Functional heteromerization of HCN1 and HCN2 pacemaker channels.

An important step toward understanding the molecular basis of the functional diversity of pacemaker currents in spontaneously active cells has been the identification of a gene family encoding hyperpolarization-activated cyclic nucleotide-sensitive cation nonselective (HCN) channels. Three of the four gene products that have been expressed so far give rise to pacemaker channels with distinct activation kinetics and are differentially distributed among the brain, with considerable overlap between some isoforms. This raises the possibility that HCN channels may coassemble to form heteromeric channels in some areas, similar to other K(+) channels. In this study, we have provided evidence for functional heteromerization of HCN1 and HCN2 channels using a concatenated cDNA construct encoding two connected subunits. We have observed that heteromeric channels activate several-fold faster than HCN2 and only a little slower than HCN1. Furthermore, the voltage dependence of activation is more similar to HCN2, whereas the cAMP sensitivity is intermediate between HCN1 and HCN2. This phenotype shows marked similarity to the current arising from coexpressed HCN1 and HCN2 subunits in oocytes and the native pacemaker current in CA1 pyramidal neurons. We suggest that heteromerization may increase the functional diversity beyond the levels expected from the number of HCN channel genes and their differential distribution.

The newly cloned hyperpolarization-activated cyclic nucleotide-sensitive cation nonselective (HCN) 1 ion channels give rise to currents that are almost indistinguishable or very similar to native pacemaker currents termed I f in heart (1) and I h or I q in brain tissue (2). To date, three of the four HCN genes have been functionally expressed, giving rise to pacemaker currents with distinct gating properties (for review see Refs. [3][4][5]. HCN1 channels display fast gating properties and are almost unaffected by cAMP (6), similar to I h recorded from some hippocampal pyramidal cells. HCN2 channels activate slowly and are strongly modulated by cAMP (7,8), similar to native pacemaker currents in some cardiac cells. HCN4 channels activate very slowly and also respond strongly to cAMP (9,10), similar to I h in thalamic relay neurons.
Expression patterns have confirmed that HCN1 is expressed restrictively in neurons of the neocortex, hippocampus, cerebellar cortex, and brainstem nuclei (6,11,12). In contrast, HCN2 is widely expressed throughout the brain with prominent labeling of thalamic and brainstem nuclei as well as in heart tissue (6,7). Finally, HCN4 is predominantly expressed in thalamus, heart, and testis (9). This distinct distribution pattern of HCN isoforms across the brain has been suggested to attribute to functional heterogeneity of native neuronal pacemaker currents (12). Nevertheless, considerable overlap was observed between HCN1 and HCN2 using in situ hybridization, namely in hippocampal pyramidal cells and some brainstem nuclei (12). Moreover, the coexistence of HCN1 and HCN2 mRNA has been demonstrated at the single-cell level for hippocampal CA1 neurons (13), raising the possibility that heteromeric channel complexes could form in some areas.
The functional heteromerization of K ϩ channels with a tetrameric subunit stoichiometry has previously been demonstrated for voltage-gated K ϩ channels (14 -19) as well as inwardly rectifying K ϩ channels (20, 21) but is currently unknown for HCN channels. HCN channel subunits have an overall structure similar to voltage-gated K ϩ channel subunits, including 6 transmembrane domains (S1-S6), a selectivity filter containing the GYG motif, and a putative voltage sensor (S4) containing a positively charged residue at every third position. Based on these structural similarities it has been suggested that HCN channels will most likely adopt the general tetrameric subunit architecture of both 6-and 2-transmembrane segment K ϩ channels (4). To investigate the possibility that functional heteromeric HCN channels may form by coassembly of different HCN isoforms, we expressed a concatenated cDNA construct encoding a covalently linked HCN1 and HCN2 subunit into Xenopus oocytes. Currents were recorded using the 2-microelectrode voltage clamp technique and were compared with the biophysical properties of channels formed upon coexpression of HCN1 and HCN2 subunits and channels formed upon expression of HCN1 and HCN2 alone. The cAMP sensitivity of the homomeric and heteromeric HCN channels was assessed by coexpression of a G-protein-coupled receptor.

EXPERIMENTAL PROCEDURES
cDNA Constructs and in Vitro Transcription-The entire coding sequence for the mouse HCN1 and HCN2 channels (originally termed mBCNG-1 and mBCNG-2, respectively) were subcloned into the vector pSD64TF and pGEMHE, respectively. Plasmids were first linearized either with BamHI (for HCN1) or with SphI (for HCN2). Next, the cRNAs were synthesized from the linearized plasmids using the large scale SP6 (for HCN1) or T7 (for HCN2) mMessage mMachine transcription kit (Ambion). A concatenated construct encoding a covalently linked HCN1 and HCN2 subunit was engineered by digesting HCN1 with PshAI (blunt) and SpeI. HCN2 was digested with SmaI (blunt) and SpeI and ligated into the corresponding sites of the previous construct. The ligation product, HCN1⅐HCN2, was linearized with XbaI for size verification on an agarose gel and in vitro transcription using the large scale SP6 mMessage mMachine transcription kit. The encoded dimer lacks the last 45 residues of the HCN1 C terminus and contains a 9-amino acid linker (GDPNSPFLA) that connects HCN1-Gln 865 stop to HCN2-M1 (see Fig. 1A).
* 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.
The human -opioid receptor (MOR) cDNA was subcloned into our custom high expression vector, pGEMHE (17) and linearized for transcription as previously described (22). The rat G␣ s Q227L cDNA clone (23) was subcloned in pcDNA1.1 and linearized with XbaI for in vitro transcription. The mouse 5-HT4 (a) R cDNA clone (24) in its original vector (pRK5) was first subcloned into the XbaI and HindIII sites of pSGEM, a modified version of pGEMHE (25). For in vitro transcription, 5-HT4 (a) R/pSGEM, was linearized with NheI. The capped cRNAs were synthesized from the linearized plasmids using the large-scale T7 mMessage mMachine transcription kit.
Determination of cAMP Sensitivity-In this study, we coexpressed either -opioid or 5-HT4 (a) receptors with HCN channels. Both receptors are representative members of the G-protein-coupled receptor superfamily that display an overlapping distribution pattern with HCN1 and HCN2 isoforms in the brain (24,26) and activate adenylyl cyclase (23,24). The pathway through which the effect occurs involves G␤␥ activation of adenylyl cyclase (27,28) and requires coexpression of G␣ s Q227L (a constitutively active mutant) in the case of the -opioid receptor (23). The effect of the 5-HT4 (a) receptor, in part caused by its constitutive activity (24), occurs directly through G␣ s activation. HCN currents were recorded before, during, and after application of an agonist (1 M DAMGO for -opioid receptors and 1 M 5-HT for 5-HT4 (a) receptors) to activate the receptor.
Data Analysis-Current-voltage relationships were obtained by measuring the current amplitude after application of 4 s hyperpolarizing test pulses and normalization to the fully activated current at Ϫ110 mV (HCN1), Ϫ115 mV (HCN1⅐HCN2 tandem), and Ϫ125 mV (HCN2). Boltzmann activation curves were constructed using peak tail current amplitudes observed upon application of subsequent hyperpolarizing test pulses. Current values were plotted as a function of the applied voltage step and fitted with a Boltzmann equation, where I represents the current, I off the offset, I max the maximal current, V test the applied test voltage, V1 ⁄2 the midpoint potential, and S the slope of the Boltzmann curve. Curves of n experiments were normalized to the maximal tail current amplitude and corrected for the offset. Time constants of activation were determined by fitting the late rising phase of the current with a double exponential function. The initial lag of the rising phase was excluded from the fitting procedure to obtain an appropriate fit. The lag phase maximally comprised the first 12 ms, 32 ms, and 280 ms of the rising current for HCN1, HCN1⅐HCN2 dimers, and HCN2, respectively. Averaged data are indicated as mean Ϯ S.E. Statistical analysis of differences between groups was carried out with Student's t test, and a probability of 0.05 was taken as the level of statistical significance.

RESULTS AND DISCUSSION
In this study, we investigated the possible heteromerization of HCN channels by expressing a concatenated cDNA construct encoding a covalently linked HCN1 and HCN2 subunit (Fig.  1A) into Xenopus oocytes. Both HCN1 and HCN2 are members of the HCN gene family (3,5,6), encoding pacemaker channels with easily distinguishable kinetic properties. For comparison, all records shown in Fig. 1B were evoked using the same voltage step protocol. Homomeric HCN1 channels activate rel-atively fast, whereas homomeric HCN2 channels activate more slowly (Fig. 1B, Table I). Expression of cRNA encoding the dimeric construct gives rise to currents that activate somewhat slower than HCN1, but still several-fold faster than HCN2 (Fig. 1B, Table I). This finding provides important evidence that the concatenated construct encodes heteromeric HCN1⅐ HCN2 channels, because the unexpectedly fast kinetics cannot be accounted for by the mathematical average expected from homomeric HCN1 and HCN2 channels. Furthermore, the heteromeric phenotype shows marked resemblance to currents that were evoked from oocytes coinjected with equal amounts of HCN1 and HCN2 cRNA (Fig. 1B). This result suggests that HCN1 and HCN2 subunits could spontaneously coassemble to form a population of channels with a subunit arrangement similar to the heteromeric channels formed by HCN1⅐HCN2 dimers.
Current-voltage relationships were constructed from the records in Fig. 1B and are compared in Fig. 1C. It should be noted that the 4-s test pulses applied to record the traces shown in Fig. 1B do not allow HCN2 channels to obtain a steady-state level of activation because of their relatively slow time course of activation (12). To make a valid comparison of the voltage dependence of activation between homomeric and heteromeric HCN1 and HCN2 channels, we calculated half-maximal values of activation (V1 ⁄2 ) from experiments with voltage steps of increasing duration (Fig. 1D). These results show that the use of relatively short hyperpolarizing pulses shifts the estimates of FIG. 1. Current traces recorded from oocytes expressing homomeric and heteromeric HCN channels. A, schematic illustration of the HCN1⅐HCN2 tandem construct. B, comparison of pacemaker currents evoked from oocytes expressing homomeric HCN1, homomeric HCN2 channels, and heteromeric channels formed by tandem-linked HCN1⅐HCN2 subunits or by coinjection of HCN1 and HCN2 cRNA (4-s test potentials range from Ϫ40 to Ϫ110 mV in 10 mV steps; holding potential and tail potential were Ϫ40 mV). Scale bars indicate 1 A and 1 s in all panels. C, current-voltage relationships constructed from experiments in A for homomeric HCN1 (OE), homomeric HCN2 (f) and tandem HCN1⅐HCN2 channels (q). D, determination of the steadystate midpoint potential of activation (V1 ⁄2 ) by application of hyperpolarizing test pulses with increasing duration. The same symbols were used as in C. Dashed line indicates the average between triangles and squares. Unfilled triangles are data from Ref. 12. Heteromeric Pacemaker Channels 6070 the V1 ⁄2 -values to more negative potentials. Comparison of calculated V1 ⁄2 -values shows that heteromeric HCN1⅐HCN2 channels activate at membrane potentials intermediate to homomeric HCN1 and HCN2 channels for pulses Յ5 s. However, heteromeric HCN1⅐HCN2 channels activate at membrane potentials more negative than the average of steady-state V1 ⁄2values for homomeric HCN1 and HCN2 channels (average indicated with a dashed line), for pulses long enough to achieve steady-state levels of activation (Ͼ10 s). This result would suggest that a voltage-dependent transition in the opening of the heteromeric HCN1⅐HCN2 channel complex is limited by the isoform that activates at the most negative membrane voltages, as in the case of HCN2.
As could be observed from the current traces in Fig. 1B, heteromeric HCN1⅐HCN2 channels activate with a faster time course than the average of homomeric HCN1 and HCN2 channels predicts. For clarification, we have superimposed traces from homomeric and heteromeric channels recorded during a hyperpolarizing step to Ϫ90 mV ( Fig. 2A). The current trace recorded from oocytes coexpressing HCN1 and HCN2 channels perfectly matches with the trace obtained from the tandem HCN1⅐HCN2 channels, suggesting that channels are formed with a subunit stoichiometry similar to the defined subunit arrangement of the tandem-linked channels. Next, we quantified the kinetics of activation by fitting current traces during hyperpolarizing voltage steps (as shown in Fig. 1B) with two exponential functions (12). The calculated fast and slow time constants of activation were plotted as a function of the applied test potential and are shown in Fig. 2, B and C, respectively. Values calculated for the traces in Fig. 2A were fast ϭ 214 Ϯ 20 ms (n ϭ 6) and slow ϭ 909 Ϯ 54 ms (n ϭ 6) for tandem HCN1⅐HCN2 channels and fast ϭ 168 Ϯ 8 ms (n ϭ 6) and slow ϭ 944 Ϯ 49 ms (n ϭ 6) for coexpressed HCN1 and HCN2 channels. Homomeric HCN1 and HCN2 channels as well as heteromeric HCN1⅐HCN2 channels have slow time constants of activation that are 5-to 10-fold higher than fast time constants. Compared with the respective time constants for homomeric HCN1 channels, homomeric HCN2 channels have fast and slow time constants that are ϳ10-fold higher. These results are in accordance with recently reported data on pacemaker channels in the mouse CNS (12). Heteromeric HCN1⅐HCN2 channels have time constants of activation that are significantly faster than the average of homomeric HCN1 and HCN2 channels predicts (indicated with a dashed line). Heteromeric HCN1⅐ HCN2 channels activate similarly to homomeric HCN1 channels, with fast and slow time constants of activation that are only about 3-fold higher. This result suggests that HCN1 subunits dominate the activation rate of the heteromeric channel complex.
Another striking feature that discriminates between homo-meric HCN1 and HCN2 channels is their differential sensitivity to cAMP (6,7). Hormones and neurotransmitters acting through the second messenger cAMP finely tune the activation of the pacemaker current by shifting the voltage dependence of activation along the voltage axis (1, 2). The underlying mechanism involves direct binding of cAMP to the channel cyclic nucleotide binding domain (29). Similar to studies using excised inside-out patches (6, 7), we found that activation of a coexpressed G-protein-coupled receptor, either the 5-HT4 (a) or -opioid receptor, shifts the V1 ⁄2 -value by 14.9 mV toward more positive potentials for homomeric HCN2 channels and no shift for homomeric HCN1 channels (Fig. 3A). Intriguingly, heteromeric HCN1⅐HCN2 channels display a cAMP sensitivity that is intermediate to homomeric HCN1 and HCN2 channels, with a 7.2 mV shift toward more positive potentials in response to activation of coexpressed G-protein-coupled receptor. Representative current traces from experiments using homomeric and heteromeric channels coexpressed with -opioid or 5-HT4 (a) receptors are shown in Fig. 3, B and C, respectively. Based on the structural similarities of the 6 transmembrane domains, the pore-forming region and cytoplasmic N and C termini of HCN channels with other K ϩ channels, it has previously been suggested that HCN channels will most likely conform to the tetrameric subunit arrangement of both the 6and 2-transmembrane domain K ϩ channels (4). With respect to our results with HCN1⅐HCN2 tandems, this idea would suggest that functional channels are likely to be formed by coassembly of two HCN1⅐HCN2 tandems, adopting a HCN1/HCN2/HCN1/ HCN2 or HCN1/HCN2/HCN2/HCN1 subunit stoichiometry. Our experimental results on the cAMP sensitivity of hetero-  2. Analysis of activation kinetics for HCN channels. A, superimposed current traces at Ϫ90 mV for homomeric HCN1 and HCN2 channels as well as heteromeric channels obtained by coinjection of HCN1 and HCN2 or tandem-linked HCN1⅐HCN2 channels. Records were normalized to the steady-state current amplitude after a 4-s pulse (30 s for HCN2). Also shown are algebraic sums of HCN1 and HCN2 currents in ratios of 1:3, 1:1, and 3:1. B and C, fast (filled symbols) and slow (unfilled symbols) time constants of activation determined for homomeric HCN1 (triangles) and HCN2 (squares) channels and heteromeric tandem HCN1⅐HCN2 channels (circles). Dashed line indicates the average between triangles and squares.
Taken together, our data raise the question about the relevance to native pacemaker currents. To address this issue, it is important to note that HCN isoforms not only differ in their biophysical properties and cAMP sensitivity, but also in their distinct distribution pattern across the brain (12,13,30). Nevertheless, considerable overlap was shown between HCN1 and HCN2 using in situ hybridization, namely in hippocampal pyramidal cells and some brainstem nuclei (12). Moreover, the coexistence of HCN1 and HCN2 mRNA has been demonstrated at the single cell level for hippocampal CA1 neurons (13), raising the possibility that heteromeric channel complexes could form in some areas. In Table I, we have compared our data to the results obtained by Franz et al. (13) based on the similar recording conditions (room temperature, whole cell configuration) and data analysis (double exponential fits to the time course of activation). The cAMP-induced shift of the V1 ⁄2 -value for I h in hippocampal CA1 neurons correlates well with the shift that we observed for heteromeric HCN1⅐HCN2 channels. This result can be expected from a population of channels that either consists of equal amounts of homomeric HCN1 and HCN2 channels or a population of channels with a heteromeric stoichiometry, similar to the tandem-linked channels. Comparison of the kinetic properties and the voltage dependence of activation for I h in hippocampal CA1 neurons with our results however favors the formation of heteromeric channels, because equally distributed homomeric channels cannot account for the fast gating properties of the native I h .
In conclusion, the cAMP dependence and gating properties of tandem-linked HCN1 and HCN2 subunits reveal that HCN channels can function as heteromers, most likely tetramers. Based on the resemblance of the tandem it's distinct current phenotype with currents obtained from coexpressed HCN1 and HCN2 subunits and native pacemaker current in neurons in which different isoforms coexist, our results suggest that heteromerization of different HCN channels may further contribute to the functional heterogeneity of native pacemaker currents. Finally, the dimeric construct may also serve as a model to elucidate the physiological role of native pacemaker currents and as a relevant pharmacological tool to develop new compounds for these channels.