Department of General Physiology and Biochemistry, Laboratory of Molecular Physiology and Neurobiology, via Celoria 26, 20133 Milano, Italy
INFM-Milano U. Unit, via Celoria 26, 20133 Milano, Italy
Corresponding author D. DiFrancesco: Department of General Physiology and Biochemistry, Laboratory of Molecular Physiology and Neurobiology and INFM- Milano U. Unit, via Celoria 26, 20133 Milano, Italy. Email: email@example.com
‘Funny-’ (f-) channels of cardiac sino-atrial node (SAN) cells are key players in the process of pacemaker generation and mediate the modulatory action of autonomic transmitters on heart rate. The molecular components of f-channels are the hyperpolarization-activated, cyclic nucleotide-gated (HCN) channels. Of the four HCN isoforms known, two (HCN4 and HCN1) are expressed in the rabbit SAN at significant levels. However, the properties of f-channels of SAN cells do not conform to specific features of the two isoforms expressed locally. For example, activation kinetics and cAMP sensitivity of native pacemaker channels are intermediate between those reported for HCN1 and HCN4. Here we have explored the possibility that both HCN4 and HCN1 isoforms contribute to the native If in SAN cells by co-assembling into heteromeric channels. To this end, we used heterologous expression in human embryonic kidney (HEK) 293 cells to investigate the kinetics and cAMP response of the current generated by co-transfected (HCN4 + HCN1) and concatenated (HCN4-HCN1 (4–1) tandem or HCN1-HCN4 (1–4) tandem) rabbit constructs and compared them with those of the native f-current from rabbit SAN. 4–1 tandem, but not co-transfected, currents had activation kinetics approaching those of If; however, the activation range of 4–1 tandem channels was more negative than that of the f-channel and their cAMP sensitivity were poorer (although that of 1–4 tandem channels was normal). Co-transfection of 4–1 tandem channels with minK-related protein 1(MiRP1) did not alter their properties. HCN1 and HCN4 may contribute to native f-channels, but a ‘context’-dependent mechanism is also likely to modulate the channel properties in native tissues.
The ‘funny’ current (If), or cardiac pacemaker current, is highly expressed in the pacemaker cells of the sino-atrial node (SAN) and in other cardiac myocytes able to pace spontaneously, such as atrio-ventricular cells and Purkinje fibres of the conduction tissue. In SAN cells, If drives slow diastolic depolarization and is responsible for generation of spontaneous activity. f-channels carry an inward current at diastolic voltages and are activated by membrane hyperpolarization and by direct binding of intracellular cAMP; the latter mechanism mediates heart rate changes induced by sympathetic and parasympathetic transmitters, which act by modifying cAMP synthesis by stimulation and inhibition of adenylate cyclase, respectively (DiFrancesco, 1993; Accili et al. 2002). The neuronal equivalent of If, the hyperpolarization-activated current (Ih), is expressed in a variety of neuronal tissues where it serves different functions, all linked to its ability to induce a depolarizing action upon appropriate stimulation (Pape, 1996). Ih can therefore control the spontaneous firing rate or modify cell excitability, this latter a function that can be especially relevant to modulation of synaptic strength or sensory detection (Demontis et al. 1999; Beaumont & Zucker, 2000; Stevens et al. 2001; Mellor et al. 2002).
In some cell types, the properties of native hyperpolarization-activated channels are similar to those of individual, heterologously expressed HCN isoforms. This happens for example in rabbit rods, where Ih has features essentially overlapping those of the HCN1 current and where HCN1 mRNA and protein, as identified by immunolabelling, are detected in retinal tissue and in isolated cells (Demontis et al. 2002).
In other cell types, however, the properties of native channels do not appear to conform to those of any of the known isoforms expressed locally. In the rabbit SAN, for example, RNase-protection assay and immunolabelling data (Shi et al. 1999; Moroni et al. 2001) indicate significant expression of HCN4 and HCN1 isoforms, whereas levels of HCN2 message are just above detection, although early work indicated that HCN2 is the predominant cardiac isoform (Santoro & Tibbs, 1999); however, the reported activation kinetics of If in the rabbit SAN (see for example DiFrancesco & Noble, 1989) are apparently rather slower than those of heterologously expressed HCN1 and faster than those of HCN4 (compare with Altomare et al. 2001). Also, homomeric HCN2 channels expressed in human embryonic kidney (HEK 293) cells have atypical features when compared to the properties of the If in the SAN or in Purkinje fibres, including the dependence of channel conductance upon permeating ion concentrations and block by Cs+ (Moroni et al. 2000).
These observations lead to the natural consideration that some of the native channel properties may be determined by co-assembly of different HCN isoforms into heteromers, with or without the contribution of auxiliary subunits. It is indeed known that different HCN isoforms can co-assemble. For example, HCN1 and HCN2 have been co-expressed individually or in concatenated form in oocytes, to yield functional channel proteins with properties distinct from those of the original channels (Ulens & Tytgat, 2001; Chen et al. 2001). Furthermore, recent evidence has indicated that minK-related protein 1 (MiRP1), a protein highly expressed in the cardiac pacemaker region, functions as a β-subunit when co-expressed with HCN1 and HCN2 in Xenopus oocytes, by enhancing expression and accelerating activation of HCN1 and HCN2 (Yu et al. 2001).
In the present work, we used heterologous expression to investigate whether co-expression of the major components of HCN channels in SAN cells, the HCN1 and HCN4 isoforms (Ishii et al. 1999; Moroni et al. 2001), generates channels with properties similar to those of native SAN If channels. To do so we measured the kinetic properties and cAMP sensitivities of the currents generated by co-transfected ((rabbit) rbHCN4 + rbHCN1)) and concatenated (rbHCN4-rbHCN1 or rbHCN1-rbHCN4) constructs expressed in HEK 293 cells and compared them with those of the native cardiac pacemaker current. In order to verify if MiRP1 subunits modify channel properties in the HEK 293 expression system, we also tested the effects of co-expressing MiRP1 with HCN4 alone or with the tandem construct.
rbHCN4 (kindly provided by Dr H. Ohmori, University of Kyoto, Japan) and rbHCN1 were inserted for functional expression into the mammalian expression vectors pCI (Promega) and pcDNA3.1+ (Invitrogen), respectively. The tandem rbHCN4- rbHCN1 (4–1 tandem) construct was prepared in the pCI vector by fusion of rbHCN1 cDNA in frame with the C-terminus of rbHCN4 cDNA. We removed the stop codon of rbHCN4 by cutting with Bsu 36I 46 amino acids (aa) upstream of the stop codon and with Eco RV in the polylinker of the pCI vector. The resulting linearized cDNA fragment (Bsu 36I/Eco RV) was filled blunt and self-ligated to verify, by functional expression, that the short deletion in the C-terminus of rbHCN4 did not modify the kinetic properties of the channel (data not shown). Otherwise, the cDNA of rbHCN1 was ligated to this fragment. We used the Nar I restriction site at the N-terminus of rbHCN1. When filled blunt, it was predicted to keep the frame between the two clones. rbHCN1 was excised from the original vector (pCDNA3.1+) with Nar I and Kpn I, blunted and ligated with rbHCN4 prepared as described above. We verified by restriction analysis that the construct was correct. By using Nar I we lost the first 9 aa of rbHCN1; we have shown previously (Moroni et al. 2001) that deletion of the first 29 aa at the N-terminus of rbHCN1 results in functional channels with the same properties as the wild-type channels.
We used a different procedure to generate the symmetrical tandem rbHCN1-rbHCN4 (1–4 tandem) construct, due to the lack of a common restriction site in the appropriate location. To fuse the two clones and remove the stop codon of rbHCN1, we used a PCR overlap method (Ho et al. 1989). We produced a fusion band containing an Eco RI restriction site at the N-terminus (corresponding to the Eco RI site at position 1749 of rbHCN1) and a Bam HI site at the C-terminus (corresponding to the Bam HI site at position 128 of rbHCN4). This fragment was cloned into a Bluescript vector (construct 1) and verified by sequencing. rbHCN4 was excised from the original vector (pCI) with Bam HI and cloned into the Bam HI site of the construct 1, to obtain construct 2. Finally, the insert was excised from construct 2 by cutting with EcoRI and cloned into the Eco RI site of rbHCN1. All constructs were verified by restriction analysis.
MinK-related protein 1 (MiRP1, kindly provided by Dr S. Goldstein, Yale University, New Haven, USA) was inserted into the mammalian expression vector pIRES-EGFP (Clontech), which allowed both the MiRP1 and the enhanced green fluorescent protein (EGFP) genes to be translated from a single bicistronic mRNA. The pIRES-EGFP vector carries a partially disabled internal ribosome entry site (IRES) sequence to reduce the rate of translation initiation at the EGFP site relative to that at the site of the inserted MiRP1 clone. Visual selection of fluorescent cells thus ensured that in electrophysiology recordings the cells under study expressed the MiRP1 protein at high levels.
Modified HEK 293 cells (Phoenix cells) were transiently transfected according to a standard calcium phosphate protocol, as described previously (Vaccari et al. 1999). The constructs were co-transfected into HEK 293 cells (Kinsella & Nolan, 1996) together with a GFP-containing vector. For each 35 mm Petri dish of 70 % confluent cells, 4 μg of channel plasmid DNA (rbHCN4, rbHCN1, 4–1 tandem or 1–4 tandem) was employed, irrespective of the plasmid length, together with 2 μg of GFP-containing plasmid. When two channels (rbHCN4 and rbHCN1) were co-transfected, the total amount of channel DNA was 4 μg per Petri dish and the relative amounts of the two were calculated in order to transfect equimolar amounts.
When MiRP1 was co-transfected with rbHCN4 (or 4–1 tandem), variable amounts (2–3 μg) of pIRES-EGFP plasmid containing the MiRP1 gene were mixed with 4 μg of rbHCN4 (or rbHCN4- rbHCN1) in order to ensure that the two genes were in equimolar ratio. GFP expression, provided by the pIRES-EGFP plasmid, was taken to indicate that both MiRP1 translation and cell transfection had occurred.
Western blot analysis
Membrane proteins were extracted from rbHCN- or MiRP1- transfected cells and 50 μg protein per lane was run on an 8 % SDS-PAGE gel and then electroblotted onto nitrocellulose filters (Schleicher & Schuell, Germany). Blotted proteins were exposed to blocking solution (Tris, 1 mm; NaCl, 50 mm; Tween 20, 0.05 %; non-fat dried milk, 5 %; pH 7.4) and then treated with the appropriate antibodies diluted (1:200) in the same solution; anti-HCN1 was purified in our laboratory (Moroni et al. 2001), anti-MiRP1 and anti-HCN4 were purchased from Alomone Labs (Jerusalem, Israel). Goat anti-rabbit secondary antibodies conjugated to horseradish peroxidase were used at 1:3000 dilution and visually detected with the ECL method (Amersham Biosciences, Italy).
Transfected HEK 293 cells were plated on Polysine microscope slides (Menzel-Glasser), allowed to settle for 1–2 h and subsequently rinsed with phosphate buffer saline (PBS) solution. Cells were fixed in paraformaldehyde (4 %) for 6–10 min and glass slides were then rinsed for 20 min with PBS containing 0.1 m glycine. Permeabilization was obtained by treating for 30 min with PBS solution containing 1 % BSA and 0.15 % Triton X-100. Primary and secondary antibodies were diluted in a similar solution containing a lower concentration of Triton X-100 (0.05 %). Incubation with the primary antibody (1:200 dilution) was carried out either overnight at 4 °C or for 2 h at room temperature. Secondary antibody (1:1000) incubation was performed for 1 h after thorough washout in PBS solution. A final 2 h washout in PBS solution was carried out before mounting with 100 % glycerol. In control experiments, HCN4- and HCN1-transfected cells were exposed to anti-HCN1 and anti-HCN4, respectively and no signal was detected, thus excluding cross-reactivity as well as non-specific labelling. Anti-MiRP1 specificity was verified by lack of detectable signal in mock-transfected cells.
In Fig. 1, data from Western blot analysis and immunofluorescence staining are shown for co-transfected rbHCN1 + rbHCN4 proteins (A and B) and for MiRP1 proteins co-expressed with GFP (C and D). The data clearly indicate that rbHCN1, rbHCN4 and MiRP1 proteins all reach the cell membrane, where the highest fluorescence signals are visible. rbHCN1 and rbHCN4 Western blot bands and fluorescence signals show comparable levels of expression for the two proteins, suggesting a similar degree of efficiency of the cell transcription/translation machinery for the two clones. Notice also that whereas GFP is expressed ubiquitously in the whole cell, in agreement with its cytoplasmic location, the MiRP1 signal is concentrated on the cell membrane.
Isolation of rabbit cardiac SAN cells
We used standard methods to dissect the SAN region from rabbit hearts and to isolate SAN myocytes (DiFrancesco et al. 1986; Accili et al. 1997b). Hearts were extracted from young animals (about 0.8 to 1 kg in weight) deeply anaesthetized by intramuscular injection of xylazine (4.6 mg kg−1) and ketamine (60 mg kg−1) and killed by cervical dislocation and exanguination. The procedures conformed with guidelines for the care and use of laboratory animals as established by State (D.L. 116/1992) and European directives (86/609/CEE). Following enzymatic dissociation, SAN myocytes were stored at 4 °C and used for the day.
HEK 293 cells were incubated at 37 °C in 5 % CO2 for 1–5 days after transfection and plated at low density in 35 mm plastic Petri dishes 3–5 h before the experiments (Moroni et al. 2000). A Petri dish was placed under the stage of an inverted microscope and GFP-expressing cells were selected under fluorescent light. Aliquots of SAN cells were dispersed after isolation into plastic Petri dishes and allowed to settle for up to 1 h before they were placed under the microscope. Cells (either SAN or HEK 293) were superfused at room temperature (25–26 °C) with Tyrode solution containing (mm): 140 NaCl, 5.4 KCl, 1.8 CaCl2, 5.5 d-glucose and 5 Hepes-NaOH, pH 7.4. The intracellular-like solution used to fill whole-cell pipettes contained (mm): 10 NaCl, 130 KCl, 1.0 EGTA, 5 Hepes-KOH, 0.5 MgCl2, 2 ATP (sodium salt), 0.1 GTP (sodium salt) and 5 phosphocreatine, pH 7.
During voltage-clamp recordings, perfusion of Tyrode solution was switched off and test solutions were delivered by a fast-perfusion device positioned on top of the cell under study. The extracellular solution used in whole-cell experiments contained (mm): 110 NaCl, 30 KCl, 1.8 CaCl2, 0.5 MgCl2 and 5 Hepes- NaOH (pH 7.4); 1 mm BaCl, 2 mm MnCl2, 100 μm NiCl2 and 20 μm nifedipine were added when necessary to reduce possible interference from Ca2+ and K+ conductances.
In macro-patch experiments, cells (either SAN or HEK 293) were superfused with a high-K+ solution containing (mm): 130 KCl, 10 NaOH, 5 Hepes-NaOH (pH 7.2, pCa = 7). Following patch excision and exposure of the inside-out patch to the perfusing solution, this was switched to one containing (mm): 130 potassium aspartate, 10 NaCl, 2 CaCl2, 5 EGTA-KOH, 10 Hepes-KOH (pH 7.2, pCa = 7). Large-tipped pipettes (resistance < 1 MΩ) were filled with a solution containing (mm): 70 NaCl, 70 KCl, 1.8 CaCl2, 1 MgCl2, 1 BaCl2, 2 MnCl2, 5 Hepes-NaOH (pH 7.4). cAMP (10 μm) was added to the control solution when required.
Fractional activation curves of HCN- and f-currents were obtained by standard activation-deactivation two-step protocols (DiFrancesco et al. 1986). Hyperpolarizing steps of variable duration, sufficient to reach steady-state activation at all voltages (see Fig. 3), were applied from a holding potential of −35 mV and current tails were measured upon return to a fixed voltage. After normalization to maximum amplitude, tails were plotted to obtain the activation variable. Experimental data were fitted by the Boltzmann equation:
where y is fractional activation, V is voltage, V1/2 is the half-activation voltage and s is the inverse slope factor.
The time constants of current activation were obtained by fitting the activation time course on hyperpolarization with a single exponential function after an initial delay (Altomare et al. 2001).
Shifts induced by cAMP in the activation curves of HCN currents or of If were obtained by a method not requiring measurement of the conductance-voltage relationship. This involved applying hyperpolarizing steps to near the mid-point of the activation curve and adjusting the holding potential (-35 mV in the control solution) until the cAMP-induced change in current was compensated and the control current size fully restored. Measurement of the displacement of the holding potential (in mV) allowed the estimation of the shift in the activation curve (Accili & DiFrancesco, 1996; Bois et al. 1997).
Experimental data were compared using Student's independent t test and significance level was set to P= 0.05. Comparison between activation time constants (see Figs 3, 5, 7 and 8) was performed by fitting time constant/voltage (τV) curves by the function:
where the voltage dependence of rate constants α and β is described by the equations:
Equation (1) derives from an allosteric model of voltage dependence of HCN channels, based on the hypothesis that channels are tetramers, with each subunit carrying a voltage sensor that switches between ‘reluctant’ and ‘willing’ states (Altomare et al. 2001). Here, the parameter a represents the multiplying factor of equilibrium constants of closed/open transitions upon movement of one voltage sensor from the ‘reluctant’ to the ‘willing’ state (a= 0.2), z is the equivalent unit of electronic charge moving in the electric field during channel opening, and r=RT/F (= 25.85 mV), with R, T and F having their usual thermodynamic meaning (T= 300 °K).
Fixed values of z and β0 were selected by fitting mean τV curves for rbHCN4, rbHCN4 + rbHCN1 and 4–1 tandem channels to eqn (1) and averaging the three values thus obtained. We obtained z= 1.195 and β0= 22.63 s−1. Time constant data from single cells expressing any of the three clones were then fitted to eqn (1) to determine the rate constant α0, which was used to compare data by Student's t test analysis.
Data points were plotted as means ±s.e.m. values. For shortness, in the following text the species prefix will be dropped and rbHCN1 and rbHCN4 will be termed simply HCN1 and HCN4.
Kinetic properties of co-expressed HCN1 + HCN4 channels
In Fig. 2A and B, sample current traces recorded during hyperpolarizing steps to −105 mV from a holding potential of −35 mV from cells expressing HCN1 (A) or HCN4 (B) are superimposed, after normalization to the same amplitude, on a representative If trace recorded with the same protocol from a native rabbit SAN.
The time course of the native current was clearly slower than that of HCN1 and faster than that of HCN4 current. Fitting the time course of each trace with a single exponential following an initial delay (Altomare et al. 2001) yielded time constants of 0.792, 0.223 and 1.677 s for native If, HCN1 and HCN4 currents, respectively. In Fig. 2C, the same native If trace is superimposed onto a sample current trace recorded at −105 mV from a HEK 293 cell co-expressing HCN1 and HCN4. The time course of the HCN1 + HCN4 current was still slower than that of If (time constant of 1.324 s).
Extending the analysis of activation time courses to a wider range of voltages yielded the results shown in Fig. 3. The time constant curve for the SAN If current was intermediate between those of HCN1 and HCN4. According to existing data, If activation can be properly described by a single exponential time course once a brief delay is allowed for (Altomare et al. 2001); this and other evidence, including single-channel recording (DiFrancesco, 1986; DiFrancesco & Mangoni, 1994), argue against the existence of a mixed population of channels with clearly separable kinetic components. The observation in Fig. 3 that If has activation kinetics intermediate between those of HCN1 and HCN4 therefore may suggest that f-channels represent a homogeneous population of heteromeric HCN1 and HCN4 channels. However, co-transfection of HCN1 and HCN4 subunits did not reproduce the native f-channel kinetics and yielded a current whose activation was much slower than that of If. Indeed, as shown in Fig. 3B, the time constants for the HCN1 + HCN4 current were only slightly smaller than those for the HCN4 current. A comparison between HCN4 and HCN1 + HCN4 τV curves performed by fitting data in Fig. 3B to eqn (1) (see Methods) did not reach significance (P > 0.05).
To further analyse the kinetic properties of co-transfected channels, we measured the mean activation curves for the HCN1, HCN4 and HCN1 + HCN4 currents and compared them with that of the native If in SAN cells, as shown in Fig. 4. All curves were measured, as described in the Methods, by using test steps of sufficiently long duration to achieve full current activation at all voltages. As apparent from Fig. 4B, the position of the native channel activation curve was more positive than that of either HCN1 or HCN4. When fitted by the Boltzmann relation, the half activation voltages (V1/2) and the inverse slope factors (s) were: V1/2=−63.7, −73.0 and −79.5 mV and s= 8.9, 8.6 and 10.0 mV for native, HCN1 and HCN4 channels, respectively. These values are comparable with previously published data (Accili et al. 1997; Viscomi et al. 2001; Altomare et al. 2001). Co-transfection of HCN1 + HCN4 yielded a curve which was close to, if slightly more negative than, the HCN4 curve (V1/2=−82.9 mV, s= 10.6 mV). The difference between the V1/2 and s values of HCN4 and co-transfected HCN1 + HCN4 curves was not statistically significant (P > 0.05).
The data shown in Figs 3 and 4 lead to a twofold consideration: firstly, the co-transfection of two HCN isoforms does not necessarily lead to channels with properties that can be obtained by averaging those of the individual components. Similar behaviour has been observed upon the co-transfection of HCN1 and HCN2, which led to channels whose properties could not be derived simply by linear summation of properties of individual components (Ulens & Tytgat, 2001; Chen et al. 2001). Secondly, co-transfection of HCN1 + HCN4 clearly did not reproduce either the activation kinetics or the relatively depolarized position of the activation curve of native f-channels. This may indicate the presence of a ‘context’ dependence of kinetic properties (i.e. a dependence upon modulatory conditions, such as phosphorylation or interaction with cytoskeletal proteins) that can vary between expression systems (Qu et al. 2002), or dependence upon auxiliary subunits.
Kinetic properties of concatenated HCN4-HCN1 channels
Although the modest acceleration of the HCN4 activation time course obtained by co-transfecting the HCN1 isoform with HCN4 (Fig. 3), even if not statistically significant, may suggest the formation of heteromers, the co-transfection data as a whole indicate a predominance of the HCN4 phenotype. On the other hand, the HEK 293 expression system does not allow identification of the stoichiometric ratio of the components of expressed channels, leading to uncertainty as to the molecular basis for the behaviour of co-transfected channels relative to isoforms 1 and 4.
We therefore restricted the study to a fixed stoichiometric ratio of 1:1 by using concatenated HCN4-HCN1 (4–1 tandem) constructs (see Methods). Using the same protocols shown above in Figs 2, 3 and 4, we measured time constants of activation and the activation curve in HEK 293 cells expressing 4–1 tandem channels. As shown in Fig. 5, activation of the 4–1 tandem channels was significantly faster than that of co-transfected heteromeric HCN1 + HCN4 channels (P < 0.05), although still slower than that of native f-channels. On the other hand, the activation curve was nearly superimposable on that of HCN1 + HCN4. The Boltzmann fit of the 4–1 tandem data yielded V1/2=−86.4 mV and s= 9.6 mV, values which were not statistically different from those of the activation curve of co-expressed HCN1 + HCN4 channels (P > 0.05).
In the hypothesis that HCN channels are tetramers, the difference between kinetic properties of 4–1 tandem and co-expressed channels suggests that the prevailing stoichiometric ratio of co-transfected HCN1 + HCN4 channels is not 2:2. The prevalence of the HCN4 phenotype may reflect a preferential assembly of heteromers with 3:1 HCN4:HCN1 stoichiometry ratio, or a ‘dominant’ action of HCN4 subunits in the determination of heteromultimer features.
An explanation for the similarity of the activation curve of 4–1 tandem channels to those of co-transfected channels and HCN4 channels alone is less obvious and this may represent a specific feature of the 4–1 tandem construct that depends upon subunit-subunit interactions (see Discussion; Altomare et al. 2001).
Action of cAMP
The response to cAMP is a discriminative feature of HCN isoforms: sensitivity to cAMP is poor for HCN1, high for HCN2 and the highest for HCN4 channels (Kaupp & Seifert, 2001). The substantial response of native f-channels to cAMP (DiFrancesco & Tortora, 1991; DiFrancesco & Mangoni, 1994; Bois et al. 1997) is one of the important indications that HCN4 is likely to represent a major component of them (Shi et al. 1999). We therefore investigated the response to cAMP of heteromeric channels in inside-out patches from HEK 293 cells and compared it to the cAMP response of individual isoforms, as well as to that of inside-out f-channels from SAN membranes. In all cases except for HCN1, the cAMP sensitivity was evaluated by measuring shifts in the activation curve induced by 10 μm cAMP, as described in the Methods (see also Accili & DiFrancesco, 1996). Inside-out HCN1 channel recording proved to be especially demanding and we therefore measured the cAMP sensitivity of HCN1 by recording in whole-cell conditions the response to a membrane-permeant cAMP analogue (8-(4-chlorophenylthio) adenosine 3′,5′-cyclic monophosphate, pCPT-cAMP, 100 μm). The results are shown in Fig. 6. In agreement with known data, HCN1 channels responded weakly to cAMP, with a mean shift in the activation curve of 5.8 ± 0.9 mV (n= 5), whereas HCN4 channels responded more strongly, with a shift of 16.3 ± 0.9 mV (n= 15). The HCN4 shift was somewhat larger than the shift of native f-channel activation curve (10.3 ± 0.2 mV, n= 31). The activation curve of co-transfected HCN1 + HCN4 channels shifted by 10.3 ± 0.7 mV (n= 12), a value which was not significantly different from that of native channels (P > 0.05), although it differed from that of HCN4 channels (P < 0.05). Surprisingly, the cAMP-induced shift of the 4–1 tandem curve was only 6.0 ± 0.6 mV (n= 10), a value not significantly different from that of the HCN1 curve (P > 0.05).
These data show that although the kinetics of concatenated 4–1 tandem constructs approach those of native channels, their cAMP sensitivity does not, being more similar to that of HCN1 channels. Since the C-terminus determines the sensitivity to cAMP (Viscomi et al. 2001), these data suggest that the action of cAMP is preferentially mediated by cAMP binding to the C-terminus of HCN1.
To check if this assumption is correct, we used 1–4 tandem channels, obtained by ligation of the C-terminus of HCN1 to the N-terminus of HCN4 (see Methods), and analysed their response to cAMP. Indeed, as shown in Fig. 6F and G, 1–4 tandem channels had a sensitivity to cAMP which was significantly larger than that of the 4–1 tandem and similar to that of co-transfected HCN1 + HCN4 and native channels (not significantly different, P > 0.05). These data agree with the hypothesis that, in C-termini which are fused to N-termini and may therefore have a restricted mobility, the function of the cyclic nucleotide-binding domain in mediating cAMP-dependent stimulation is at least partly impaired.
Co-expression with MiRP1
It has recently been shown that MiRP1 is expressed in SAN and ventricular tissues and, when co-injected into oocytes with HCN1 and HCN2 clones, enhances their level of expression and speeds activation kinetics (Kupershmidt et al. 1999; Yu et al. 2001). We reasoned that the faster time course of activation of native f-channels relative to 4–1 tandem channels expressed in HEK 293 cells (Fig. 5) could be at least partly explained by the action of MiRP1. Since HCN4 is, among the cardiac HCN isoforms, the one with the slowest activation kinetics and that most abundantly expressed in the pacemaker region of rabbit heart (Shi et al. 1999; Altomare et al. 2001), we first checked if MiRP1 modulates the HCN4 current. The results shown in Fig. 7 argue against this hypothesis. Upon hyperpolarization to the voltages indicated, sample traces recorded from representative cells expressing HCN4 (left) and HCN4 + MiRP1 (right) are shown in Fig. 7A and have a similar time course.
A more complete kinetic analysis confirmed that neither the mean activation time constant curve (Fig. 7B) nor the mean fractional activation curve of HCN4 (Fig. 7C) were modified significantly by co-expression with MiRP1. Fitting of the latter yielded half-activation voltages of −83.7 and −79.1 mV and inverse slope factors of 12.0 and 13.4 mV, respectively, for HCN4 alone (n= 5) or in combination with MiRP1 (n= 4); these values were not significantly different (P > 0.05). Furthermore, MiRP1 co-expression did not alter the mean current density. At −145 mV, this was 49.3 ± 10.4 pA pF−1 (n= 12) for HCN4 alone and 53.8 ± 12.1 pA pF−1 (n= 7) for HCN4 + MiRP (not significantly different, P > 0.05).
Having verified the lack of effect of MiRP1 on HCN4, we next evaluated whether MiRP1 could exert a modulatory action on heteromeric assemblies of fixed stoichiometry such as the 4–1 tandem constructs. Again, as shown in Fig. 8, co-expression of the 4–1 tandem with MiRP1 in HEK 293 cells did not have substantial effects on channel kinetics. The time constant and activation curves for 4–1 tandem channels were nearly superimposable on those obtained by co-expression with MiRP1. Boltzmann fits of the latter yielded V1/2=−90.7 and −84.7 mV and s= 12.3 and 10.4 mV for the 4–1 tandem channel without (n= 6) and with MiRP1 (n= 5), respectively (not significantly different, P > 0.05). The current density was also unaffected. At −145 mV, the current recorded from cells expressing 4–1 tandem channels was 21.7 ± 4.8 pA pF−1 (n= 12) and that recorded from cells co-expressing the 4–1 tandem and MiRP1 was 35.0 ± 6.4 pA pF−1 (n= 12); these values were not significantly different (P > 0.05).
The results shown in Figs 7 and 8 rule against MiRP1 playing a major role as a β-subunit, able to modulate the degree of expression and/or kinetics of homomeric HCN4 and heteromeric 4–1 tandem channels.
Since the cloning in the late 1990s of the HCN family of channels (Clapham, 1998), investigators have tried to identify the molecular constituents of native pacemaker channels in the tissues where they are expressed. Differences between properties of hyperpolarization-activated channels in different tissues have long been known. For example, some of the properties of the cardiac If current, as originally described and later analysed in SAN myocytes (Brown et al. 1979; Brown & DiFrancesco, 1980; DiFrancesco et al. 1986; DiFrancesco, 1993), clearly differ from those of the related Ih current in some types of neuronal cells, such as the CA1 hippocampal neurons (Maccaferri et al. 1993; Pape, 1996). Differences typically include slower kinetics and higher cAMP sensitivity; however, a wide range of variable properties, including activation and deactivation kinetics, voltage thresholds of activation, the positions and slopes of activation curves and responses to cAMP (DiFrancesco, 1993; Pape, 1996; Santoro & Tibbs, 1999), is found throughout the tissues in which pacemaker currents are expressed.
Based on the observation that HCN4 and HCN1 are the major isoforms found in the rabbit cardiac SAN region (Ishii et al. 1999; Shi et al. 1999; Moroni et al. 2001) (although HCN2 is more highly expressed than HCN1 in mouse SAN, Moosmang et al. 2001), here we have expressed HCN1 + HCN4 and 4–1 or 1–4 tandem constructs in order to verify if co-assembly of these subunits (with or without the inclusion of MiRP1) resulted in channels with properties similar to those of native f-channels.
The properties of HCN1 and HCN4 are distinct from those of native f-channels
We collected data on channel kinetics in HEK 293 cells and in SAN myocytes using the same experimental conditions for recordings (i.e. temperature, intracellular and extracellular solutions) so as to allow a direct comparison. We found that the rate of activation of f-channels was intermediate between those of HCN1 and HCN4 in the whole voltage range investigated, as shown in Figs 2 and 3. The response of f-channels to cAMP was also intermediate between those of HCN1 and HCN4 (Fig. 6). These results clearly indicate that the native If current of the SAN does not result from the contribution of either HCN1 or HCN4 α-subunits alone and, together with the evidence for co-localized expression of HCN1 and HCN4 in the SAN, suggest that f-channels might result from the combination of these two isoforms.
On the other hand, this appears to contrast with the evidence that both HCN1 and HCN4 currents activate at voltages more negative than that required to activate If (Fig. 4). Recent evidence, however, suggests that the position of the activation curve is modulated by context-dependent processes. Indeed, when HCN channels are overexpressed in ventricular myocytes, the activation curves of both HCN2 and HCN4 are shifted more positive by some 16–19 mV with respect to the curves of the same isoforms expressed in HEK 293 cells, although their relative position remains essentially unaltered (Qu et al. 2002).
An interpretation of this phenomenon that would account for the negatively directed shift of the activation curve during run-down (DiFrancesco et al. 1986) or upon inside-out patch excision (DiFrancesco & Mangoni, 1994), is that unknown cellular factors present or functional in native, but not HEK 293 cells, may interact with intracellular regions of the channels and inhibit a tendency of the protein toward negative activation (Qu et al. 2002). This would also explain why cleavage with pronase of inhibitory intracellular regions of channels determines a large positive shift (56 mV) in the activation curve of If in SAN cells (Barbuti et al. 1999).
Our data are in agreement with the idea that a similar situation may apply to If in the SAN cells. According to this view, the more negative position of the HCN1 and HCN4 activation curves with respect to f-channels could be attributable to a context dependence involving unknown intracellular factors. It is interesting to note that the f-channel curve would be intermediate between the HCN1 and HCN4 curves, if these were to be shifted by the same amount (16–19 mV) separating curves in HEK 293 and ventricular cells according to Qu et al. (2002) (see Figs 4 and 5).
Co-transfection of HCN4 and HCN1 generates channels with kinetics similar to those of HCN4
As shown in Figs 3 and 4, co-expression of HCN1 and HCN4 subunits resulted in channels with activation kinetics and activation curve similar to those of HCN4 alone. In the simplistic hypothesis that HCN1 and HCN4 subunits are equivalent in the processes of co-assembly and membrane insertion and that they affect kinetic properties in proportion to the stoichiometric ratio, one would expect their co-expression to yield channels with properties intermediate between those of the individual subtypes. The results shown in Figs 3 and 4 may thus indicate either that (1) the subunits do not co-assemble uniformly and channels with 3:1 or 4:0 HCN4:HCN1 stoichiometry are favoured or (2) regardless of whether subunits co-assemble uniformly or not, HCN4 has a ‘dominant’ role in determining the channel kinetics. As is discussed below, data obtained with concatenated 4–1 tandem channels suggest that the former hypothesis is more likely to be correct.
The possibility suggested by our data that uneven heteromeric HCN4:HCN1 assemblies are favoured, with HCN4 prevailing over HCN1, is in agreement with existing data suggesting an RNA ratio of 1:4 between HCN1 and HCN4 (Shi et al. 1999), although there is no guarantee that the message ratio is maintained at the protein level. Clearly, however, our results do not allow a precise quantitative determination of which HCN4:HCN1 preferred stoichiometry characterizes native f-channels.
4–1 tandem channels have activation kinetics approaching those of f-channels
In tetrameric assemblies, concatenated 4–1 tandem constructs restrict the stoichiometry to the 2:2 HCN1:HCN4 ratio. As shown in Fig. 5B, 4–1 tandem channels have activation kinetics intermediate between those of HCN4 and HCN1; though still slower, they approximate the activation kinetics of native f-channels. On the other hand, the position of the activation curve (Fig. 5C) is not intermediate between those of the two subtypes, but lies close to that of HCN4. We do not have a ready explanation for this result. However, the position of the activation curve of HCN channels is a complex function of the voltage dependence of voltage sensors and of open-closed transitions (Altomare et al. 2001) and this feature may simply reflect specific kinetic properties of the 4–1 tandem construct that are not simply the average of those of the individual isoforms. As an additional possibility, the ‘context’ dependence of activation curves on cytoplasmic factors, although limited in HEK 293 cells (Qu et al. 2002), may act differently for 4–1 tandem channels and single isoforms. Since C-termini of HCN4 subunits are fused to N-termini of HCN1 subunits, two of the four available C-termini in 4–1 tandem tetrameric channels have restricted mobility. This might for example imply that removal of the inhibitory action that C-termini exert on 4–1 tandem channels (Viscomi et al. 2001; Wainger et al. 2001) is impaired, resulting in a negative shift of the activation range.
4–1 tandem, but not 1–4 tandem, channels have a lower sensitivity to cAMP than co-transfected channels
As shown in Fig. 6, the response to cAMP of co-transfected HCN1 + HCN4 channels is intermediate between those of HCN1 and HCN4. This result appears to contrast with Figs 3 and 4, which suggest that the HCN4 phenotype prevails in the determination of kinetic features.
A possible interpretation, however, arises from previous observations indicating that the shift induced by cAMP is due to the partial removal of an inhibitory action exerted by the C-termini on HCN channels (Barbuti et al. 1999; Viscomi et al. 2001; Wainger et al. 2001). If any of the four C-termini can contribute to this inhibition, then the presence of even one C-terminus from the HCN1 subunit could determine a reduction in sensitivity to cAMP. According to this interpretation, the C-termini of HCN1 subunits would have a ‘dominant’ effect on the response to cAMP. The results shown in Fig. 6 would then be compatible with the kinetic data shown in Figs 3, 4 and 5, according to which co-transfection of the two isoforms results in the preferential formation and/or membrane insertion of heteromeric channels with a higher proportion of HCN4 subunits.
4–1 tandem channels, on the other hand, have a poor sensitivity to cAMP, similar to that of HCN1. This result also agrees with the hypothesis that HCN1 subunits have a ‘dominant’ effect and confer cAMP sensitivity to HCN4- HCN1 heteromers. A limited cAMP sensitivity may also result, as has been mentioned above to explain the negative position of the activation curve, from restricted mobility of the C-termini of HCN4 subunits, which are fused to N-termini of HCN1 subunits in 4–1 tandem channels. This hypothesis is further supported by the evidence that 1–4 tandem channels respond to cAMP more strongly than 4–1 tandem channels, with a sensitivity similar to that of co-transfected HCN1 + HCN4 channels (Fig. 6F and G).
MiRP1 subunits have no effect on HCN4 and 4–1 tandem channel kinetics and density in HEK 293 cells
The data in Figs 7 and 8 rule against the idea that MiRP1 acts as an HCN subunit in our experimental conditions. This contrasts with evidence from Xenopus oocytes, where MiRP1, but not minK, enhances expression of and accelerates activation kinetics of HCN1 and HCN2 (Yu et al. 2001). We do not have an explanation for this discrepancy, although a possibility relates to the different cell types used for heterologous transfection.
In line with this hypothesis, we found that in HEK 293 cells, MiRP1 co-expression was ineffective also on the density of HCN1 current (HCN1: 17.44 ± 4.8 pA pF−1, n= 7; HCN1 + MiRP1: 18.9 ± 5.1 pA pF−1, n= 5; not significantly different, P > 0.05), although current density is a parameter most sensitive to MiRP1 in Xenopus oocytes according to Yu et al. (2001).
In conclusion, transfection into HEK 293 cells of concatenated, but not of free, HCN1 and HCN4 isoforms generates hyperpolarization-activated channels with activation kinetics approaching those of native f-channels in the SAN. Other properties of native channels such as the position of the activation curve and the response to cAMP, however, are not reproduced by 4–1 tandem channels, although freely co-transfected HCN1 + HCN4 channels do simulate the cAMP sensitivity of f-channels. Interestingly, normal cAMP sensitivity is also found in 1–4 tandem channels, which supports the view that restricted mobility of C-termini impairs cAMP-dependent channel stimulation.
Our data are compatible with the idea that native f-channels are heteromers of HCN1 and HCN4 subunits, but only in the context provided by the intracellular environment of cardiac cells and mediated by still unknown intracellular factors.
We should like to thank Dr Ohmori (Kyoto University, Japan) for providing rbHCN4 and Dr S. Goldstein (Yale University, USA) for providing MiRP1. We also thank Dr G. Consalez (Dibit HSR, Milano, Italy) for assistance in development of PCR overlap protocols. This work was supported by the MIUR (Cofin 2000 to D.D.) and partly by Telethon (P. 971 to D.D.).