The nitric oxide (NO)/cyclic guanosine monophosphate (cGMP) signaling cascade participates in the modulation of synaptic transmission. The effects of NO are mediated by the NO-sensitive cGMP-forming guanylyl cyclases (NO-GCs), which exist in 2 isoforms with indistinguishable regulatory properties. The lack of long-term potentiation (LTP) in knock-out (KO) mice deficient in either one of the NO-GC isoforms indicates the contribution of both NO-GCs to LTP. Recently, we showed that the NO-GC1 isoform is located presynaptically in glutamatergic neurons and increases the glutamate release via hyperpolarization-activated cyclic nucleotide (HCN)-gated channels in the hippocampus. Electrophysiological analysis of hippocampal CA1 neurons in whole-cell recordings revealed a reduction of HCN currents and a hyperpolarizing shift of the activation curve in the NO-GC2 KOs associated with reduced resting membrane potentials. These features were mimicked in wild-type (WT) neurons with an NO-GC inhibitor. Analysis of glutamate receptors revealed a cGMP-dependent reduction of NMDA receptor currents in the NO-GC2 KO mice, which was mimicked in WT by HCN channel inhibition. Lowering extracellular Mg2+ increased NMDA receptor currents in the NO-GC2 KO and allowed the induction of LTP that was absent at physiological Mg2+. In sum, our data indicate that postsynaptic cGMP increases the N-methyl-d-aspartate (NMDA) receptor current by gating HCN channels and thereby is required for LTP.
The signaling molecule nitric oxide (NO) has been proposed to play a role in synaptic transmission and in long-term potentiation (LTP; Garthwaite 2008). In the neuronal system, NO is formed physiologically by the endothelial and neuronal NO synthases (eNOS and nNOS) and has been proposed to be generated in the course of N-methyl-D-aspartate (NMDA) receptor activation and as a retrograde messenger to increase neurotransmitter release (Snyder and Bredt 1991). Yet, NO's precise physiological functions and the underlying molecular mechanisms are unknown.
The NO effects are mediated by the NO receptor guanylyl cyclases (NO-GCs), 2 highly homologous, cyclic guanosine monophosphate (cGMP)-generating enzymes (Koesling et al. 2004). Both NO-GCs are heterodimers consisting of a common β-subunit dimerized to different α subunits (NO-GC1, α1β1 and NO-GC2, α2β1; Russwurm et al. 1998). The NO-GC isoforms do not differ in enzymatic or regulatory properties (Russwurm et al. 1998) and are expressed in similar amounts in the central nervous system (Mergia et al. 2003). Yet, the NO-GC2 isoform interacts with the PDZ, protein domain found in post synaptic density protein 95 (PSD-95), Drosophila disc large tumor suppressor 1, and zonula occludens-1 protein, domains of PSD-95 (Russwurm et al. 2001), which results in a special subcellular localization of this isoform.
For cGMP, different receptor molecules exist such as kinases (Hofmann et al. 2006), phosphodiesterases (Bender and Beavo 2006), and ion channels (Biel et al. 2009). Although in general considered to display a higher affinity for 3′,5′-cyclic adonosine monophosphate (cAMP), one class of ion channels, the hyperpolarization-activated cyclic nucleotide (HCN)-gated channels are potentially also modulated by cGMP. HCN1, the major HCN channel isoform in the hippocampal CA1 region, had been described as poorly sensitive to cyclic nucleotides. However, Chen et al. (2001) reported a very high affinity of HCN1 for cAMP (K1/2 of 0.06 µM) and proposed that HCN1 is already modulated by basal cAMP levels in intact Xenopus oocytes used for expression. In a recent study, HCN1's cyclic nucleotide-binding domains were shown to exhibit comparable affinities toward cAMP and cGMP (Lolicato et al. 2011). The HCN channels conduct a current (Ih), which contributes to multiple membrane properties governing cellular excitability (Ludwig et al. 2003; Robinson and Siegelbaum 2003; Frère et al. 2004; Craven and Zagotta 2006; Lewis et al. 2010).
To address the impact of 2 similarly regulated NO-GCs, we generated mice deficient for either one of the NO-GC isoforms (Mergia et al. 2006). Surprisingly, LTP measured in the hippocampus (Taqatqeh et al. 2009) and visual cortex (Haghikia et al. 2007) was abolished in both knock-out (KO) strains, indicating that both NO-GC isoforms are required for LTP. Recently, we showed that cGMP formed presynaptically by NO-GC1 via the increase of HCN channel currents facilitates glutamate release in the hippocampus (Neitz et al. 2011).
Here, we used NO-GC2 KO mice to study the postsynaptic function of NO-GC2 in whole-cell recordings in the CA1 region of hippocampal slices. In accordance with the hypothesis of cGMP gating HCN channels postsynaptically, voltage of half-maximal activation (V1/2) was shifted by 8 mV to more negative values in the NO-GC2 KO than that in the wild-type (WT), and Ih was reduced in NO-GC2 KOs and reconstituted to WT levels by the application of cGMP. In addition, inhibition of cGMP formation in WT caused a leftward shift of V1/2 and a reduction of Ih, demonstrating that cGMP is permanently formed and able to affect the activation curve of HCN similar as has been reported for cAMP. Analysis of glutamate receptors revealed greatly reduced NMDA receptor currents in NO-GC2 KOs, which increased almost to WT levels by intracellularly applied cGMP. HCN channel inhibition (ZD7288 or DK-AH269) reduced NMDA receptor currents to NO-GC2 KO-like levels in WT mice, indicating that the cGMP effect on NMDA receptors was mediated by HCN channels. Lowering extracellular Mg2+, which increased the NMDA receptor current, allowed the induction of LTP in the NO-GC2 KO. In summary, we propose that NO/cGMP acts as a feed-forward loop increasing NMDA receptor currents.
Materials and Methods
For electrophysiological experiments, NO-GC1KO and WT mice (24–28 days) generated and analyzed as described previously (Mergia et al. 2006) were anesthetized by halothane or isoflurane inhalation. Subsequently, brains were rapidly removed, and 350-µm thick horizontal slices were cut from the ventral part of the hippocampus using a vibratome (VT 1000S; Leica, Germany) in ice-cold, standard artificial cerebrospinal fluid (ACSF) containing (mM): 125 NaCl, 2.5 KCl, 25 NaHCO3, 2 CaCl2, 1.5 MgCl2, 1.25 NaH2PO4, and 25 d-glucose, pH 7.4 (gassed with 95% O2 and 5% CO2). Individual slices were used after 60 min of recovery at room temperature in standard ACSF solution, and recordings were conducted at 34 ± 2°C.
A modified ACSF was used for the measurements of Ih and NMDA receptor currents with the following composition (mM): 115 NaCl, 2 KCl, 25 NaHCO3, 2 CaCl2, 1.5 MgCl2, 1.25 NaH2PO4, 25 d-glucose, 10 tetraethylammoniumchloride (TEA), and 2 4-aminopyridine.
Recordings were obtained with borosilicate glass capillaries (5–6 MΩ; GB 150F-8P, Science Products, Germany) filled with a potassium gluconate-based solution containing (mM): 135 potassium gluconate, 20 KCl, 2 MgCl2, 10 N-(2-hydroxyethyl)piperazine-N'-(2-ethanesulfonic acid), and 10 ethylene glycol-bis(2-aminoethylether)-N,N,N′,N′-tetraacetic acid. For recordings of NMDA receptor-mediated currents, 5-mM N-(2,6-dimethylphenylcarbamoylmethyl)triethylammonium bromide (QX-314) was added to the potassium gluconate solution, thereby avoiding the generation of action potentials.
Hippocampal pyramidal neurons were visualized with differential interference contrast optics and a ×40 objective (Olympus, Japan), and somatic whole-cell recordings were measured using an Axopatch 200 B amplifier (Axon Instruments, Union City, CA, USA). Signals were filtered at 2 kHz and digitized at 5 kHz using a Digidata 1200 system with pClamp10 software (Molecular Devices, Sunnyvale, CA, USA). Access resistance was constantly monitored and ranged from 12 to 15 MΩ, whereas seal resistance was >1 GΩ. Cells in which one of the parameters changed >20% were discarded. Voltage-clamp recordings were not corrected for liquid junction potentials and series resistance.
Recordings of Ih were carried out in the modified ACSF solution as mentioned above in the presence of tetrodotoxin (0.5 µM), 6,7-dinitroquinoxaline-2,3(1H,4H)-dione (DNQX) (20 µM), d(–)-2-amino-5-phosphonopentanoic acid (d-AP5) (25 µM), picrotoxin (PTX) (50 µM), and (2S)-3-[[(1s)-1-(3,4-Dichlorophenyl)ethyl]amino-2-hydroxypropyl](phenylmethyl)phosphinic acid hydrochloride (CGP-55845) (2 µM). The design of the protocol assessing Ih is described in Kanyshkova et al. (2009) and shown in Figure 1A as a scheme. Briefly, hyperpolarizing voltage step commands from −40 to −130 mV (ΔV = −10 mV) were applied, followed by a constant step to −110 mV. To increase recording stability and taking into account the increasing fast activation kinetics, pulse length was shortened by Δt = 500 ms per increasing hyperpolarization step (3.5-s pulse length at −130 mV).
The steady-state amplitude of Ih was isolated from the instantaneous current by fitting the current traces with a single-exponential function and plotted against the step potential (Kilb and Luhmann 2000). The density of Ih was calculated by dividing the Ih amplitude at Imax (−130 mV) by the membrane capacitance determined by a hyperpolarizing step (−30 pA) in the current-clamp mode during whole-cell recordings. The time constants of activation were analyzed at a potential of −130 mV using a single-exponential fit.
For quantitative analysis of the Ih activation, p(V), tail current amplitudes were estimated 50–60 ms after stepping to the constant potential −110 mV from a variable voltage step using the following equation:
with Imax being the tail current amplitude for the voltage step from −130 to −110 mV and Imin for the voltage step from −40 to −110 mV. The resulting p(V) data were further fitted by a Boltzmann equation:
In experiments carried out with cGMP (50 µM) added to the recording pipette, the steady-state activation, amplitude of Ih, and NMDA receptor current were estimated 20 min after obtaining whole-cell configuration. Resting membrane potential (RMP) was determined under current-clamp conditions in standard ACSF. The input resistance (Ri) was calculated from a membrane hyperpolarization step (−30 pA, 1 s duration) of neurons at a holding potential of −80 mV.
Spontaneous action potentials (APs) were recorded in the presence of high K+-containing ACSF solution (5 mM KCl), with the cells current-clamped to −60 ± 3 mV. Only neurons that started firing APs at these membrane potentials were used for further analysis.
2-Amino-3-(3-hydroxy-5-methyl-isoxazol-4-yl)propanoic acid (AMPA) receptor-dominated spontaneous excitatory postsynaptic currents (sEPSCs) were recorded at −80 mV in the presence of 50 µM PTX, 25 µM D-AP5, and 2 µM CGP-55845 for at least 5 min. The events were analyzed offline, and the threshold for event detection was set at 2.5 times root-mean-square noise level.
EPSCs were evoked by a glass electrode (5–7 MΩ), placed 300–400 µm away from recorded neurons on Schaffer Collateral fibers in the stratum radiatum in the presence of 50 µM PTX and 2 µM CGP-55845. The initial AMPA receptor-mediated EPSCs were recorded at −80 mV, and stimulus intensity (0.05 ms, 160–200 µA, 0.1 Hz) was set for amplitudes of 250–300 pA. The stimulation was kept constant in all groups and during all recordings of isolated NMDA receptor-mediated EPSCs, achieved by the subsequent bath application of 20 µM DNQX. I–V relationships of NMDA receptor-mediated EPSCs were studied at different holding potentials in steps of +20 mV in the range of −80 to −20 mV. Peak amplitudes obtained at each holding potential are displayed as mean amplitudes from 6 evoked responses. In some experiments, the Mg2+ concentration of the ACSF solution was reduced to 0.15 mM. NMDA receptor EPSCs were confirmed by the bath application of D-AP5 (25 µM).
In experiments carried out with 8-bromoguanosine-3′, 5′-cyclic monophosphate (8-Br-cGMP) (100 µM), the properties of NMDA receptor-mediated EPSCs were determined 15 min after bath application.
The kinetics and charge transfer of AMPA- and NMDA receptor EPSCs were determined at −80 and −20 mV, respectively. The decay time constant was calculated for AMPA receptor EPSCs by fitting single traces with a single-exponential function and for NMDA receptor EPSCs by using a double-exponential function.
LTP was tested in recordings of evoked excitatory postsynaptic potentials (EPSPs) under whole-cell patch-clamp conditions in the current-clamp mode. Extracellular presynaptic stimulations were performed by placing a glass electrode (resistance 3 MΩ) within the Schaffer collateral fibers, which project onto the recorded neuron. The stimulation intensity was kept constant throughout each experiment. Baseline EPSPs were induced at a frequency of 0.033 Hz and recorded for the duration of 10 min. The amplitude of these signals reached 5.5 ± 0.5 mV (n = 34) and was not different between the groups. Next, an electrical theta-burst stimulation (TBS) was applied to the Schaffer collaterals. The TBS protocol consisted of 5 synaptic trains (at 20-s intervals) of 5 bursts (at 5 Hz) each providing 4 stimuli at 100 Hz. Each TBS was paired with an intracellular depolarization delayed by 5 ms (1100 pA, 45 ms duration), which induced spikes in all recorded neurons. The TBS was followed by 40 min of EPSP recordings at baseline stimulation. The input resistance was monitored throughout the experiment by a negative current injection of 30 pA through the recording pipette after each test stimulation. Only cells with an input resistance changing <20% during the experiment were used for further analysis. Changes in long-term synaptic plasticity were analyzed by averaging the EPSP amplitudes evoked by the last 15 stimuli in each recorded neuron. These data were compared between slices from the WT and NO-GC2 KO ± 8-Br-cGMP groups.
Picrotoxin (50 µM), DNQX (20 µM), D-AP5 (25 µM), 4-AP (2 mM), TEA (20 mM), CGP-55845 (2 µM), 8-Br-cGMP (100 µM, bath-applied 15 min beforehand), and ZD7288 (30 µM, bath-applied 15 min beforehand) were obtained from Tocris Biozol (Eching, Germany). 1H-[1,2,4]oxadiazolo[4,3-a]quinoxalin-1-one (ODQ) (10 µM, bath-applied 15 min beforehand), cGMP (50 µM, applied intracellularly), QX-314 (5 mM), and DK-AH269 (10 µM, bath-applied 15 min beforehand) are purchased from Sigma-Aldrich (Munich, Germany).
Summary data are expressed as mean ± SEM and analyzed using analysis of variance (ANOVA) with repeated measurements, paired or unpaired Student's t-test. n refers to the number of cells measured. Cells from at least 3 different animals of the mentioned genotype were used in each experiment.
In a recent study on presynaptic properties of hippocampal CA1 neurons, we found that cGMP formed presynaptically by NO-GC1 facilitates glutamate release via HCN channels (Neitz et al. 2011). Here, we analyzed a possible postsynaptic role of NO-GC2 in the CA1 region of the hippocampus using whole-cell recordings in NO-GC2 KO mice.
cGMP is Gating HCN Channels Postsynaptically
As presynaptically cGMP was found to act via HCN channels, we studied a possible impact of cGMP on postsynaptic hyperpolarization-activated current (Ih) by whole-cell voltage-clamp recordings in WT and NO-GC2 KO pyramidal neurons. Ih was activated from a holding potential of −40 mV by applying hyperpolarizing voltage steps of increasing amplitudes (▵V = −10 mV) and shortened duration (▵t = −500 ms per step) followed by a constant step to −110 mV (Fig. 1A). The steady-state current amplitudes (Ih) determined as described under Materials and Methods are plotted versus the step potential and shown in Figure 1B. Steady-state current amplitudes obtained in NO-GC2 KO were significantly reduced compared with WT neurons with the maximal current amplitudes at −130 mV being reduced by 40% (−365 ± 11 pA, n = 22 and −217 ± 12 pA, n = 34, respectively; Fig. 1B). The results are compatible with the assumption that cGMP present in the postsynaptic WT neurons increases Ih, whereas in the NO-GC2 KOs, the cGMP responsible for gating HCN channels is missing. To test this hypothesis, we used the inhibitor of the cGMP-forming NO-GC, ODQ (10 µM, bath-applied 15 min beforehand), in WT neurons and found a reduction of Ih (maximal amplitudes −365 ± 11 pA in WT, n = 22 vs. −210 ± 13 pA in WT + ODQ, n = 18; Fig. 1C), which is comparable with the current measured in the NO-GC2 KO neurons. Conversely, the application of cGMP (50 µM, applied intracellular) increased Ih in the NO-GC2 KOs to WT-like level (−217 ± 12 pA in NO-GC2 KO, n = 34 vs. −383 ± 15 pA in NO-GC2 KO + cGMP, n = 19; Fig. 1D). The reduction of Ih in NO-GC2 KOs as well as in ODQ-treated WT slices argues for a permanent effect of cGMP on Ih.
In accordance, the density of Ih at −130 mV was higher in the WT than in NO-GC2 KO neurons (3.4 ± 0.11, n = 22 vs. 1.9 ± 0.13 pA/pF, n = 34), decreased by ODQ in WT (1.8 ± 0.11 pA/pF, n = 17), and conversely increased by cGMP in the NO-GC2 KO neurons (3.6 ± 0.22 pA/pF, n = 19; Fig. 1E). In addition to the reduced Ih in the NO-GC2 KO, the time constant of activation determined at a potential of −130 mV was higher in NO-GC2 KO than in WT (WT, τ = 98 ± 8 ms, n = 21 vs. NO-GC2 KO, τ = 136 ± 10 ms, n = 24), restored to the WT-like level by cGMP in the NO-GC2 KO (NO-GC2KO + cGMP, τ = 97 ± 5 ms, n = 15), and decreased to NO-GC2 KO-like values with ODQ in WT (WT + ODQ, τ = 137 ± 12 ms, n = 12; Fig. 1F).
Next, we determined activation curves using tail current analysis performed as described under Material and Methods. Mean steady-state activation curves obtained by plotting normalized tail current amplitudes against step potential are shown in Figure 2B–D. The activation curve determined in the NO-GC2 KO neurons is clearly leftwards shifted to more negative potentials compared with the WT ones. In accordance, the V1/2 value is more negative in the NO-GC2 KO (−95 ± 1.2 mV, n = 31) than in WT neurons (−87 ± 0.9 mV, n = 23). As cyclic nucleotides have been shown to alter voltage dependence of Ih activation, these results are in favor of the assumption that the channel is gated by cGMP in WT and is partly cyclic nucleotide-free in the NO-GC2 KO. Accordingly, the NO-GC inhibitor, ODQ (10 µM, bath-applied 15 min beforehand), shifted V1/2 to NO-GC2 KO-like values in WT (−97 ± 1.4 mV, n = 17) and did not cause any shift in the NO-GC2 KO neuron (−96 ± 1.1 mV, n = 12). Conversely, cGMP (50 µM, applied intracellularly 20 min beforehand) restored WT-like V1/2 in the NO-GC2 KO neurons (−87 ± 1.4 mV, n = 19).
We tried to increase the effect of endogenous cGMP on Ih in WT with the unspecific phosphodiesterase inhibitor, 3-isobutyl-1-methylxanthine (IBMX) (100 µM; bath-applied 15 min beforehand). In the presence of IBMX, the activation curve of Ih in WT was slightly, although not significantly, shifted to the right with a slightly depolarized V1/2 (−84 ± 0.6 in WT + IBMX, n = 11 vs. −87 ± 0.9 mV in WT cells, n = 23; P = 0.07; Supplementary Fig. 1). Additional experiments with other PDE inhibitors are required to increase the effect of endogenous cGMP in WT.
Electrophysiological Alterations in the NO-GC2 are Mimicked by HCN Inhibition in WT
The HCN channel-mediated current contributes to the RMP in the hippocampal neurons. In accordance with the proposal of cGMP gating HCN channels, the RMP of the NO-GC2 KO pyramidal neurons (Fig. 3A) was more negative than that of WT neurons (−74 ± 1 mV in NO-GC2 KO, n = 14 vs. −67 ± 1 mV in WT, n = 26) and reconstituted to the WT level by the intracellular application of cGMP (−67 ± 2 mV, n = 8). Conversely, the application of the NO-GC inhibitor, ODQ (10 µM, bath-applied 15 min beforehand), and the HCN channel inhibitor, ZD7288 (30 µM, bath-applied 15 min beforehand), shifted the RMP in WT neurons to more negative NO-GC2 KO-like values (−73 ± 2 mV in WT + ODQ, n = 15; −78 ± 2 mV in WT + ZD7288, n = 14). ZD7288's maximal effect is reached after 10 min as shown by the time-course of ZD7288 action on Ih in Supplementary Figure 2. These findings support the concept of a tonic effect of cGMP on HCN channel conductance. ZD7288 had no effect on NO-GC2 KO neurons (n = 13). All together, these results indicate that NO-GC2 is permanently activated in the WT neurons at least under our recording conditions, and the resulting increase in cGMP allows a depolarizing current through HCN channels.
In addition to the alteration of the RMP, Ih inhibition has been reported to increase input resistance. Indeed, the input resistance was approximately 20% higher in the NO-GC2 KO than that in WT neurons upon a hyperpolarizing current pulse (−30 pA, 1 s duration; 117 ± 6 MΩ in NO-GC2 KO, n = 26 vs. 95 ± 5 MΩ in WT, n = 17; Fig. 3B). Further, input resistance was decreased by 8-Br-cGMP in the NO-GC2 KO (n = 9) and increased by inhibition of NO-GC by ODQ in WT (n = 12), which is consistent with a reduced constitutive activation of Ih in the NO-GC2 KOs. Accordingly, inhibition of Ih with ZD7288 (30 µM, bath-applied 15 min beforehand) increased input resistance in WT slices (n = 16) and did not affect NOGC2 KO slices (n = 14, data not shown).
Besides a more hyperpolarized RMP and an increased membrane resistance, HCN channel inhibition either pharmacologically by ZD7288 or by genetic deletion alters neuronal excitability. Therefore, we measured spontaneous AP firing in pyramidal neurons current-clamped at −60 mV. As seen in Figure 4, AP firing was increased in the NO-GC2 KO compared with WT neurons (1.8 ± 0.2 Hz, n = 12 vs. 1 ± 0.1 Hz, n = 18). ZD7288 treatment (30 µM, bath-applied 15 min beforehand) increased the frequency of AP firing in WT neurons (1.8 ± 0.2 Hz, n = 18) almost to NO-GC2 KO levels while not causing any increase in NO-GC2 KO neurons (1.8 ± 0.2 Hz, n = 12). The amplitude of APs did not differ between neurons of WT and NO-GC2 KO mice (Fig. 4C). In sum, the neuronal properties measured in the NO-GC2 KO slices can be mimicked upon Ih inhibition in WT slices, which is compatible with the assumption that the function of the HCN channel is altered by cGMP.
NMDA Receptor Currents are Increased by cGMP
So far, NO-GC2 KO hippocampal neurons exhibited several features that had been described for the absence of Ih before. In a recent study, we found a lack of hippocampal LTP in the NO-GC2 KO mice (Taqatqeh et al. 2009). To explain the absence of LTP in the NO-GC2 KOs, we studied glutamate receptor properties. Neither spontaneous nor evoked AMPA receptor-mediated EPSCs revealed any differences between WT and NO-GC2 KO pyramidal cells in terms of kinetic properties as frequency, peak amplitude, rise time, and time constants of decay as well as charge transfer at −80 mV (Fig. 5). Next, we studied the NMDA receptor currents and normalized the stimulation conditions by applying a stimulation intensity that evoked AMPA receptor-mediated EPSCs of similar amplitude. While the stimulation conditions were kept constant, the AMPA receptor antagonist, DNQX, (20 µM) was bath-applied to pharmacologically isolate NMDA receptor-mediated EPSCs, which were measured at different holding potentials (−80 to −20 mV, Fig. 6). In contrast to the unaltered AMPA receptor currents, NO-GC2 KO pyramidal cells displayed a pronounced reduction of NMDA receptor currents compared with WT ones at the different membrane potentials (e.g. −34 ± 5 pA for NO-GC2 KO, n = 17 vs. −69 ± 7 pA for WT at −60 mV, n = 18; Fig. 6A).
Besides the reduction of NMDA receptor currents, the respective receptor kinetics did not significantly differ between the NO-GC2 KO and WT neurons (Supplementary Fig. 3A–E). In addition, expression of NMDA receptors did not differ between WT and NO-GC2 KO as the NR2A and 2B subunit content in WT and NO-GC2 KO hippocampi were found to be comparable (Supplementary Fig. 3F). To ensure that the reduced NMDA receptor currents in the NO-GC2 KOs are due to the absence of cGMP, WT slices were treated with the NO-GC inhibitor, ODQ (10 µM, bath-applied 15 min beforehand). Indeed, in the presence of ODQ, NMDA receptor currents of WT neurons were reduced and indistinguishable from those in NO-GC2 KOs (−38 ± 4 pA in WT + ODQ, n = 6 vs. −34 ± 5 pA at −60 mV in NO-GC2 KOs, n = 17; Fig. 6B vs. A). These findings indicate that cGMP has an impact on NMDA receptor-mediated currents. The assumption was confirmed by the treatment of NO-GC2 KO slices with 8-Br-cGMP (100 µM, bath-applied 15 min beforehand), which significantly increased NMDA receptor amplitudes even above those of WT slices (−107 ± 20 pA in NO-GC2 KO + 8-Br-cGMP, n = 6 vs. −69 ± 7 pA in WT at −60 mV, n = 18; Fig. 6C vs. A). To exclude that bath-applied 8-Br-cGMP solely acts presynaptically, we also tested the effect of intracellularly applied cGMP in NO-GC2 KO neurons (n = 8) and found a significant increase in the NMDA receptor current (Fig. 6D). The increase in the NMDA receptor current by intracellularly applied cGMP is reduced compared with the one induced by bath-applied 8-Br-cGMP, a finding compatible with an additional presynaptic effect of 8-Br-cGMP and/or enhanced degradation of cGMP compared with 8-Br-cGMP. Nevertheless, the finding with intracellularly applied cGMP supports the notion of an impact of cGMP on NMDA receptor current.
NMDA Receptor Currents are Influenced by Ih
To find out whether the observed cGMP effect on NMDA receptor currents results from cGMP gating HCN channels, we tested the HCN channel inhibitor, ZD7288 (30 µM, bath-applied 15 min beforehand; Fig. 7A–D). In the presence of the HCN channel inhibitor, NMDA receptor currents of WT slices were reduced similarly as those in ODQ-treated WT or NO-GC2 KO slices (−29 ± 7 pA in WT + ZD7288 at −60 mV, n = 5). Intracellular application of ZD7288 to WT neurons (n = 5) yielded comparable effects as the bath-applied compound (data not shown). The results show that inhibition of HCN channels has a comparable effect as genetic deletion of NO-GC2 or its pharmacological inhibition by ODQ in WT. Hence, the results are compatible with the assumption that cGMP via gating of HCN channels increases NMDA receptor currents. As HCN channels are highly enriched in dendrites, Ih conceivably exerts local effects on NMDA receptor currents. The hypothesis is underlined by the finding that the cGMP-induced increase of NMDA receptor currents in the NO-GC2 KO was reversed by ZD7288 (30 µM, bath-applied 15 min beforehand; see Fig. 7D), whereas in the untreated NO-GC2 KO, the HCN inhibitor had no effect (−28 ± 4 pA, n = 5 vs −30 ± 4 pA, n = 5 in NO-GC2 KO in the absence or presence of ZD7288; Fig. 7B). To reinsure the involvement of the HCN channels within the observed cGMP effect, another HCN inhibitor, DK-AH269 (10 µM, bath-applied 15 min beforehand; Fig. 7E,F), with a different mode of action was used (Raes et al. 1998). Yet again, the HCN inhibitor reduced NMDA receptor-mediated currents of WT slices (−101 ± 10 pA, n = 6 vs. −69 ± 4 pA, n = 6 at −60 mV) to those obtained in the NO-GC2 KO slices (−34 ± 5 pA at −60 mV, n = 17), whereas the NMDA receptor currents in the NO-GC2 KO slices remained unchanged (−33 ± 18 pA at −60 mV, n = 7). Thus, the above results suggest that cGMP acting via HCN channels is increasing NMDA receptor-mediated currents.
Mg2+ Sensitivity of NMDA Receptor Currents is Preserved in NO-GC2 KO Mice
The I–V relation of NMDA receptors greatly depends on the Mg2+ block, which is removed in a voltage-dependent manner. NMDA receptor-mediated currents recorded at a lower extracellular Mg2+ concentration (0.15 instead of the regular 1.5 mM) were increased in WT neurons (Fig. 8) as expected with up to 6-fold larger peak amplitudes at different holding potentials (e.g. −405 ± 57 pA at −60 mV, n = 7). Also in NO-GC2 KO neurons, recordings revealed an up to 6-fold increase of peak amplitudes (−212 ± 48 pA at −60 mV, n = 10) at low extracellular Mg2+, but currents remained overall lower than in WT (Fig. 8B). This finding demonstrates that the physiologically relevant Mg2+ block of the NMDA receptors by extracellular Mg2+ still occurs in the NO-GC2 KO. Again, the HCN channel blocker, ZD7288 (30 µM, bath-applied 15 min beforehand), decreased NMDA receptor currents in WT neurons (−405 ± 57 pA for WT, n = 7 vs. −157 ± 27 pA for WT + ZD7288 at −60 mV, n = 7; Fig. 8C) and had no effect in NO-GC2 KO neurons (−212 ± 48 pA for NO-GC2 KO, n = 10 vs. −208 ± 28 pA for NO-GC2 KO + ZD7288 at −60 mV, n = 9; Fig. 8D), showing that Ih has an impact on NMDA receptor currents even at low extracellular Mg2+ conditions.
LTP can be Restored in the NO-GC2 KOs at Low Extracellular Mg2+
Is the reduction of the NMDA receptor-mediated currents in NO-GC2 KO slices responsible for the lack of LTP shown in our previous field-potential recordings (Taqatqeh et al. 2009)? To investigate this issue, we performed LTP experiments in whole-cell recordings (Fig. 9). In accordance with our previous field-potential recordings, LTP was almost abolished in NO-GC2 KO slices (NO-GC2 KO 110 ± 15%, n = 8 vs. WT 176 ± 11% relative EPSP amplitude, n = 7; Fig. 9C,E). Interestingly in NO-GC2 KO slices, LTP could be reconstituted with a cGMP analog (100 µM 8-Br-cGMP, bath-applied 15 min beforehand), resulting in a relative EPSP amplitude of 196 ± 9% (n = 6, Fig. 9C,E). We repeated the experiment at low extracellular Mg2+ in order to increase NMDA receptor currents and, indeed under these conditions, the level of evoked LTP in NO-GC2 KO was indistinguishable from that in WT, suggesting that the increase in NMDA receptor currents is sufficient for the induction of LTP (NO-GC2 KO 254 ± 21%, n = 8 vs. WT 255 ± 23% relative EPSP amplitude, n = 5; Fig. 9D,F).
In sum, our data indicate that cGMP acts postsynaptically via an enhancement of HCN channel-mediated currents, thereby providing a depolarizing current that affects NMDA receptor currents and promotes induction of LTP.
The NO/cGMP signaling cascade has been proposed to play a role in synaptic transmission already 20 years ago (Snyder and Bredt 1991; Garthwaite 2008). In the central nervous system, 2 NO synthases and 2 NO receptor GCs exist. Whereas the occurrence of NOS isoforms is widely accepted, the existence of 2 NO-GC isoforms has hardly been noticed. Partially, this may be due to the fact that the NOS isoforms differ in regulatory properties, while the 2 NO-GCs, NO-GC1 and NO-GC2 are functionally indistinguishable.
LTP was abolished in the double NOS Kos, which is compatible with a role of NO in synaptic plasticity (Son et al. 1996). A role of NO in synaptic plasticity is also supported by analysis of the NO-GC isoform KOs. In these mice, LTP in the hippocampal CA1 region or visual cortex depends on the presence of both NO-GC1 and NO-GC2, since it was already abolished on deletion of one of the NO-GCs (Haghikia et al. 2007; Taqatqeh et al. 2009). These results are compatible with a pre- and postsynaptic role of the NO-GCs. In a recent study on the presynaptic glutamate release in the hippocampus of the NO-GC KOs, we were able to show that cGMP formed by NO-GC1 in response to eNOS-derived NO enhances glutamate release via HCN-blocker-sensitive channels (Neitz et al. 2011). Here, we analyzed the postsynaptic properties of the glutamatergic neurons of the CA1 region of NO-GC2 deficient mice and unravel an additional postsynaptic role of cGMP in synaptic transmission.
cGMP is Gating HCN Channels Postsynaptically
To analyze a postsynaptic effect of cGMP on HCN channels activity, we monitored somatic hyperpolarization-induced Ih in the pyramidal neurons of NO-GC2 KOs. Indeed, Ih in the NO-GC2 KO neurons was found to be reduced compared with that in WT, indicating that cGMP has an impact on the open-probability of HCN channels. Accordingly, cGMP application enhanced Ih in the NO-GC2 KO neurons to almost WT-like levels and conversely ODQ, the NO-GC inhibitor, reduced Ih in the WT neurons to NO-GC2 KO-like levels. The ODQ effect underlines that NO-induced cGMP formation is continuously occurring under the experimental conditions tested. Typically, cyclic nucleotides shift the voltage dependence of activation of the HCN channels. Indeed, a more negative V1/2 value was determined in the NO-GC2 KO compared with WT neurons (−97 vs. −87 mV). The application of cGMP yielded depolarized WT-like V1/2 in the NO-GC2 KO, whereas ODQ led to a more hyperpolarized V1/2 in WT neurons reaching NO-GC2 KO-like levels. The results strongly indicate that cGMP is gating HCN channels in WT and in the NO-GC2 KOs, a substantial fraction of the HCN channels appears to be in a cyclic nucleotide-free state. The findings are further supported by the time constant of activation that was higher in NO-GC2 KO compared with WT, restored to WT-like level by cGMP, and decreased to NO-GC2 KO-like values with ODQ in WT.
These findings are unexpected at first as HCN channels are supposed to be gated mainly by cAMP. In the hippocampal CA1 region, HCN1 is the major HCN isoform (Ludwig et al. 2003). In contrast to HCN2 and HCN4, HCN1 has been reported to have a poor sensitivity toward cyclic nucleotides (Ludwig et al. 1998). However, by comparing HCN1 channel properties in intact ooyctes and inside-out patches, Chen et al. (2001) suggested that basal cAMP levels in Xenopus oocytes are sufficient for channel modulation. Their finding is supported by a high affinity of HCN1 for cAMP (K1/2 of 0.06 µM), whether HCN1 is also affected by cGMP has not been addressed in that study. A high affinity of HCN1 for cAMP has also been described in a more recent report in which the CNDBs of HCN1 were purified in the cyclic nucleotide-bound state (Lolicato et al. 2011). Possibly, HCN1 is not poorly sensitive to cyclic nucleotides, but conversely has such a high affinity that the channels in cells used for recombinant expression are already in the nucleotide-bound state, which explains why additional cyclic nucleotides do not have a pronounced effect in whole-cell measurements. For the CNDBs of HCN1, the authors report comparable affinities of HCN1 for cAMP and cGMP. Conceivably, a fraction of HCN1 is gated by cAMP, yet, the more negative V1/2 in the NO-GC2 KO argues against a completely cAMP-bound state of HCN1, which can be explained by cAMP being strictly locally controlled by phosphodiesterases.
Of course, another HCN channel than HCN1 may be modulated by cGMP. In addition to HCN1, HCN2 has been reported to occur in the hippocampal CA1 region. HCN2 is nucleotide sensitive as the activation curve is shifted to the right by 5–25 mV by cAMP and cGMP. As cAMP binds with 10-fold higher affinity than cGMP, cGMP is usually not considered as an endogenous HCN ligand (1 vs. 10 µM; Ludwig et al. 1998; Craven and Zagotta 2006). However, NO-GC2 is activated up to 200-fold by NO and targeted to postsynaptic densities next to the NO-generating nNOS by the interaction with PDS-95; therefore, locally very high cGMP concentrations are possible. On the other hand, HCN channel properties have been mainly analyzed with recombinantly expressed HCN channels; in their physiological environment, HCN channels may exhibit different features, that is, an increased sensitivity toward cGMP due to post-translational modifications (Pian et al. 2006), heterodimerization or interactions with additional proteins (Santoro et al. 2009).
Evidence that HCN channels act as target molecules for cGMP has been also obtained before. Pape and Mager (1992) have shown already in 1992 that NO via cGMP controls oscillatory activity in thalamocortical neurons by shifting voltage dependence of Ih to more positive values. In more previous studies, NO was shown to persistently depolarize axonal membranes through cGMP acting on HCN channels in rat optic nerve, and a role of cGMP gating HCN channels has been suggested in deep cerebellar nuclei (Garthwaite et al. 2006; Wilson and Garthwaite 2010).
HCN channels have been reported to be open partially under resting conditions, thereby contributing to the RMP and membrane resistance (Ludwig et al. 2003; Nolan et al. 2004). Moreover, the channels have been shown to play a role in synaptic excitability (Huang et al. 2009). Our experimental analysis of these parameters in the NO-GC2 KO yielded characteristics comparable with those determined in the HCN blocker-treated WT neurons: 1) reduced RMPs (−73 vs. −68 in WT), 2) increased membrane resistance, 3) enhanced spontaneous AP firing. Comparable features have been obtained by others using a HCN blocker or HCN deficient mice (Ludwig et al. 2003; Nolan et al. 2004, Huang et al. 2009). Together, our results indicate that cGMP acts postsynaptically by promoting HCN channel activity. The finding that the HCN blocker did not cause any changes in the NO-GC2 KO mice shows that, without cGMP, the HCN channels do not contribute to the tested neuronal properties. In this context, it should be pointed out that an indirect action of cGMP on HCN channels, for example, phosphorylation of HCN by cGMP-dependent kinases cannot be completely excluded.
cGMP Increases NMDA Receptor Currents via HCN Channels
As the NO-GC2 KO was reported to lack LTP in the CA1 region, hippocampal glutamate receptors were studied in whole-cell recordings. Whereas the properties of AMPA receptors did not differ between WT and NO-GC2 KOs, analysis of NMDA receptors revealed greatly reduced amplitudes in the NO-GC2 KO neurons at physiological membrane potentials. The kinetic properties of the NMDA receptors did not show any significant variations between WT and NO-GC2 KO, which argues against a difference in subunit composition. The results are confirmed by the finding of comparable expression of the 2 major NMDA receptor subunits, NR2A and NR2B, in western blot analysis.
As ODQ, the NO-GC inhibitor, decreased the NMDA receptor currents in WT neurons to NO-GC2 KO-like levels and conversely, the reduced NMDA receptor currents of the NO-GC2 KO were increased by cGMP to WT levels, the magnitude of the NMDA receptor amplitudes apparently depends on the level of cGMP. In contrast to our data, Boulton et al. (1995) did not notice an effect of ODQ on NMDA receptor-mediated EPSPs, differences in the experimental conditions may account for the divergent findings.
We hypothesized that the cGMP effect on NMDA receptors is mediated via HCN channels. Indeed, NMDA receptor currents of WT neurons were reduced by 2 different HCN blockers to those in the NO-GC2 KO neurons, in which the HCN blocker did not cause any alterations. These data suggest that cGMP indirectly via HCN channels modulates NMDA receptor currents. Conceivably, cGMP shifts the voltage dependence of HCN channels to more depolarized values thereby enhancing Ih that contributes to membrane depolarization required for proper NMDA receptor activation. As NO-GC2, via PDS-95, is located in close proximity to NMDA receptors, and the HCN channels are highly enriched in dentrites (Lörincz et al. 2002), local effects of cGMP and Ih on NMDA receptor currents appear reasonable. The specificity of HCN blockers has been questioned (Chevaleyre and Castillo 2002). In case, the observed inhibition of NMDA receptor current by the HCN blockers is not a consequence of HCN channel inhibition, the cGMP-dependent modulation of NMDA receptor current detected in the NO-GC2 KO and ODQ-treated WT neurons may involve other cGMP targets than the HCN channels, that is, cGMP-dependent protein kinases or cGMP-regulated phosphodiesterases.
As NMDA receptor activity is of utmost importance for synaptic plasticity, we considered the reduced NMDA receptor current in the NO-GC2 KO to be of significance for the reported lack of LTP. NMDA receptor current in the NO-GC2 KO neurons was found to be increased at low extracellular Mg2+ and indeed under these conditions, LTP was inducible in the NO-GC2 KO. The results suggest that the increase in NMDA receptor current at low extracellular Mg2+ is sufficient to induce LTP in the NO-GC2 KO, that is, in the absence of postsynaptic cGMP.
In line with the finding of cGMP increasing NMDA receptor currents, NO-GC2 has been reported to be localized in close proximity to the NMDA receptor via its interaction with PSD-95 (Russwurm et al. 2001). By interacting with the third PDZ domain of PSD-95, NO-GC2 is placed closed to the NO-generating nNOS and the NMDA receptor binding to the PSD-95's second and first PDZ domain, respectively (Christopherson et al. 1999). Overall, we postulate a postsynaptic setting in which NO/cGMP formation is triggered by NMDA receptor-mediated Ca2+ influx, which via nNOS activation and the subsequently formed cGMP via HCN channels promotes NMDA receptor current representing a so far unknown feed-forward loop.
In concerns of NO/cGMP signaling in glutamatergic neurons, we identify here the postsynaptically working members of the signaling cascade (NO-GC2 and nNOS) and with an effect on NMDA receptor current demonstrate their participation in postsynaptic responsiveness. Together with our previous finding that presynaptic NO/cGMP (NO-GC1 and eNOS) increases the neurotransmitter release probability, we now provide evidence that 2 NO/cGMP pathways exist that enhance the strength of synaptic transmission on either sides of the synaptic cleft.
In the light of our current study, the lack of any conclusive concept for NO's action up to now can be explained 1) by the continuous NO-stimulated cGMP formation occurring under various experimental conditions and 2) by the unexpected complexity of the pre- and postsynaptic localization of the NO-GCs. The NO-GC isoform KOs now provide clear cut information as due to the lack of membrane permeability, the cGMP made by one isoform is not able to make up for the cGMP formed by the other isoform.
With the HCN channels (most likely HCN1), we identify novel target molecules for cGMP and supply evidence that gating of the channels by cGMP increases NMDA receptor currents. The effect of cGMP on NMDA receptor currents can now explain how cGMP is required for LTP and provides a new mechanistic insight into the molecular basis of synaptic plasticity.
This work was supported by the Deutsche Forschungsgemeinschaft to (D.K. and T.M.), by the DFG-Graduiertenkolleg Grant 736 (to U.E. and T.M.), and by a MAIFOR grant of the UMC of the Joh.-Gut. Univ. Mainz.
We thank Simone Dahms-Pritorious, Medah Özcan, and Ursula Krabbe for excellent technical assistance. Conflict of Interest: None declared.