The murine HCN3 gene encodes a hyperpolarization-activated cation channel with slow kinetics and unique response to cyclic nucleotides.

Hyperpolarization-activated cation channels of the HCN gene family are crucial for the regulation of cell excitability. Importantly, these channels play a pivotal role in the control of cardiac and neuronal pacemaker activity. Dysfunction of HCN channels has been associated with human diseases, including cardiac arrhythmia, epilepsy, and neuropathic pain. The properties of three HCN channel isoforms (HCN1, HCN2, and HCN4) have been extensively investigated. By contrast, due to the lack of an efficient heterologous expression system, the functional characteristics of HCN3 were by and large unknown so far. Here, we have used lentiviral gene transfer to overexpress HCN3 in HEK293T cells. HCN3 currents revealed slow activation and deactivation kinetics and were effectively blocked by extracellular Cs+ and the bradycardic agent ivabradine. Cyclic AMP and cGMP had no significant impact on activation kinetics but induced a 5-mV shift of the half-maximal activation voltage (V0.5) to more hyperpolarized potentials. A negative shift of V0.5 induced by cyclic nucleotides is an unprecedented feature within the HCN channel family. The expression of HCN3 in mouse brain was examined by Western blot analysis using a specific antibody. High levels of protein were detected in olfactory bulb and hypothalamus. In contrast, only very low expression was found in cortex. Using reverse transcriptase PCR transcripts of HCN3 were also detected in heart ventricle. In conclusion, the distinct expression pattern in conjunction with the unusual biophysical properties implies that HCN3 may play an unique role in the body.

Hyperpolarization-activated cation channels of the HCN gene family are crucial for the regulation of cell excitability. Importantly, these channels play a pivotal role in the control of cardiac and neuronal pacemaker activity. Dysfunction of HCN channels has been associated with human diseases, including cardiac arrhythmia, epilepsy, and neuropathic pain. The properties of three HCN channel isoforms (HCN1, HCN2, and HCN4) have been extensively investigated. By contrast, due to the lack of an efficient heterologous expression system, the functional characteristics of HCN3 were by and large unknown so far. Here, we have used lentiviral gene transfer to overexpress HCN3 in HEK293T cells. HCN3 currents revealed slow activation and deactivation kinetics and were effectively blocked by extracellular Cs ؉ and the bradycardic agent ivabradine. Cyclic AMP and cGMP had no significant impact on activation kinetics but induced a 5-mV shift of the half-maximal activation voltage (V 0.5 ) to more hyperpolarized potentials. A negative shift of V 0.5 induced by cyclic nucleotides is an unprecedented feature within the HCN channel family. The expression of HCN3 in mouse brain was examined by Western blot analysis using a specific antibody. High levels of protein were detected in olfactory bulb and hypothalamus. In contrast, only very low expression was found in cortex. Using reverse transcriptase PCR transcripts of HCN3 were also detected in heart ventricle. In conclusion, the distinct expression pattern in conjunction with the unusual biophysical properties implies that HCN3 may play an unique role in the body.
The hyperpolarization-activated cation current, termed I h or I f , is widely expressed in heart cells and neurons. The current is best known for its prime role in the generation of rhythmic activity in cardiac and neuronal pacemaker cells (1,2). I h is also present in several types of non-pacing neurons where it contributes to various physiological functions, including the setting of the resting membrane potential, synaptic transmission, and dendritic integration. Recent evidence suggests that I h is involved in diseases making the channel a promising target for drug therapy. For example, the dysfunction of cardiac I h was identified in patients suffering from sick sinus brady-cardia (3,4). Furthermore, it was proposed that overexpression of I h in heart ventricle is associated with cardiac hypertrophy (5). Ivabradine (S-16257-2), a blocker of I h , is currently considered as a novel drug in the therapy of tachycardic arrhythmia and angina pectoris (6). I h is also likely to participate in neurological diseases. In particular, there is accumulating evidence that the current is involved in epileptogenesis (7,8). Moreover, overexpression of I h was observed in rat models of peripheral nerve injury suggesting a potential role of this current in driving neuropathic pain (9). I h is encoded by a family of four hyperpolarization-activated cyclic nucleotide-gated (HCN 1 1-4) channels (1,10). HCN channels are members of the 6-transmembrane superfamily of cation channels (11). A structural hallmark of all HCN channels is a cyclic nucleotide-binding domain in the C terminus that confers sensitivity to cAMP (12,13). So far, the functional properties of three members of the HCN channel family (HCN1, HCN2, and HCN4) have been extensively studied using heterologous expression. Moreover, the specific physiological relevance of these channels has been defined using genetargeting approaches in mice (8,14,15). By contrast, there is only sparse information on the properties of the HCN3 channel (9,16). In situ hybridization and immunocytochemistry indicated that this channel is expressed at low levels in rat and mouse brains (17)(18)(19). The biophysical and pharmacological properties of the channel are by and large unknown.
In the present study we set out to address this important issue. We investigated the expression levels of murine HCN3 in mouse brain regions using Western blot analysis. Moreover, we achieved robust expression of the channel in HEK293T cells using lentiviral expression vectors. We show that HCN3 reveals some properties shared by other HCN channel types but is unique among these channels by being rather inhibited than activated by cyclic nucleotides.

Lentiviral HCN2 and HCN3 Expression Vectors-
The mouse HCN3 (mHCN3) coding sequence was excised as a 2.7-kb HindIII-SpeI fragment from the plasmid mHCN3/pcDNA3 (16). The mHCN3 was cloned via XbaI and SpeI sites into pBluescript II KS plasmid containing the IRES-EGFP coding sequence (Clontech). Lentiviral plasmid LV-HCN3 was prepared by replacing the GFP coding sequence with the XbaI-SalI HCN3-IRES-EGFP fragment. A schematic representation of the HCN3 lentiviral ex-* This work was supported by the Deutsche Forschungsgemeinschaft. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
RT-PCR-Total RNA was isolated using TRIzol (Invitrogen) and subsequently treated with DNase I (Roche Applied Science). First strand cDNA was synthesized from 5 g of RNA with the Superscript II H Ϫ -Kit (Invitrogen) using oligo(dT) primers. HCN3 was amplified from 0.5 l of cDNA using following primers and conditions: 5Ј-GTCCGC-CGGGGCCTGGAT-3Ј (forward), 5Ј-CCTCCCACTGGTGTATGTAGC-3Ј (reverse); 40 cycles at 60°C. Amplicons were separated on 5% polyacrylamide gels, stained with ethidium bromide, and visualized on a Gel Doc 2000 system (Bio-Rad). The primer pairs were intron-spanning to avoid amplification of genomic DNA.
Generation of Anti-HCN3 Antibody and Western Blot-Polyclonal rabbit antibody directed against the C-terminal (amino acids 552-779) region of murine HCN3 was generated by immunization (Gramsch Laboratories; Schwabhausen, Germany) with a His-Tag fusion protein expressed and purified using the QiaExpress-Kit (Qiagen, Germany) and affinity-purified using the Amino-Link-Kit (Pierce).
To determine the specificity of the anti-HCN3 antibody, membrane proteins were isolated from HEK293 cells transfected with mHCN1, mHCN2, mHCN3, or hHCN4 as described previously (22) and subjected to Western blot analysis with anti-HCN3. HEK293T cells transduced with LV-HCN3 were analyzed using the same protocol.
Tissue from various mouse brain regions was dissected and snap frozen in liquid nitrogen. Samples were homogenized on dry ice using a mortar and pestle, boiled in lysis buffer (2% SDS, 50 mM Tris) for 10 min, and centrifuged (15 min at 16,000 ϫ g) to remove cell debris. Proteins were separated using 7% SDS-PAGE followed by Western blot analysis according to standard procedures. The antibodies anti-HCN3 (1:2,000) and anti-tubulin (1:400; Dianova, Germany) were used.
For deglycosylation, homogenates or membrane proteins were first denatured in the presence of 0.5% SDS/0,1 M 2-mercaptoethanol for 3 min at 95°C and then incubated in 50 mM PBS/1% Nonidet P-40 and 500U of peptide/N-glycosidase F (Roche) for 1 h at 37°C. Subsequently, proteins were subjected to Western blot analysis.
Electrophysiology-Currents were measured 2-3 days after infection with recombinant lentivirus using whole-cell patch clamp technique. Patches were equilibrated for at least 2 min before starting experiments to minimize run down of current. The standard extracellular solution was composed of 110 mM NaCl, 0.5 mM MgCl 2 , 1.8 mM CaCl 2 , 5 mM HEPES, 30 mM KCl, pH ϭ 7.4, adjusted with NaOH. When required, ivabradine (S 16257-2) and Cs ϩ were added to the extracellular solution by dissolving a stock 10 mM solution to the final 30 M and a stock 1 M to the final 2 mM, respectively. The intracellular solution contained 130 mM KCl, 10 mM NaCl, 3 mM MgATP, 0.5 mM MgCl 2 , 1 mM EGTA, 5 mM HEPES, pH ϭ 7.4, adjusted with KOH. For determining the cAMP and cGMP sensitivity of HCN channels intracellular solution was supplemented with 0.5 mM cAMP or cGMP. The extracellular solutions were exchanged by a local solution exchanger. The different solutions reached the cell membrane within less than 100 ms. All recordings were obtained at room temperature. Data were acquired at 10 kHz using an Axopatch 200B amplifier and pClamp 8 (Axon Instruments). Voltage clamp data were stored on the computer hard disk and analyzed off-line by using Clampfit 8 (Axon Instruments) and Origin (Origin Lab Corporation). For determination of the voltage of half-maximal activation (V 0.5 ) currents were elicited by hyperpolarizing the membrane for 3 s to voltages ranging from Ϫ140 to Ϫ20 mV (in 10-mV increments) from a holding potential of Ϫ40 mV followed by a 500 ms step to Ϫ140 mV. Amplitude of tail currents, determined immediately after the disappearance of the capacitance transient, were normalized to the maximal current at Ϫ140 mV (I max ) and plotted as a function of the preceding membrane potential. The data points were fitted with the Boltzmann function, ( where I min is an offset caused by a non-zero holding current, V m is the test potential, V 0.5 is the voltage of half-maximal activation, and k is the slope factor. Time constants of channel activation ( act ) of HCN2 and HCN3 channels were determined by monoexponential function fitting the current evoked during hyperpolarizing voltage pulses to appropriate volt-age. As has been described earlier the initial lag in the activation of HCN channel currents was excluded from the fitting procedure (23). For determination of contribution of instantaneous current (I inst ) to the total I h current, the size of instantaneous as well as steady-state current elicited by a 3-s pulse to Ϫ140 mV was determined and corrected for the leak, and the amplitude of instantaneous current was normalized with respect to the amplitude of the steady-state current.
Statistics-All values are given as mean Ϯ S.E., and n is the number of experiments. An unpaired Student's t test was performed for the comparison between two groups. Values of p Ͻ 0.05 were considered significant.

Expression of HCN3 in Mouse Heart and Brain-RT-PCR
revealed that HCN3 transcripts are present in the mouse brain and heart ventricle (Fig. 1A). To study the expression levels of the channel in distinct regions of the brain we raised a polyclonal antibody against the unique C terminus of the protein.
The antibody recognized a 86-kDa protein (corresponding to the theoretical M w of mHCN3) in membrane fractions of HEK293 cells transfected with mHCN3 (Fig. 1B). This band was not observed in non-transfected cells and in cells transfected with HCN1, HCN2, or HCN4, demonstrating the specificity of the anti-HCN3 antibody (Fig. 1B). In mouse brain the antibody detected a slightly bigger protein (ϳ90 kDa) than in HEK293 cells transfected with mHCN3 (Fig. 1C). This size difference was because of N-linked glycosylation because it disappeared after incubation with N-glycosidase F (Fig. 1D). The expression levels of HCN3 were high in the olfactory bulb and hypothalamus, intermediate in amygdala and hippocampus, and low in retina and cortex (Fig. 1C). In cerebellum, HCN3 was only faintly, if at all, detectable (Fig. 1C). Similarly, the expression of HCN3 protein in heart tissue was below detection level (not shown).
Lentiviral Expression of HCN3 in HEK293T Cells-In previous attempts using chemical transfection methods only a small percentage of cells exhibited I h , with current densities too low for a detailed biophysical analysis. To circumvent this problem we made use of a state-of-the-art lentiviral expression system (21). The vector design allows the coexpression of the gene of interest together with EGFP (via an IRES sequence) to facilitate the screening of the positive cells. Fig. 2A shows an example of HEK293T cells infected with LV-HCN3. About 30% of the cells revealed green fluorescence. The successful expression of the HCN3 protein was confirmed by Western blot. When a membrane preparation from infected cells was used, a specific band of ϳ86 kDa was observed, corresponding to the nonglycosylated HCN3 protein (Fig. 2B).
HCN3 Encodes a Hyperpolarization-activated Current with Slow Kinetics-EGFP-positive cells were used for electrophysiological analysis. HCN3 currents were compared in side-byside experiments with I h from HEK293T cells infected with a lentiviral HCN2 expression vector. The current density obtained for HCN3 was approximately five times smaller than the current density in HCN2-expressing cells (27 Ϯ 5 pA/pF, n ϭ 29 for HCN3 versus 139 Ϯ 26 pA/pF, n ϭ 26 for HCN2). Fig.  3, A and B show representative whole-cell current traces elicited by a family of hyperpolarizing steps from a holding potential of Ϫ40 mV. Both currents are composed of a fast I inst and a slowly activating sigmoidal component. The relative amplitude of the I inst was consistently larger in HCN3 (13 Ϯ 2% of total current, n ϭ 19) than in HCN2 (5.0 Ϯ 0.7%, n ϭ 20). HCN3 activated and deactivated with significantly slower kinetics than HCN2. The activation time constants at a fully activating membrane potential (Ϫ140 mV) were act ϭ 470 Ϯ 30 ms (n ϭ 13) for HCN3 and act ϭ 330 Ϯ 30 ms (n ϭ 9) for HCN2. By contrast, HCN2 and HCN3 did not differ from each other, with respect to their voltage of half-maximal activation (V 0.5 ϭ Ϫ95 mV, Fig. 3D). In addition, the reversal potential obtained from the I-V curve of the fully activated channel was not different between the two channels (Ϫ27 mV, data not shown) indicating that both channels share the same ion selectivity.
HCN3 currents revealed the typical pharmacological profile of native and heterologously expressed I h channels. The channel was readily blocked by 2 mM extracellular Cs ϩ (Fig. 4A). Likewise, the bradycardic drug ivabradine (6) almost completely inhibited the fully activated current at a concentration of 30 M. This concentration was used previously to block specifically I h in sino-atrial node cells (24) (Fig. 4B).
Cyclic Nucleotides Shift the Activation Curve of HCN3 to More Negative Voltages-We next went on to test the effect of cAMP on lentivirally expressed HCN channels. Fig. 5, A and B show representative normalized current traces of fully activated HCN2 and HCN3 channels (at Ϫ140 mV) obtained either with control pipette solution or after perfusion with 0.5 mM cAMP. In accordance with previous studies using chemical transfection methods (25) cAMP speeded up the activation kinetics of HCN2 ( act (ϪcAMP) ϭ 330 Ϯ 30 ms (n ϭ 9); act (ϩcAMP) ϭ 165 Ϯ 7 ms (n ϭ 12); Fig. 5E). By contrast, perfusion with 0.5 mM cAMP did not significantly alter the kinetics of the HCN3 current ( act (ϪcAMP) ϭ 470 Ϯ 30 ms (n ϭ 13); act (ϩcAMP) ϭ 510 Ϯ 30 ms (n ϭ 11)). Cyclic AMP shifts the voltage dependence of all HCN channels characterized so far to more depolarized potentials. Indeed, the voltage curve of lentivirally expressed HCN2 was

FIG. 3. Current-voltage (I-V) characteristics of HCN3 and HCN2 overexpressed in HEK293T cells using lentiviral vectors.
A and B, family of current traces of HCN3 (A) and HCN2 (B) elicited by stepping from a holding potential of Ϫ40 mV to hyperpolarizing voltages from Ϫ140 to Ϫ20 mV for 3 s; followed by a 500-ms step to Ϫ140 mV. C, voltage dependence of activation time constant act for HCN3 (•) and HCN2 (E). D, I-V characteristics of HCN3 (•) and HCN2 (E) as derived from the tail current analysis of currents like those in A and B, respectively. Tail currents immediately after the disappearance of the capacitance transient (arrow) were normalized to maximal current at Ϫ140 mV (I max ) and plotted as a function of the preceding membrane potential. Solid lines are fits to the Boltzmann function. The fitting parameters were obtained as follows: HCN3, V 0.5 ϭ Ϫ95 Ϯ 1 mV; k ϭ 9.6 Ϯ 0.6 mV (n ϭ 18); HCN2, V 0.5 ϭ Ϫ95 Ϯ 1 mV; k ϭ 7 Ϯ 1 mV (n ϭ 13).
Cyclic GMP modulated HCN3 currents in a similar manner as cAMP did. Intracellular application of 0.5 mM cGMP did not significantly alter the kinetics of the fully activated HCN3 current ( act (ϪcGMP) ϭ 470 Ϯ 30 ms (n ϭ 13); act (ϩcGMP) ϭ 520 Ϯ 50 ms (n ϭ 17) at Ϫ140 mV; Fig. 6A). At less hyperpolarizing conditions cGMP tended to slow down HCN3 current kinetics (Fig. 6B). However, this effect was not statistically significant (p Ͼ 0.05). In the presence of cAMP the same behavior was observed (data not shown). Like cAMP, cGMP shifted the voltage dependence of HCN3 activation slightly to more negative values (V 0.5 ϭ Ϫ95 Ϯ 1 mV versus Ϫ101 Ϯ 1 mV (n ϭ 18 and 17 in the absence and presence of cGMP, respectively); Fig. 6C). This effect was statistically significant (p Ͻ 0.05) and resulted from the change in the steepness of the I-V curve (k (ϪcGMP) ϭ 9.6 Ϯ 0.6 mV (n ϭ 18); k (ϩcGMP) ϭ 11.3 Ϯ 0.6 mV (n ϭ 17)).
Thus, with respect to V 0.5 values, cyclic nucleotides regulated HCN2 and HCN3 in opposite direction. Because the voltage dependence of HCN2 activation determined in our study was entirely consistent with previous data on that channel the unique behavior of HCN3 was genuine and not an artifact resulting from the lentiviral expression system used.

DISCUSSION
Our knowledge of the HCN channel family has increased dramatically since the first cloning of the channels 7 years ago (20,26,27). Despite the principal progress made, one member of the family, HCN3, was enigmatic up to now. In this study, we set out to determine for the first time the basic characteristics of this channel.
Using lentiviral gene transfer we achieved robust expression of the channel in HEK293T cells. Although we consistently found I h in cells that were successfully infected (as indicated by green fluorescence of coexpressed EGFP), current densities of HCN3-expressing cells were about five times smaller than those seen with HCN2. By contrast, protein levels of HCN3 were not significantly different from those of HCN2 (data not shown). Thus, the smaller amplitude of HCN3 currents probably reflects an intrinsic property of the HCN3 channel molecule (e.g. lower open probability or smaller single-channel conductance) or alternatively, may point to the absence of channel components (e.g. auxiliary subunits) that are missing in our expression system and are necessary for normal channel activation. Finally, the possibility that the discrepancy observed is because of lower cell surface expression of HCN3 must be also considered.
HCN3 currents activated significantly slower than HCN2 currents expressed under the same conditions. The HCN3 activation constant was well within the range of that observed for HCN4 (25). It should be noted that the act determined for HCN2 in the present study (330 ms at Ϫ140 mV) was also somewhat larger than previously found (20) suggesting that the expression system may exert some minor influence on the channel kinetics.
The most surprising feature of HCN3 was its unique re- sponse to cyclic nucleotides. In contrast to all other HCN channel types described so far, the activity of HCN3 was not enhanced by these second messengers. Rather, cAMP and cGMP changed the slope of the voltage dependence of activation, inducing a slight but significant shift (ϳ5 mV) of the activation curve to more negative voltages. The effect of cyclic nucleotides was apparent at membrane voltages more negative to Ϫ90 mV. Hence, cAMP and cGMP inhibited rather than activated the HCN3 current. According to a cyclic allosteric model (28) the free energy available from ligand binding should be coupled to enhance gating, which in the case of HCN channels, should result in the shift of I-V curve to more positive values and faster kinetics (29). Such a phenomenon was indeed observed for HCN2 and HCN4 (30). However, in the case of HCN3, despite the requirement of rather strong hyperpolarization and slow kinetics of gating, the binding of cyclic nucleotides resulted in the reverse effect, shifting the I-V curve to more negative values. This apparent discrepancy could be reconciled in terms of the recent ligand-gating model (12,31,32). According to this model, the binding of cAMP controls the transition in the oligomerization state of cyclic nucleotide-binding domain between the 4-fold symmetric tetrameric gating ring (enhancing the gating) and 2-fold symmetric dimer of dimers (suppressing gating). This transition should be facilitated through the interactions in the C-linker, connecting the cyclic nucleotide-binding domain with the gate. Interestingly, it has been shown recently that a tripeptide mutation in the C-linker of HCN2 reversed the polarity of ligand gating (33). The mutation converted cAMP from an agonist that facilitated channel opening into an inverse agonist that inhibited channel opening. Therefore, it is plausible that the observed inhibitory effect of ligand binding on gating of HCN3 could result from the formation of the 4-fold symmetric tetrameric gating ring in the absence of cAMP because of the amino acid composition of its C-linker or its cyclic nucleotide-binding domain. Similarly to the situation observed in the mutated HCN2 channel (33), ligand binding would facilitate the transition into 2-fold symmetric dimer of dimers, suppressing gating.
Using a polyclonal antibody raised against the C terminus of the channel we demonstrated expression of the HCN3 protein in restricted areas of the brain. The protein was not detected in heart ventricle although the HCN3 mRNA was readily identified in this tissue. This finding suggests that HCN3 is expressed only at low levels in heart. Like other HCN channels, native HCN3 is N-linked glycosylated (22,34). In HCN2, the glycosylation site has been recently determined (22). The identified residue (Asn-380) is highly conserved in the HCN channel family and, hence, may also confer glycosylation of HCN3.
The highest expression levels of HCN3 were found in the olfactory bulb and hypothalamus, intermediate levels were found in the amygdala and hippocampus, and low levels were found in the retina and cortex. This expression pattern is in line with a recent report describing the localization of HCN channels in the same regions of the rat brain (19). The expression of HCN3 in hypothalamus is of particular pharmacological interest because neurons of this brain region are involved in the regulation of important physiological functions, including the tone of the autonomous nervous system, circadian rhythm, and hormone secretion (35). To our knowledge, native I h currents displaying the specific properties of HCN3 (slow kinetics and inhibitory rather than stimulatory effect of cyclic nucleotides) have not yet been identified in vivo. Several reasons could account for this discrepancy. For example, it seems possible that HCN3 heteromerizes with other HCN isoforms (22,36,37) or interacts with additional auxiliary subunits (1, 38) that could modulate current properties. Alternatively, the over-lapping expression of HCN channel isoforms in the brain (19) opens the possibility of co-localization of different homomeric HCN channels within the same cell. The macroscopic current from such cells would be composed of contributions from different homomeric HCN channels. Genetic deletion of HCN3 in mice will be required to define the exact physiological relevance of HCN3 in distinct brain regions.
Finally, we demonstrate that HCN3 is inhibited by ivabradine, a member of a novel class of I h blockers that is considered for clinical use as bradycardic agent. Like the structurally related cilobradine (39), ivabradine is quite unselective among I h channels because it also blocks HCN1, HCN2, and HCN4. 2 It will be desirable to develop isoform specific HCN channel blockers. The lentiviral HCN3 expression system described in this study provides an important tool to achieve this goal.