Presynaptic HCN1 channels regulate Cav3.2 activity and neurotransmission at select cortical synapses.

The hyperpolarization-activated cyclic nucleotide-gated (HCN) channels are subthreshold, voltage-gated ion channels that are highly expressed in hippocampal and cortical pyramidal cell dendrites, where they are important for regulating synaptic potential integration and plasticity. We found that HCN1 subunits are also localized to the active zone of mature asymmetric synaptic terminals targeting mouse entorhinal cortical layer III pyramidal neurons. HCN channels inhibited glutamate synaptic release by suppressing the activity of low-threshold voltage-gated T-type (Ca(V)3.2) Ca²(+) channels. Consistent with this, electron microscopy revealed colocalization of presynaptic HCN1 and Ca(V)3.2 subunit. This represents a previously unknown mechanism by which HCN channels regulate synaptic strength and thereby neural information processing and network excitability.

Hyperpolarization-activated Cyclic Nucleotide-gated (HCN) channels are low threshold, voltage-gated ion channels with very unusual biophysical properties1. These channels are open at potentials more negative to −50 mV and are important for regulating neuronal resting membrane potential (RMP). In addition, the channels are permeable to Na+ and K+ and form an inward current at rest, thereby depolarizing RMP. Interestingly, in the hippocampus and cortex, they are highly localized to pyramidal cell dendrites2, 3. Here, by regulating the RMP and in this way the biophysical properties of other ion channels as well as by modulating the membrane resistance, they influence excitatory post-synaptic potential (EPSP) kinetics and integration4-7. These effects are likely to contribute substantially to synaptic plasticity and thus, processes such as information storage8-12. Indeed a reduction in HCN channel expression has been associated with enhanced learning8, 9.

Emerging evidence, though, suggests that in addition to their dendritic localization, HCN1 channels may be present on certain axons and synaptic terminals in the hippocampus, cortex and other regions of the brain3, 13-15. Indeed, pre-synaptic HCN channel current, Ih, has been indicated to affect inhibitory synaptic transmission in the rodent cerebellum, hippocampus and basal ganglia14, 16, 17, though the mechanism by which this occurs is unknown. However, in contrast to invertebrate neurons where Ih has been shown to influence excitatory synaptic release18, 19, an explicit role for Ih in regulating excitatory basal synaptic transmission in mammals has not yet been revealed. This is despite HCN subunit expression being detected in some immature mammalian glutamatergic pathways13 and functional Ih recorded from pre-synaptic immature terminals in the calyx of Held20. In this study, we demonstrate that HCN1 channels are present at the active zone of mature glutamatergic cortical synaptic terminals establishing contact with entorhinal cortical (EC) layer III cells where they decrease neurotransmission by restricting Ca2+ entry via pre-synaptic T (CaV3.2) -type Ca2+ channels. Since changes in synaptic strength are pivotal to induction and maintenance of synaptic plasticity21 and given that EC layer III neurons are likely to be involved in memory formation and spatial navigation22, 23, our results indicate that this is a novel mechanism by which HCN1 subunits may contribute to such physiological processes.

Results

HCN1 channels modulate glutamatergic synaptic release onto EC layer III cells exclusively

Synaptic release plays a central role in regulating individual neuronal as well as neural network excitability. Thus, elucidating the factors regulating synaptic transmission is critical for understanding how neural circuits function. In an earlier study4, we had observed that a loss of Ih enhances synaptic transmission onto mature EC layer III pyramidal neurons, the principal cell type within layer III24. This is likely to have been, at least partly, due to enhanced feed forward excitation resulting from pyramidal cell hyperexcitability. However, since HCN subunits may be present at certain synaptic terminals13-17, 20, it is possible that this effect could be due to changes in synaptic release per se by altered pre-synaptic HCN channel function.

To test this, we initially recorded non-evoked miniature excitatory post-synaptic currents (mEPSC) in the presence of tetrodotoxin (TTX, 1 μM) and GABA receptor inhibitors at a fixed potential of −70 mV from mouse EC layer III neurons (Fig 1a) using the whole-cell voltage clamp technique (as described previously14, 16, 17, see Methods). To exclude the effects of post-synaptic HCN channels12 and to reduce errors due to space-clamp25, 15 μM ZD7288 was incorporated in the patch pipette in all our experiments, unless otherwise stated. Intracellular ZD7288 enhanced the outward holding current, indicating that the post-synaptic neuronal resting membrane potential (RMP) was hyperpolarized (Supplementary Table 1). The mEPSC rise and decay times were also slowed down as would be expected from blocking post-synaptic HCN channels (see Refs.4, 5, 24). Including ZD7288 in the patch pipette, though, had no effect on mEPSC frequency indicating this did not affect synaptic release per se (Supplementary Table 1). Further, substituting KMeSO4 with CsMeSO4 in the intracellular solution did not alter either somatic or dendritic mEPSC frequency, amplitude and kinetics (Supplementary Table 1), indicating that the errors due to space-clamp were not reduced further by blocking post-synaptic K+ channels25.

Figure 1

Pharmacological block or deletion of HCN channels enhances mEPSC frequency in EC layer III pyramids

a Morphology of typical mouse EC layer III pyramidal neuron. The scale represents 20 μm. b and c Example mEPSC recordings from wildtype (Wt) and HCN1−/− soma and dendrites before and after 15 min bath application of the HCN channel blocker, ZD7288 (ZD; 15 μM). The cumulative probability curves for each recording are displayed above the trace and on the right respectively. The average normalized mEPSCs obtained from the recordings are also presented. The scale shown in the upper panel of b and c applies to all traces within those panels. Graphs depicting the mean (filled squares) and standard error of Wt and HCN1−/− soma and dendritic mEPSC frequency in the absence and presence of ZD7288 are also shown. Open squares illustrate the mEPSC frequency from individual experiments. On the far left of each panel are the amplitude histograms for the total number of mEPSCs obtained from Wt soma (n=7), Wt dendrites (n=5), HCN1−/− soma (n=6) and HCN1−/− dendrites (n=5) with and without ZD7288. Superimposed on the histograms are Gaussian fits to demonstrate the peak amplitude of mEPSCs.

To determine if functional HCN channels are present pre-synaptically, we bath-applied ZD7288 (15 μM) for 15 min only to exclude any potential non-specific effects of the compound on synaptic transmission26. This concentration was deemed specific as it enhances wildtype dendritic excitability whilst having little effect on HCN1−/− dendrites (see Ref. 4). Intriguingly, external application of ZD7288 enhanced wildtype somatic mEPSC frequency by 113.8 ± 19.1% (n=7, p < 0.01; Fig 1b). The outward holding current as well as mEPSC amplitude, rise-times and decay time constants were not further affected (Fig 1b). Further, 15 min external application of another Ih blocker, zatebradine27, 28 (10 μM) also enhanced mEPSC frequency in a comparable manner without affecting amplitude or kinetics (Supplementary Fig 1a). In the above experiments, ZD7288 was included in the internal patch pipette solution. However, even when this was not present in the internal solution, external application of ZD7288 still amplified wildtype mEPSC frequency by 108.3 ± 14.8% (n=6). In these experiments, the decay time constant increased significantly from 7.9 ± 0.6 (n=6) to 10.0 ± 0.5 (n=6, p < 0.05), as would be predicted by blocking post-synaptic HCN channels.

Since dendrites receive the majority of excitatory synaptic inputs29, we also recorded mEPSCs from EC layer III apical dendrites at a distance of 120 -150 μm from the soma (Fig 1c). The frequency, but not amplitude or kinetics, of dendritic mEPSCs was increased by application of ZD7288 by 69.7 ± 5.3% (n=5; Fig 1c). This effect was not significantly different (p=0.1) to that observed at the soma. The above results indicate that synapses targeting EC layer III neurons contain pre-synaptic Ih.

HCN1 subunits are likely to be predominantly expressed in the EC3, 4. We, therefore, compared HCN1−/− and wildtype mEPSCs recorded from layer III neurons. Since ZD7288 was present in the internal pipette solution, wildtype and HCN1−/− outward holding current and mEPSC kinetics were similar (Supplementary Table 1). Interestingly, the average somatic, but not dendritic, HCN1−/− mEPSC amplitudes were smaller than in wildtypes (Supplementary Table 1). Importantly, though, HCN1−/− mEPSC frequency was significantly greater than in wildtypes (Fig 1b, c, Supplementary Table 1). Moreover, omitting ZD7288 from the internal pipette solution did not affect HCN1−/− mEPSC frequency or kinetics (Supplementary Table 1). Additionally, 15 min bath application of ZD7288 or zatebradine had no effect on somatic or dendritic HCN1−/− mEPSC frequency, amplitude or shape (Fig 1b, c; Supplementary Fig 1b). These findings confirm that the enhanced wildtype mEPSC frequency caused by pharmacological blockers of Ih is likely to be due to HCN channel block and not non-specific effects of the compounds26. These results also indicate that pre-synaptic HCN1 channels are likely to be involved in regulating excitatory spontaneous release from synapses localised to EC layer III neurons.

We then recorded miniature inhibitory synaptic currents (mIPSCs) to determine if HCN channels were also present pre-synaptically on interneurons. In contrast to mEPSCs, somatic EC layer III mIPSCs were unaffected by 15 min bath-application of ZD7288 (Supplementary Fig 2a). Moreover, mIPSC frequency, amplitude and kinetics recorded from wildtype and HCN1−/− neurons did not differ (Supplementary Fig 2b). These findings suggest that in the EC, HCN1 channels selectively regulate glutamatergic synaptic release.

We next asked whether HCN1 channels were present at excitatory synapses targeting other principal EC neurons, particularly layer II stellate (Fig 2a) and layer V pyramids (Fig 2c). Intriguingly, application of ZD7288 had no effect on wildtype mEPSC frequency, amplitudes or kinetics recorded from the soma of these neurons (Fig 2b, d). In addition, there were no differences between HCN1−/− and wildtype layer II and layer V mEPSCs (Fig 2). Since layer III pyramid dendrites are present in layer II and mEPSC frequency onto these is altered by pharmacological block of Ih or HCN1 deletion (Fig 1c), these results suggest that HCN1 channels are predominantly located at glutamatergic synaptic terminals contacting EC layer III pyramids.

Figure 2

HCN channel block or deletion has no effect on mEPSCs onto EC layer II and layer V neurons

a and c Typical morphologies of mouse EC layer II stellate cell and layer V neuron. The scale bar in a and b represent 20 μm and 50 μm respectively. b and d Representative mEPSC recordings from wildtype (Wt) and HCN−/−layer II stellate and layer V pyramid soma before and after application of ZD7288 (ZD; 15 μM). The values above the traces indicate that outward holding current at −70 mV. The average mEPSC obtained from the recordings are displayed below the traces. The cumulative probability curves for each of the traces are also illustrated on the right of the recordings. The scale shown in the upper trace for b and d applies to all traces within b and d respectively. Graphs depicting the individual (open squares) and average (closed squares) mEPSC frequencies recorded from wildtype and HCN−/− stellate and layer V pyramid neurons before and after application of ZD7288 are shown at the far right in each panel.

HCN1 subunits are present at the active zone of selective terminals

To determine if HCN1 subunits are located on EC synaptic terminals, we carried out electron microscopy studies using a pre-embedding immunogold method (see Methods). We counted 2752 particles in the medial EC from sections obtained from 3 wildtype mice. 72.2 ± 0.8% of HCN1 immunogold particles were observed along the extra-synaptic plasma membrane of dendritic shafts and spines (Fig 3a, e), confirming that HCN channels are largely present in EC dendrites. Interestingly, 27.8 ± 1.0% of immunogold particles were located at pre-synaptic sites (Fig 3b, e).

Figure 3

HCN1 subunits are localized to active zones of synaptic terminals

a, b Immunoparticles for HCN1 along the extrasynaptic plasma membrane (e.g. arrows) of dendritic spines (s) and shafts (Den) establishing synapses with excitatory terminals (b) in EC layer III. Pre-synaptic HCN1 subunits in excitatory terminals were mainly located in the active zone (b). c Although at lower proportion, immunoparticles for HCN1 were also observed in putative inhibitory axon terminals (it) (e.g. arrowheads), recognised by the shape of synaptic vesicles and the lack of a prominent postsynaptic density in the postsynaptic element, establishing synaptic contact with somata. d Immunoreactivity for HCN1 in EC layer III of HCN1 null mice as revealed at the electron microscopy level. No labelling could be detected in the null mice. ax, axon terminal; s, dendritic spine. e Quantitative analysis on the percentage of immunoparticles (n=2572) for HCN1 at post- and pre-synaptic sites in EC layer III, showed that 72% were localised at postsynaptic sites and 28% at presynaptic sites. Cyt, cytoplasm; N, nucleus. Scale bars: a, c and d, 0.5 μm; b, 0.2 μm.

We further analysed the distribution of pre-synaptic particles. Of 153 immunogold particles that were detected on synaptic terminals, the majority ((129/153, 84%) of HCN1 immunogold particles were localized at terminals establishing asymmetrical excitatory synapses with pyramid dendritic spines (Fig 3b). Of these, 71% (92/129) were present in the active zone and 29% (37/129) in pre-synaptic terminal membranes. In contrast, only 16% (24/153 particles in 8 synapses) were located in the axonal plasma membranes establishing symmetrical synapses (Fig 3c). Interestingly, very few immunogold particles were observed in layer V (data not shown). No staining occurred if either the primary or secondary antibodies were omitted. Further, no signal was detected in HCN1−/− sections obtained from 4 mice under either light (Supplementary Fig 3) or electron microscopy (Fig 3d). Thus, these results together with the electrophysiological results suggest that HCN1 subunits are present at excitatory pre-synaptic terminals synapsing onto EC layer III neurons. These data are also consistent with our findings that ZD7288 only alters spontaneous excitatory synaptic transmission, and not inhibitory synaptic transmission, at layer III neurons (Fig 1, Supplementary Fig 1 and Supplementary Fig 2).

Enhanced Ca2+ dependence of HCN1−/− mEPSCs

How do pre-synaptic HCN1 subunits regulate synaptic transmission at synapses targeting EC layer III neurons? Since acute treatment of ZD7288 in wildtype neurons results in enhanced mEPSC frequency, it suggests that pre-synaptic HCN1 subunits do not affect synaptogenesis. Nevertheless, we compared the number of synapses in superficial EC layers (layers I-III) between wildtypes and HCN1−/− mice using electron microscopy. From sections obtained from 3 wildtype and 4 HCN1−/− mice, no obvious differences in the number of synapses were detected (Supplementary Fig 4), indicating that HCN1 subunits on synaptic terminals are likely to reduce synaptic transmission by directly affecting release probability.

One possible mechanism by which pre-synaptic HCN1 channels may regulate synaptic transmission is by limiting Ca2+ influx, as has been suggested to occur in dendrites7. Recent studies suggest that spontaneous as well as action potential-driven synaptic release may be dependent upon Ca2+ entry via voltage-gated Ca2+ channels (VGCC)30-34. We, therefore, tested if the mEPSCs recorded from layer III soma were Ca2+-dependent. Replacement of Ca2+ in the extracellular solution with an equal concentration of Mg2+ reversibly reduced wildtype mEPSC frequency by 56.3 ± 5.1% (n=9, p < 0.01) within 5 min (Fig 4a, b). Interestingly, substitution of Ca2+ with Mg2+ lowered HCN1−/− mEPSC frequency by 70.1 ± 3.6% (n=9, p < 0.01, Fig 4a) in a reversible manner (Fig 4a, b). This decrease in HCN1−/− mEPSC frequency was significantly (p = 0.04) greater than that in wildtypes. Neither the amplitudes nor the time constants of wildtype or HCN1−/− mEPSCs were affected by replacing external Ca2+ with Mg2+. The outward holding current, though, was reduced significantly to a similar extent in wildtype and HCN1−/− neurones (average holding current in the presence of Ca2+ in wildtypes and HCN1−/− cells = 30.0 ± 7.9 pA (n=9) and 39.4 ± 20.0 pA (n=9) respectively; holding current without Ca2+ in wildtypes and HCN1−/− pyramids = 5.0 ± 7.1 pA (n=9, p < 0.05) and 10.0 ± 20.3 (n=9, p < 0.05)). Given the larger Ca2+ dependence of HCN1−/− mEPSCs, these results imply that pre-synaptic HCN1 channels regulate synaptic release by constraining Ca2+ influx into terminals. Indeed, in the absence of extracellular Ca2+, bath application of ZD7288 no longer increased mEPSC frequency in wildtypes (Fig 4c).

Figure 4

Ca2+ dependence of EC layer III mEPSCs

a Traces from wildtype (Wt) and HCN1 null neurons respectively in the absence, presence and washout of Ca2+ - free external solution. The outward holding current values are displayed above the traces. The mean and standard error of 9 individual experiments are shown in the bar graphs on the right. b Time course of the effects of applying Ca2+ free solution. The average frequency per minute during 5 Wt and 4 HCN−/− recordings was calculated and normalised. c Example recordings from Wt neurons under control conditions, following application of Ca2+-free solution and Ca2+-free solution containing ZD7288. The outward holding current values at −70 mV are indicated above the traces. The graph on the right depicts the average frequency from 4 experiments. The scale bar shown in (a) applies to all traces within the figure. Significance (p < 0.05) is indicated by an asterisk.

HCN1 channels inhibit mEPSC frequency by limiting T-type Ca2+ channel activity

We next asked if Ca2+ entry through specific VGCCs contributes to generation of spontaneous mEPSCs and whether pre-synaptic HCN1 channels limit the activity of particular VGCCs. To investigate this, we first examined the effects of various pharmacological inhibitors of VGCCs on wildtype mEPSC frequency. The P/Q Ca2+ channel blocker, ω -agatoxin IVA35, 36 (100 nM) and the N-type Ca2+ channel inhibitor, ω-conotoxin GVIA35, 36 (100 nM) reduced wildtype mEPSC frequency by 47.3 ± 3.1% (n=11, Supplementary Fig 5) and 22.2 ± 4.7% (n=9, Supplementary Fig 5) respectively, without affecting amplitude or kinetics (Supplementary Table 2). The effects occurred within 10 min of bath application. ω -agatoxin IVA and ω-conotoxin GVIA co-application decreased mEPSC frequency by 48.6 ± 5.5% (n=6, Supplementary Fig 5). In contrast, neither the L-type Ca2+ channel inhibitor, nifedipine35, 36 (2 μM), the R-type Ca2+ channel antagonist, SNX-48237 (200 nM) nor the selective T-type Ca2+ channel blockers, mibefradil36 (10 μM), TTA-P2 (1 μM)38, TTA-A239, 40 (500 nM) had any effect on wildtype mEPSCs (Supplementary Fig 5, Supplementary Table 2). Since wildtype mEPSC frequency inhibition in the presence of ω -agatoxin IVA and ω-conotoxin GVIA together is comparable to that resulting from removal of extracellular Ca2+ (Fig 4a, b), it suggests that in these neurones, Ca2+- influx via P/Q- and N-type generates Ca2+-dependent mEPSCs.

Next we tested whether the ZD7288 induced enhancement in wildtype mEPSC frequency (Fig 1) could be occluded by pre-treatment with either ω -agatoxin IVA (100 nM) or ω-conotoxin GVIA (100 nM). Co-application of ZD7288 with either ω-agatoxin IVA or ω-conotoxin GVIA or both increased mEPSC frequency in a comparable manner to that produced by ZD7288 alone (Fig 5c). Moreover, in the presence of either nifedipine or SNX-482, ZD7288 also augmented mEPSC frequency (Fig 5c). However, intriguingly, the effects of ZD7288 on mEPSC frequency were prevented by pre-treatment with three different inhibitors of the low voltage-activated T-type Ca2+ channels: mibefradil (10 μM) as well as the more potent and selective inhibitors TTA-P238-40 (1μM) and TTA-A239, 40 (500 nM) (Fig 5a, c). The latter two have been shown to have no effects on other VGCC in either heterologous expression systems39, 40 or neurons38. These findings suggest that inhibition of HCN1 channels enhances excitatory synaptic transmission by altering T-type Ca2+ channel activity.

Figure 5

T-type Ca2+ channel blockers reduce the increase in mEPSC frequency caused by pharmacological block of Ih or deletion of HCN1 subunits

a and b Recordings obtained in wildtype (Wt) and HCN−/− neurons in the absence and presence of the potent and selective T-type Ca2+ channel blocker, TTA-A2 (500 nM). In Wt neurons, ZD7288 was additionally applied after TTA-A2 neurons too. The outward holding current at a holding potential of −70 mV are indicated above the traces. The cumulative probability curves for all the recordings are displayed on the right in each panel. The scale bar shown in a and b applies to all traces within a and b respectively. c Bar graph to demonstrate the effects of co-application of ZD7288 (15 μM) and Ca2+ channel blockers for T- (TTA-A2 (500 nM), TTA-P2 (1 μM) and mibefradil (10 μM; Mib)), P/Q – (ω-agatoxin IVA (100 nM)), N- (ω -conotoxin GVIA (100 nM)), L- (nifedipine (2 mM)) and R- (SNX-482 (200 nM)) type Ca2+ channels. The effects of treatment with ZD7288 alone are also shown. d Graph to illustrate the effects of the T-type (TTA-A2, TTA-P2 and mibefradil (Mib)) and R-type (SNX-482) Ca2+ channel blockers on mEPSC frequency in HCN1 null EC layer III neurons. In c and d, the numbers of observations for each treatment are indicated above the individual bars. Asterisks indicate significance at p < 0.05.

To investigate this further, we explored the effects of T-type Ca2+ channel blockers on HCN1−/− mEPSC frequency. In contrast to the results obtained in wildtypes (Supplementary Fig 5), mibefradil (10 μM), TTA-A2 (500 nM) and TTA-P2 (1 μM) significantly reduced HCN1−/− somatic mEPSC frequency by 35.7 ± 4.8% (n=6, p < 0.05), 49.0 ± 4.8% (n=7, p < 0.05) and 50.4 ± 4.4% (n=6. P < 0.05) respectively (Fig 5b, d), whilst having no effect on amplitudes or time constants (Supplementary Table 3). As mibefradil may inhibit R-type Ca2+ channels too41, we tested the effects of SNX-482 (200 nM) on HCN1−/− mEPSCs. 15 min application of SNX-482 had no effect on these (Fig 5d). To verify that T-type Ca2+ channels expression is not modified in HCN1−/− mice, we did Western blot analysis using CaV3.1 and CaV3.2 antibodies. We were unable to detect any CaV3.1 protein expression in wildtype or HCN1−/− EC. However, bands corresponding to the size of CaV3.2 proteins expressed in tsA-201 cells were present in wildtype and HCN1−/− EC samples (Supplementary Fig 6a, 7a). Semi-quantitative analysis indicated that there was no difference in expression of this between wildtype and HCN1−/− tissue (Supplementary Fig 6a, 7a). These findings re-affirm that a decrease in pre-synaptic HCN channel function leads to increased contribution of T-type Ca2+ channel activity in modulating basal glutamatergic synaptic release.

Co-localisation of HCN1 and CaV3.2 subunits on EC synaptic terminals

Since T-type Ca2+ channel inhibitors prevent the effects of ZD7288, it raises the possibility that these and HCN1 channels co-exist on the same pre-synaptic terminals. To investigate this, we took advantage of CaV3.1 and CaV3.2 antibodies to examine the localization of CaV3 and HCN1 subunits in wildtype EC. For these particular experiments, the anti-HCN1 antibody was visualized using an immunoperoxidase reaction and the anti-Cav3 antibody by a silver-intensified immunogold reaction (see Methods). Consistent with the Western blot data, CaV3.1 labelling could not be detected. However, remarkably, CaV3.2 immunogold particles were present on asymmetrical synaptic terminals (Fig 6a) in sections obtained from 3 wildtype mice. Approximately 84% of these were present on axons pre-labelled with HCN1. HCN1 and CaV3.2 co-labelling was noticeable on spines and dendrites too (Supplementary Fig 8). No CaV3.2 signal was identified on sections obtained from 3 Ca mice under either electron (Fig 6a) or light microscopy (Supplementary Fig 9). Moreover, no specific labelling was observed if the primary antibody was omitted or replaced with normal serum. In addition, the selective location of signals in structures labelled with only one or the other of the signalling products within the same section, as well as having side by side double-labelled structures, showed that our procedures did not produce false-positive double-labelling results. These results, thus, indicate that CaV3.2 and HCN1 subunits co-localise on synaptic terminals.

Figure 6

CaV3.2 co-localise with HCN1 at EC synaptic terminals and cause the increase in excitatory synaptic transmission following HCN1 channel inhibition

a Two typical electron micrographs showing immunoreactivity for HCN1 (peroxidase reaction product) and Cav3.2 (immuoparticles) in wildtype (Wt) sections as detected using pre-embedding double-labelling methods at EM level. Immunogold particles for CaV3.2 were absent in CaV3.2 null (Ca) sections. Axon terminals and dendritic spines are indicated by b and s respectively. Scale bars: 0.2 μm. b Example recordings showing the effects of 15 min bath application of ZD7288 (ZD, 15 μM) on mEPSCs recorded from CaV3.2 null (Ca) EC layer III neurons and Wt neurons. The vertical and horizontal scale bars on the traces represent 10 pA and 2 s respectively. In values above the recordings indicate the outward holding currents at −70 mV. The cumulative probability curves are shown on the right. The average normalised mEPSC traces obtained from the traces before and after ZD7288 application are shown below the recordings. The graphs on the far right depict the individual (open squares) and mean (filled squares) frequency of mEPSCs obtained from CaV3.2 Wt (n=5) and null (n=6) in the absence and presence of ZD7288. Significance at p < 0.05 is indicated by an asterisk.

mEPSCs onto Ca neurons were unaffected by HCN channel inhibition

Since T-type Ca2+ channels and HCN channels can co-exist on synaptic terminals (Fig 6a) and T-type Ca2+ channel blockers prevented the increase in mEPSC frequency by the HCN channel inhibitor, ZD7288 (Fig 5a, c), we hypothesized that HCN channel blockers may not affect excitatory synaptic transmission in Ca tissue. We, therefore, recorded mEPSCs from EC layer III neurons present in slices obtained from wildtype and Ca mice42. As in previous experiments, ZD7288 (15 μM) was included in the patch pipette to inhibit post-synaptic HCN channels. Wildtype and Ca mEPSC frequency, amplitude and kinetics were very similar (Supplementary Table 4), confirming that CaV3.2 channels do not modulate basal excitatory synaptic transmission. Interestingly, 15 min application of ZD7288 (15 μM) had little effect on Ca mEPSCs (Fig 6b) but enhanced wildtype mEPSC frequency (Fig 6b). HCN1 protein levels were similar CaV3.2 wildtype and null EC (Supplementary Fig 6b, 7b), indicating that the lack of effect of ZD7288 on Ca mice is not likely to be due to altered HCN1 expression. These results further support the notion that block of pre-synaptic HCN channels increases excitatory synaptic transmission by enhancing Ca2+ entry via pre-synaptic CaV3.2 Ca2+ channels.

Hyperpolarization enhances pre-synaptic T-type Ca2+ channel activation

Next, we wondered how a loss in HCN channel function could augment T-type Ca2+ channel activity. T-type Ca2+ channels are low threshold VGCCs that typically activate at potentials more positive to −75 mV43, 44. Interestingly, native as well as heterogously-expressed CaV3.1- 3.3 channels inactivate at potentials below −55 mV−60 mV43, 44. Hence, there is a small “window” during which these channels are active. Since T-type Ca2+ channel blockers had no effect on basal wildtype mEPSC frequency (Supplementary Fig 5) and the basal frequency between Ca and wildtype was similar (Supplementary Table 4), despite CaV3.2 immunogold particles being present on wildtype synaptic terminals (Fig 6b), it implies that these channels are inactive at rest. A reduction in HCN channel activity hyperpolarizes the membrane1, 4 and thus one possibility is that RMP hyperpolarization relieves T-type Ca2+ channel inactivation. To investigate this hypothesis, we lowered external K+ concentration from 2.5 mM to 1.75 mM, which, using the Nernst equation, is estimated to hyperpolarize the RMP by ~10 mV. Indeed, reducing external K+ concentration resulted in EC layer III neuronal hyperpolarization by 7.3 ± 0.9 mV (n=3). Further, switching from 2.5 mM to 1.75 mM K+ significantly enhanced wildtype mEPSC frequency within 10 min by 91.6 ± 14.1% (n=6, p < 0.01, Fig 7a, c). This was restored to near control levels by adding mibefradil41 (10 μM; Fig 7a). Interestingly, changing the external K+ from 2.5 mM to 1.75 mM in HCN1−/− slices, where the RMP is expected to be approximately −80 mV4, had little effect on mEPSC frequency (% increase = −1.5 ± 3.6%, n=5, Fig 7b), presumably as T-type Ca2+ channel activity was not further altered. The outward holding current, mEPSC amplitude or kinetics were not affected by reducing external K+ concentration in wildtype or HCN1−/− neurons (Fig 7a, b). These results provide additional support for the notion that T-type Ca2+ channels are present on pre-synaptic terminals and can contribute to synaptic release under certain conditions.

Figure 7

Hyperpolarization enhances mEPSC frequency by activating T-type Ca2+ channels

a and b Representative traces and their cumulative probability curves illustrating the effects of lowering the external K+ concentration from 2.5 mM to 1.75 mM in wildtype (Wt) and HCN1 null neurons. The additional effects of mibefradil (Mib; 10 μM) on Wt mEPSCs in 1.75 mM K+ are also presented. The outward holding currents at −70 mV are shown above the traces. An overlay of the average normalised EPSCs from Wt and HCN1−/− neurons under control conditions (black), with 1.75 mM K+ (light grey) and in the presence of 1.75 mM K+ and mibefradil (dark grey) are also displayed. The average mEPSC frequency (filled squares) as well as the individual frequency values obtained from each experiment (open squares) under these conditions in Wt and HCN1 null neurons are shown on the far right in each panel. The scale bar shown in a and b applies to all traces within a and b respectively. Asterisks represent significance at p < 0.05. c Graph showing the time course of the effects of reducing the K+ concentration from 2.5 mM to 1.75 mM in Wt and HCN1−/− neurons. The average mEPSC frequency for each minute of the recording before and after 15 min bath application of 1.75 mM KCl is shown.

HCN channels modulates evoked release onto EC layer III neurons

We also asked if HCN channels affect evoked release too. To determine this, we stimulated EC layer III distal dendrites to evoke pairs of EPSCs at 20 Hz in the absence and presence of 15 μM ZD7288 (see Methods). ZD7288 (15 μM) was included in the intracellular solution to block post-synaptic HCN channels. The amplitude of the first EPSC was adjusted between 50 to 100 pA (average amplitude = 71.3 ± 9.8 pA). Single EPSCs were also elicited in between pairs of EPSCs. Subtraction of these from the pairs of EPSCs was then used to elucidate the amplitudes of the first and second EPSCs within the pairs (Fig 8). Paired pulse ratio (PPR) was calculated as the amplitude of the second EPSC divided by the amplitude of the first. Under control conditions, the average PPR was 0.9 ± 0.1 (n=5, Fig 8). 15 min bath application of ZD7288 reduced PPR significantly to 0.7 ± 0.1 (n=5, p < 0.01, Fig 8) and increased the amplitude of the first evoked EPSC by 16.7 ± 5.8% (n=6, p < 0.01). In addition, square of the coefficient of variation (CV2) decreased significantly to 0.7 ± 0.1 (p < 0.01). In contrast, ZD7288 (15 μM) had no effect on PPR in HCN1−/− neurons (PPR ratio before and after ZD7288 = 1.1 ± 0.2 and 1.1 ± 0.1, n=6, p = 0.92). The lowered PPR and CV2 in wildtypes indicate that synaptic release was enhanced by pharmacological block of HCN channels. These findings, therefore, demonstrate that pre-synaptic HCN channels affect multiple modes of synaptic transmission.

Figure 8

Effects of pharmacologically blocking HCN channels on evoked synaptic release

a Representative traces in the absence and presence of ZD7288 (ZD, 15 μM) obtained at the soma are shown above the graph. In between each paired pulse, a single stimulus was used to elicit an EPSC. By subtracting this EPSC from the paired EPSCs, the amplitude and shape of the individual EPSCs were obtained. The insets show the overlayed black (second EPSC) and grey (first EPSC) subtracted traces before and after application of ZD7288. The vertical and horizontal scale bars for the paired pulse traces as well as the subtracted traces represent 40 pA and 50 ms respectively. b Graph showing the effects of bath applying ZD7288 on the paired pulse ratio (PPR) in EC layer III neurons. Pairs of EPSCs were obtained every minute by stimulating the distal dendrites of EC layer III neurons.

Discussion

The significant findings of this study are that functional HCN1 channels are present on glutamatergic synaptic terminals synapsing onto EC layer III neurons where they reduce spontaneous and evoked neurotransmitter release. These effects are at least partly due to altered availability of low-threshold, voltage-gated T-type Ca2+ channel. In support, electron microscopy showed that HCN1 and CaV3.2 subunits co-exist on EC pre-synaptic terminals. As EC layer III neurons have been suggested to play significant roles in processes such as spatial navigation22 and learning and memory23 and HCN channel function has been associated with modification of neural network rhythms and memory formation8, the dynamic modulation of synaptic inputs onto these neurons by HCN and T-type Ca2+ channels may contribute to the regulation and modulation of these physiological states.

Although HCN1 and HCN2 subunits are both expressed in the EC3, our results suggest that HCN1 subunits are likely to predominantly contribute to the formation of pre-synaptic HCN channels in glutamatergic synapses in the EC. Indeed, HCN1 subunit immunoreactivity was detected on the active zone of wildtype excitatory synaptic terminals in the adult EC (Fig 3). Spontaneous glutamatergic synaptic transmission was also augmented in HCN1−/− slices and was not further affected by the non-selective HCN channel blockers, ZD7288 (Fig 1) and zatebradine (Supplementary Fig 1). These agents, on the other hand, enhanced wildtype mEPSC frequency (Fig 1, Supplementary Fig 1). Further, ZD7288 reduced the wildtype EPSC paired pulse ratio (Fig 8). These findings indicate that pharmacological inhibition of HCN channels in normal tissue enhances glutamtergic synaptic transmission. Since ZD7288 has also been suggested to depress synaptic transmission independently of affecting HCN channels26, it is possible that the observed effects of ZD7288 might be underestimated (Fig 1, Fig 8). Further, the effects of zatebradine are dependent upon HCN channels being open27, 28 and therefore, the increase in mEPSC frequency in the presence of this compound may also be undervalued (Supplementary Fig 1). This might account for the increase in spontaneous release in HCN1−/− slices being much greater than that obtained with application of pharmacological blockers on wildtype slices (Fig 1, Supplementary Fig 1). Notwithstanding, our results provide strong support for the involvement of HCN1 channels in modulating EC excitatory synaptic transmission.

Remarkably, our results suggest that the presence of pre-synaptic HCN channels depolarizes the RMP, thereby restricting Ca2+ entry via pre-synaptic T-type Ca2+ channel activity and preventing glutamate release (Figs 4 –7). This suggests that pre-synaptic HCN channels may provide an additional regulatory mechanism for controlling synaptic release at these synapses. Hence, by affecting T-type Ca2+ channel activity, modulation of HCN channel activity by protein kinases or phosphatases will alter synaptic transmission at these synapses. Indeed, EC synaptic terminals may have functional metabotropic glutamate receptors45, activation of which could potentially reduce HCN channel activity by altering kinase activity and induce plasticity10.

Altered T-type Ca2+ channel activity by HCN channels has been suggested to occur in hippocampal dendrites too7 and it is, therefore, tempting to speculate that this might be a universal mechanism by which these channels affect neuronal excitability. Fascinatingly, though, in invertebrate neurons, Ih has been shown to increase excitatory synaptic release in a “non-ionic” way via a pathway involving actin depolymerization18. It is possible that this might occur in some mammalian interneurons too16, 17. This therefore suggests that HCN channels in synaptic terminals may have multiple roles: a “non-ionic” purpose whereby they modulate release by processes that involve changes in the cytoskeleton and a more conventional function in which they influence synaptic transmission by modifying the terminal RMP as well as the membrane resistivity, which has been shown to occur post-synaptically in dendrites4, 24. Which role they adopt may depend on their location within a synapse as well as their proximity to other proteins and ion channels.

Interestingly, the expression of HCN channel subunits and their effects on synaptic release were restricted to a subset of EC glutamatergic synaptic terminals (Figs 1-3). This suggests that there may be factors that control expression of HCN channels in specific synapses and not others. These synapses may belong to a particular cell type. Indeed, synapses targeting EC layer III neurons can originate from a number of sources, including other EC layer III neurons, layer V and pre-subicular neurons. Since HCN subunits have not been detected in the temporoammonic (TA) pathway, EC layer III pyramid axons, or terminals synapsing onto CA1 dendrites (which belong to the TA pathway)13, it is likely that HCN channels are present on those synaptic terminals originating from EC layer V or pre-subiculum. Further work is, therefore, required to resolve whether HCN subunit expressing synapses belong to a particular cell type and whether these synaptic terminals contain or lack proteins that regulate HCN channel expression. In addition, it remains to be determined if other cortical synapses express functional pre-synaptic HCN and T-type Ca2+ channels. Nonetheless, our results show that functional HCN channels can be present pre-synaptically as well as on pyramid cell dendrites in the mature cortex. Thus there may be multiple means by which HCN channels can affect neural network excitability and ultimately processes such as learning and memory.