Targeted Deletion of Kcne2 Impairs HCN Channel Function in Mouse Thalamocortical Circuits

Background Hyperpolarization-activated, cyclic nucleotide-gated (HCN) channels generate the pacemaking current, Ih, which regulates neuronal excitability, burst firing activity, rhythmogenesis, and synaptic integration. The physiological consequence of HCN activation depends on regulation of channel gating by endogenous modulators and stabilization of the channel complex formed by principal and ancillary subunits. KCNE2 is a voltage-gated potassium channel ancillary subunit that also regulates heterologously expressed HCN channels; whether KCNE2 regulates neuronal HCN channel function is unknown. Methodology/Principal Findings We investigated the effects of Kcne2 gene deletion on Ih properties and excitability in ventrobasal (VB) and cortical layer 6 pyramidal neurons using brain slices prepared from Kcne2 +/+ and Kcne2 −/− mice. Kcne2 deletion shifted the voltage-dependence of Ih activation to more hyperpolarized potentials, slowed gating kinetics, and decreased Ih density. Kcne2 deletion was associated with a reduction in whole-brain expression of both HCN1 and HCN2 (but not HCN4), although co-immunoprecipitation from whole-brain lysates failed to detect interaction of KCNE2 with HCN1 or 2. Kcne2 deletion also increased input resistance and temporal summation of subthreshold voltage responses; this increased intrinsic excitability enhanced burst firing in response to 4-aminopyridine. Burst duration increased in corticothalamic, but not thalamocortical, neurons, suggesting enhanced cortical excitatory input to the thalamus; such augmented excitability did not result from changes in glutamate release machinery since miniature EPSC frequency was unaltered in Kcne2 −/− neurons. Conclusions/Significance Loss of KCNE2 leads to downregulation of HCN channel function associated with increased excitability in neurons in the cortico-thalamo-cortical loop. Such findings further our understanding of the normal physiology of brain circuitry critically involved in cognition and have implications for our understanding of various disorders of consciousness.


Introduction
The pacemaker current I h , which is generated by hyperpolarization-activated, cyclic nucleotide-gated (HCN) channels, regulates intrinsic excitability, synaptic integration and rhythmic oscillatory activity [1][2][3]. There are four Hcn genes, each coding for a distinct isoform (HCN1-4) (reviewed by Biel et al. 2009 [3]), which are variably distributed in the brain [4]. Although permeable to both Na + and K + , HCN channels are members of the voltage-gated potassium channel superfamily. HCN channels are not inhibited by the inwardly rectifying K + channel blockers Ba 2+ or tetraethlylammonium, nor the voltage-gated K + channel blocker 4-aminopyrindine, although they are inhibited by several different organic blockers, including ZD7288 [1,3].
KCNE2, originally named MinK-related protein 1 (MiRP1), is a single transmembrane-spanning protein that acts as an ancillary (b) subunit for a number of potassium channel pore-forming a subunits, regulating channel conductance, voltage dependence, gating kinetics, trafficking and pharmacology [5][6][7][8][9] (for review see [10]). Studies using heterologous or over-expression systems have shown that co-expression of KCNE2 with HCN1, 2 or 4 significantly alters the amplitude and kinetics of I h with variable effects on voltage-dependent gating [11][12][13][14]. KCNE2 also increases HCN1, HCN2, and HCN4 single channel conductance, further suggesting a direct interaction [14]. Despite these observations, however, the impact of KCNE2 expression on brain HCN channel function is unknown.
Kcne2 mRNA is present in many brain regions [15] where HCN isoforms are strongly expressed [16][17][18], raising the possibility that KCNE2 could directly influence the function of HCN channels in central neurons. KCNE2 is also highly expressed in the apical membrane of the choroid plexus epithelium, where it influences cerebrospinal fluid composition by regulating KCNQ1 and Kv1.3 K + channel a subunits [19], potentially also indirectly influencing neuronal excitability. Thalamic neurons express HCN2 and HCN4, with HCN2 being the major functional isoform [20,21] while cortical pyramidal neurons strongly express HCN1 [4].
Dysregulation of HCN channel function is strongly implicated in various experimental seizure models [22,23] as well as in human epilepsy [24]. Changes in cellular excitability within corticothalamic circuits can result in seizure activity [25][26][27]. The corticothalamocortical circuit consists of reciprocal connections between the cortex and thalamus such that thalamic VB neurons project to layer 4 and 6 of the somatosensory cortex [28], and layer 6 pyramidal neurons in turn send axons to thalamic neurons, including those in VB [28,29]. Thus, the thalamus and cortex are ideal regions to study the effects of KCNE2 on HCN channel function.
Here, using Kcne2 +/+ and Kcne2 2/2 mice, we have discovered that targeted Kcne2 deletion alters I h properties and neuronal excitability in VB and somatosensory cortex layer 6 neurons and reduces HCN1 and HCN2 protein expression in the brain. Preliminary results have been previously reported [30].

Ethics statement
All experiments were performed following approval by, and in accordance with, Weill Cornell Medical College, University of California, and US federal guidelines.
Access resistance and capacitance were compensated after a whole-cell configuration was established, and were monitored throughout recordings; data were discarded if either of the two parameters changed by .20% of the original values. Liquid junction potentials were calculated and corrected off-line [35]. For recordings of I h , neurons were voltage-clamped at 250 mV; 10and 5-s hyperpolarizing voltage steps respectively were applied to VB and cortical neurons from 250 to 2120 mV (10 mV/step).

Data Analysis
Data processing and analysis including the construction of I h activation curves, fit of time constants, and measurements of temporal summation, burst and tonic spike firing were performed using MiniAnalysis (Synaptosoft, Decatur, GA) or Clampfit 10 (Molecular Devices, Foster City, CA) as previously described [34]. Steady-state activation curves of I h currents were determined from the tail current (for VB neurons, tail currents were analyzed upon returning to 250 mV while in cortical neurons, the tail current was measured at 260 mV); normalized tail current values were plotted as a function of the voltage steps, and were fitted with the Boltzmann function. Activation time constants were determined by fitting 4-s and 1-s segments of the current trace using a double exponential function [18], while tail current traces were fitted with a single exponential to obtain deactivation kinetics. Data are presented as mean 6 SEM; statistical significance was determined using Student's t test or one-way ANOVA with pairwise comparisons, as appropriate.

Kcne2 deletion impairs HCN channel function in VB neurons
VB neurons predominantly express HCN2 (and to a lesser extent HCN4) subunits in the soma and generate a large I h current which slowly activates [20,21], making the VB an excellent region in which to gain insights into whether KCNE2 influences native HCN function. We therefore compared properties of I h currents recorded from VB neurons in brain slices prepared from Kcne2 +/+ and Kcne2 2/2 mice. Families of I h current traces were elicited by a series of 10 s hyperpolarizing voltage steps (Fig. 1A). Other voltage-gated ion channels (Na + , Ca 2+ and K + ) were blocked by concomitant application of tetrodotoxin, Ni 2+ , Ba 2+ and 4aminopyridine (see Methods). To determine the dependence of HCN channel activation on voltage, we measured tail current amplitudes at 250 mV following application of hyperpolarizing voltage steps. Tail currents were normalized to maximum amplitude and normalized values were fitted with the Boltzmann function ( Fig. 1B) to construct activation curves. Group data demonstrate that there was a significant difference in the midpoint voltage of steady-state activation (V 1/2 in mV): 284.360.7 for Kcne2 +/+ and 291.860.9 for Kcne2 2/2 (P,0.001, t-test, n = 37/genotype); slope (mV) was not altered by the deletion.
All VB neurons recorded demonstrated a robust I h (up to 5,500 pA) in both genotypes. Currents were normalized to cell capacitance to obtain I h density (pA/pF, Fig. 1C). Although the maximum I h density at 2120 mV was not different between Kcne2 +/+ and Kcne2 2/2 mice, Kcne2 deletion reduced I h density in Kcne2 2/2 neurons at physiological potentials (280 and 290 mV, Fig. 1C) by virtue of the hyperpolarizing shift in the voltage dependence of activation ( Fig. 1 B). Instantaneous I h currents were not significantly changed by Kcne2 deletion (not shown).
To examine whether Kcne2 deletion altered the ion selectivity of VB HCN channels, I h reversal potential was measured as previously described [33] from tail currents elicited by a series of voltage steps and plotted as a function of pre-pulse voltage. I h reversal potential, and therefore HCN ion selectivity, was not significantly changed by Kcne2 deletion (Kcne2 +/+ , 234.561.2 mV; Kcne2 2/2 , 233.460.8 mV, n = 3 cells/genotype) (Fig. 1D).
We also analyzed the kinetics of channel activation by using fits of traces elicited at 2120 mV with two exponential components. The fast time component (Tau fast ) was 0.3560.026 s for Kcne2 +/+ and 0.70760.05 s for Kcne2 2/2 , and the slow time component (Tau slow ) was 1.8460.21 s for Kcne2 +/+ and 3.0760.29 for Kcne2 2/2 ( Fig. 2A). Tail currents measured at 250 mV following a voltage step to 2120 mV were fitted with a single exponential to quantify deactivation kinetics. Kcne2 deletion also increased the deactivation time constant (0.89860.02 s vs. 1.9760.045 s, Fig. 2B). Thus, Kcne2 deletion slowed the kinetics of both activation and deactivation, doubling the time constants for both processes.
To support the hypothesis that the altered VB currents in Kcne2 2/2 mice arose from shifts in I h rather than other currents, we compared the functional attributes of VB currents pharmacologically isolated using the I h blocker ZD7288 in Kcne2 +/+ and Kcne2 2/2 mice. Thus, ''net'' I h , was calculated by subtracting current traces obtained in the presence of ZD7288 (50 mM) from ''control'' traces (no ZD7288). Comparison of control I h with net I h demonstrated no significant difference in HCN channel properties between these two groups within a given genotype, yet the genotype-dependent differences we observed using control I h , i.e., a shift in voltage dependence (Fig. 1B) (Fig. 3C). The relationship of currentvoltage curves was also the same for control versus net I h within a given genotype (Fig. 3D). The data support the hypothesis that changes in I h properties arising from Kcne2 deletion result from altered HCN channel function.

Kcne2 deletion markedly down-regulates I h in cortical pyramidal neurons
Kcne2 mRNA is detected at relatively high density in cortical dendritic regions [15], where HCN1 is also predominantly expressed [4,37]. We therefore examined the potential influence of Kcne2 deletion on I h properties in cortical layer 6 pyramidal (corticothalamic) neurons. Families of I h current traces were obtained (Fig. 4A) using the same I h isolation solution as for VB neurons. Analysis of steady-state activation curves (Fig. 4B) revealed that the deletion markedly shifted voltage dependence to more hyperpolarized potentials by 10.260.8 mV, with the V 1/2 (in mV) being 281.260.6 for Kcne2 +/+ pyramidal neurons and 291.460.7 for Kcne2 2/2 neurons (P,0.001, t-test, n = 10/ genotype). There was no significant change in slope (9.260.6 vs. 8.960.5 mV, respectively). Unlike the scenario in VB, I h current density (pA/pF) measured in these pyramidal neurons was small, and the maximal density detected at 2120 mV was 2.860.26 pA/ pF for Kcne2 +/+ and 1.960.15 pA/pF for Kcne2 2/2 . The low I h density in either genotype is consistent with uneven distribution of HCN subunits along the somatodendritic axis, with very low density over the somata and extremely high density in distal dendrites [4,37,38].
Unlike VB neurons, Kcne2 2/2 pyramidal neurons displayed a significant decrease in I h current density across the 270 to 2120 mV range (Fig. 4C). While activation time constants for I h in pyramidal cells were fast in both genotypes compared to those in VB neurons, Kcne2 deletion again increased time constants in pyramidal neurons (Kcne2 +/+ : Tau fast , 0.0660.004 s, Tau slow , 0.4160.02 s; Kcne2 2/2 : Tau fast , 0.09760.007, Tau slow , 0.7160.07). The kinetics of both activation and deactivation for I h were slowed in Kcne2 2/2 pyramidal neurons (Fig. 4D). The kinetic data strongly suggest that I h recorded from cortical pyramidal neurons is primarily generated by the HCN1 isoform [39][40][41][42] and slowed by Kcne2 deletion.

Kcne2 deletion enhances excitability and susceptibility to 4-AP in VB neurons
Down-regulation of HCN channel function by Kcne2 deletion would be predicted to alter intrinsic and synaptic excitability, and this possibility was investigated here. The resting membrane potential (RMP) was 268.660.7 mV for Kcne2 +/+ VB neurons compared to 272.160.8 mV for Kcne2 2/2 VB neurons (n = 20). A small voltage response was elicited for measurement of input resistance by intracellular injection of a hyperpolarizing current pulse (230 pA, 500 ms); Kcne2 deletion significantly increased input resistance from 182614 to 288612 MV (n = 12/genotype; Fig. 5A). Intrinsic temporal summation of subthreshold voltage response was evoked by intracellular injection of an EPSC-shaped current train [35,43]. As shown in Fig. 5B, temporal summation (%) was 204612 in Kcne2 +/+ neurons, and was significantly increased to 288618 (P,0.01, n = 8/genotype) in neurons from Kcne2 2/2 mice. Thus, Kcne2 deletion increased intrinsic excitability in VB neurons, raising the question as to whether excitatory synaptic transmission was altered.
Spontaneous EPSPs were evident in both genotypes (Fig. 5C, and see below for additional information) and could be blocked by CNQX and AP5 (not shown). As very few spontaneous bursts were observed in either genotype under control conditions, we used a low concentration of 4-aminopyridine (4-AP; 0.1 mM) to study burst firing and neuronal sensitivity to convulsant challenge. Bathapplication of 4-AP induced low-threshold bursts in both genotypes. Fast single action potentials occurred between bursts in Kcne2 +/+ neurons but few appeared in Kcne2 2/2 cells. Group data indicate that the burst frequency was significantly higher in Kcne2 2/2 neurons than that in Kcne2 +/+ neurons (Fig. 5 D) although no significant change in burst duration or spikes/burst was observed. These data indicate that Kcne2 deletion increases intrinsic excitability and facilitates low-threshold burst firing in thalamocortical VB neurons.

Loss of kcne2 produces hypersusceptibility to 4-AP in layer 6 pyramidal neurons
Since I h density was significantly reduced across the voltage range in Kcne2 2/2 layer 6 pyramidal neurons (Fig. 4C), we also examined whether excitability in these cells was altered. Input resistance significantly increased in Kcne2 2/2 pyramidal neurons compared to those from Kcne2 +/+ mice (395618 vs. 172611 MV, n = 8/genotype, P,0.001, Fig. 6C). Temporal summation was tested in both genotypes using the same protocol for VB neurons (traces not shown). Kcne2 deletion significantly increased summation by 4563.5% (220614% vs. 320616%, n = 8). As was the case with VB neurons, spontaneous EPSPs were evident in both genotypes in the absence (control) of 4-AP; however, the frequency was higher in Kcne2 2/2 than in Kcne2 +/+ pyramidal neurons (Fig. 6  A-B). Bath application of 4-AP (0.1 mM) induced rhythmic low threshold Ca 2+ spike (LTS) burst firing patterns in both genotypes; such bursting could last for more than 90 min. Group data for cortical burst properties are summarized in Fig. 6D. Compared to Kcne2 +/+ , the burst frequency in  In Kcne2 2/2 mice, cortical burst duration was also much longer than that seen in VB counterparts (3.460.3 s vs. 0.6 s60.04, P,0.001), a 5.6-fold increase. The magnitude of the effects on voltage responses in Kcne2 2/2 neurons was much larger in corticothalamic pyramidal neurons than that observed in thalamocortical VB neurons, implying that corticothalamic excitatory transmission was increased in brain slices from Kcne2 2/2 mice.

Kcne2 deletion had little effect on glutamate release
As noted above, the frequency of spontaneous EPSPs in layer 6 pyramidal neurons in the absence of 4-AP was increased by Kcne2 deletion, suggesting a possible alteration in glutamate release machinery. To investigate this possibility, miniature excitatory postsynaptic currents (mEPSCs) were recorded in both genotypes in the absence of 4-AP but the presence of TTX (500 nM) and bicuculline (20 mM), thereby blocking spike-driven events and fast GABAergic transmission. Fast mEPSCs were readily detected in pyramidal neurons from both genotypes and could be blocked by co-application of CNQX and AP-5 ( Fig. 7A-B), confirming that they were mediated by ionotropic glutamate receptors. mEPSC frequency (8.361.1 vs. 9.861.5 Hz, Kcne2 +/+ and Kcne2 2/2 , respectively, n = 6 cells/genotype), and amplitude (29.861.3 vs. 31.261.4 pA) were similar between the two genotypes. mEPSC decay time, however, was significantly prolonged in Kcne2 2/2 neurons (1.5860.2 vs. 0.5460.05 ms) (Fig. 7C-D), and the prolongation in the decay time was consistent with the observed increase in R in (cf. [44]). The data suggest that Kcne2 deletion does not alter glutamate release from presynaptic glutamatergic neurons, and that the enhanced excitability observed above is likely due to the reduction of shunting produced by I h , and appears similar to what has been described for HCN1-null cortical pyramidal neurons [45].

Kcne2 deletion reduces brain HCN1 and HCN2 protein expression
Co-IP experiments were employed to determine whether KCNE2 forms protein complexes with HCN1 and HCN2 in mouse brain. Western blots using HCN1 or HCN2 antibodies to probe KCNE2 antibody-precipitated fractions did not yield specific signal (data not shown), suggesting either these complexes do not form in mouse brain, that the amount of complex formation is below our detection limit, or our co-IP protocol did not preserve native complexes.
As Kcne2 deletion reduced native VB and pyramidal neuron I h , we also quantified HCN1, HCN2 and HCN4 protein in Kcne2 +/+ and Kcne2 2/2 whole-brain lysates to determine if Kcne2 deletion altered HCN1, HCN2 and HCN4 protein expression. Strikingly, neural expression of HCN2 protein was significantly reduced in Kcne2-deleted mice compared to Kcne2 +/+ mice (P = 0.02) and there was a trend toward reduction of HCN1 protein expression (P = 0.07) (Fig. 8A-B). This change appears to be specific as HCN4 protein expression was unchanged (Fig. 8A-B) and wholebrain expression of two other membrane proteins, the Kv2.1 K + channel a subunit and the KCC1 K + /Cl 2 co-transporter, was unaffected by Kcne2 deletion (data not shown).

Discussion
KCNE2 is a voltage-gated potassium channel ancillary subunit that also regulates heterologously expressed HCN channels. We present novel data which demonstrate that KCNE2 is required for normal HCN channel activity in central neurons.

Kcne2 deletion impairs neuronal I h
Previous studies have provided evidence that the potassium channel b subunit KCNE2 regulates heterologously expressed HCN1, 2 and 4 channels to modulate the activation kinetics and amplitude of I h [11,14]. Previous investigations of the ability of KCNE2 to modulate HCN gating in heterologous expression systems have yielded variable results. Thus, in Xenopus oocytes expressing HCN1 or HCN2, KCNE2 co-expression produces a small (4 mV) depolarizing shift in V 1/2 [11] while co-expression of KCNE2 with HCN4 produced a significant hyperpolarizing shift (8 mV) in V 1/2 [12]. In Chinese Hamster Ovary (CHO) cells, KCNE2 co-expression with HCN1, 2, or 4 had no effect on the V 1/2 , although other I h properties were markedly altered [14]. Similarly, over-expression of KCNE2 with HCN2 failed to alter HCN gating in neonatal ventricular myocytes [13]. Given the contradictory results obtained using different expression systems it is therefore necessary, as previously noted [3], to study the effects of native KCNE proteins on native HCN channel function.

Kcne2 acceleration of current activation
The activation time constants for I h in Kcne2 +/+ VB neurons was much slower than that of layer 6 pyramidal neurons (the fast component being 5.6-fold slower and the slow component being 4.5-fold slower), consistent with previous reports of predominant expression of HCN2 in VB and HCN1 in the cortex [16,17,41,42,[46][47][48]. Deletion of Kcne2 led to a significant slowing of activation time constants in both VB and cortical neurons  [13,14], although there is one report indicating that KCNE2 co-expression with HCN4 slows activation [12].
A comparison of neuronal and recombinant channel properties raises an important question with respect to direct regulation of gating kinetics. When HCN channels are heterologously expressed, the presence of KCNE2 accelerates the kinetics of current activation resulting from homomeric HCN channel expression, and this likely reflects a direct protein-protein interaction [12]. The kinetics of I h in CNS neurons, however, appear to be strongly influenced by HCN channel composition in different cell populations, and expression of heteromultimeric channels may also contribute to differences in the kinetics of neuronal I h [18]. VB neurons primarily express HCN2 channels, with a smaller population of HCN4 channels also present [20,21]. Given that KCNE2 accelerates I h activation in heterologous expression systems (as discussed above), the slower kinetics of I h observed in kcne2 2/2 VB neurons may be due to loss of direct regulation by KCNE2, but could also be due to alterations in HCN subunit expression. As shown in Fig. 8, there is a significant decrease in overall HCN2 expression in the brain when Kcne2 is knocked out, while overall brain HCN4 expression remains constant. This could theoretically increase the relative contribution of the more slowly activating HCN4 (compared to HCN2) to native I h [48,49], thereby giving rise to a slowly activating I h current in VB neurons.
KCNE2 protein has been detected in a range of tissues, and its deletion impacts the function of the stomach, heart and thyroid, all of which normally express KCNE2 in mice and humans [10,19,31,32,50]. While Kcne2 mRNA has been detected in a range of neural tissues [15], in a recent study we did not observe specific KCNE2 protein staining by immunohistochemistry in neuronal populations of mouse brain. In contrast, robust, highly specific KCNE2 protein staining was apparent in the apical membrane of the choroid plexus epithelium, which lines the fourth and lateral ventricles of the brain and secretes cerebrospinal fluid [19]. It is possible, then, that Kcne2 deletion alters neuronal I h characteristics indirectly by, for example, altering the cerebrospinal fluid composition in a manner that leads to electrical remodeling such as the reduction in HCN2 protein expression we observed here. This could not only reduce I h density, but also lead to slower-activating I h because of a shift in the balance between HCN2 and HCN4.

Kcne2 critically regulates excitability and burst firing
Alteration of I h is associated with a marked change in intrinsic excitability in thalamic neurons [33,34,43]. Here we found that downregulation of HCN channel function by Kcne2 deletion resulted in the increase of input resistance and temporal summation of subthreshold voltage response in both thalamic and cortical neurons, indicating an increase in intrinsic excitability (Figs. 5 and 6). As a result, burst firing also increased in Kcne2-null brain slices. Previous studies have shown that genetic downregulation or pharmacological diminishment of the I h conductance is tightly linked to the occurrence of seizure activity [20,[51][52][53][54] and the decrease of seizure threshold during convulsant challenge [20,45]. We found that a low concentration of 4-AP induced longlasting burst firing, which is similar to epileptiform activity [55][56][57][58]. The burst firing frequency was significantly increased by Kcne2 deletion in both thalamocortical and layer 6 pyramidal neurons, suggesting increased susceptibility to the convulsant challenge. Intriguingly, the duration of bursts in Kcne2-null pyramidal, but not thalamocortical, neurons was much longer, with a marked increase in fast action potentials riding on bursting calcium spikes. Such augmented excitability did not seem to result from a direct alteration of the glutamatergic release machinery as neither mEPSC amplitude nor frequency were changed by the deletion (Fig. 7), but was likely mediated through shunting reduction caused by downregulation of I h (Fig. 6).
HCN1 channels show a 60-fold increase from somatic to distal dendritic membrane in layer 5 pyramidal neurons [37] and not surprisingly, dendritic I h plays a pivotal role in controlling pyramidal cell excitability [59]. A recent study has shown that cortical layer 6 pyramidal dendrites also possess membrane properties similar to those of other layer pyramidal neurons [38]. Hence, downregulation of cortical dendritic I h leads to reduction of dendritic shunting and consequently constraints on local spike propagation to the soma, thereby facilitating synaptic integration and excitatory synaptic transmission. This effect may account for the occurrence of cortical hypersusceptibility to 4-AP observed here, and suggests that Kcne2-null mice may be more susceptible to chemical-induced seizures because of dysregulation of HCN channel function. Since 4-AP-induced synchronized oscillations (epileptiform activity) originate in cortical layer 6 and propagate to thalamus through corticothalamic projections [56], the increase in apparent EPSP frequency and burst firing in thalamocortical VB neurons (Fig. 5) may result in part from increased corticothalamic excitatory synaptic input [56], which in turn leads to an enhancement of thalamocortical output to layer 4 and 6 [28], creating an imbalance between excitation and  inhibition in the cortico-thalamo-cortical loop. Our results ( Fig. 4  and 6) support the notion that dendritic HCN1 plays a critical role for the regulation of cortical pyramidal excitability [45].
KCNE2 can modulate the function of other ion channels [10,60], including the 4-aminopyridine (4-AP)-sensitive rapidly activated, transient outward potassium current I A , which is mediated by the KCND3 gene-product, Kv4.3 [61]. KCNE2 is a b subunit for Kv4.3 [62], and Kv4.3 mRNA is present in the cortex and thalamus [63]; in a thalamic relay neuron model, I A slows the rate of rise and reduces the peak amplitude of the lowthreshold Ca 2+ spike and reducing I A to 0 results in a decrease in the inter-spike interval and in increase in the response of the cell to a depolarizing current injection [64]. Those observations raise the theoretical possibility that the observed changes in excitability observed in thalamic (Fig. 5) and cortical (Fig. 6) neurons resulted from 4-AP block of I A .
We do not think that the enhanced excitability in Kcne2 2/2 mice is the result of changes in Kv4.3 function for the following reasons. Firstly, at membrane potentials less than 265 mV, Kcne2 deletion by itself does not produce spontaneous spike firing in either thalamic (Fig. 5C, top current trace) or cortical (Fig. 6A, top current trace) neurons, indicating that I A -dependent control of spontaneous spike firing is not altered by the deletion. This is not entirely unexpected as I A and I h co-vary and have opposing and complementary effects [65][66][67]. Secondly, there is little if any measurable I A at membrane potentials less than 260 mV, as its V 1/2 is quite positive (,236 mV) [68], and the excitability experiments in the present study were performed at membrane potentials more negative than 265 mV. Thirdly, 4-AP blocks neuronal I A in the low millimolar range, with an IC 50 of 1 to 2 mM [68,69] (similar to the IC 50 of ,1.5 mM reported in HEK293 cells expressing Kv4.3 channels [70]), and the concentration used to induce burst firing in the present study was 0.1 mM; at this concentration, I A is only blocked by 7% [69]. Finally, absence of I A (as would occur in the presence of full block) produces a very different pattern of activity (at least in a model thalamic cell [64]) than what we observed. In total, these studies strongly indicate that the burst firing patterns observed in the present study did not result from 4-AP modulation of Kv4.3 mediated I A currents. Limitations of the present study Our I h data from VB and pyramidal neurons demonstrate that native HCN gating is KCNE2-dependent inasmuch as Kcne2 deletion alters native HCN gating, but the lack of native coimmunoprecipitation between KCNE2 and HCN1 or HCN2 from the brain in our study leaves the door open for at least several potential mechanisms for this functional dependence. First, KCNE2 mRNA is clearly detected in cortical and thalamic neurons [15], and it is possible that KCNE2 directly regulates HCN channels in the brain but our biochemical studies did not provide sufficient resolution to detect this, or the necessary conditions to preserve complex stability. Thus, the slowed gating and reduced current density of I h we observe upon Kcne2 deletion could stem from the loss of KCNE2 from HCN channel complexes, which would be predicted to slow gating and reduce current density via reduced single channel conductance [11][12][13][14]. Second, KCNE2 could function as an ancillary subunit and regulate HCN trafficking to specific regions of the cell membrane in a manner analogous to that of TRIP8B [71][72][73][74], so in the absence of KCNE2 perhaps HCN surface expression is impaired. Third, it is possible that KCNE2 indirectly impacts neuronal HCN function due to modulatory effects on other ionic currents, including those mediated by a number of voltage-gated potassium channels [10,60], and that those changes indirectly alter the HCNmediated current. Fourth, we previously observed that KCNE2 levels were highest in the choroid plexus [19], and KCNE2 is present in many other non-neuronal tissues [50,75], so the observed effects could be indirect from one or more of a variety of sources within the mouse. These caveats notwithstanding, Kcne2 deletion is the root cause of the observed effects reported here.

Conclusion
In summary, voltage-dependent gating of native HCN channels in CNS neurons appears to be controlled by multiple factors, including in some shape or form the b subunit KCNE2, as shown here. The present findings have revealed for the first time that KCNE2 exerts an important role in the maintenance of brain pacemaking function at physiological membrane potentials. Loss of KCNE2 leads to downregulation of HCN channel function associated with increased excitability in neurons in the corticothalamo-cortical loop. Thus, KCNE2 strongly influences HCN channel activity crucial for homeostatic regulation of a dynamic balance of excitation and inhibition [76] in the interconnected circuitry. In future work, the specific mechanisms for this functional link will be further investigated. Figure 8. Kcne2 deletion down-regulates HCN1 and HCN2 protein expression in the brain. A, Exemplar chemiluminescence signals from western blots of whole brain lysates from Kcne2 +/+ and Kcne2 2/2 mice, normalized to total protein concentration and probed with antibodies raised against HCN1, HCN2, HCN4 or GAPDH, as indicated. B, Mean chemiluminescence intensities for bands corresponding to known molecular weights for HCN1, HCN2, HCN4 and GAPDH from blots as in panel A, n = 3-4 mice per genotype. The cumulative GAPDH data on the left were obtained concomitantly with the HCN1 and HCN2 samples while the comparable GAPDH data on the right were obtained concomitantly with the HCN4 samples. *Significant difference between genotypes at 95% confidence interval. Error bars indicate SEM. doi:10.1371/journal.pone.0042756.g008