A Novel Mechanism of Modulation of Hyperpolarization-activated Cyclic Nucleotide-gated Channels by Src Kinase*[boxs]

Hyperpolarization-activated cyclic nucleotide-gated channels (HCN1-4) play a crucial role in the regulation of cell excitability. Importantly, they contribute to spontaneous rhythmic activity in brain and heart. HCN channels are principally activated by membrane hyperpolarization and binding of cAMP. Here, we identify tyrosine phosphorylation by Src kinase as another mechanism affecting channel gating. Inhibition of Src by specific blockers slowed down activation kinetics of native and heterologously expressed HCN channels. The same effect on HCN channel activation was observed in cells cotransfected with a dominant-negative Src mutant. Immunoprecipitation demonstrated that Src binds to and phosphorylates native and heterologously expressed HCN2. Src interacts via its SH3 domain with a sequence of HCN2 encompassing part of the C-linker and the cyclic nucleotide binding domain. We identified a highly conserved tyrosine residue in the C-linker of HCN channels (Tyr476 in HCN2) that confers modulation by Src. Replacement of this tyrosine by phenylalanine in HCN2 or HCN4 abolished sensitivity to Src inhibitors. Mass spectrometry confirmed that Tyr476 is phosphorylated by Src. Our results have functional implications for HCN channel gating. Furthermore, they indicate that tyrosine phosphorylation contributes in vivo to the fine tuning of HCN channel activity.

Although identified only recently, the family of hyperpolarizationactivated cyclic nucleotide-gated (HCN) 3 channels has generated great interest because it represents the molecular correlate of the hyperpolarization-activated cation current, termed I h (syn. I f or I q ) (1)(2)(3)(4). This current plays a crucial role in the control of important biological functions, including cardiac and neuronal pacemaker activity, determination of resting membrane potential, dendritic integration, and synaptic transmission. Dysfunction of HCN channels has been linked to human diseases, including cardiac arrhythmia (5)(6) and epilepsy (7)(8). Structurally, the four members of the HCN channel family (HCN1-4) belong to the 6TM ion channel superfamily (9 -11). In the plasma membrane, HCN channel subunits assemble to homo-or heterotetrameric complexes, thereby generating a large variety of channel subtypes with distinct biophysical properties (12). Further complexity is probably generated in vivo by the interaction of HCN channels with auxiliary subunits (13), interacting proteins (14 -16), and by post-translational modifications (e.g. N-linked glycosylation) (17).
Whereas our knowledge of the structure and function of HCN channels has significantly increased over the last couple of years, there is only sparse information on the cellular regulation of these channels. It is well established that hormones and neurotransmitters can modulate I h activity via G-protein pathways that modulate the cAMP concentration (1). Cyclic AMP enhances channel activity by direct binding to a cyclic nucleotide binding domain (CNBD) present in the C terminus of HCN channels. The region linking the last membrane-spanning domain (S6) to the CNBD (the C-linker) has been shown to play a key role in coupling cAMP binding with channel opening (18 -20). Unlike in many other ion channels the modulation by cAMP does not seem to involve protein kinase A-mediated serine/threonine phosphorylation (21). By contrast, several experimental observations suggest that nonreceptor protein-tyrosine kinases (PTKs), especially members of the Src family, may regulate HCN channels (1). The HCN1 channel was originally identified in a yeast two-hybrid screen using the SH3 domain of Src as bait. However, it was not reported whether this interaction also occurs in native tissue and whether Src influences channel activity (22). Based on experiments with genistein, a broad spectrum PTK inhibitor, it was suggested that cardiac I h is regulated by tyrosine phosphorylation (23)(24)(25). However, this finding was challenged by another group reporting that the genistein effect is caused through nonselective interactions with the channel molecule (26).
Src is widely expressed in neurons and heart cells and has been shown to be an important regulator of voltage-and ligand-gated ion channels (27). We therefore asked in the present study whether or not this kinase is involved in HCN channel modulation. Using genetic, biochemical, and electrophysiological methods, we demonstrate that Src specifically binds to the C terminus of HCN channels, phosphorylates the channels and thereby affects the activation kinetics of I h . Moreover we identify a specific tyrosine residue in the C-linker region of HCN channels that is the molecular target of Src.
Expression in HEK293 Cells-The cDNAs of wild-type and mutant mHCN2 and hHCN4 were subcloned into the mammalian expression vector pcDNA3 (Invitrogen). A Myc-tagged c-Src was constructed in the pcDNA3.1/Myc-His vector (Invitrogen). The dominant-negative mutant of chicken c-Src (Src-K295M) was subcloned in the pIRES2-EGFP vector (BD Biosciences Clontech). Mutant HCN channels were generated by standard PCR techniques. All plasmid constructs were verified by automated DNA sequencing.

Yeast Two-hybrid Screen
Yeast strain Saccharomyces cerevisiae EGY48 was used for YTH assays. pEG202 and pJG4 -5 fusion plasmids, together with the lacZ reporter plasmid pSH18 -34 (BD Biosciences, Clontech) were transformed into yeast by the lithium acetate method. To identify specific interactions transformants were grown on four different selective plates: 1) X-gal plates lacking uracil, histidine, and tryptophan, and containing galactose and raffinose as carbon source; 2) X-gal plates lacking uracil, histidine, and tryptophan, and containing glucose as carbon source; 3) plates lacking uracil, histidine, tryptophan, and leucine, and containing galactose and raffinose as carbon source; 4) plates lacking uracil, histidine, tryptophan, and leucine, and containing glucose as carbon source. An interaction was considered to be specific if a given transformant turned blue on plate condition 1, stayed white on plate condition 2, was able to grow on plate condition 3, and did not show growth on plate condition 4. Fig. 4B shows examples from selective plates 1 and 2.

Expression and Purification of GST Fusion Proteins
GST fusion proteins were expressed in the protease-deficient BL21 (DE3) strain of Escherichia coli by induction with 0.1 mM isopropyl-␤thiogalactopyranoside (IPTG), for 1 h at 37°C. Bacteria were pelleted by centrifugation (10 min, 4°C, 5,000 ϫ g) and resuspended in 4 ml of PBS supplemented with protease inhibitor mix (PI) containing 1 g/ml leupeptin, 1 M pepstatin A, 1 g/ml antipain, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM orthophenanthroline, 1 mM benzamidine, 1 mM iodacetamide. After a freeze-thaw cycle, lysozyme was added to a final concentration of 5 mg/ml. After incubation for 30 min on ice, 1% Triton X-100 was added, and the cell suspension was shaken for 30 min to improve solubility of the fusion proteins. Cell debris was pelleted by centrifugation (30 min, 10,000 ϫ g, 4°C). The supernatant was aliquoted and stored at Ϫ80°C. For purification of GST fusion proteins, supernatant was incubated with 1 volume of 50% slurry glutathione-Sepharose beads (Amersham Biosciences) at 4°C overnight. The loaded beads were pelleted by centrifugation (5 min, 500 ϫ g, 4°C) and washed twice with ice-cold PBS supplemented with PI. Beads were resuspended in 1 volume of PBS/PI and stored at 4°C.

GST Pull-down Assay
About 3 g of purified GST fusion protein was diluted into 1 ml of cell lysis buffer (50 mM HEPES (pH 7.5), 150 mM NaCl, 10% glycerol, 5 mM EDTA (pH 8.0), 1% Triton X-100). 10 l of glutathione-Sepharose beads were added, and the mixture was incubated for 3 h at 4°C. Beads were centrifuged at 10,000 ϫ g for 2 min and washed three times with 1ϫ HNTG buffer (50 mM HEPES (pH 7.5), 150 mM NaCl, 10% glycerin, 5 mM EDTA (pH 8.0), 0.1% Triton X-100). After removing the supernatant, beads were incubated with total lysates from transfected or untransfected HEK293 cells for 4 h at 4°C. The beads were pelleted after repeating the steps of centrifugation and washing. 20 l of SDS loading buffer were added, and the mixture was heated 5 min at 100°C before loading onto an SDS-PAGE gel.

In Vitro Kinase Assay
Beads preloaded with about 2 g of GST fusion proteins (see GST pull-down assay) were washed three times with 1ϫ HNTG buffer and once with kinase buffer (20 mM HEPES (pH 7.5), 10 mM MgCl 2 , 1 mM dithiothreitol, 200 M pervanadate, 0.35 mM ATP). The beads then were mixed with 40 l of reaction buffer (kinase buffer supplemented with 0.05 mg/ml bovine serum albumin and 0.1% ␤-mercaptoethanol) with or without 1 unit of purified c-Src (Oncogene Research Products). The reaction was carried out by shaking at room temperature for 15 min. After centrifugation and discarding the supernatant, the reaction was stopped by adding SDS loading buffer. All samples were then subjected to SDS-PAGE and Western blotting. The monoclonal anti-phosphotyrosine antibody 4G10 (anti-pY) (Upstate Biotechnology) was used to detect tyrosine-phosphorylated proteins.

Preparation of Mouse Brain Membranes and HEK293 Cell Lysates
Membrane Fractions-Freshly isolated mouse brain tissue was washed with PBS and homogenized in ice-cold MOPS lysis buffer (0.3 mM sucrose, 20 mM MOPS, 1 mM EDTA) supplemented with 200 M pervanadate and PI, using a glass cylinder and a Teflon plunger. Homogenates were centrifuged for 10 min at 5,000 ϫ g. The pellet was rehomogenized and spun once more for 10 min. The combined supernatants were then centrifuged (45 min, 4°C, 100,000 ϫ g). The pellet comprising the total membrane fraction was resuspended in MOPS lysis buffer supplemented with 200 M pervanadate and stored at Ϫ80°C.
Cell Lysates-Three days after transfection, HEK293 cells were washed twice with PBS and lysed with 500 l of cell lysis buffer supplemented with 200 M pervanadate and PI. After 10 min on ice the lysed cells were scraped off the dish and transferred to a reaction tube. The lysate was centrifuged (15 min, 4°C, 12,000 ϫ g), and the supernatant was subject to coimmunoprecipitation or stored at Ϫ80°C.

Coimmunoprecipitation
Total brain membrane fractions or cell lysates of HEK293 cells were incubated overnight at 4°C with 25 l of protein A-Sepharose (Sigma) and a specific antibody directed against one of the two examined proteins and 500 l of HNTG buffer supplemented with 200 M pervanadate and PI. Beads were pelleted by centrifugation (15 min, 4°C, 12,000 ϫ g) and washed three times with cold HNTG buffer supplemented with 200 M pervanadate. Interacting proteins were visualized after boiling for 5 min in Laemmli sample buffer by SDS-PAGE and Western blot analysis. The following antibodies used were: anti-HCN2 (Alomone), anti-Src (GD11, Upstate Biotechnology Inc.), anti-GST (Amersham Biosciences), anti-Myc (9E10); anti-phosphotyrosine (4G10 (anti-pY), Upstate Biotechnology Inc. and p-Tyr-100, Cell Signaling Technology). An antibody against the cyclic nucleotide-gated channel CNGB3 (28) was used as control.

Mass Spectrometry (MS) Analysis
The GST fusion protein containing the whole C terminus of HCN2 (amino acids 448 -863) was in vitro phosphorylated by Src. Thereafter, the protein was boiled in SDS-PAGE sample buffer, separated on a 10% Tris-glycine gel, and stained with Coomassie Blue. The piece of gel containing the HCN2 fusion protein was excised and in-gel-digested with trypsin according to standard procedures. Tryptic peptides were extracted with 5% formic acid/50% acetronitrile, dried, and stored at Ϫ20°C until analysis by mass spectrometry.
A surveyor liquid chromatography system (ThermoFinnigan, San Jose, CA), consisting of degasser, MS Pump, and autosampler, equipped with a C18 trap column (RP, 300 m ϫ 5 mm, Agilent Technologies) and PicoFrit TM column: 75 m ϫ 100 mm, 15-m tip packed with a 5-m Aquasil C18 (ThermoFinnigan, San Jose, CA) was used. The samples were loaded onto the column with an RP gradient of 2-98% B over 180 min. RP solvents were 0.1% formic acid in either water (A) or acetonitrile (B). The flow rate was 200 nl/min. A Finnigan LTQ linear ion trap mass spectrometer equipped with an ESI microspray source was used for MS/MS experiment with ion transfer capillary of 160°C and NSI voltage of 1.8 kV. The mass spectrometer was set that one full MS scan was followed by ten MS/MS scans on the ten most intense ions from the MS spectrum. Spectra from each fraction were searched by SEQUEST algorithm against the mHCN2 sequence. In these searches, differential modifications of 80 daltons to tyrosine residues were selected. For all sequences reported here, spectra were manually validated and contained sufficient information to assign not only the sequence, but also the site of phosphorylation. All output results were combined using home made software to delete the redundant data.

Cell Culture and Heterologous Expression
HEK293 cells (DMSZ, Braunschweig, Germany) were maintained in DMEM medium (Invitrogen, Life Technologies, Inc.) supplemented with 10% fetal bovine serum, 100 units/ml penicillin and 100 g/ml streptomycin, and incubated at 37°C with 10% CO 2 . For transfection, HEK293 cells were seeded on 6-well plates (diameter 3.5 cm) at a density of 700,000 cells per well. After 6 h, cells were transfected with expression plasmid DNA (0.6 g of each plasmid per well) using the FuGENE 6 transfection reagent (Roche Diagnostics). For electrophysiological measurements, transfected cells were detached using 0.05% trypsin/0.5 mM EDTA (Invitrogen, Life Technologies, Inc.) and replated onto 12-mm poly-L-lysine-coated coverslips in 24-well plates.
HL-1 cardiomyocytes were obtained from Dr. W. C. Claycomb (Lousiana State University Health Science Center, New Orleans, LA) and maintained in Claycomb Medium (JRH Biosciences, Andover, UK), supplemented with 10% fetal bovine serum, 4 mM L-glutamine, 10 M noradrenaline, and penicillin-streptomycin as previously described (29). Single HL-1 cells were detached from confluent cultures using trypsin-EDTA. Isolated cells were either replated or directly used for patch-clamp experiments.
Dorsal root ganglion (DRG) neurons were isolated from adult mice as described by Wu et al. (30). Briefly, thoracic and lumbar DRGs were dissected and transferred immediately into DMEM medium. After removal of attached nerves and connective tissues ganglion fragments were placed in a microtube containing 1 ml of DMEM supplemented with trypsin (type I, 0.5 mg/ml, Sigma), collagenase (type I, 1 mg/ml, Sigma) and DNase (type I, 0.1 mg/ml, Sigma) and were incubated at 34°C for 30 min. Thereafter, soybean trypsin inhibitor (type II-S, 1.25 mg/ml, Sigma) was added to stop trypsin digestion. The cell suspension was centrifuged (500 rpm, 3 min) to remove the supernatant and replenished with DMEM. Cells were then plated onto a 35-mm culture dish precoated with poly-L-lysine and kept in an incubator (37°C, 10% CO 2 ) for at least 1 h before electrophysiological recordings. Medium sized DRG neurons (30 -45 m) were selected for recordings. Pacemaker cells were prepared from sino-atrial node of adult mice as described by us previously (31).

Electrophysiology
Currents of heterologously expressed HCN channels were measured at room temperature 2-3 days after transfection using whole cell voltage clamp technique. The extracellular solution was composed of (in mM): 110 NaCl, 0.5 MgCl 2 , 1. The I h currents of sinoatrial node cells were measured as described previously (31). Pipettes were pulled from borosilicate glass capillaries (GC150TF, Harvard Apparatus LTD) and had resistances of 2-3 M⍀ when filled with the intracellular solution. Src kinase inhibitors (PP2 and genistein) and their respective inactive derivatives (PP3, daidzein) were purchased from Calbiochem (Germany). Stock solutions of these substances were prepared in Me 2 SO and were freshly diluted at least 1:1,000 in bath solution before using in experiments. The effects of Src inhibitors were determined after incubation of cells with the substances for at least 10 min.
Data were acquired at 10 kHz using an Axopatch 200B amplifier and pClamp 8 (Axon Instruments). Voltage clamp data were stored on the computer hard disk and analyzed off-line by using Clampfit 8 (Axon Instruments). Steady-state activation curves were determined by hyperpolarizing voltages of Ϫ140 mV to Ϫ30 mV from a holding potential of Ϫ40 mV for 3.2 s (for wild-type and mutant HCN2) and 4.8 s (for HCN4) followed by a step to Ϫ140 mV. Tail currents measured immediately after the final step to Ϫ140 mV, were normalized by the maximal current (I max ) and plotted as a function of the preceding membrane potential. The data points were fitted with the Boltzmann function: ( where I min is an offset caused by a nonzero holding current, V m is the test potential, V 0.5 is the membrane potential for half-maximal activation, and k is the slope factor. Time constants of channel activation ( act ) of wild-type and mutant HCN2 and HCN4 channels were determined by monoexponential (or biexponential in the case of HCN2-Y476F and I h of DRG neurons) function fitting the current evoked during hyperpolarizing voltage pulses to Ϫ140 mV unless otherwise specified. As has been described earlier (33) the initial lag in the activation of HCN channels was excluded from the fitting procedure. Statistical analyses were performed with Origin6.1 (OriginLab). Data are presented as mean Ϯ S.E. (n ϭ number of recorded cells).

Inhibition of Src Slows Down Kinetics of Expressed HCN2
Channel-We investigated the effect of Src on HCN2 channels expressed in HEK293 cells. These cells endogenously express substantial amounts of Src making them a suitable system for our purpose (Fig. 1A). Preincubation of HCN2-expressing cells with the Src kinase inhibitor PP2 (10 M) led to a profound deceleration of HCN2 activation kinetics (Fig. 1,  B and F). The activation time constant ( act ) at Ϫ140 mV increased from 277 Ϯ 14 ms (n ϭ 13) at control conditions to 538 Ϯ 29 ms (n ϭ 19) in the presence of PP2. In contrast, PP2 had no effect on current densities (pA/pF: PP2, Ϫ78.9 Ϯ 12.4 (n ϭ 25); control, Ϫ84.4 Ϯ 12.5 (n ϭ 27), p Ͼ 0.05). The effect on act was present over the whole voltage range (supplemental Fig. S1). By contrast, PP3, an inactive analogue of PP2, did not affect current kinetics (Fig. 1, C and F, supplemental Fig. S1). Similarly, another Src kinase inhibitor, genistein (Fig. 1, D and F), but not its inactive analogue daidzein (Fig. 1F), also slowed down the HCN2 current.
The finding that two structurally unrelated PTK inhibitors affected HCN2 in the same fashion argued against a nonspecific action of these agents. More likely, the effect of PP2 and genistein was caused by specific inhibition of Src and subsequent channel dephosphorylation by cellular tyrosine phosphatases. To further strengthen this hypothesis, we cotransfected HEK293 cells with HCN2 and a catalytically inactive Src mutant, Src-K295M (Src-KM) (34). Indeed, HCN2 currents obtained from these cells activated with slower kinetics than control currents ( act ϭ 353 Ϯ 7.4 (n ϭ 6); Fig. 1, E and F). The effect induced by Src-K295M was somewhat weaker than that of PP2, probably because of the high levels of endogenous wild-type Src. Inhibition of Src also led to inhibition of deactivation kinetics (Fig. 2, A and B). In contrast, neither PP2 nor PP3 did affect the voltage dependence of activation (Fig. 2, C  and D). Regulation of HCN2 by cAMP was principally preserved in the presence of PP2. Cyclic AMP shifted the half-maximal activation voltage (V 0.5 ) by about ϩ10 mV and speeded up activation kinetics (TABLE  ONE). Interestingly, however, the activation kinetics in the presence of saturating cAMP ϩ PP2 was consistently slower than that observed with cAMP alone ( act(PP2, cAMP) ϭ 272 Ϯ 23 ms (n ϭ 12); act(cAMP) ϭ 163 Ϯ 6.84 ms (n ϭ 8)) indicating that the decelerating effect of PP2 persisted in the presence of cAMP.
HCN2 Binds Src and Undergoes Tyrosine Phosphorylation-We performed a series of immunoprecipitation experiments to determine whether HCN2 is a substrate of Src (Fig. 3A). In the first set of experiments, cells expressing HCN2 were preincubated with the tyrosine phosphatase inhibitor pervanadate (PV) prior to immunoprecipitation with anti-HCN2. A specific band corresponding to the molecular mass of HCN2 (105 kDa) was detected with a specific anti-phosphotyrosine antibody (anti-pY). This band was significantly weaker when cells were pretreated with both PV and PP2. Similarly, the intensity of the band was reduced in HCN2/Src-KM-coexpressing cells pretreated with PV.  Taken together, these findings suggested that HCN2 undergoes tyrosine phosphorylation by Src and is dephosphorylated by cellular phosphatases if Src is inhibited. We next studied whether an interaction between Src and HCN also occurs in native tissue. To this end, immunoprecipitations with mouse brain membrane fractions were performed (Fig. 3B). Src was present in immunoprecipitates obtained with anti-HCN2 but not with a control antibody (anti-CNGB3) indicating that Src is bound to HCN2 in vivo. Moreover, HCN2 was immunoprecipitated with anti-phosphotyrosine antibodies from brain membranes prepared in the presence of PV, indicating that HCN2 is tyrosine-phosphorylated in brain tissue (Fig. 3C).
Identification of Interacting Sites of HCN2 and Src-We performed a YTH screen to determine the domains conferring interaction between HCN2 and Src. Specific interaction was seen between the full-length C terminus of HCN2 and Src (Fig. 4, A and B). Using a series of C-terminal bait vectors, we identified the sequence reaching from the CЈ-helix of the C-linker to the C terminus of the CNBD as the minimal interaction site required for binding of Src. We obtained the same result when only the SH3 domain instead of the full-length Src was tested for interaction (Fig. 4B). To confirm YTH data, we performed GST pull-down assays with full-length Src and fusion proteins corresponding to the bait vectors used in YTH (Fig. 4C). Again, the C-linker/CNBD (L-BD) region was required for specific interaction. In a reverse experiment, specific interaction between the SH3 domain of Src and the full-length HCN2 was demonstrated (Fig. 4D). Finally, in vitro kinase assays showed that Src does not only bind the L-BD region but that it also phosphorylates

Effects of cAMP on HCN2 current in the absence and presence of PP2
Cells were preincubated for at least 10 min with 10 M PP2. Whole cell currents were determined using standard intra-and extracellular solutions or with a pipette solution supplemented with 1 mM cAMP as indicated. act was determined at Ϫ140 mV.   this domain (Fig. 4E). In contrast, a GST fusion protein covering the distal C terminus neither bound Src nor was phosphorylated by this kinase (Fig. 4, C and E). Src Regulates HCN2 by Phosphorylation of Tyrosine 476-HCN2 contains seventeen tyrosine residues in cytosolic domains, all of which could principally serve as phosphorylation sites for Src (Fig. 5A). We systematically mutated each of these residues individually to phenylalanine, and in addition constructed mutants containing multiple exchanges. All mutants were tested for their sensitivity to PP2. Fig. 5B summarizes act values obtained for some of the mutants. A mutant lacking the N terminus (⌬N130), and mutants carrying mutations in the cytosolic loops (Y259F, Y331F) or distal C terminus (Y670/766F) maintained sensitivity to PP2. By contrast, a 6-fold mutant covering all tyrosines present in the C-linker (Y449 -481F) and a 9-fold mutant covering the last three tyrosines of the C-linker and all six tyrosines of the CNBD (Y476 -618F) were no longer affected by this agent. We reasoned that one of the three replacements present in both mutants (Y476F, Y477F, and Y481F) was crucial for the PP2 effect. Indeed, mutation of Tyr 476 to phenylalanine was sufficient to completely abolish sensitivity of activation kinetics to PP2 and genistein (Fig. 5, C and D) and deactivation kinetics (Fig. 5, E and F) to PP2. In contrast, Y477F and Y481F currents were slowed down by Src inhibitors to a similar extent as wild-type HCN2 (Fig. 5D). Unlike the wild-type HCN2 current, which could be well fitted by a monoexponential function, the Y476F current revealed two distinct kinetic components (Fig. 5C). The predominant component (about 80% of total current) was significantly faster than wild-type HCN2 ( act,fast(Y476F) ϭ 162 Ϯ 6.9 ms (n ϭ 24), act(HCN2) ϭ 277 Ϯ 14 ms (n ϭ 12)) whereas the minor slow component had activation constants in the range of several seconds. In summary, these findings suggested that Tyr 476 confers sensitivity to PP2 and that this residue is involved in principal channel gating.
Next, we determined the localization of phosphotyrosines in HCN2 by mass spectrometry (Fig. 6). Analysis of peptides obtained after trypsin digestion of an in vitro phosphorylated GST fusion protein containing the whole HCN2 C terminus revealed two phosphotyrosine peptides. The two peptides contained phosphate groups at position Tyr 459 (Fig. 6A) and Tyr 476 (Fig. 6B), respectively. No other phosphopeptide was detected. Together with the electrophysiological data the identification of phosphorylated Tyr 476 strongly suggested that this residue confers the modulatory action of Src. By contrast, phosphorylation of Tyr 459 was very likely not involved in this process since the Y459F mutant was fully sensitive to PP2 ( act(control) ϭ 164 Ϯ 17 ms, n ϭ 5; act(ϩPP2) ϭ 271 Ϯ 12 ms, n ϭ 7).

Mutation of Tyr 554 in HCN4 Prevents
Regulation by PP2-Tyr 476 is localized in the BЈ-helix of the HCN2 C-linker ( Fig. 7 and supplemental  Fig. S2). In the crystal structure of the HCN2 C-linker-CNBD tetramer, Tyr 476 contributes to intersubunit interaction by forming a hydrogen bond with a carbonyl backbone group (Glu 494 ) present in the CЈ-helix of a neighboring subunit (Fig. 7A and supplemental Fig. S2). Importantly, Tyr 476 is highly conserved within the HCN channel family (supplemen-  OCTOBER 7, 2005 • VOLUME 280 • NUMBER 40 tal Fig. S2A), suggesting that regulation by Src may be a general feature of HCN channels. Indeed, the HCN4 current also was slowed down by PP2 (Fig. 7B). Mutation of the tyrosine residue analogous to Tyr 476 in HCN4 (HCN4-Y554F) completely abolished the effect of PP2 (Fig. 7, C  and D). Native I h Is Regulated by Src-We finally asked whether native I h is modulated by Src. To this end we studied the effect of PP2 on I h in HL-1 cardiomyocytes (32), in sino-atrial node (SAN) pacemaker cells and in DRG neurons (Fig. 8). We selected these cell types because they endogenously express Src (not shown) and because their respective I h reveals fast (DRG), intermediate (HL-1), and slow (SAN) kinetics. Like heterologously expressed channels native I h was profoundly slowed down by PP2, confirming that regulation by Src is a general feature of I h channels.

DISCUSSION
The activation and deactivation of I h channels is tightly controlled in the cell. Importantly, a rise in cellular cAMP level speeds up these channels whereas hydrolysis of cAMP has the opposite effect (1). Here, we show that Src-mediated tyrosine phosphorylation is another crucial mechanism controlling kinetics. Inhibition of Src by pharmacological blockers as well as transfection with a catalytically inactive Src mutant profoundly slowed down activation kinetics of native and expressed I h channels indicating that the phosphorylation/dephosphorylation state determines channel kinetics. Our data indicate that the dephosphorylated channel activates significantly slower than the phosphorylated channel.
The Src kinase binds via its SH3 domain to the HCN2 channel. As a minimal requirement for high affinity binding of Src, we identified the sequence reaching from the CЈ-helix of the C-linker to the C terminus of the CNBD. This region is highly conserved among HCN1-4, suggesting that binding of Src is a common feature of HCN channels. Interestingly, Santoro et al. (22) reported binding of the SH3 domain of n-Src to the full-length C terminus of HCN1 whereas no interaction was seen for c-Src, the Src splice variant used in our study. n-Src differs from c-Src only by 6 or 11 amino acid insertions (35). The structural determinants conferring the isoform specificity of Src binding to HCN1 versus HCN2 remain to be determined.
We were unable to further narrow down the target sequence of Src, probably because the residues that directly interact with the SH3 domain are distributed in both C-linker and CNBD. It is very likely that only in the context of the correctly folded C-linker/CNBD domain these key residues are in the right spatial position required for SH3 binding. Notably, the identified sequence does not contain classic proline-rich (PXXP) SH3 binding motifs (36). However, there are several studies showing that SH3 domains can also bind to non-PXXP ligand sequences (37-39). HCN2 represents another example of a protein interacting with SH3 via a non-classic binding sequence.  . The Y554F mutations abolishes sensitivity of HCN4 to PP2. A, structure of the C-linker (green) and the CNBD (gray) tetramer of HCN2 (45). The four Tyr 476 residues (corresponding to Tyr 554 in HCN4) are indicated in red. Ch, transmembrane core of the channel. The crystal model was produced using the software package DINO (www.dino3d.org). B, PP2 slows down activation kinetics of the expressed HCN4. C, PP2 has no effect on HCN4-Y554F currents. Currents were evoked by hyperpolarizing pulses to Ϫ140 mV. D, summarized data of activation time constants ( act ) of wild-type HCN4 and HCN4-Y554F currents at Ϫ140 mV in the absence and the presence of PP2. HCN4: act(control) ϭ 480 Ϯ 28 ms (n ϭ 9), act(ϩPP2) ϭ 663 Ϯ 30 ms (n ϭ 10). HCN4-Y554F: act(control) ϭ 521 Ϯ 22 ms (n ϭ 10), act(ϩPP2) ϭ 529 Ϯ 24 ms (n ϭ 12).
Mass spectrometry identified two tyrosine residues (Tyr 459 and Tyr 476 ) in the C-linker of HCN2 that are phosphorylated by Src. The electrophysiological analysis revealed that only one of these residues, Tyr 476 , is important in controlling activation kinetics. When this residue was mutated to phenylalanine activation kinetics were no longer sensitive to Src inhibitors. We cannot totally exclude that the loss of Src regulation in Y476F was caused by some nonspecific steric effect. However, it should be noted that mutation of any other of the seventeen intracellular tyrosine residues of HCN2 (including Tyr 459 ) did not affect regulation by Src. Moreover, mutation of the analogous tyrosine residue in HCN4 (Y554F) also completely abolished sensitivity to Src inhibitors. The functional role of phosphorylation of Tyr 459 , if there is any, remains to be elucidated. Tyrosine phosphorylation has also been identified in the A subunit of the rod cyclic nucleotide-gated channel (CNGA1). In this channel phosphorylation of Tyr 498 , which is located in the ␤1 strand of the CNBD, leads to a decrease of the apparent affinity for cGMP (40). Interestingly, the corresponding residue of HCN2 (Phe 538 ) is localized in close proximity to Tyr 476 (less than 10 Å). The molecular identity of the kinase regulating CNGA1 is not known so far. However, given the structural similarity of C-linker-CNBD regions of HCN and CNG channels, interaction between PTKs and CNGA1 may also involve a SH3 domain.
How may phosphorylation of Tyr 476 regulate activation kinetics of HCN channels? Tyr 476 is localized in the BЈ-helix of the C-linker (supplemental Fig. S2, A and B). Previous experiments suggest that the C-linker controls coupling of ligand binding to channel gating in both CNG and HCN channels (20,(41)(42)(43)(44). In the crystal structure of the HCN2 C terminus, the hydroxyl group of Tyr 476 forms a hydrogen bond to the carbonyl backbone group of glutamate residue (Glu 494 ) present in the CЈ-helix of the neighboring subunit of the channel tetramer, thereby contributing to intersubunit contacts. One may speculate that the introduction of the bulky phosphate group by Src or the loss of the hydroxyl group (as occuring when tyrosine is replaced by phenylalanine) will weaken or even abolish the interaction between residue 476 and its counterpart in the CЈ-helix and, hence, destabilize the interaction between neighboring C-linkers. According to this model, the phosphorylated channel activates faster because its C-linkers are interacting somewhat less tightly with each other than C-linkers of dephosphorylated channels. This model is obviously at odds with the original report on the HCN2 crystal structure claiming that the densely packed tetrameric C-linker ring ("gating ring") represents the open state of the channel (45). However, in support of our data two recent studies (46 -47) provide strong evidence that the C-linker conformation in crystallized HCN2 indeed represents the closed state of the channel. More structural data will be required to define the exact relation between the HCN2 activation state and the conformation of its C-linker.
The identity of the cellular phosphatase that opposes Src remains to be determined. Unfortunately, we were not able to reliably measure I h after preincubation with the phosphatase inhibitor pervanadate because this agent induced profound cell detachment and led to a destabilization of the plasma membrane. Moreover, the effect of Src inhibitors was only seen when intact cells were preincubated with these agents, but was not observed in excised patches or when cells were perfused after establishing the whole cell modus. These findings suggest that the phosphatase required for dephosphorylation is either readily lost and/or that channel regulation requires the intact intracellular milieu including the presence of cytoskeletal and anchoring proteins. High sensitivity to cellular factors and experimental parameters is a well known property of I h . For example, activation curves of I h are shifted by 30 -40 mV to more hyperpolarized voltages when currents are measured in excised patches instead as in whole cell mode (48). Recently, Yu et al. (25) reported effects of genistein on HCN channels expressed in Xenopus oocytes. Like in HEK293 cells, genistein slowed down HCN2 and HCN4 activation kinetics. However, in contrast to our findings, genistein also reduced current densities of both channels and induced in HCN2 a shift of V 0.5 to more negative voltage. In addition, current modulation was seen when genistein was applied after establishing the whole cell modus, whereas we observed modulation only after preincubation with blockers prior to establishing the whole cell modus. As pointed out above, the cellular environment has profound impact on the properties of I h . Thus, the differences in both studies may reflect the intrinsic diversity of the expression systems used. In our study, regulation of native I h in sinoatrial pacemaker cells, DRG neurons and HL-1 cardiomyocytes was consistent with that found for HCN2 and HCN4 in HEK293 cells. Thus, HEK293 represent a valid system to study the effect of tyrosine phosphorylation on HCN channels.
We also studied the functional interaction between tyrosine phosphorylation and cAMP-dependent modulation of HCN2. Cyclic AMP shifted the activation curve of HCN2 in the presence of Src blockers by the same value (about ϩ10 mV) as at control conditions. Moreover, cAMP speeded up activation kinetics when Src was inhibited. However, in the presence of Src inhibitors the act values obtained at saturating cAMP were consistently bigger than in the absence of these agents. Collectively, these findings suggest that regulation of HCN channel activity by cAMP operates principally independent of the phosphorylation state of Tyr 476 . However, the phosphorylation state of Tyr 476 determines the extent by which a given cAMP concentration can accelerate activation kinetics. The complex modulation of channel activation resulting from the interplay between tyrosine phosphorylation and cel- Src is ubiquitously expressed in brain and heart cells. Thus, regulation of I h channels by this kinase may be a general feature of this ion channel class. In agreement with this notion we demonstrated that HCN2 binds Src and is tyrosine-phosphorylated in mouse brain. Moreover, we provide evidence that Src regulates I h from murine DRG neurons and sinoatrial node cells which mainly express HCN1/HCN2 (49 -50) and HCN4 (51), respectively. Native I h shows a wide range of activation kinetics (52). It was assumed that this diversity results from the subunit composition of the individual channels, the cAMP levels and physical parameters like pH, ionic milieu, and temperature. The control of the phosphorylation state of Tyr 476 by Src and not yet defined phosphatases represents another mechanism controlling kinetics. Importantly, unlike the cAMP system, Src affects activation speed without altering the voltage dependence of activation. This finding suggests that the control of activation kinetics by Src may reflect an important regulatory mechanism to adjust the properties of I h to the individual requirements of different types of neurons and heart cells.
Src is regulated by a complex network of upstream signals. For instance, pathways involving Ras, ephrinB, cytokine, and integrin receptors were shown to regulate the activity of Src bound to the NMDA receptor (53). Ion channel activity may also be regulated by pathways that involve protein phosphatases opposing Src. IGF-1 was shown to regulate the activity of the rod CNG channel via such a pathway (54). Intriguingly, Src can be also stimulated by direct interaction with G␣ s (55)(56). As pointed out before, many neurotransmitters and hormones affect I h currents via G protein-coupled receptors (GPCRs) and G␣ s / G␣ i that control the activity of adenylyl cyclase and, hence, the cAMP level. A link between G proteins and Src could provide a fine tuning mechanism to control the extent of I h modulation by GPCRs.