‐rank test (AAV‐control ± EC18 n = 12 wells, AAV‐HCN4 shRNA ± EC18 n = 11 wells). NB, network burst; MFR, mean firing rate

4. DISCUSSION AND CONCLUSIONS

Here, we provide evidence that HCN4 channels are important arbiters of neuronal network excitability. The conditional brain knockout of HCN4 channels in adult mice reduced seizure susceptibility in pro‐convulsant assays and also reduced low dose PTZ‐induced spiking recorded with ECoG. Only minor behavioural changes were observed in the conditional HCN4 knockout (cHCN4KO) mouse, consistent with those observed in the Nestin‐Cre conditional HCN4 channel knockout mouse (Zobeiri et al., 2019). We also demonstrate that the preferentially selective HCN4 channel blocker, EC18, reduced proconvulsant‐mediated seizure susceptibility and reduced PTZ‐induced ECoG spiking in male wild‐type mice. The anti‐seizure effect of EC18 was occluded in the cHCN4KO mouse. Furthermore, EC18 did not reduce the firing rate or bursting in cortical cultures treated with AAV‐HCN4 shRNA. We used established rodent proconvulsant models that can predict human efficacy and are considered useful tools for screening compounds in early stages of drug discovery (Loscher, 2011; Yuen & Troconiz, 2015). However, testing EC18 on other acquired and genetic models of epilepsy will be important. Importantly, the sex‐specific impact of pharmacological and molecular ‘block’ of HCN4 channels on seizure susceptibility has not been systematically studied here and requires further investigation. Overall, these data support the notion that HCN4 channels are important arbiters of brain excitability and that blockers of these channels may be effective anti‐seizure drugs.

EC18 is a compound that has approximately sixfold increased selectivity for HCN4 over HCN1 and HCN2 isoforms based on heterologous expressed channels (Del Lungo et al., 2012; Romanelli et al., 2016, 2019). The selectivity found in recombinant systems is maintained in tissues expressing different HCN isoforms. EC18 blocked I_h in guinea pig sinoatrial cells and dog cardiac Purkinje fibres that express HCN4 channels, while having minimal impact in dorsal root ganglia neurons that express HCN1 as the major isoform (Del Lungo et al., 2012). Within the CNS, EC18 blocks I_h on thalamic neurons with an EC₅₀ of about 10 μM (Romanelli et al., 2019). EC18 also weakly blocks slow delayed rectifier K⁺ channels at 10 μM (Romanelli et al., 2019). In neuronal cultures, the reduction in both mean firing and network burst rates was occluded by knocking down HCN4 mRNA using an shRNA strategy. This evidence indicates that EC18 is acting primarily through HCN4 channels to reduce network excitability and consequently seizure susceptibility.

HCN4 expression is limited in the CNS with highest levels seen in the thalamus, medial habenula and olfactory bulb (Santoro et al., 2000). The thalamus is well positioned to contribute to seizure generalization. Thalamic relay neurons in particular are thought to be a key requirement for the propagation of generalized seizures (Blumenfeld, 2005). Thalamic relay neurons from a genetic rodent model of generalized seizures have increased I_h, implicating changes in this current as part of the pathogenic process (Cain, Tyson, Jones, & Snutch, 2015). Furthermore, in their genetic rodent model, David et al. (2018) demonstrated that direct broad‐spectrum pharmacological block of HCN channels in the thalamus decreased thalamocortical neuron firing and abolished spontaneous seizures. Molecular knock‐down of HCN2 channels, also highly expressed in the thalamus, similarly reduced seizures. This is in contrast with other studies that indicate that reductions in HCN2 channel activity in thalamocortical neurons induce generalized seizures (Chung et al., 2009; Hammelmann et al., 2019; Ludwig et al., 2003). Interestingly, while specific HCN2 channel knockout in the ventrobasal nuclei was sufficient to induce generalized seizures, HCN4 channel knockout was without effect (Hammelmann et al., 2019). This suggests that HCN2 and HCN4 subunits play distinct roles within thalamic neurons and that targeting each independently may produce different outcomes on excitability. Zobeiri et al. (2019) have recently demonstrated that brain‐specific HCN4 channel knockout reduces thalamocortical neuron rebound burst firing and network bursting behaviour in brain slice preparations. The cellular changes observed in the HCN4 KO mouse are associated with a slowing of thalamic and cortical oscillations in awake mice (Zobeiri et al., 2019) providing a plausible cellular basis for seizure protection. Therefore, our hypothesis is that HCN4 inhibition reduces thalamocortical bursting behaviour and consequently reduces the propagation of seizures although confirming this will require additional study.

HCN4 channel blockers are likely to be well tolerated. HCN4 channels are highly expressed in cardiac tissue, especially in the sinoatrial node, making them important regulators of heart rate (Herrmann, Hofmann, Stieber, & Ludwig, 2012). The broad‐spectrum HCN channel blocker, ivabradine, causes bradycardia and is used clinically for the treatment of angina pectoris and systolic heart failure (Koruth, Lala, Pinney, Reddy, & Dukkipati, 2017). Importantly, ivabradine has a very favourable safety profile with bradycardia and self‐limiting visual disturbances as side effects (Koruth et al., 2017). Therefore, systemic HCN4 channel blockers are expected to have, at worst, a similar favourable peripheral side‐effect profile to ivabradine. It is difficult to draw conclusions about potential CNS side effects as ivabradine does not cross the blood brain barrier well (Savelieva & Camm, 2006, 2008). However, in mice, EC18 had minimal side effects, showing only a reduction in locomotion at early time points. Furthermore, seizure protection in the cHCN4KO mouse also occurred in the context of minimal behavioural changes. These data suggest that both peripheral and central side effects of selective HCN4 channel blockers are likely to be minimal.

Efforts to develop more selective HCN4 blockers are warranted. EC18 has provided a good proof‐of‐concept small molecule for testing the impact of HCN4 channels on neuronal excitability. However, EC18 only has partial selectivity for HCN4 channels (Del Lungo et al., 2012; Romanelli et al., 2016, 2019). Therefore, despite demonstrating that the anti‐bursting and anti‐seizure impact of EC18 is occluded when HCN4 channels are knocked down, we cannot exclude that EC18 is not having these effects through actions on HCN1 and/or HCN2 channels. Moreover, poor brain penetration and rapid elimination from brain and plasma make EC18 a less than ideal therapeutic drug. Small molecules with improved pharmacokinetics and selectivity are needed. Gabapentin causes a shift of the voltage of activation, which is selective for HCN4 over HCN1 and HCN2, arguing that absolute HCN subunit selectivity may be possible (Tae et al., 2017). Development of strategies geared at reducing HCN4 protein through molecular methods may also be worth pursuing. These could include the development of viral tools that deliver shRNA, or antisense oligonucleotides designed to specifically knock down the HCN4 protein.

In conclusion, we provide strong evidence that HCN4 channels modulate neuronal excitability and seizure susceptibility. We also propose that HCN4 channels may be pharmacologically relevant targets for anti‐seizure drugs.

AUTHOR CONTRIBUTIONS

C.A.R. conceived and organized the study. Q.K., A.M.P. and C.A.R. designed the study. A.L., E.C. and M.R. provided materials. Data were acquired by Q.K., A.M.P., L.E.B., E.M., J.O., L.J1., L.J2. and J.A.N. Q.K., A.M.P., E.M., L.J2. and J.A.N. analysed data. Q.K., A.M.P. and C.A.R. interpreted the data. The manuscript was drafted and revised by Q.K., A.M.P. and C.A.R. All authors commented and discussed the manuscript.

CONFLICT OF INTEREST

The authors declare no conflicts of interest.

DECLARATION OF TRANSPARENCY AND SCIENTIFIC RIGOUR

This declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research as stated in the BJP guidelines for Design & Analysis, Immunoblotting and Immunochemistry and Animal Experimentation, and as recommended by funding agencies, publishers and other organizations engaged with supporting research.

Supporting information

Figure S1. Exploratory HCN protein expression levels in the hippocampal region of the cHCN4KO brain relative to the mean of the SLICK‐H control (n = 3).

Figure S2. HCN mRNA expression levels in the hippocampal region of the cHCN4KO brain relative to the mean of the SLICK‐H control (n = 5).

Figure S3. Typical uncropped Western blot image showing HCN4 antibody signals from thalamus–enriched protein samples of SLICK‐H (+) and cHCN4KO (−) mice. Anti‐HCN4 + anti‐mouse antibodies were applied to protein transferred from a 7% SDS‐PAGE gel as described in Methods Section 2.5.3. Box indicates position of the HCN4 protein at ~140 kDa. M represents the molecular marker ladder.

Figure S5. Typical uncropped Western blot image showing HCN2 antibody signals from thalamus–enriched protein samples of SLICK‐H (+) and cHCN4KO (−) mice. Anti‐HCN2 + anti‐rabbit antibodies were applied to protein transferred from a 7% SDS‐PAGE gel as described in Methods Section 2.5.3. Box indicates position of the HCN2 protein at ~98 kDa. M represents the molecular marker ladder.

Figure S6. Typical uncropped Western blot image showing HCN1 antibody signals from thalamus–enriched protein samples of SLICK‐H (+) and cHCN4KO (−) mice. Anti‐HCN1 + anti‐rabbit antibodies were applied to protein transferred from a 7% SDS‐PAGE gel as described in Methods Section 2.5.3. Box indicates the position of the presumed HCN1 protein at ~120 kDa, but no strong HCN1 signal could be detected. M represents the molecular marker ladder.

Figure S7. Typical uncropped 7% SDS‐PAGE western blot image stained with Ponceau S. Figure shows similar protein transfer from thalamus–enriched protein samples of SLICK‐H (+) and cHCN4KO (−) mice at the positions on the blot where proteins of similar molecular weights to those of HCN proteins would be located. M represents the molecular marker ladder.

Figure S8. Immunocytochemistry of cultured cortical primary neurons for MEA analysis. Wide field view of primary neurons after infection with AAV‐HCN4 shRNA (a‐d) or AAV‐control (e‐h). Panels (a) and (e) GFP expression indicating viral expression; (b) and (f) showing neuron specific antibody NeuN; (c) and (g) DAPI staining of cell nuclei of all cells; (d) and (h) merged images showing co‐localization of GFP with NeuN. Methodology and antibodies used as in Methods Section 2.9.4. (Scale bar = 50 μm).

Figure S9. MEA raster plots. (a, b) Raster plots showing 30s of spontaneous firing of AAV‐control before (a) and after (b) 10 μM EC18 treatment. (c, d) Raster plots showing 30s of spontaneous firing of AAV‐HCN4 shRNA before (c) and after (d) 10 μM EC18 treatment.

Data S1 Supporting Information

Table S1: Accuracy and precision of the LC MS/MS assay used for EC18 quantification. All data mean ± SD.

Table S2: Seizure susceptibility of the SLICK‐H and cHCN4KO mouse models.

Table S3: Behavioural assessment of the SLICK‐H and cHCN4KO mouse models. All data mean ± SEM.

Table S4: Effect of EC18 (10 mg kg⁻¹, i.p.) on seizure susceptibility of the C57BL/6J mouse strain.

Table S5: Effect of EC18 (10 mg kg⁻¹, i.p.) on seizure susceptibility of the SLICK‐H and cHCN4KO mouse models.

Kharouf Q, Phillips AM, Bleakley LE, et al. The hyperpolarization‐activated cyclic nucleotide‐gated 4 channel as a potential anti‐seizure drug target. Br J Pharmacol. 2020;177:3712–3729. 10.1111/bph.15088