Proteolytic Processing of HCN2 and Co-assembly with HCN4 in the Generation of Cardiac Pacemaker Channels*

In sino-atrial and atrio-ventricular nodal cells, hyperpolarization-activated cyclic nucleotide-gated (HCN) inward current carrying cationic channels, If, are expressed that contribute importantly to the diastolic depolarization critical for cardiac pacemaker activity. Although previous studies have demonstrated myocardial expression of both the HCN2 and HCN4 subunits, the specific roles of these subunits in the generation of functional myocardial If channels remain unclear. To explore the molecular compositions of functional cardiac If channels, antibodies targeted against specific C- and N-terminal sequences in HCN2 and HCN4 were exploited to examine HCN2 and HCN4 subunit expression in adult (mouse) heart and to immunoprecipitate endogenous HCN-encoded cardiac If channel complexes. Western blot experiments revealed that although the full-length HCN2 (105 kDa) and HCN4 (160 kDa) proteins are readily detected in transiently transfected HEK-293 cells and in adult (mouse) brain, the molecular mass of the HCN2 protein in the myocardium is ∼60 kDa. In addition, the myocardial 60-kDa HCN2 protein lacks the C terminus, which contains the cAMP binding domain. In heterologous cells, the C-terminal-truncated HCN2 protein co-assembles with HCN4 to form functional heteromeric HCN channels, which activate faster than homomeric HCN2 or homomeric HCN4 channels, and display properties similar to endogenous myocardial If channels Taken together, these results suggest that functional myocardial If channels reflect the heteromeric assembly of HCN2 and HCN4 subunits and further that the HCN4 subunit underlies the cAMP-mediated regulation of cardiac If channels.

In sino-atrial and atrio-ventricular nodal cells, hyperpolarization-activated cyclic nucleotide-gated (HCN) inward current carrying cationic channels, I f , are expressed that contribute importantly to the diastolic depolarization critical for cardiac pacemaker activity. Although previous studies have demonstrated myocardial expression of both the HCN2 and HCN4 subunits, the specific roles of these subunits in the generation of functional myocardial I f channels remain unclear. To explore the molecular compositions of functional cardiac I f channels, antibodies targeted against specific C-and N-terminal sequences in HCN2 and HCN4 were exploited to examine HCN2 and HCN4 subunit expression in adult (mouse) heart and to immunoprecipitate endogenous HCN-encoded cardiac I f channel complexes. Western blot experiments revealed that although the full-length HCN2 (105 kDa) and HCN4 (160 kDa) proteins are readily detected in transiently transfected HEK-293 cells and in adult (mouse) brain, the molecular mass of the HCN2 protein in the myocardium is ϳ60 kDa. In addition, the myocardial 60-kDa HCN2 protein lacks the C terminus, which contains the cAMP binding domain. In heterologous cells, the C-terminal-truncated HCN2 protein co-assembles with HCN4 to form functional heteromeric HCN channels, which activate faster than homomeric HCN2 or homomeric HCN4 channels, and display properties similar to endogenous myocardial I f channels Taken together, these results suggest that functional myocardial I f channels reflect the heteromeric assembly of HCN2 and HCN4 subunits and further that the HCN4 subunit underlies the cAMP-mediated regulation of cardiac I f channels.
Hyperpolarization-activated cyclic nucleotide-gated (HCN) 2 cationic currents are responsible for generating spontaneous pacemaker potentials in the heart (1-3) and in the central nervous system (4). Four members of the HCN family, HCN1-4, have been identified and extensively characterized in heterologous cells (5)(6)(7)(8)(9). Similar to voltage-gated K ϩ (Kv) channels, each HCN channel subunit contains six transmembrane domains and a pore region with a GYG signature motif (5,6). Heterologous expression of the various HCN subunits reveals hyperpolarization-activated cationic inward currents with distinct voltage-and time-dependent properties, as well as differential sensitivities to cAMPdependent modulation (8,9). In the myocardium, HCN4 is the most abundantly expressed transcript (5), whereas HCN1 is the primary HCN transcript in brain (5) and HCN2 is readily detected in both heart and brain (7). Previous studies have shown that the conserved N-terminal 52-amino acid sequence in each of the HCN proteins plays critical roles in channel assembly and trafficking (10). In addition, there is a cyclic nucleotide binding domain (120 amino acids) in the C termini of the HCN subunits (8,11,12), although the C-terminal amino acid sequences of each (HCN1-4) of the HCN proteins are distinct.
Although the molecular compositions of functional cardiac and neuronal I f channels have not been determined, it is clear that the HCN1-4 subunits can generate functional homomeric or heteromeric channels in heterologous expression systems (11)(12)(13)(14). HCN1, for example, co-assembles with HCN2 (11,13) or with HCN4 (14) to form heteromeric channels with properties distinct from each of the homomeric (HCN1, HCN2, or HCN4) channels. It has also been reported that an HCN2 pore mutant produces dominant-negative suppression of HCN4-encoded (as well as HCN2-encoded) currents in heterologous cells and, in addition, reduces cardiac I f densities (15). Studies focused on exploring directly the molecular correlate(s) of functional cardiac I f channels have taken advantage of mice (Hcn2 Ϫ/Ϫ or Hcn4 Ϫ/Ϫ ) harboring targeted disruptions of the Hcn2 or Hcn4 locus (16 -18). I f are lower in cells isolated from Hcn2 Ϫ/Ϫ mice, although, interestingly, the modulatory effects of cAMP on I f are similar to those measured in wild-type cells (16). These results were interpreted as suggesting that HCN2 is primarily responsible for the cAMP-insensitive component of the cardiac pacemaker current (16). In contrast, targeted deletion of Hcn4 is embryonic lethal (17), suggesting that HCN4 is required for early development. Subsequent studies, however, demonstrated that the loss of HCN4 in adult mouse heart is not lethal and, in addition, that I f densities are reduced markedly (but not eliminated) in Hcn4 Ϫ/Ϫ sinoatrial nodal cells (18).
Taken together, the analyses of these genetically engineered mouse models suggest roles for both HCN4 and HCN2 in the generation of cardiac I f channels. Expression of HCN2 or HCN4 (alone) in heterologous (HEK-293) cells, however, produces (homomeric) HCNX-encoded currents with activation thresholds more negative than those observed for (native) I f in cardiac cells (19). In addition, the currents produced on coexpression of HCN1 and HCN2 display gating kinetics similar to cardiac pacemaker currents (20,21). These observations suggest the interesting possibility that functional myocardial HCN channels reflect the heteromeric assembly of HCN subunits. Consistent with this suggestion, resonance energy transfer experiments suggest that homomeric and heteromeric HCN channels are present in native tissues (22).
Exploiting immunoprecipitation and Western blot analyses, the experiments here reveal that HCN2 co-immunoprecipitates with HCN4 from adult (mouse) heart, suggesting that cardiac HCN channels reflect the heteromeric assembly of HCN2 and HCN4 subunits. In addition, the biochemical data presented reveal that HCN2 protein expressed in adult (mouse) heart is a C-terminal-truncated protein, lacking the cAMP binding domain, further suggesting that cAMP modulation of functional cardiac HCN channels is mediated by HCN4.

MATERIALS AND METHODS
cDNA Constructs and Transient Transfections-Mouse HCN2 and HCN4 cDNAs in pCI (Invitrogen, Carlsbad, CA) were obtained from Dr. Richard Robinson (Columbia University College of Physicians and Surgeons). The HCN2 and HCN4 cDNAs were subcloned into the eukaryotic expression vector pcDNA3 (Invitrogen). A hemagglutinin (HA) epitope tag was added at the C terminus of the HCN2 protein with the following pair of primers: 5Ј-TAT GAC CGT GAGA TGG TGC-3Ј and 5Ј-GC TCT AGA GC TCA AGC GTA ATC TGG AAC ATC GTA TGG GTA CAA GTT GGA AGA GAG GCG-3Ј, and this construct was also subcloned into the pcDNA3. The C-terminal HCN2 deletion construct, HCN2⌬C, was generated using primers 5Ј-CAC CAG TGG GAA GAG ATT TTC-3Ј and 5Ј-TCT AGA TCA GAC GAA GTT GGG GTC-3Ј. After the subcloning, the sequences of all constructs were confirmed by automated DNA sequencing (Biotech Center of the University of Wisconsin). The HCN2⌬C cDNA was then subcloned into the 5Ј-end of a bicistronic vector (pIRES-GFP1, kind gift from David Johns, Johns Hopkins University). This construct allowed the expression of HCN2⌬C and EGFP under control of the cytomegalovirus (CMV) promoter.
HEK-293 cells, obtained from the American Tissue Culture Collection (ATCC) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 100 units/ml penicillin, and 100 units/ml streptomycin in a 37°C 5% CO 2 , 95% air incubator as previously described (23). Cells were transfected at ϳ80% confluence using Fugene (Roche Applied Science, Indianapolis, IN) according to the manufacturer's instructions. For electrophysiological experiments, ϳ1 ϫ 10 5 HEK-293 cells were seeded on 35-mm dishes with 1.5-ml culture media. For biochemical experiments, ϳ6 ϫ 10 5 HEK-293 cells were plated on 100-mm dishes with 10 ml of culture media. The generation of the stable HCN4-expressing HEK-293 cell line was described previously (24). Approximately 0.7 g (for electrophysiology experiments) or 4 g (for biochemical experiments) of total DNA was transiently transfected to wild-type HEK-293 cells or into HEK-293 cells stably expressing HCN4 (24).
Immunoblots and Immunoprecipitations-Polyclonal anti-peptide antibodies, generated against N-terminal residues in human HCN2 (residues 147-161) and HCN4 (residues 119 -155) were obtained from Alomone Labs, Ltd. (Jerusalem). These antibodies are referred to here as anti-N2-HCN2 and anti-N-HCN4, respectively. A monoclonal antibody against a C-terminal peptide (residues 761-883) in rat HCN2 (referred to here as anti-C2-HCN2) was developed by and obtained from the UC Davis/NIH Neu-romab Facility, supported by National Institutes of Health Grant U24NS050606 and maintained by the University of California at Davis. The polyclonal anti-Na v 1.5 antibody was purchased from Upstate, Inc. (Lake Placid, NY). In addition, rabbit polyclonal antipeptide antibodies were generated against residues 43-59 and 847-863 in the N and C termini, respectively, of HCN2 (Zymed Laboratories Inc. Antibodies, Inc., Burlingame, CA). These antibodies, referred to here as anti-N1-HCN2 and anti-C1-HCN2, were affinity-purified and screened for specificity and cross-reactivity against HCN2 and HCN4. Similarly, a rabbit polyclonal antibody (anti-C-HCN4) was generated against C-terminal residues 758 -779 in HCN4 (Zymed Laboratories Inc. Antibodies, Inc., Burlingame, CA), purified and characterized. For Western blots, cell lysates were prepared as described previously (25,26) and heart protein samples were isolated from whole mouse hearts. Protein concentrations in all lysates were measured using a DC protein assay kit (Bio-Rad) and a ELx808 microplate reader (BioTek, Winooski, VT) at a wavelength of 750 nm. As also described previously (23,24), proteins were fractionated on SDS-PAGE gels, transferred to polyvinylidene difluoride membranes and immunoblotted with one of the antibodies described above, followed by incubation with a goat anti-mouse or anti-rabbit horseradish peroxidase-conjugated secondary antibody. A fullrange Rainbow molecular weight marker was applied to assess the molecular weight of protein samples (GE Healthcare, Piscataway, NJ). Following washing, membranes were incubated with the SuperSignal West Dura Extended duration substrate (Pierce) and exposed to a Bio-imaging system (UVP, Upland, CA) for visualization.
Electrophysiological Recordings-Whole cell HCN-encoded inward currents were recorded from wild type and HCN4-expressing HEK-293 cells 24 h after transient transfections with the HCN2 or HCN2⌬C cDNA constructs; all recordings were obtained at room temperature. The extracellular recording solution contained (in mM): NaCl 110, MgCl 2 0.5, KCl 30, CaCl 2 1.8, and HEPES 5. Recording pipettes contained (in mM): NaCl 10, MgCl 2 0.5, KCl 130, HEPES 5, and EGTA 1. EGFP-expressing cells were identified under epifluorescence illumination prior to recordings. Electrophysiological experiments were controlled, and data were collected using an Axopatch 200B amplifier interfaced to a Digidata 1322A data acquisition system using the pCLAMP 9 software package (Molecular Devices, Sunnyvale, CA). Recording electrodes were fabricated from borosilicate glass (WP Instruments) on a two stage (P-87; Sutter Instrument) vertical puller. Whole cell HCN-encoded currents were evoked in response to hyperpolarizing voltage steps to potentials between Ϫ60 and Ϫ130 mV from a holding potential of Ϫ40 mV; voltage clamp protocol is illustrated below the experimental data in Fig. 5. A one-way analysis of variance with Bonferroni correction was performed to determine the statistical significance of differences among groups of cells (mean data); where appropriate, p values are presented in the text.

RESULTS
Co-Immunoprecipitation of HCN2 and HCN4 in Heart-Previous studies have shown that HCN2 and HCN4 can co-assemble in heterologous cells, raising the interesting possibility that functional myocardial HCN-encoded I f channels are heteromeric. To test directly the hypothesis that HCN2 and HCN4 associate in the myocardium in situ, (mouse) heart proteins were isolated and immunoprecipitated with specific anti-HCN2 and anti-HCN4 antibodies. Parallel experiments using a rabbit IgG as a negative control were also performed. Proteins precipitated with the anti-N-HCN4 antibody (see "Materials and Methods") were fractionated on SDS-PAGE gels and immunoblotted with the anti-N2-HCN2 antibody. A parallel immunoblot of the heart protein lysate (without immunoprecipitation) was also probed with the anti-N2-HCN2 antibody. A single ϳ60-kDa protein band was detected with the anti-N2-HCN2 antibody in the Western blot of the fractionated total heart protein lysate (Fig. 1A, lane 4). The same band was identified in the sample immunoprecipitated with the anti-N-HCN4 antibody (Fig. 1A, lane 2), but this band was not present in the sample immunoprecipitated with the rabbit IgG (Fig. 1A, lane 1) or in the wash following the immunoprecipitation (Fig. 1A, lane 3).
Co-immunoprecipitation of HCN2 and HCN4 was also observed in a "reverse" (immunoprecipitation) experiment in which the heart protein lysates were precipitated with the anti-N2-HCN2 antibody and were then immunoblotted with the anti-N-HCN4 antibody. A single ϳ160-kDa band was detected in the heart protein lysate prior to immunoprecipitation (Fig. 1B, lane 4).
A protein band at the same molecular mass was identified in the blot of fractionated proteins following immunoprecipitation with the anti-N2-HCN2 antibody (Fig. 1B, lane 2). In contrast, the 160-kDa band was not evident in the blots of fractionated mouse heart proteins following immunoprecipitations with rabbit IgG (Fig. 1B,  lane 1) or in the wash (Fig. 1B, lane 3). The results were confirmed by three independent experiments using different heart samples.

C-terminal Modification of HCN2
in Heart-Although the results described above clearly suggest association of HCN2 and HCN4 in (adult mouse) heart, the Western blots probed with the anti-N2-HCN2 antibody (Fig. 1A) revealed a single protein band at a much lower molecular mass (ϳ60 kDa) than was expected (ϳ 105 kDa) based on the predicted amino acid sequence of the fulllength HCN2 protein. Interestingly, examination of Western blots probed with anti-HCN2 antibodies that have appeared in several published reports also revealed the presence of a low molecular mass (ϳ60 kDa) HCN2 protein in canine and rat heart lysates (27)(28)(29). To explore further the hypothesis that this 60-kDa protein reflects HCN2 expression in adult mouse heart, we generated two additional polyclonal anti-peptide antibodies targeting the N and C termini of HCN2 and used these, together with the Neuromab monoclonal anti-C2-HCN2 antibody (see "Materials and Methods"), to further examine HCN2 expression. One antibody, referred to as anti-N1-HCN2, was generated against amino acids (43-59) in the N terminus of HCN2 and the other, referred to as anti-C1-HCN2, was targeted against amino acids (847-863) in the C terminus of HCN2.
Western blots of heart and brain proteins with these antibodies (and with the anti-N2-HCN2 used in Fig. 1) were completed on isolated mouse brain and heart proteins, as well as on protein lysates from HEK-293 cells transiently transfected with HCN2 cDNA. As illustrated in Fig. 2, these experiments revealed that both the anti-N1-HCN2 and anti-N2-HCN2 antibodies detected the expected full-length (ϳ105 kDa) HCN2 protein in total brain lysates and in lysates from HEK-293 cells transfected with the fulllength HCN2 construct (Fig. 2, A and B). In the heart lysates, however, both the anti-N1-HCN2 and the anti-N2-HCN2 antibodies detected only a ϳ60-kDa band (Fig. 2, A and B). On the other hand, nothing was detected in the heart lysates using either the anti-C1-HCN2 antibody or the monoclonal anti-C2-HCN2 antibody (Fig.  2, C and D). Importantly, the full-length ϳ105-kDa HCN2 protein was detected with both the anti-C1-HCN2 and the anti-C2-HCN2 antibodies in the Western blots of the protein lysates prepared from mouse brain and from HEK-293 cells transiently transfected with the HCN2 cDNA (Fig. 2, C and D). Similar results were obtained in six (6) independent experiments using different heart (brain) samples; the unique presence of the 60-kDa (HCN2) protein band in heart is highly reproducible. In addition, only the 60-kDa protein is identified in the heart samples; there is no evidence for the expression of a high molecular mass (ϳ105 kDa) HCN2 protein in the heart (as is identified in brain).  A3 and B3). As is also evident, neither HCN2 (A) nor HCN4 (B) was immunoprecipitated using rabbit IgG (lanes A1 and B1).
Although all protein isolation steps were handled on ice and in the presence of protease inhibitors to reduce protease-dependent effects and to minimize protein degradation, it is certainly possible that the presence of the 60-kDa HCN2 protein in the heart samples does, in fact, reflect protein degradation. To explore this hypothesis directly, the same heart protein samples were used to examine HCN4 and Nav1.5 protein expression. Western blots probed with the anti-N-HCN4 or anti-C-HCN4 antibody revealed the presence of a single protein band at 160 kDa corresponding to full-length HCN4 (Fig. 3, A and B). Similarly, a single protein band at ϳ250 kDa was observed with the anti-Nav1.5 antibody (Fig. 3C). Although these observations clearly suggest that protein degradation is not a general problem, it remains possible that the HCN2 protein is specifically degraded during one or more steps in the procedures used. To explore the possible role of HCN2 protein degradation during sample handling, an HA-tagged HCN2 construct was generated and expressed in HEK-293 cells. Lysates from the HA-tagged HCN2-expressing HEK-293 cells were prepared and mixed with heart lysates at a (mass) ratio of 1:10 for 1 h in ice. Western blots were then run and probed with the anti-HA or the anti-N2-HCN2 antibody.
As illustrated in Fig. 4A (lane 2), full-length HA-tagged HCN2 (ϳ105 kDa) was detected with the anti-HA antibody in samples prepared by mixing lysates from the HA-tagged HCN2-expressing cells with the heart lysates, as well as from the HEK-293 cells transiently transfected with the HA-tagged HCN2 cDNA (Fig. 4A,  lane 3). Control (antibody) experiments revealed that nothing was detected with the anti-HA antibody in Western blots of fractionated proteins from heart (Fig. 4A, lane 1) or from HEK-293 cells transfected with untagged HCN2 (Fig. 4A, lane 4). The membranes were then stripped and immunoblotted with the anti-N2-HCN2 antibody. With this antibody, the full-length HCN2 (ϳ105 kDa) protein was again detected in samples prepared by mixing lysates of HEK-293 cells expressing the HA-tagged HCN2 (Fig. 4B, lane 3) or the untagged HCN2 (Fig. 4B, lane 4) and the heart lysates (Fig. 4B, lane 2). In contrast, as observed in previous experiments (Fig. 1), only the ϳ60-kDa band was detected in the heart lysates (Fig. 4B, lane 1). In addition, the ϳ60-kDa HCN2 protein was detected in the samples prepared by mixing lysates of HEK-293 cells expressing HA-tagged HCN2 with the heart lysates (Fig. 4B, lane 2). The full-length (ϳ105 kDa) HCN2 protein is also detected with the anti-N2-HCN2 antibody in the mixed (heart and HEK-293 cell) lysates. These observations, together with the results presented in Fig. 3, clearly suggest that the finding of the ϳ60-kDa HCN2 protein in heart lysates ( Figs. 1 and 2) is not an artifact of protein degradation that occurs during the protein isolation and/or handling stages. Rather, these results indicate that the mature myocardial HCN2 protein lacks the C terminus (see "Discussion"). In addition, as illustrated in Fig. 1, the mature HCN2 protein associates with HCN4 in the myocardium, suggesting a functional role in the generation of cardiac HCN channels (see "Discussion").
Properties of HCN Currents Encoded by the C-terminal-truncated HCN2-Previous studies have shown that pacemaker currents are reduced, although the effects (on the currents) of cAMP are the same in cells isolated from mice harboring a targeted disruption in the HCN2 locus as in wild-type myocytes (16). Results from the Western blot analyses here revealed that the mature HCN2 protein in the myocardium has a molecular mass of about FIGURE 2. Post-translational modification at the C terminus of the HCN2 protein in the heart. Proteins isolated from HEK-293 cells transiently transfected with (full-length) HCN2 and from mouse brain or heart were fractionated and immunoblotted (IB) with the anti-N1-HCN2 (A), anti-N2-HCN2 (B), anti-C1-HCN2 (C), or anti-C2-HCN2 (D) antibody, as indicated. HCN2 protein bands were indicated by a closed arrow. As expected, the full-length (ϳ105 kDa) HCN2 protein was readily and reliably detected in the samples prepared from the HCN2-expressing HEK-293 cells. In addition, a single ϳ105-kDa band was identified in brain. In contrast, only a much lower molecular mass, ϳ60-kDa band was detected in the heart protein samples using either the anti-N1-HCN2 or the anti-N2-HCN2 antibody. In addition, there were no proteins identified in the heart samples using either the anti-C1-HCN2 or the anti-C2-HCN2 antibody, although both antibodies detected the fulllength HCN2 protein in brain and in HCN2-expressing HEK-293 cells.
60 kDa and lacks the C terminus, suggesting that the mature HCN2 protein in the myocardium does not contain a cAMP binding domain (8). A C-terminal-truncated HCN2 variant (HCN2-⌬C) was generated with a deletion of C terminus (from Thr-527) to explore this hypothesis further. The HCN2⌬C construct lacks the C-terminal cAMP binding domain and has a predicted molecular mass of 59 kDa. Functional studies revealed that no HCNencoded currents were detected in HEK-293 cells expressing HCN2⌬C (Fig. 5A). In contrast, currents were readily recorded in HEK-293 cells co-expressing HCN2⌬C with HCN4 (Fig. 5A), as well as in cells expressing HCN2 or HCN4 alone (Fig. 5A).
Analyses of the properties of the evoked currents also revealed that co-expression of HCN2⌬C with HCN4 shifted the voltage dependence of current activation to more positive potentials (Fig. 5B), and increased the half-maximal activation voltage, V 1/2 , by 4 mV (Fig. 5C), compared with cells expressing HCN4 alone. The most dramatic change in the current waveforms seen in cells co-expressing HCN2⌬C with HCN4, however, was the marked decrease in the time constant of current activation. Co-expression of HCN2⌬C with HCN4 markedly accelerated the kinetics of current activation, compared with the currents produced on expression of HCN4 alone (Fig. 5D). In addition, heteromeric HCN channels produced by HCN2⌬C/HCN4 co-expression display faster activation kinetics than homomeric HCN channels produced on expression of wildtype HCN2 alone (Fig. 5D). It is also interesting to note that heteromeric HCN2⌬C/HCN4 channels display activation time constant similar to the time constant of endogenous I f channel activation (16 -18, 30). Taken together, these results suggest that post-translationally modified HCN2 in the heart co-assembles with HCN4 to form functional heteromeric HCN channels.

DISCUSSION
Heteromeric Assembly of HCN2 and HCN4 in the Myocardium-The results of the biochemical experiments completed and presented here demonstrate that the HCN2 and HCN4 subunits co-immunoprecipitate from (adult mouse) heart, suggesting that these subunits co-assemble in situ to form functional cardiac HCN-encoded pacemaker channels. This conclusion is in accord with previous immunohistochemical findings showing that HCN2 and HCN4 are co-localized in (embryonic) mouse heart (22). The results here importantly extend these earlier findings with the demonstration that HCN2 and HCN4 can be co-immunoprecipitated from heart. Interestingly, the results presented here are also consistent with several lines of evidence suggesting functional roles for both HCN2 and HCN4 in the generation of functional cardiac I f channels. Mice (Hcn4 Ϫ/Ϫ ) harboring a targeted disruption in the Hcn4 locus lack HCN4 and die early during embryonic development (17). Electrophysiological recordings from embryonic Hcn4 Ϫ/Ϫ myocytes, however, revealed that pacemaker currents were barely detectable and could not be stimulated by application of cAMP (17), suggesting a critical role for HCN4. Consistent with this suggestion, cardiac specific loss of HCN4 reportedly resulted in marked reductions in I f densities (18). In animals in which the Hcn2 locus has been deleted by homologous recombination, in contrast, I f amplitudes/densities were reportedly reduced by only about 30% (16). In addition, although Hcn2 Ϫ/Ϫ mice survive, I f in Hcn2 Ϫ/Ϫ myocytes activates more slowly than I f measured in wild-type cells (16). The responses of Hcn2 Ϫ/Ϫ cells to cAMP stimulation appear to be indistinguishable fromthosemeasuredinwild-typecells (16).Inlightofthebiochemical findings here, it seems reasonable to suggest that the electrophysiological findings in these targeted deletion animals are consistent with thesuggestionthatfunctionalcardiacI f channelsreflecttheco-assem- . Detection of full-length channel proteins in heart lysates. Proteins, isolated from HEK-293 cells stably expressing HCN4 and from adult mouse heart, were fractionated on SDS-PAGE gels and immunoblotted (IB) with the anti-N-HCN4 antibody (A) or the anti-C-HCN4 antibody (B). HCN4 and Nav1.5 protein bands were indicated by a closed arrow. A ϳ160-kDa band was detected in samples from the HEK-293 cell line stably expressing HCN4 and in mouse heart. Nothing was detected in untransfected (wild type) HEK-293 cells with either of these (anti-HCN4) antibodies. C, similarly, a single, high molecular mass protein (at ϳ250 kDa) was detected in lysates from adult mouse heart and from HEK-293 cells stably expressing Nav1.5, but not in extracts from untransfected HEK-293 cells. In contrast, and as expected, nothing is detected with this antibody in the heart lysates (lane 1) or in lysates from untransfected HEK-293 cells (lane 4). Following stripping, the membrane was re-probed with the anti-N2-HCN2 antibody (B). As is evident, the full-length HCN2 and the 60-kDa HCN2 proteins were detected in the sample (lane 2) prepared by mixing the heart lysate with the extract from HCN2-expressing HEK-293 cells, whereas only the fulllength HCN2 is detected in HEK-293 cells (lane 3), and only the 60-kDa protein is detected in the heart (lane 1).
bly of HCN2 and HCN4. In addition, the presence of HCN2 plays an important role in determining the kinetics of I f activation.
It is certainly possible that there are homomeric HCN4 channels that are expressed, i.e. together with heteromeric HCN2: HCN4 channels, in the myocardium (22). Consistent with previously reported findings (15,22), results are also presented here showing that both HCN2 and HCN4 can form homomeric HCN channels in heterologous cells when expressed alone and that the time-and voltage-dependent properties of the (homomeric) currents are distinct (Fig. 5A). It is also possible that the subunit composition(s) of functional cardiac HCN channels is altered (and that homomeric channels play important roles) in myocardial disease. It has been reported, for example, that I f density is increased in heart failure in both rats and in humans (31,32). It is possible that this up-regulation of functional I f densities reflects changes in the relative numbers (and importance) of homomeric versus heteromeric HCN channels and/or alters the subunit stoichiometry of these channels. Further experiments will be necessary to explore these hypotheses directly.
C-Terminal Modification of HCN2 in the Heart-Unlike HCN4, the apparent molecular mass of the HCN2 protein detected in the adult mouse heart (ϳ60 kDa) was much smaller than expected, based on the amino acid sequence of the full-length (ϳ105 kDa) HCN2 protein. Interestingly, the presence of a short form of the HCN2 protein (ϳ60 kDa) in the heart is suggested by previously published Western blot data from other laboratories (28,29), although there was no discussion of these observations and no effort appears to have been made to determine the reason for the (exclusive?) presence of the low molecular weight protein identified. The results presented here revealed that only the short variant of HCN2 is detected in the heart. Northern blot analyses have demonstrated the expression of full-length HCN2 RNA in heart and in brain (5). This observation suggests that the fulllength HCN2 protein is likely made in the heart and modified post-translationally. It might also be suggested that cardiac HCN2 is selectively modified by proteases and/or is particularly sensitive to protein degradation during sample handling/sample preparation. However, this seems unlikely to account for the observations here for several reasons. Using specific antibodies targeting the N and C termini of HCN2, for example, the full-length (HCN2) protein is routinely identified in brain and in extracts from transiently transfected HEK-293 cells expressing HCN2 (Fig. 2). In addition, the same heart protein samples were also analyzed here for the expression of other high molecular mass proteins (Ͼ150 kDa) by Western blot, and these analyses reliably revealed expression of the ϳ160-kDa HCN4 (Fig. 3, A and B) and ϳ250-kDa Nav1.5 (Fig.  3C) proteins. The ability to detect full-length HCN4 and Nav1.5 in (all of) the cardiac protein samples suggests that the finding of the 60-kDa HCN2 protein does not reflect nonspecific proteolysis in the samples studied. In addition, the full-length, HA-tagged HCN2 protein was detected in Western blots of HEK-293 cell lysates mixed with heart lysates (Fig. 4). Taken together, the results of the FIGURE 5. A C-terminal-truncated HCN2 (HCN2⌬C) protein forms heteromeric channels with HCN4 in HEK-293 cells. A, representative recordings obtained from (wild type or HCN4) expressing HEK-293 cells following transient transfection with cDNA constructs encoding the full-length HCN2 or the C-terminal HCN2 truncation in mutant, HCN2⌬C. Representative currents in HEK-293 cells stably expressing HCN4 are also illustrated. In each case, currents evoked during 3 s hyperpolarizing voltage steps to test potentials between Ϫ130 and Ϫ60 mV from a holding potential of Ϫ40 mV are illustrated; the voltage clamp paradigm is illustrated below the records. B, peak currents at each test potential in each cell were measured and normalized to the maximal peak current recorded in the same cell, and mean Ϯ S.E. normalized peak currents are plotted as a function of the test potential; the individual curves were fitted to the Boltzmann equation (see "Materials and Methods"). Mean Ϯ S.E. voltages (V 1/2 ) of half-maximal current activation (C) and activation time constants (D) are illustrated; n values are presented above the bars (*, p Ͻ 0.05). experiments here suggest that the low molecular mass HCN2 protein detected in heart is not an experimental artifact that can be attributed to protein degradation, but rather suggests that the HCN2 protein is modified post-translationally in the myocardium. The immunoblot data also reveal that the mature cardiac HCN2 is a C-terminal-truncated protein, although the exact location of the amino acid truncation site remains to be identified.
The mature cardiac HCN2 protein is not detected with antibodies targeting the HCN2 C terminus that reliably detect full-length HCN2 protein in brain and in heterologous cells (Fig. 2) indicating that this protein is modified in the C-terminal domain. In addition, the low molecular mass (ϳ60 kDa) of the identified protein suggests that the mature cardiac HCN2 protein lacks the cAMP binding domain and will, therefore, be insensitive to cAMP stimulation. It is probably not surprising, therefore, that I f in Hcn2 Ϫ/Ϫ myocytes is modulated by ␤-adrenergic (cAMP) stimulation and that the observed effect of ␤-stimulation is indistinguishable from that measured in wild-type cells (12). Previous studies in Xenopus oocytes suggest that removal of the cAMP binding domain in HCN2 modifies the activation kinetics of HCN-encoded currents, although HCN-encoded currents are not observed in mammalian cells expressing the C-terminal-truncated HCN2 protein (Fig. 5), presumably due to trafficking defects (9). The HCN2 protein lacking the C-terminal cAMP binding domain, can, however, co-assemble with HCN4 to form functional heteromeric HCN channels in mammalian cells (Fig. 5) and interestingly, the time-dependent property of the heteromeric currents closely resemble those of endogenous cardiac I f channels.
In summary, the results of immunoprecipitation, immunoblot, and patch clamp analyses here show that a C-terminal modified HCN2 protein is expressed in adult mouse heart and that this protein co-immunoprecipitates with HCN4, suggesting that functional cardiac I f channels are heteromeric, comprising HCN2 and HCN4 subunits. The stoichiometry of functional cardiac I f channels remains to be determined. The mature cardiac HCN2 protein appears to lack the C-terminal cAMP binding domain, based on the observed molecular mass and the fact that no proteins are detected in heart lysates specific C-terminal anti-HCN2 antibodies that reliably detect fulllength HCN2 protein in extracts from (adult mouse) brain and HCN2-expressing heterologous cells. These observations suggest that the HCN4 subunit underlies cAMP-mediated modulation of cardiac I f . The results presented also demonstrate that the properties of heteromeric HCN2⌬C:HCN4 currents more closely resemble cardiac I f channels than do the currents produced on expression of HCN4 (or HCN2) subunits alone.