HCN1 hyperpolarization-activated cyclic nucleotide–gated channels enhance evoked GABA release from parvalbumin-positive interneurons

Significance

Here, we report, using a combination of genetic and pharmacological approaches, that HCN1 hyperpolarization-activated cation channels expressed in the axons and synaptic terminals of parvalbumin-positive (PV+) interneurons enhance the strength of evoked inhibitory synaptic transmission onto hippocampal CA1 pyramidal cells. Two-photon calcium imaging showed that these channels increase the mean calcium transient in individual presynaptic boutons of PV+ interneurons in response to a train of electrically evoked action potentials by elevating the probability with which a single evoked action potential elicited a calcium response. As the PV+ interneurons are essential for balancing excitation in the hippocampal neuronal network, our findings may help explain how HCN1 loss-of-function variants give rise to severe forms of epilepsy in early childhood.

Abstract

Hyperpolarization-activated, cyclic nucleotide–gated (HCN) channels generate the cationic I_h current in neurons and regulate the excitability of neuronal networks. The function of HCN channels depends, in part, on their subcellular localization. Of the four HCN isoforms (HCN1-4), HCN1 is strongly expressed in the dendrites of pyramidal neurons (PNs) in hippocampal area CA1 but also in presynaptic terminals of parvalbumin-positive interneurons (PV+ INs), which provide strong inhibitory control over hippocampal activity. Yet, little is known about how HCN1 channels in these cells regulate the evoked release of the inhibitory transmitter GABA from their axon terminals. Here, we used genetic, optogenetic, electrophysiological, and imaging techniques to investigate how the electrophysiological properties of PV+ INs are regulated by HCN1, including how HCN1 activity at presynaptic terminals regulates the release of GABA onto PNs in CA1. We found that application of HCN1 pharmacological blockers reduced the amplitude of the inhibitory postsynaptic potential recorded from CA1 PNs in response to selective optogenetic stimulation of PV+ INs. Homozygous HCN1 knockout mice also show reduced IPSCs in postsynaptic cells. Finally, two-photon imaging using genetically encoded fluorescent calcium indicators revealed that HCN1 blockers reduced the probability that an extracellular electrical stimulating pulse evoked a Ca²⁺ response in individual PV+ IN presynaptic boutons. Taken together, our results show that HCN1 channels in the axon terminals of PV+ interneurons facilitate GABAergic transmission in the hippocampal CA1 region.

A wide range of GABAergic inhibitory interneurons contribute to shaping neuronal network activity throughout the brain. Their axons commonly target various compartments of their postsynaptic targets, resulting in distinct modulatory effects on neural activity (1, 2). Thus, in the hippocampus and neocortex, somatostatin-expressing interneurons target distal dendritic regions of excitatory pyramidal neurons (PNs), where they regulate synaptic integration and plasticity (3–6). In contrast, parvalbumin-expressing interneurons (PV+ INs) target the perisomatic subdomain and axon initial segment of PNs, thereby regulating spike output and network oscillation (3, 7–9).

While the synaptic connections that interneurons make with their targets define network activity that determines behavioral output (10, 11), the efficacy of these inhibitory connections depends on the expression of diverse voltage-gated ion channels in somatic, dendritic, axonal, and presynaptic compartments. The clinical importance of voltage-gated channels is indicated by findings that mutations in interneuron channels can lead to different forms of epileptic encephalopathies (12). Perhaps the best understood example is Dravet’s syndrome, where a loss-of-function mutation in the SCN1A excitatory voltage-gated sodium channel, which is strongly expressed in PV+ INs, causes seizure activity due to a reduction in PV+ IN excitability, resulting in decreased inhibition (13, 14). Mutations in the HCN1 subtype of the hyperpolarization-activated, cyclic nucleotide-modulated (HCN) channel family, which is strongly expressed in PV+ IN axons and presynaptic boutons, have recently been found to underlie certain cases of early infantile epileptic encephalopathy (EIEE; 15–17). However, our understanding of how HCN1 channels contribute to interneuron function remains relatively unexplored.

HCN channels generate the hyperpolarization-activated cationic current I_h, which dynamically controls membrane resting potential, input resistance, and synaptic integration at postsynaptic sites, thereby regulating the excitability of neuronal networks (18–21). The channels are encoded by four closely related HCN genes (HCN1-4), which display distinct patterns of neuronal expression, with HCN1 expressed in select brain regions, in particular in the neocortex, hippocampus, and cerebellum (18, 22–24). In the hippocampus, HCN1 is strongly expressed in distal apical dendrites of CA1 PNs (22, 25, 26), where it acts as an inhibitory constraint on the temporal integration of perforant path excitatory inputs (27).

In contrast, HCN channels in parvalbumin-positive inhibitory neurons (PV+ INs), including hippocampal basket cells (26, 28), are enriched in axons and presynaptic processes (29), where the channels enhance action potential (AP) initiation and propagation during sustained high-frequency firing, decreasing the latency with which a basket cell somatic action potential evokes an inhibitory postsynaptic current (IPSC) in their target neurons, and help maintain persistent axonal firing (30).

To date, there is conflicting evidence as to whether and how I_h and HCN channels regulate inhibitory synaptic transmission. Thus, whereas pharmacological blockers of HCN channels decrease the frequency of both miniature (31) and spontaneous (32) inhibitory postsynaptic currents (IPSCs) in hippocampal neurons, the same antagonists increase miniature and spontaneous IPSCs in prefrontal cortex neurons (33). The conflicting conclusions may reflect potential off-target effects of these compounds on synaptic transmission (34) and/or differential contributions of different HCN isoforms in different classes of interneurons. Thus, the role of HCN channels, and HCN1 in particular, in PV+ IN function remains uncertain. Using a combination of pharmacological, genetic, and calcium imaging-based approaches, we here report that HCN1 both regulates PV+ IN somatic membrane properties and enhances the efficacy of inhibitory synaptic transmission from PV+ INs onto their CA1 PN targets.

Results

HCN1 Is Expressed in PV+ IN Presynaptic Processes.

We first examined the possible expression of HCN subunits in PV+ IN presynaptic processes in the CA1 region of the hippocampus using immunofluorescence colabeling with antibodies against synaptotagmin-2 (SYT2), a calcium sensor localized on synaptic vesicles of PV+ INs (35, 36), and antibodies against either the HCN1 or HCN2 subunits. Both HCN1 and HCN2 were expressed in CA1, mainly in stratum lacunosum moleculare (SLM), reflecting their high density within the apical dendrites of CA1 PNs (Fig. 1). However, HCN1 was also expressed in stratum pyramidale (SP), the site of PV+ IN input, in a perisomatic pattern that colocalized with SYT2 staining (Fig. 1 A and B). In contrast, immunostaining for HCN2 subunits was weak in SP of CA1 and not colocalized with SYT2. A low level of HCN2 in SP of CA1 has been reported previously in both mice (37) and rats (26), in contrast to the relatively high levels of HCN2 in SP of CA3 seen in these studies. Overall, our results indicate that it is primarily HCN1 subunits, which form the HCN channels present in the presynaptic terminals in SP of CA1 (Fig. 1 C and D).

Fig. 1.

Confocal images of double and triple immunofluorescence staining in the hippocampus. (A) Low-magnification overlay image of dual staining for synaptotagmin 2 (SYT 2, purple) and HCN1 (green) of the entire hippocampus. The scale bar represents 400 µm. (B) High-magnification images of area CA1, showing stratum oriens (SO), stratum pyramidale (SP), and stratum radiatum (SR); Left: stained for synaptotagmin 2 (SYT2, purple); Middle: stained for HCN1 (green); Right: overlay of the two stainings (colocalization shown in white). The scale bar represents 50 µm. (C) Low-magnification overlay image of dual staining for synaptotagmin 2 (SYT 2, purple) and HCN2 (green) of the entire hippocampus. The scale bar represents 400 µm. (D) High-magnification images of area CA1, showing SO, SP, and stratum radiatum (SR); Left: stained for synaptotagmin 2 (SYT2, purple); Middle: stained for HCN2 (green); Right: overlay of the two stainings (colocalization shown in white). Note: The cell soma, visible in the HCN2 labeling, is a putative oligodendrocyte. The scale bar represents 50 µm. (E) High-magnification images of area CA1 of a wild-type mouse, showing SP; far Left: stained for parvalbumin (PV, orange); Middle Left: stained for HCN1 (green); Middle Right: overlay of the PV and HCN1 stainings; far Right: overlay of the three stainings for PV, HCN1, and synaptotagmin 2 (SYT2, purple, three-way colocalization shown in white). The scale bar represents 50 µm. (F) Same as (E), but images are taken from a HCN1 knockout mouse. (G) Mander’s overlay coefficient of HCN1 and PV in SP of CA1 in the wild-type (WT) and HCN1 knockout mice. n = 12 slices from 2 mice for WT and 20 slices from 3 animals from Hcn1^−/−, P < 0.0001 with the unpaired t test. (H) Mander’s overlay coefficient of HCN1 and SYT 2 in the same images as (G). P < 0.0001 with the unpaired t test. (I) Mander’s overlay coefficient of PV and SYT 2 in the same images as (G). P = 0.45 with the unpaired t test.

To confirm the localization of SYT2 and HCN1 to PV+ IN presynaptic processes, we performed triple labeling for PV, SYT2, and HCN1. Indeed, we found that all three markers were colocalized in SP within filamentous structures, which are likely axon processes (26). In contrast, PV+ IN somas, defined by PV antibody staining, lacked detectable staining for either HCN1 or SYT2 (Fig. 1E).

To confirm the specificity of the HCN1 antibody labeling, we also performed the triple labeling experiments in homozygous HCN1 knockout mice (Hcn1^tm2Kndl/J, 18, Fig. 1F). We quantified coexpression using Mander’s colocalization coefficient [MCC (38, 39)] for all three possible pairs of labels in wild-type and knockout mice (n = 12 brain slices from 2 wild-type (WT) mice and 20 brain slices from 3 Hcn1^−/− mice). Colocalization of HCN1 and PV signal was significantly higher in WT mice (MCC = 0.36 ± 0.01) than HCN1 knockout mice (MCC = 0.07 ± 0.02; P < 0.0001 with the unpaired t test; Fig. 1G). Similarly, HCN1 and SYT2 signals were more colocalized in WT (MCC = 0.46 ± 0.03) than Hcn1^−/− mice (MCC = 0.10 ± 0.03; P < 0.0001 with the unpaired t test; Fig. 1H). In contrast, there was no effect of HCN1 deletion on colocalization of PV and SYT2 signals (MCC = 0.54 ± 0.15 in WT mice and 0.58 ± 0.09 in Hcn1^−/− mice; P = 0.45 with the unpaired t test; Fig. 1I).

Although in CA1 pyramidal cells, HCN1 is primarily expressed in distal apical dendrites (22), light-microscopic immunocytochemistry cannot rule out the possibility that the HCN1 signal in SP reflects channels in the pyramidal neuron soma. We thus created a conditional knockout mouse line in which HCN1 was deleted specifically in PV+ INs by crossing the floxed HCN1 line (19) with a PV-Cre line ( Hcn1^flox/flox: Pvalb^tm1(cre)Arbr/J). There was a marked reduction in HCN1 fluorescence in the CA1 SP of the conditional knockout mice compared to wild-type littermate controls, with no change in signal in the stratum radiatum (SR) or SLM (Fig. 2 A and B). Costaining for PV in the same animals confirmed that PV+ IN axons are mainly present in SP, in both wild-type mice and mice with the conditional HCN1 deletion (Fig. 2B).

Fig. 2.

Confocal images of HCN and PV labeling in wild-type and conditional PV+ In-HCN1 knockout mice. (A) Low-magnification image, showing HCN1 labeling (green) throughout the hippocampus in wild-type (Left) and PV+-IN-specific HCN1 knockout mouse (Right). SP: stratum pyramidale, SR: stratum radiatum, SLM: stratum lacunosum moleculare. The white arrow shows the area at high magnification in panel B. The scale bar represents 400 µm. (B) Top: High-magnification image, showing HCN1 labeling (green) in SP of CA1 in wild-type (WT, Left) and PV+-IN-specific conditional HCN1 knockout (cKO, Right) mice. The scale bar represents 50 µm. Middle: Parvalbumin (PV) labeling (purple) in the same images as seen on the Top. Bottom: Merge of the Middle and Top images. Colocalization of HCN1 and PV labeling can be seen in white. (C) Average fluorescence of immunostaining for the HCN1 channel, using the PV fluorescence as a mask, normalized by HCN1 fluorescence in SLM for each slice (n = 8 slices of 2 WT mice and 9 slices of 3 cKO mice, P < 0.0001 with the unpaired t test).

We quantified the changes in HCN1 expression confined to the region of the PV+ IN processes using the PV immunohistochemical signal to generate a mask over each image. We then calculated the average fluorescence of the HCN1 staining within this mask. To account for variation of antibody labeling in different brain slices, we then normalized this fluorescence value by the average fluorescence signal in a region within SLM of the same tissue sample. This analysis revealed a nearly three-fold decrease in normalized HCN1 fluorescence signal under the PV mask in the conditional knockout mice (0.23 ± 0.02; n = 8 slices in 3 conditional knockout mice) compared to wild-type littermates (0.61 ± 0.04, n = 9 slices in 2 wild-type mice; P < 0.0001, mixed effects model analysis; Fig. 2C). While these data do not rule out the possibility that HCN1 is also expressed in the axons of other interneuron types, we conclude that HCN1 is indeed strongly expressed in the axons and/or presynaptic terminals of PV+ INs within the CA1 SP layer.

HCN Channels have Little Impact on PV+ IN Somatic Membrane Properties.

To determine the impact of HCN channels on cellular excitability and membrane properties in PV+ INs, we crossed the mouse line expressing Cre recombinase selectively in PV+ INs (Pvalb^tm1(cre)Arbr/J) with a reporter mouse line that expressed the fluorescent marker tdTomato in a Cre-dependent manner (Gt(ROSA)^{26Sortm14(CAG-tdTomato)Hze}/J, or Ai14). This resulted in reliable tdTomato expression in PV+ INs specifically (Fig. 3A), which allowed us to perform whole-cell patch-clamp recordings from identified PV+ INs in the CA1 region of acute hippocampal slices. These experiments revealed two populations of PV+ INs based on their electrophysiological properties and response to the HCN channel blocker ZD7288 under whole-cell current-clamp conditions. We applied ZD7288 at 10 µM and limited the duration of our recordings to within 10 min of its application to minimize off-target effects (40).

Fig. 3.

Impact of HCN channel block on somatic properties of PV+ INs. (A) Confocal section of parvalbumin immunofluorescence staining and tdTomato expression in PV-Cre mice crossed to the tdTomato reporter line. The scale bar represents 200 µm. (B) Example voltage traces from two cells in response to a series of hyperpolarizing current steps ranging from 0 to −175 pA (in steps of −25 pA, upper traces) and to a single +250 pA depolarizing current step (lower traces). Top traces are from a cell exhibiting no voltage sag during hyperpolarization. Bottom traces are from a cell exhibiting voltage sag. Voltage responses shown before (black/red traces) and during (blue/pink traces) bath application of ZD7288. (C–F) Summary graphs of the effects of ZD 7288 (before: black/red; after: blue/pink) on indicated passive membrane parameters, for “sag” and “non-sag” cells. (G) AP firing frequency in response to positive current steps (F–I curve) before (black/red) and after (blue/pink) ZD7288. (H–J) Summary graphs of the effects of ZD7288 (before: black/red; after: blue/pink) on action potential frequency at a stimulus current of 350 pA, AP threshold and average AP peak respectively, for sag (red controls) and non-sag (black controls) cells. For all panels, small circles show individual cells, large circles show means, error bars show standard error. (**** means P < 0.0001 with the paired t test after KS normality test.)

The majority of PV+ cells (33 out of 48) showed no voltage sag—a hallmark of HCN channel activation—in response to hyperpolarizing current steps (“no sag” cells). The remainder of PV+ INs (15 out of 48) exhibited a moderate voltage sag (“sag” cells, Fig. 3B). Bath application of the HCN blocker ZD7288 (10 µM) blocked the sag in the sag cell population, with sag ratio (defined in Methods) = 0.84 ± 0.01 in the absence and 0.97 ± 0.01 in the presence of ZD7288 (P < 0.0001 with the paired t test after KS normality test; n = 10; Fig. 3C). ZD7288 had no effect on the sag ratio of the non-sag cells (0.97 ± 0.01 in the absence and 0.97 ± 0.01 in the presence of ZD7288; P = 0.58; n = 13). We confirmed that effects of ZD7288 were not due to drift or rundown in cell parameters during the 10 min drug application (SI Appendix, Fig. S1). In addition, the almost complete abolishment of the voltage sag—the hallmark of HCN channel function—indicates that application of 10 µM ZD7288 for 10 min is effective in blocking HCN channels.

Somewhat unexpectedly, ZD7288 had no effect in sag cells on resting membrane potential (−69.6 ± 0.9 mV in the absence and −70.4 ± 0.7 mV in the presence of ZD7288; P = 0.37; n = 10; Fig. 3D) or input resistance measured with small (−50 pA) hyperpolarizing current steps from the resting potential (94.4 ± 4.6 MΩ in the absence and 96.6 ± 4.4 MΩ in the presence of ZD7288; P = 0.49; n = 10; Fig. 3E). This contrasts with findings in many cell types (31, 41) that HCN channels depolarize the resting membrane and lower input resistance, when measured with small voltage deviations. However, HCN channel block did reduce input resistance using larger hyperpolarizing current steps sufficient to elicit voltage sag (94.5 ± 4.9 MΩ in the absence and 105.9 ± 4.9 MΩ in the presence of ZD7288; P < 0.0001; n = 10; Fig. 3F). These results suggest that the voltage dependence of HCN channel activation was shifted to values negative to the resting potential in the sag cells, which could result from electrotonic attenuation of the somatic hyperpolarization at an axonal site of HCN channel expression.

Bath application of ZD7288 also had no impact on somatic action potential (AP) properties in sag or non-sag cells, including firing frequency (sag cells: 88.09 ± 7.25 Hz before compared to 88.64 ± 8.26 Hz after ZD7288 with a +350 pA current step; P = 0.92; n = 10, Fig. 3 G and H), voltage threshold (−45.5 ± 1.5 mV before compared to −45.6 ± 1.8 mV after ZD7288; P = 0.92; n = 10, Fig. 3I), or peak AP voltage (43.76 ± 1.7 mV before compared to 43.1 ± 1.8 mV after ZD7288; P = 0.31; n = 10, Fig. 3J). Thus, HCN channels do not have a significant influence on somatic resting membrane properties or excitability in CA1 PV+ INs.

HCN Channels Enhance the PV+ IN Evoked Inhibitory Postsynaptic Current in CA1 PNs.

The fact that we saw strong HCN1 expression in the synaptic terminals of PV+ INs in CA1 SP (Fig. 1) suggests that the main impact of HCN channels in this cell type may be on synaptic transmission. We thus measured the effect of HCN channel blockade on IPSCs recorded in CA1 PNs evoked by electrical stimulation in CA1 SP. To examine monosynaptic IPSCs we blocked fast excitatory synaptic transmission by applying AMPA and NMDA receptor antagonists to the bath solution. Of note, we included the intracellular HCN channel blocker QX-314 (5 mM) in the internal solution of the patch pipette to block any potential effects of postsynaptic HCN channels in the CA1 PN membrane. IPSCs were recorded while the cell was clamped to +10 mV before and after bath application of ZD7288 (Fig. 4A). We applied ZD7288 at 10 µM and limited the duration of its application to 10 min to minimize off-target effects (40).

Fig. 4.

Block of HCN channels decreases IPSCs evoked by stimulation of PV+ INs. (A) (Left and Top) Schematic of voltage clamp recordings from CA1 PNs with a stimulating electrode in the pyramidal cell layer. (Bottom) Example trace of an IPSC evoked by a 35 V 0.2 ms extracellular stimulus before (back) and after (blue) ZD7288 (10 µM) application. Excitatory transmission was blocked with CNQX (25 µM) and D-APV (50 µM). QX-314 (5 mM) was used in the recording pipette to block postsynaptic HCN channels. (B) Amplitude of evoked IPSCs using extracellular electrical stimulation before (black) and after (blue) bath application of ZD7288. (C) (Top) Schematic of voltage clamp recordings from CA1 PNs optogenetically stimulating ChR2-expressing PV+ INs using blue light pulses (2 ms). (Left) neuron filled with biocytin in red and ChR2, expressed in PV INs in green. The scale bar represents 150 µm. (Bottom) Example trace of an IPSC, elicited by light pulse stimulation of PV+ IN axons before (back) and after (blue) ZD7288 (10 µM) application. QX-314 (5mM0 was used in the recording pipette to block postsynaptic HCN channels. (D) Light pulse stimulation of PV+ IN axons expressing ChR2 before (black) and after (blue) bath application of ZD7288. (**** means P < 0.0001 with the paired t test after KS normality test.)

We found that application of the HCN channel blocker ZD7288 caused a significant ~35% reduction in IPSC amplitude (from 640.5 ± 50.3 pA before to 391.9 ± 34.8 pA after ZD7288 at a stimulation amplitude of 35 V; P < 0.0001; n = 32; Fig. 4B). The effect was strongest at highest stimulus intensities tested and was not due to drift or rundown (SI Appendix, Fig. S2). The more selective HCN channel blocker ivabradine exerted a similar inhibitory effect (IPSC reduced from 663.1 ± 57.2 pA to 469.8 ± 39.3 pA after ivabradine application; P < 0.0001; n = 17; SI Appendix, Fig. S3 A and B), indicating that the reduction of the IPSC is likely mediated by the block of HCN channels rather than an off-target effect.

As extracellular electrical stimulation recruits a variety of IN subtypes, we used an optogenetic approach to selectively activate the PV+ INs by crossing the PV-Cre mouse line with a line that expresses channelrhodopsin-2 (ChR2) in a Cre-dependent manner (Gt(ROSA)26Sor^{tm32(CAG-COP4*H134R/EYFP)Hze}/J, or Ai32) (Fig. 4C). We then stimulated these cells and their axons with a 2-ms light pulse at 470 nm, which evoked large IPSCs in CA1 PNs. Similar to its effect on the electrically evoked IPSC, bath application of ZD-7288 caused a significant decrease in the optogenetically evoked IPSC, from 626 ± 76.8 pA in the absence of drug to 405.8 ± 52.5 pA in the presence of drug (P < 0.0001; n = 14; Fig. 4D), confirming that HCN channels enhance PV+ IN inhibitory synaptic transmission.

Primary Importance of the HCN1 Isoform in Regulating Inhibitory Synaptic Transmission.

To examine the specific role of HCN1 in controlling synaptic inhibition, we examined the effect of ZD7288 on IPSCs recorded in an unconditional HCN1 knockout mouse line (Hcn1^tm2Kndl/J, 18). IPSCs were recorded from CA1 PNs in hippocampal slices from the Hcn1^−/− mice in response to electrical stimulation in the SP layer, in the presence of CNQX and APV in the extracellular solution as well as QX-314 in the internal solution.

As expected, application of ZD7288 caused a reduction in IPSC amplitude in wild-type (Hcn1^+/+) littermates (IPSC = 940.5 ± 92.4 pA before and 721.8 ± 92.8 pA after ZD7288; P = 0.0029; n = 8; Fig. 5A). The HCN blocker also produced a significant reduction in the IPSC in heterozygous mice missing one HCN1 allele (Hcn1^+/−; IPSC = 912.4 ± 85.0 pA before and 693.6 ± 76.2 pA after ZD7288; P = 0.0001; n = 9; Fig. 5B). In contrast, application of ZD7288 caused only a small ~5% decrease in the IPSC in homozygous knockout mice (Hcn1^−/−; IPSC = 626.4 ± 65.8 pA before and 597.9 ± 63.9 pA after ZD7288; P = 0.056; n = 14; Fig. 5C). This result demonstrates that the action of the HCN channel blocker to reduce the IPSC was mediated by a specific effect on channels containing the HCN1 subunit.

Fig. 5.

The impact of HCN channel block on IPSCs is occluded in homozygous HCN1 knockout mice. (A) Evoked IPSCs using extracellular electrical stimulation before (black) and after (blue) bath application of ZD7288 in wild-type (Hcn1^+/+) littermates of HCN1 KO mice. (B) Evoked IPSCs using extracellular electrical stimulation before (yellow) and after (blue) bath application of ZD7288 in heterozygous (Hcn1^+/−) mice. (C) Evoked IPSCs using extracellular electrical stimulation before (red) and after (blue) bath application of ZD7288 in homozygous (Hcn1^−/−) mice. (D) Comparison of the evoked IPSC amplitudes of wild-type (black), Hcn1^+/−(yellow), and Hcn1^−/− (red) mice before bath application ZD7288.(* means P < 0.05, ** means P < 0.01, and *** means P < 0.001; statistics in (A–C) with the paired t test after KS normality test and statistics in (D) with one-way ANOVA with the Tukey multiple comparison test after KS normality test.)

If HCN1 does indeed play an important role in controlling the strength of synaptic inhibition, then homozygous loss of this subunit should lead to a reduction in the amplitude of the IPSC relative to that in wild-type animals. Indeed, we found that the amplitude of the IPSC in Hcn1^−/− animals (626.4 ± 65.8 pA; n = 14) was significantly smaller than that in either wild-type littermates (940.5 ± 92.4 pA; n = 8; P = 0.023; one-way ANOVA with the Tukey multiple comparison test) or heterozygotes (912.4 ± 85.0 pA; P = 0.034; n = 9). In contrast, there was no difference in IPSC size between wild-type and heterozygous mice (P = 0.74; Fig. 5D). Thus, one Hcn1 allele appears sufficient to produce a normal-sized IPSC.

An HCN Antagonist Increases the Paired-pulse Ratio Due to Blockade of HCN1.

To examine whether HCN channels may act presynaptically to enhance the probability of GABA release, we measured the effect of ZD7288 on IPSCs evoked by paired-pulse stimulation (50 ms interpulse-interval, Fig. 6A), as the magnitude of the paired-pulse ratio (PPR; second IPSC/first IPSC) is thought to be inversely proportional to the probability of transmitter release. In the absence of ZD7288, the IPSCs showed significant paired-pulse depression, with a PPR of 0.28 ± 0.03. Blockade of HCN channels by application of ZD7288 caused a significant decrease in paired-pulse depression, with the paired-pulse ratio increasing to 0.43 ± 0.02 (n = 18; P < 0.0001 compared to the absence of ZD7288; Fig. 6B), consistent with a reduction in the initial probability of vesicle release. To selectively measure the PPR from PV+ INs, we used a pair of light pulses to activate ChR2-expressing PV+ INs selectively. Again HCN blockade reduced paired-pulse depression (PPR = 0.30 ± 0.03 in the absence and 0.47 ± 0.06 in the presence of ZD7288; n = 8; P = 0.011; Fig. 6 C and D), suggesting that HCN channels enhance the probability of transmitter release from PV+ IN synaptic terminals.

Fig. 6.

HCN channel blockade increases paired-pulse ratio of IPSCs evoked by PV+ IN stimulation. (A) (Left) Schematic of voltage clamp recordings from CA1 PNs with a stimulating electrode in the pyramidal cell layer. (Right) Example trace of IPSCs in response to 20 Hz paired 35 V electrical pulse stimulation before (black) and after (blue) ZD7288 (10 µM) application. Excitatory transmission was blocked with CNQX (25 µM) and APV (50 µM). QX-314 (5 mM) was used in the recording pipette to block postsynaptic HCN channels. (B) Paired-pulse ratio (PPR) before (black) and after (blue) bath application of ZD7288 in response to 20 Hz extracellular electrical stimulation. (C) (Left) Schematic of voltage clamp recordings from CA1 PNs with optogenetic stimulation using blue (470 nm) light pulses (2 ms). (Right) Example trace of IPSCs in response to 20 Hz paired light pulse stimulation (blue triangles) before (black) and after (blue) ZD7288 (10 µM) application. QX-314 (5 mM) was used in the recording pipette to block postsynaptic HCN channels. (D) Paired-pulse ratio (PPR) before (black) and after (blue) bath application of ZD7288 in response to 20 Hz light stimulation of ChR2-expressing PV+ IN axon terminals. (E) PPR before (yellow) and after (blue) bath application of ZD7288 in response to 20 Hz extracellular electrical stimulation in Hcn1^+/− mice. (F) PPR before (red) and after (blue) bath application of ZD7288 in response to 20 Hz extracellular electrical stimulation in Hcn1^−/− mice.

To examine whether HCN1 was responsible for the regulation of PPR, we examined the effect of ZD7288 application in HCN1 knockout mice. HCN channel blockade had no effect on PPR for IPSCs elicited by electrical stimulation in Hcn1^−/− mice (PPR = 0.31 ± 0.03 in the absence and 0.33 ± 0.03 in the presence of ZD7288; P = 0.10; n = 14; Fig. 6E and SI Appendix, Fig. S4), supporting a primary role of HCN1 in regulating presynaptic function. In contrast, HCN blockade was able to increase PPR in mice heterozygous for Hcn1 (PPR= 0.24 ± 0.02 in the absence and 0.30 ± 0.02 in the presence of ZD7288; n = 9; P = 0.012; Fig. 6D), consistent with the conclusion that a single allele of Hcn1 is sufficient to produce a normal level of inhibitory neuron presynaptic function. Overall, these results indicate that presynaptic HCN1 channels in PV+ INs likely enhance the magnitude of the IPSC by enhancing the probability of GABA release.

HCN Channels Promote Activation of Individual Presynaptic Boutons in PV+ INs.

Next, we turned to calcium imaging to determine the mechanism by which HCN1 enhances PV+ IN presynaptic function. We injected a Cre-dependent AAV to express an axon-targeting genetically encoded fluorescent calcium indicator (AAV5-hSynapsin1-FLEX-axon-GCaMP6s) into area CA1 of PV-Cre mice (Fig. 7A). We then used two-photon microscopy to image calcium transients, measured as ΔF/F, within the PV+ IN axon varicosities (presumed presynaptic boutons) within the CA1 SP layer in acute brain slices (Fig. 7B). We first activated the presynaptic axons using three separate trains of electrical stimuli, with each train consisting of 5 pulses applied at 30 Hz, with a 15 s interval between trains. All recordings were performed in the presence of CNQX and APV, first in the absence and then in the presence of ZD7288.

Fig. 7.

Two-photon imaging of PV+ IN axon-specific GcaMP6s in acute brain slices. (A) Schematic representation of viral injection site of axon-GCaMP6s and the location of in vitro extracellular electrical stimulation. (B) Sample GCaMP6s fluorescence images of boutons before (Left) and during (Right) a stimulus response; active boutons are marked by pink circles. (C) ΔF/F of two example boutons in response to 5 pulses at 30-Hz extracellular electrical stimulation in the pyramidal cell layer of CA1 at t = 2 s; one bouton was classified as responding (black) and the other as nonresponding (gray). (D) Average ΔF/F of responding boutons before (black) and 10 min after (blue) start of ZD7288 application. (E) Average ΔF/F of responding boutons shortly (<2 min) after start of imaging (black) and 12 min later (green). (F) ΔF/F of example bouton in response to one single extracellular electrical stimulation in the pyramidal cell layer before (red) and after (blue) ZD7288 bath application. Individual traces represent repetitions of the stimulus. (G) Distribution of peak ΔF/F amplitudes in response to single extracellular electrical stimuli in the pyramidal cell layer (n = 135 stimuli in 15 boutons from 6 slices from 3 animals) before (red) and after (blue) ZD7288 bath application. Lines show best fits with two Gaussian components. (H) ΔF/F of example bouton in response to a 5-pulse train of extracellular electrical stimulation in the pyramidal cell layer at 30 Hz before (black) and after (blue) ZD7288 bath application. Individual traces represent repetitions of the pulse train. (I) Distribution of peak ΔF/F amplitudes in response to a 5-pulse train of extracellular electrical stimulation in the pyramidal cell layer at 30 Hz (n = 486 stimuli in 54 boutons from 8 slices from 4 animals) before (black) and after (blue) ZD7288 bath application. Lines show best fits with two Gaussian components.

The trains of electrical stimulation elicited a measurable transient increase in fluorescence intensity in 20.4 ± 2.6% of GCaMP6s-expressing boutons (range 7 to 42%; n = 1076 boutons from 16 slices from 8 animals) in a field of view (Fig. 7C), presumably boutons whose axons were activated by the stimulating electrode. The fluorescence response of certain boutons fluctuated from train to train, with some trains eliciting a noticeable ΔF/F transient and other trains failing to elicit a response. We defined those boutons in which the ΔF/F response (measured within a 1.6 s window after stimulation) elicited by at least one of the three trains of stimuli was greater than three times the SD of the baseline fluorescence as “responsive boutons.”

Application of ZD7288 caused a significant decrease in the mean peak ΔF/F signal in the responsive boutons to 68% of its value before drug application, from 4.1 ± 0.3 to 2.8 ± 0.2 (P = 0.0004; n = 219 boutons from 16 slices from 8 animals; Fig. 7D). This decrease was not due to a rundown of the response during the second train as there was no significant decrease in the size of the calcium transient when we gave two trains of stimuli over an identical time frame in the absence of ZD7288 (ΔF/F = 2.95 ± 0.38, compared to 2.66 ± 0.37 after 10 min; P = 0.36; n = 123 boutons from 11 slices from 4 animals; Fig. 7E). These results indicate that HCN1 blockade reduces Ca²⁺ influx into the PV+ IN presynaptic boutons.

The effect of ZD7288 could result from a reduction in the size of the Ca²⁺ transient elicited by a single action potential or a decrease in the number of action potentials that invade the presynaptic bouton in response to the brief train of electrical stimulation. We therefore examined the effect of ZD7288 application on the responses of individual boutons to a single electrical stimulus pulse. We applied 9 single pulses, separated from each other by >15 s before and after ZD7288 application. We found that the amplitude of the Ca²⁺ transients elicited by single stimuli fluctuated between successes and failures, in a roughly all-or-none manner (Fig. 7F). The ΔF/F response amplitude histogram had two well-defined peaks, one centered around 0, reflecting failures, and a second broader peak shifted to positive values, corresponding to the successes (Fig. 7G). The histograms were well fit by two Gaussian components (R² > 0.98), allowing us to define the mean ΔF/F values, SD, and the fractional area under each component, with the fractional area of the success component providing the probability of observing a success. Application of ZD7288 caused little change in the ΔF/F value of the peak of the success component (ΔF/F = 1.09 before ZD7288, compared to 1.19 after ZD7288), but decreased the probability of successes by 24.4% (Fig. 7G).

Single-bouton responses to individual 5-pulse stimulus trains produced very few failures. After ZD7288 application, most boutons maintained their high rate of successful responses but showed a decrease in the amplitude of the successes (Fig. 7H). The amplitude histogram was again well fit by two Gaussian components (R² > 0.91, Fig. 7I); however, ZD7288 application did not noticeably change the fraction of successes or failure (71.3% success rate before ZD7288, compared to 69.5% after ZD7288) but instead caused a decrease in the ΔF/F value of the success peak by 30.3% (from 4.17 before ZD7288 to 2.86 after ZD7288, Fig. 7I), which likely reflects a reduction in the probability that a single stimulus within the train elicits a success. Together, these data support the view that blockade of HCN channels reduced the probability that an AP evoked by a single electrical stimulus evoked a calcium response in a bouton.

In principle, the ZD7288 could decrease the probability of a success by either decreasing the probability that an axonal action potential activated a bouton or by decreasing the probability that our stimulating electrode elicited an axonal action potential. However, we found that ZD7288 did not alter the firing or properties of antidromic action potentials elicited by the same stimulation protocol and recorded in the PV+ IN soma (SI Appendix, Fig. S5). Thus, using the same strong suprathreshold current pulse used to measure the effect of ZD7288 on the IPSC or bouton calcium signal, the train of stimuli reliably elicited an antidromic spike in 100% of the stimuli of each train, both in the absence and presence of ZD7288. We thus conclude that the effect of ZD7288 to decrease the probability of observing a calcium signal in the PV+ IN presynaptic boutons likely results from an increased failure of the axonal action potential to propagate into the finer presynaptic processes and/or presynaptic boutons of the PV+ INs.

Discussion

In this study, we found that HCN1 channels are enriched in the presynaptic axonal processes of PV+ interneurons surrounding CA1 PNs of the mouse hippocampus, where they function to enhance inhibitory synaptic transmission by promoting the evoked release of GABA. Prior immunohistochemical labeling experiments, both at the light microscopic and electron microscopic levels, have shown that HCN1 subunits are expressed at very high levels in the axons and axonal terminals of PV+ INs throughout the brain, including the hippocampus and neocortex (26, 42, 43). However, prior to our study, there has been little direct examination of the role of HCN1 in presynaptic function of PV+ INs and none in area CA1 of the hippocampus.

Here, we used a combination of pharmacological, general and cell type–specific genetic deletion, and optogenetic approaches to examine the physiological role of HCN1 in PV+ INs, in both determining somatic membrane properties and in modulating evoked inhibitory synaptic transmission onto CA1 PNs. Our whole-cell patch-clamp recordings from PV+ IN somata confirmed that HCN channels have a minimal impact on somatic passive and active membrane properties (Fig. 3). This is consistent with our triple immunohistochemical labeling, which showed that HCN1 and PV colocalization was restricted to the axon processes and presynaptic terminals. It also confirms previous studies that reported predominantly axonal and presynaptic expression of HCN channels in hippocampal PV+ INs (29–31).

PV+ INs are not a uniform population but have distinct anatomical and functional properties (2, 5, 44, 45). Our findings and those of previous studies suggest that HCN channels have distinct actions on intrinsic membrane properties in distinct populations of PV+ INs. Previous patch-clamp recordings of PV+ INs in the dentate gyrus of the rat hippocampus showed that, whereas these neurons did not have a noticeable voltage sag, a characteristic electrophysiological signature of HCN channels, block of HCN channels with ZD7288 caused hyperpolarization of the resting membrane potential as well as an increase in input resistance (30, 31). In contrast to these results, but in agreement with a study performed in rat CA1 (46), we found that HCN channel block with ZD7288 did not alter resting potential or input resistance of PV+ INs in the CA1 pyramidal layer of mice. A variety of factors may account for the differing results, including the hippocampal subregion examined, species, animal age, and recording temperature. Moreover, although most PV+ INs did not show a prominent somatic voltage sag in response to hyperpolarizing current steps, a minority of PV+ neurons did show a modest ZD7288-sensitive somatic sag. As antidromic APs were recorded in both classes of cells (SI Appendix, Fig. S5), we conclude that these variations in sag were not an artifact of cutting the proximal axons of non-sag cells during slice preparation. Rather, our results suggest the presence of two distinct subclasses of PV+ INs in the CA1 pyramidal cell layer with distinct roles of HCN channels in regulating somatic membrane properties. Whether these cells correspond to known subclasses of CA1 PV+ INs (2), including soma-targeting basket cells, axon initial segment targeting chandelier cells, and dendrite targeting bistratified cells, will require further investigation.

Perhaps the most significant contribution of our study is in elucidating the importance of HCN channels, in general, and HCN1 in particular, in the presynaptic regulation of inhibitory synaptic transmission from PV+ INs to CA1 PNs. By including the intracellular HCN channel blocker QX-314 in our whole cell patch pipette recordings of IPSCs from CA1 PNs, we were able to focus on presynaptic effects of HCN channel blockade with ZD7288 on synaptic inhibition mediated by PV+ INs. Measures of the paired-pulse ratio in the presence of QX-314 and Ca²⁺ imaging from PV+ IN presynaptic boutons both indicate that presynaptic HCN1 channels function to enhance GABA release. In addition, our imaging experiments further indicate that HCN1 does not enhance the presynaptic Ca²⁺ response to a presynaptic action potential, but rather acts to enhance the probability that an action potential evoked by an electrical stimulus evokes a measurable presynaptic calcium response.

Our findings are in broad agreement with prior studies that found that HCN channels regulate axonal function, although there are also important differences. Whereas we found that HCN blockade did not alter action potential firing in response to a brief train of 5 stimuli, HCN channels have been found to be important for maintaining persistent firing in both soma and axons during longer trains of high-frequency stimulation (29, 30, 34, 47, 48), largely through an action to oppose membrane hyperpolarization in response to a train of spikes (29). HCN channels also have been found to play an important role in maintaining AP fidelity. Thus, in dentate gyrus PV+ basket cells, blockade of HCN channels increases the threshold for antidromic action potentials evoked by even a single electrical stimulus (31). In contrast, blockade of HCN channels at the axon initial segment decreases the threshold for spike initiation (37). In our study of PV+ INs in the CA1 region of the hippocampus, we found that HCN1 blockade had little effect somatic on action potential characteristics, including threshold, although it did appear to impair orthodromic spike propagation into the presynaptic processes (Fig. 7).

Studies on the effect of pharmacological inhibition of HCN channels on inhibitory synaptic transmission have also been conflicting. Thus, blockade of HCN channels has been found to decrease the frequency, but not amplitude, of miniature inhibitory postsynaptic currents (mIPSCs) recorded in dentate gyrus granule cells (31). Whereas HCN channel blockade did not affect mIPSC frequency recorded in CA1 PNs, it decreases the frequency of spontaneous, action-potential-dependent IPSCs, with no effect on IPSC amplitude (32). In the cerebellum, HCN channel block decreases both the frequency and amplitude of spontaneous IPSCs (49). In contrast to the above findings, HCN channel block increases mIPSC frequency in globus pallidus (50) and increases the frequency of both mIPSCs and sIPSCs in the medial prefrontal cortex (33), suggesting that HCN channels suppress inhibitory synaptic transmission. An inhibitory effect of HCN channels has also been reported in subsets of excitatory synaptic terminals in the entorhinal cortex (51, 52), where HCN channels inhibit glutamate release by suppressing the activity of low-threshold voltage-gated T-type calcium channels in the presynaptic terminal (53). In the hippocampus, presynaptic HCN channels have been found to also modulate excitatory synaptic transmission, particularly at high stimulus frequencies, and to enhance short-term depression (42, 54, 55).

Such diverse effects of pharmacological inhibition of HCN channels on action potentials and synaptic transmission could reflect a differential role of the distinct subtypes of HCN channel subunits that may be expressed in different neurons or neuronal compartments within the same neuron. However, some of the conflicting results may also arise from potential off-target effects of the HCN channel blocker ZD7288 on transmitter release (40). To guard against such off-target effects we limited our application of ZD7288 to a concentration of 10 µM for no more than 10 min (40). Moreover, by examining mice with a genetic deletion of HCN1, we verified that the effects we observed with this drug are specific to blockade of HCN1. Finally, using the PV-Cre mouse line to selectively delete HCN1 from PV+ INs allowed us to verify the presence of HCN1 in these cells. Future studies in other systems using such genetic-based specific approaches may help resolve some of the discrepant findings.

Our results also have interesting implications for understanding PV+ interneuron function in regulating neural circuits and behavior. These inhibitory neurons are known to play an important role in regulating and balancing excitability in the hippocampal network through feed-forward and feedback inhibition (3, 7–9). Therefore, the changes in the efficacy of inhibitory GABA release we have observed during pharmacological blockade or genetic deletion of HCN1 are likely to impact network function and animal behavior. Changes in inhibitory control of network activity are also likely to induce neuropathology. Indeed, impairments in PV+ IN function have been associated with a number of neuronal and cognitive disorders (56, 57). For example, alterations in PV+ IN firing due to mutations in the SCNA1 gene that encodes the PV+ IN Nav1.1 voltage-gated Na⁺ channel have been shown to underlie the seizures and intellectual disability seen in individuals with Dravet syndrome (14, 58, 59).

Recent human genetic studies have identified more than 40 de novo mutations in the HCN1 gene in patients suffering from a range of early childhood onset seizures, ranging from severe early infantile epileptic encephalopathy (EIEE), to generalized epilepsy with febrile seizures (GEFS+), to milder febrile seizure (FS) phenotypes (16, 17). Such mutations can result in a variety of effects on HCN1 channel function, including shift or loss in voltage dependence and impaired channel expression (60–62). Patients carrying the more severe HCN1 mutations also show cognitive impairments and developmental delay. While recent studies have shown the impact of function-impairing HCN channel mutations in pyramidal cells in the cortex and hippocampus (60, 62) on epileptiform activity, the reduction of inhibition, reported in our study is expected to contribute to excessive excitability in the hippocampal network. Further investigations may yield insights into how alterations in synaptic inhibition from PV+ INs contribute to pathological patterns of neural activity.

Methods

Animals.

Experiments were conducted on male C57BL/6 J mice (Jackson Labs, stock #000664) aged 2 to 10 mo. All animal experiments were conducted in accordance with policies of the NIH Guide for the Care and Use of Laboratory Animals and the Institutional Animal Care and Use Committee (IACUC) of Columbia University. The following commercially available mouse lines were used: Pvalb^tm1(cre)Arbr/J (Jackson Labs, stock #017320); Gt(ROSA)26Sor^{tm14(CAG-tdTomato)Hze}/J (Jackson Labs, stock #007914); Gt(ROSA)26Sor^{tm32(CAG-COP4*H134R/EYFP)Hze}/J (Jackson Labs, stock #024109); C57BL/6-Tg(Pvalb-tdTomato)15Gfng/J (Jackson Labs, stock # 027395, used for additional recordings in PV+ INs in SI Appendix, Fig. S1). and Hcn1^tm2Kndl/J (general HCN1 knockout line, Jackson Labs, stock #016566), Hcn1^tm1Kndl/J (cre-dependent HCN1 knockout line, Jackson Labs, stock # 028299).

Immunohistochemistry.

Animals were perfused with phosphate-buffered saline (PBS) followed by 4% paraformaldehyde in PBS, and brains postfixed overnight at 4 °C. After several washes in PBS, 40 μm coronal slices were cut using a vibratome, and free-floating sections permeabilized in PBS + 0.1% Triton, followed by incubation in blocking solution (PBS + 5% normal donkey serum) for 1 h at room temperature. Primary antibody incubation was carried out in blocking solution overnight at 4 °C. Antibodies used were mouse monoclonal anti-HCN1 (clone N70-28, NeuroMab 75-110, dilution 1:300; Davis, CA); mouse monoclonal anti-HCN2 (clone N71-37, NeuroMab 75-111, dilution 1:250; Davis, CA); mouse monoclonal anti-Syt2 (Znp-1, Developmental Studies Hybridoma Bank, dilution 1:250; Iowa City, IA); rabbit anti-Parvalbumin (Synaptic Systems 195002, dilution 1:700). Secondary antibody incubation was performed in blocking solution for 2 h at room temperature. All secondary antibodies were used at 1:500 dilutions: goat anti-mouse IgG1 cross-adsorbed (Alexa Fluor 488, Life Technologies A21121; Eugene, OR), goat anti-mouse IgG2a cross-adsorbed (Alexa Fluor 647, Life Technologies, A21241; Eugene, OR), and goat anti-rabbit IgG (H + L) cross-adsorbed (Alexa Fluor 568, Life Technologies, A11011; Eugene, OR). Images were acquired on a Zeiss LSM 700 laser scanning confocal microscope with Zen 2012 SP5 FP3 black edition software, using either a Zeiss Fluar 5x/0.25 objective (0.5 zoom, pixel size: 2.5 × 2.5 µm²) or a Zeiss Plan-Apochromat 20X/0.8 objective (1.0 zoom, pixel size: 0.3126 × 0.3126 µm²).

Slice Preparation.

Mice were anesthetized by inhalation of isoflurane (5%) for 7 min, subjected to cardiac perfusion of ice-cold oxygenated artificial cerebrospinal fluid, modified for dissections (d-ACSF; 195 mM sucrose, 10 mM glucose, 10 mM NaCl, 7 mM MgCl₂, 0.5 mM CaCl₂, 25 mM NaHCO₃, 2.5 mM KCl, 1.25 mM NaH₂PO₄, and 2 mM Na-pyruvate, pH 7.2) for 30 s before decapitation according to the procedures approved by the IACUC of Columbia University. The skull was opened, and the brain was removed and immediately transferred into ice-cold carbogenated d-ACSF. The hippocampus was dissected in both hemispheres. Each hippocampus was placed in the groove of an agar block and 400 µm thick hippocampal slices were cut using a vibrating tissue slicer (VT 1200, Leica, Germany) and transferred to a chamber containing a carbogenated mixture of 50% d-ACSF and 50% ACSF (22.5 mM glucose, 125 mM NaCl, 1 mM MgCl₂, 2 mM CaCl₂, 25 mM NaHCO₃, 2.5 mM KCl, 1.25 mM NaH₂PO₄, 3 mM Na-pyruvate, and 1 mM ascorbic acid, pH 7.2) at 35 °C, where they were incubated for 40 to 60 min. Thereafter, slices were held at room temperature (21 °C) until transfer into the recording chamber.

Slice Electrophysiology.

Slices were transferred from the incubation chamber into the recording chamber of an Olympus BX51WI microscope (Olympus, Japan), where they were held in place by a 1 mm grid of nylon strings on a platinum frame. Slices were continually perfused with ACSF at 34 ± 1 °C, maintained by a thermostat-controlled flow-through heater (Warner Instruments, CT). Healthy somas of CA1 PNs were identified visually under 40X (20 × 2) magnification and patched under visual guidance using borosilicate glass pipettes (I.D. 0.75 mm, O.D. 1.5 mm, Sutter Instruments, UK) with a tip resistance of 4 to 5.5 MΩ, connected to a Multiclamp 700B amplifier (Molecular Devices, CA) and filled with intracellular solution, containing (in mM): 135 K-gluconate, 5 KCl, 0.1 EGTA, 10 HEPES, 2 NaCl, 5 MgATP, 0.4 Na₂GTP, 10 Na₂-Phosphocreatin, adjusted to a pH of 7.2 with KOH. Recordings were only accepted if the series resistance after establishing a whole-cell configuration did not exceed 25 MΩ and did not change by more than 20% of the initial value during the course of the experiment. The average series resistance in our experiments was 13.2 ± 0.7 MΩ with the highest one at 18.2 MΩ. Extracellular electrical stimulation was performed by inserting a borosilicate glass pipette (same as patch pipettes) filled with 1 M KCl into SP at least 150 µm (100 µm for antidromic AP experiments) from the patching site, connected to a stimulus isolator and triggered via a TTL pulse through the output channels of the recording software (see below). Generally, 35 V pulses lasting 0.2 ms were used to trigger IPSCs unless otherwise indicated in the Results section or supplementary materials. Light stimulation of ChR2-expressing axon terminals was achieved by using a 470 nm pE-100 excitation light source (CoolLED, UK), connected to the fluorescent light path of the microscope, also triggered via TTL pulse.

Pharmacology: Stock solution of 10 mM ZD7288 (Tocris, UK) was stored at −20 °C and diluted in ACSF to a concentration of 10 µM before bath application to the slice. In some experiments, blockers of AMPA and NMDA receptors, 25 µM CNQX and 50 µM D-APV, respectively, were added to the bath solution. Ivabradine (Cayman Chemicals, MI) was added to the bath solution at a concentration of 30 µM. QX-314 was added to the intracellular solution at 5 mM.

Stereotaxic Virus Injection.

Mice were anesthetized using isoflurane (Covetrus, Portland, ME) and provided analgesics (Carprofen, Zoetis, Troy Hills, NJ). A craniotomy was performed above the target region and a glass pipette was stereotaxically lowered to the desired depth. Injections were performed using a nano-inject II apparatus (Drummond Scientific), with 25 nl of solution delivered every 15 s until a total amount of 200 nl was reached. The pipette was retracted after 5 min. AAV5-hSynapsin1-FLEx-axon-GCaMP6s virus (Addgene, MA) (63) was injected bilaterally, at a titer of 1 × 10¹² vg/ml, with injection coordinates AP −1.55, ML ± 1.05, DV −1.5 (in millimeters with Bregma as reference). One single virus injection was performed per hemisphere, and each hemisphere counted as an independent injection site.

Two-photon Imaging.

Slices were treated in the same way as they were for electrophysiological recordings (see above) and were then transferred into a recording chamber under an Olympus BX61WI microscope at 40x magnification (2 times 20×) connected to a Prairie imaging system, using a Deep See 2-photon laser (Spectra-Physics, CA), exciting GCaMP6s at a frequency of 920 nm. Extracellular electric stimulation was performed by inserting a borosilicate glass pipette filled with 1 M KCl into SP, connected to a custom-built variable power supply and triggered via a TTL pulse through the output channels of the recording software. Electrical stimulation and 2-photon imaging were also synchronized via a TTL pulse. Slices were scanned manually for boutons, excitable by a 30 Hz 5 pulse extracellular stimulus train. A region of interest was set and plane scans were performed of a small area around the selected boutons, such that the image sequence captured at a minimum of 3 frames per second.

Data Acquisition and Analysis.

Electrophysiological recordings were digitized, using a Digidata 1322A A/D interface (Molecular Devices, CA), at a sampling rate of 20 kHz (low pass filtered at 10 kHz) and recorded pClamp 10 software (Molecular Devices, CA). The amplifier setting of the Multiclamp 700B was controlled through Multiclamp Commander (Molecular Devices, CA). After 50 ms baseline recording, 1-s current steps of –350 to +350 pA were applied to the patched cells in increments of 25 pA, after which an additional 1 s of poststep membrane potential was recorded. The trigger time between these episodes was 3 s. Voltage deflections in response to current steps of –50 to +50 pA were used to calculate the input resistance. Initial resting membrane potential (RMP) was obtained immediately upon breaking into the cell and monitored throughout the recording. Voltage sag in response to negative current steps was calculated by dividing the steady-state voltage deflection during the late phase of the –100 pA current step by the peak of the voltage deflection during the same step. Action potential (AP) threshold was determined as the membrane voltage at which the derivative of the voltage trace exceeded 40 mV/ms. AP peak values are given as absolute Vm values. AP peak after hyperpolarization (AHP) was calculated as the amplitude of the first minimum V_m peak after the peak of the AP, using the average of the V_m of 1 ms just before stimulation as a baseline. Data were analyzed using Axograph X software (Axograph Scientific, Australia), MATLAB (Mathworks, MA) as well as Microsoft Excel (Microsoft Corp., WA) or Prism 8 (GraphPad, CA) and visualized in Acrobat Illustrator (Adobe, CA). To determine statistical significance, the paired t test was used for paired data, and the unpaired t test was used for nonpaired data, unless otherwise stated. In the case where t tests were used, data have been verified to be normally distributed by the Kolmogorov–Smirnov normality test.

Confocal microscopy image analysis was performed using ImageJ 1.49v software (NIH). For measuring HCN1 fluorescence intensity values within perisomatic PV+ inhibitory axon terminals in CA1 SP, images of single optical sections displaying Parvalbumin labeling were converted to a binary format using ImageJ (version 1.53a) after manually deleting any labeled cell somas and then setting the threshold to 3.0% (thus capturing pixels within the 3.0% highest intensity window in each image to normalize for potential differences in Parvalbumin staining across samples). Next, a selection was created from the thresholded image in order to define the region of interest (ROI). The selected ROI was then applied to the corresponding colabeled images displaying HCN1 staining. Measurements were obtained for ROI area size and average fluorescence intensity within the ROI for each image (expressed as arbitrary units, a.u.). Area sizes sampled for wild-type and cKO were similar for each of the direct comparison groups. The MCC was calculated as

$$MCC=\frac{{\sum }_i\left({C1}_i\times {C2}_i\right)}{\sqrt{{\sum }_i{C1}_i^2\times {\sum }_i{C2}_i^2}},$$

where C1_i and C2_i are the fluorescence values of each pixel i in the two channels to be evaluated (38, 39).

Two-photon image sequences were validated by eye, using ImageJ software, and further analyzed using custom Matlab scripts. For bouton detection, Images were filtered, using a temporal correlation filter (64) and a 2D median filter. Region of interest (ROI) centers for each bouton were determined as local 2D peaks. ROI size was set to a circle around the peak with a radius of 3 pixels (the average size of a bouton at the magnification, used) around the local peak. ΔF/F within these ROIs was then calculated from the original image sequence.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.