Trafficking and Gating of Hyperpolarization-activated Cyclic Nucleotide-gated Channels Are Regulated by Interaction with Tetratricopeptide Repeat-containing Rab8b-interacting Protein (TRIP8b) and Cyclic AMP at Distinct Sites*

Ion channel trafficking and gating are often influenced by interactions with auxiliary subunits. Tetratricopeptide repeat-containing Rab8b-interacting protein (TRIP8b) is an auxiliary subunit for neuronal hyperpolarization-activated cyclic nucleotide-gated (HCN) channels. TRIP8b interacts directly with two distinct sites of HCN channel pore-forming subunits to control channel trafficking and gating. Here we use mutagenesis combined with electrophysiological studies to define and distinguish the functional importance of the HCN/TRIP8b interaction sites. Interaction with the last three amino acids of the HCN1 C terminus governed the effect of TRIP8b on channel trafficking, whereas TRIP8b interaction with the HCN1 cyclic nucleotide binding domain (CNBD) affected trafficking and gating. Biochemical studies revealed that direct interaction between TRIP8b and the HCN1 CNBD was disrupted by cAMP and that TRIP8b binding to the CNBD required an arginine residue also necessary for cAMP binding. In accord, increasing cAMP levels in cells antagonized the up-regulation of HCN1 channels mediated by a TRIP8b construct binding the CNBD exclusively. These data illustrate the distinct roles of the two TRIP8b-HCN interaction domains and suggest that TRIP8b and cAMP may directly compete for binding the HCN CNBD to control HCN channel gating, kinetics, and trafficking.

HCN 3 channels are encoded by the HCN1-4 gene family and mediate hyperpolarization-activated cation current, I h , which plays an important role in regulating electrical activity of neurons (for review, see Refs. [1][2][3]. HCN channel gating and trafficking to specific subcellular domains contribute critically to the multiple physiological roles of the channels, and regulation of these processes is important for normal physiological functions as well as in disease (4). Each pore-forming HCN subunit is composed of an intracellular N terminus, six transmembrane domains, and a large intracellular C-terminal tail that includes a group of 6 helices (AЈ-FЈ, C-linker) that links the transmembrane domains to a cyclic nucleotide binding domain (CNBD). Direct binding of cyclic nucleotides to the CNBD of HCN subunits facilitates channel availability by opening of channels at more physiological (less hyperpolarized) potentials and by speeding kinetics of activation (5)(6)(7).
In addition to being regulated by cAMP, HCN channels interact with numerous proteins that can influence their function (for review, see Ref. 8). Recent studies have demonstrated that in the brain, native HCN channels co-purify with TRIP8b (also called peroxin 5-like protein, PEX5L) (9), and consistent with a role as an auxiliary HCN channel subunit, TRIP8b regulates HCN channel trafficking, voltage gating, and kinetics in vitro and in vivo (9 -11). The TRIP8b C terminus is highly homologous to the peroxisomal import protein, peroxin 5 (PEX5), and is comprised of two sets of three tetratricopeptide repeat (TPR) domains separated by a linker, a C-terminal tail, and a region enriched in acidic amino acids N-terminal to the first TPR set that is conserved in both PEX5 and TRIP8b (10,12). The N terminus of TRIP8b shares no sequence homology to PEX5 and is subject to extensive alternative splicing that leads to expression of multiple distinct TRIP8b isoforms. We previously showed that the TRIP8b TPR sets together interact with the HCN1 C-terminal tripeptide, whereas a region including the conserved acidic domain and the first two TPR domains is required for independent interaction with the HCN1 channel C-linker and CNBD (10). Alternatively spliced TRIP8b isoforms differentially affect surface expression and thereby control HCN channel current density, whereas all TRIP8b isoforms exhibit negative regulation on channel gating and kinetics (10,11).
unknown. Because interactions between PEX5 and cargo C-terminal tripeptides are important for translocating PEX5 cargo from the cytoplasm to the peroxisomal matrix, we suspected that the interaction between the TRIP8b TPR domains and the HCN subunit C-terminal tripeptides could be important for the trafficking effects of distinct TRIP8b isoforms. The effects of TRIP8b on HCN channel gating and kinetics are the opposite of the effects of cAMP, and indeed TRIP8b reversibly blocks cAMP effects on HCN2 channel function (9). Whereas we found direct interaction between TRIP8b and the HCN1 C-linker and CNBD, it has remained unknown if TRIP8b and cAMP compete for the same site on the CNBD. In this study we address the distinctive roles of the two interaction sites of TRIP8b and HCN channels using biochemical, cell biological, and electrophysiological approaches.

EXPERIMENTAL PROCEDURES
Cloning-All restriction enzymes were purchased from New England Biolabs. All oligonucleotides for use in PCR amplification and oligo insertion were synthesized by Integrated DNA Technologies (Coralville, IA). The proper introduction of all mutations and deletions was verified by DNA sequencing.
cDNA Plasmids-The yeast two-hybrid and mammalian expression constructs of HCN1, HCN1-HA, and wild type TRIP8b were described previously (10). Point mutations in TRIP8b were introduced in pGAD-TRIP8b(1c). First, using two-part PCR to generate overlapping fragments, a silent XbaI site was introduced in exons encoding TPR2-C27 and -L28, and then a silent AgeI site was introduced in exons encoding TPR5-R16 and -S17. TPR domain mutations were introduced into this new TRIP8b(1c) clone containing silent XbaI and AgeI sites by PCR mutagenesis using primers and restriction enzyme sites listed in Table 1 with HindIII (blunted with Klenow), and NcoI was then inserted into yeast two-hybrid bait vector pGBK-T7 at NcoI/ SmaI. Other CNBD clones were obtained by PCR of the parent plasmids to amplify cDNA encoding the CNBD plus 17 N-terminal and 4 C-terminal amino acids followed by subcloning into yeast two-hybrid bait vector pGBK-T7. Table 2 lists the primers and restriction sites used.
Yeast Two-hybrid Assays-Directed yeast two-hybrid interaction assays were performed as previously described (13).
Preparation of Purified GST Fusion Proteins-Fusion protein was expressed in BL21 bacteria (Stratagene) and purified by glutathione-Sepharose affinity chromatography (Amersham Biosciences) according to manufacturer's instructions. Protein was quantified by Coomassie Plus Protein Assay Reagent (Thermo), separated into aliquot, and stored at Ϫ20°C.
GST Pulldown Assays-HEK293 cells were transfected with cDNA plasmids encoding HCN1, HCN1[⌬3], HCN1[R538E], or HCN1[⌬3ϩR538E] and lysed in TEEN-Tx (TEEN buffer (50 mM Tris-HCl, pH 7.4, 1 mM EDTA, 1 mM EGTA and 150 mM NaCl) containing 1% Triton X-100) as described above. After removal of an input fraction, 5 mg of GST-tagged TRIP8b fusion protein was added along with glutathione-Sepharose beads, and the reaction was gently rocked for 2 h at 4°C. Supernatant was then removed, and beads were washed six times with cold TEEN-Tx followed by the addition of 4ϫ sample buffer and boiling to elute bound proteins. For some experiments, 8-bromoadenosine 3Ј:5Ј-cyclic monophosphate or adenosine 3Ј:5Ј-cyclic monophosphate (Sigma) was added to the pulldown reaction mixture. IC 50 was calculated using GraphPad Prism software.
Western Blotting-Western blotting was performed as previously described (14). Primary antibody concentrations used were gp anti-HCN1 (1:3000) and gp anti-TRIP8b (1:3000), and the concentrations of secondary horseradish peroxidase-conjugated rabbit anti-gp antibody were 1:10,000. Densitometric quantization of band intensity was performed using Photoshop. Statistical analyses were performed by ANOVA using GraphPad Prism software, and the significance of intergroup differences was determined using 1-way ANOVA with Tukey's post hoc test. Differences at the level of p Յ 0.05 were considered statistically significant.
Flow Cytometry-The surface expression evaluation by flow cytometry was performed as described previously (9). HEK293 cells were transfected with HA-HCN1 and TRIP8b(1a-2-4) wild type and mutations with cell transfected with pEGFP as a control. For the cAMP studies, cells were incubated with 100 M forskolin (Sigma) for 1 h after 72 h of transfection. Nonpermeabilized cells were stained with mouse anti-HA primary antibody to label surface HA and then were stained with secondary antibody conjugated to Alexa-647 (Invitrogen). Cells were run in a Cyan ADP flow cytometer (Dakocytomation) at the Northwestern University Flow Cytometry Core facility, and data were analyzed by FloJo software. The presented fluorescence index was calculated as the integral of the Alexa-647 fluorescence with respect to cell number, which was normalized by eGFP-positive cells in each transfection condition. The fluorescence indices for each condition were normalized to the control transfection condition (HCN1 plus eGFP). Statistical analyses were performed using 1-way ANOVA with Tukey's post hoc-test.
Electrophysiological Measurements in HEK293 Cells-HEK293 cells for electrophysiological measurements were grown and maintained as previously described (15). 1-2 days Regulation of HCN Channels before transfection, cells were trypsinized and re-plated on 12-mm glass coverslips at 50 -60% density. Transfection with plasmid cDNA was performed 2-3 days after plating using the calcium-phosphate precipitation method. A vector encoding the wild type mouse HCN1 channel was co-transfected with various TRIP8b constructs at a 1:1 ratio (0.6 g of DNA per construct per coverslip) alongside a vector encoding eGFP (0.06 g/coverslip) to visualize transfected cells.
Whole-cell patch clamp recording was performed 24 -48 h post-transfection. A coverslip containing transfected HEK293 cells was placed in a bath chamber of an inverted Olympus IX-70 microscope and was continuously perfused with an extracellular solution containing 110 mM NaCl, 5 mM KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , and mM 10 HEPES (pH set to 7.30). Recording pipettes were pulled from borosilicate glass and had a resistance of 2-4 megaohms when filled with an intracellular solution containing 105 mM potassium gluconate, 30 mM KCl, 2 mM Mg-ATP, 5 mM EGTA, 0.5 mM CaCl 2 , and 10 mM HEPES (pH set to 7.30). Series resistance was Ͻ15 megaohms and was compensated for at least 70%. Under these conditions, voltage errors originating from uncompensated series resistance were in almost all cases lower than 1 mV and never higher than 2 mV.
Whole-cell voltage clamp was realized with an Axopatch 200B amplifier (Molecular Devices) controlled by a custommade program written in MATLAB (Mathworks). Currents were low-pass Bessel filtered at 2 kHz. The currents were sampled at 5 kHz by a custom-made program written in MATLAB that also controlled the amplifier and generated the voltage commands. Voltages were offline corrected for liquid junction potentials (calculated as 13 mV). I h was evoked by a series of hyperpolarizing steps from a holding potential of Ϫ40 to Ϫ130 mV in decrements of Ϫ10 mV. Analysis of electrophysiological data was performed using custom-written procedures in MATLAB. The amplitude of I h was quantified as the difference between the current measured immediately after the beginning of the hyperpolarizing step and the steady state current determined at the end of the pulse. Current density was calculated using the cell capacitance as read from the compensation dial on the amplifier. The cell capacitance (quantified at 14.4 Ϯ 0.4 picofarads, n ϭ 111 cells) did not differ across the various experimental conditions (oneway ANOVA).
Activation kinetics were best fit to a single exponential function, and all current traces included in the final analysis had a goodness-of-fit value above 0.98. We investigated whether adding a second or third exponential to the fit would lead to a considerable improvement of the fit, but this was hardly ever the case in our preparation. As would have complicated statistical comparison, we did not include these results in the further analysis.
The voltage-dependent current as measured above results from the voltage-dependent activation and the driving force as best described by the Goldman-Hodgkin-Katz equation, is the voltage-dependent conductance of I h , F the Faraday constant, R the gas constant, and T represents the absolute temperature; for Ion in and Ion out we used concentrations that yielded the correct Nernst equilibrium potential for I h (E rev ϭ Ϫ36 mV). The voltage dependence of g(V) was well described by a Boltzmann equation, where g max is the maximal conductance, V is the holding potential, V 50 is the voltage of half-maximal activation, and V s is a slope factor (in mV).
All data points are given as the means Ϯ S.E. Statistical comparison between groups was performed using an ANOVA followed by a Dunnett post-hoc test or by the non-parametric Kruskal-Wallis ANOVA with multiple testing using Dunn-Sidak correction. Statistical analysis was performed using SPSS software (Version 18.0, SPSS Inc.) and MATLAB. p Ͻ 0.05 was considered to indicate a significant difference.

Two Distinct Interaction Domains of TRIP8b and HCN1-In
previous work, we demonstrated that TRIP8b interactions with HCN1 occur at distinct sites in both TRIP8b and HCN1 (10).
Binding to the C-terminal tripeptide of HCN1 (SNL) required the C-terminal portion of TRIP8b, including both sets of TPR domains. Additionally, we found a second interaction site with HCN1 that required a 58-amino acid segment in the N terminus upstream of the TPR sets. This second binding domain interacted with an HCN1 construct containing only the intracellular C-linker and cyclic nucleotide binding domain (HCN1-(386 -591); HCN1-CNBDϩ). We performed a further deletion of 4 -6 of the C-linker helices (HCN1-(453-591); HCN1-CNBD) and found that an intact CNBD, but not C-linker, is required for this interaction (Fig. 1A). For clarity, we refer to these two interactions hereafter in this manuscript as NC (for TRIP8b N terminus, HCN CNBD) and TT (for TRIP8b TPR domains, HCN Tripeptide).
To evaluate the structural architecture of these distinct interaction sites, we performed mutagenesis to disrupt NC or TT interactions in a full-length TRIP8b isoform known to dramatically up-regulate HCN1, TRIP8b(1a-2-4). As a preliminary screen we utilized directed yeast two-hybrid screens to characterize effects of mutations in disrupting either NC or TT interactions. We found that wild type TRIP8b(1a-2-4) interacted with HCN1 C-terminal fragments containing the C-terminal tripeptide (with no CNBD) or the CNBD (with no tripeptide) (Fig. 1A). Wild type TRIP8b(1a-2-4) also interacted with fragments containing either the C-terminal tripeptide or the CNBD from HCN2, HCN3, and HCN4, indicating both the NC and TT interactions occur with all HCN isoforms. We next made a full-length TRIP8b(1a-2-4) construct containing a deletion of the 58 amino acid residues upstream of the TPR domains required for the HCN1 NC interaction, TRIP8b(1a-2-4)[⌬58]. TRIP8b (1a-2-4)[⌬58] did not interact with any HCN CNBD but did interact with all HCN subunit constructs containing the C-terminal tripeptide (Fig. 1A). Thus, the mutation of TRIP8b(1a-2-4)[⌬58] selectively eliminates interaction with the HCN(1-4) CNBD.
We next sought to create mutations in TRIP8b to disrupt the TT interaction with the HCN subunit C-terminal tripeptide. We took advantage of the homology of TRIP8b and PEX5 combined with the knowledge of PEX5 structure (12) and extensive information about functional mutagenesis (16,17) to design TRIP8b constructs predicted to disrupt binding to the terminal three amino acids (SNL). By aligning TRIP8b and PEX5 sequences, residues that are important for PEX5 binding to SKL were identified in TRIP8b (supplemental Fig. 1). We then created PEX5 homologous mutations in full-length TRIP8b. Using yeast two-hybrid assays, we identified four point mutations in four distinct TRIP8b TPR domains that disrupted TT interaction (TPR2-E15K, TPR3-N13A, TPR5-N5K, and TPR6-R2A), none of which affected binding to HCN constructs containing the CNBD (Fig. 1A). The PEX5 C-loop at the C terminus after the TPR domains is important for some PEX5 cargo binding (12), and we similarly found that deletion of the TRIP8b residues C-terminal to the sixth TPR domain also disrupted HCN tripeptide binding (data not shown). We next generated a TRIP8b construct with mutations predicted to completely disrupt interactions with HCN subunits. As shown in Fig. 1A, the double mutant construct TRIP8b(1a-2-4)[⌬58ϩTPR3-N13A] was unable to interact with CNBD-or tripeptide-containing HCN constructs in yeast two hybrid assays. The interaction between TRIP8b mutations and HCN1 was next verified by coimmunoprecipitation in HEK293T cells. Consistent with yeast two-hybrid data, wild type TRIP8b(1a-2-4) coimmunoprecipitated robustly with HCN1. Furthermore, both TRIP8b(1a-2-4)[⌬58] and TRIP8b(1a-2-4)[TPR3-N13A] immunoprecipitated HCN1. As predicted, coimmunoprecipitation was abolished in the TRIP8b construct harboring both mutations, TRIP8b(1a-2-4)[⌬58ϩTPR3-N13A] (Fig. 1, B and C).
Interaction with Both the CNBD and the C-terminal Tripeptide of HCN1 Contribute to TRIP8b Regulation of I h Density-Acting as an auxiliary subunit of the HCN channel complex, TRIP8b interaction influences channel current density and gating (9 -11). Utilizing our TRIP8b constructs harboring specific defects in HCN subunit binding, we sought to determine which binding sites between TRIP8b and HCN subunits contributed to this function. We first evaluated TRIP8b effects on current density. Expression of HCN1 alone in HEK293 cells resulted in a mean current density of 7.4 Ϯ 1.7 pA/pF (measured at Ϫ110 mV, n ϭ 24, Table 3, Fig. 2). Co-expression of HCN1 with wild type TRIP8b(1a-2-4), a TRIP8b isoform that dramatically up-regulates HCN current (10), markedly enhanced current density (33.8 Ϯ 5.4 pA/pF, p Ͻ 0.001, Table 3, Fig. 2). Transfection of TRIP8b(1a-2-4) harboring the ⌬58 mutation also significantly up-regulated current density as compared with HCN1 alone (24 Ϯ 4.3 pA/pF, n ϭ 17; Table 3, Fig. 2), but values were lower than those obtained with full-length TRIP8b(1a-2-4). Interestingly, expression of HCN1 with the TRIP8b(1a-2-  Table 3; Fig.  2). Co-transfection of the double mutant TRIP8b(1a-2-4)[⌬58ϩTPR3-N13A], which cannot interact with either the CNBD or the C-terminal tripeptide domain of HCN1, generated an HCN1 current density that was comparable with that found with HCN1 alone (9.2 Ϯ 0.9 pA/pF, n ϭ 22, p Ͼ 0.05, Table 3, Fig. 2). These results suggest that interaction between TRIP8b and both the CNBD and the C-terminal tripeptide SNL contribute to the TRIP8b effect on HCN current.

TRIP8b Interaction with CNBD or C-terminal Tripeptide Regulates Current Density through the Effects on HCN Channel
Trafficking-TRIP8b controls current density via regulation of HCN1 surface expression (10,11). Therefore, we expected that the ⌬58 and TPR3-N13A mutants, both enhancing current density, would also enhance surface expression. To measure surface expression, we performed flow cytometry using non-

TABLE 3 Biophysical properties of I h in the presence of various TRIP8b(1a-2-4) mutants
Biophysical properties of I h in HEK293 cells co-transfected with HCN1 and various TRIP8b(1a-2-4) mutants are shown. Differences in current density were assessed using Kruskal-Wallis test (p Ͻ 0.001) followed by post hoc Mann-Whitney tests with a Dunn-Sidak correction. Other comparisons were made using ANOVA (p Ͻ 0.001 for ; p Ͻ 0.01 for both V 50 and k) with a post hoc Dunnett test. All data points are the mean Ϯ S.E. permeabilized HEK293T cells cotransfected with eGFP and different combinations of TRIP8b mutant constructs and HCN1. As shown in Fig. 3, A and B  had no effect (***, p Ͻ 0.001, 1-way ANOVA with Tukey's post hoc-test, n ϭ 3). Input was 6% of total lysate, and immunoprecipitation was 15%. Molecular mass markers are shown in kDa. channel gating that is manifested in slower activation time constants () and a hyperpolarizing shift of the voltage-activation curve (assessed as the voltage of half-maximal activation, V 50 ) (9 -11). Because the CNBD domain is known to play an important role in the regulation of HCN channel gating (5,6) and because TRIP8b antagonizes effects of cAMP on HCN2 channel function (9), we tested if direct and specific interactions with the CNBD domain mediate the effects of TRIP8b on HCN channel voltage dependence and kinetics.

Construct
Confirming previous results (10,11), co-transfection of HEK293 cells with HCN1 and wild type TRIP8b(1a-2-4) resulted in significantly slowed I h activation kinetics (Fig. 4) accompanied by a hyperpolarizing shift of the activation curve (Fig. 5, see also Table 3). The ⌬58 mutation, which eliminates TRIP8b binding to the CNBD domain, abolished the inhibitory effect of TRIP8b on I h gating as both the activation kinetics and the voltage-activation were comparable with I h in cells that were transfected with HCN1 alone (Figs. 4 and 5). In contrast, co-transfection of HCN1 with TRIP8b(1a-2-4)[TPR3-N13A] exerted an inhibitory effect on both I h kinetics and voltage-dependent activation, similar to the effect of wild type TRIP8b(1a-2-4). Finally, co-transfection of HCN1 with the double mutant TRIP8b(1a-2-4)[(⌬58ϩTPR3-N13A] yielded similar kinetics and voltageactivation properties to those obtained when HCN1 was expressed either alone or along with the ⌬58 mutant. Taken together, these results indicate that whereas current density can be up-regulated by interaction of TRIP8b with either of the two interaction sites on the HCN channel, the effects on channel activation and kinetics are mediated solely by interactions with the CNBD domain. cAMP Inhibits TRIP8b Interaction with the HCN CNBD-The direct interaction of TRIP8b with the HCN1 CNBD to affect channel activation is consistent with prior findings suggesting that TRIP8b is a competitive antagonist of the functional effects of cAMP on HCN2 activation (9). To determine whether cAMP directly antagonizes the interaction between TRIP8b and the HCN CNBD, we performed GST pulldown assays in the presence of cAMP concentrations from 0 to 1 mM, a range that includes concentrations below and above saturation of cAMP effects on I h voltage gating (18) (Fig. 6, A  and B). Graded concentrations of cAMP did not significantly impair the interaction of TRIP8b with an HCN1 construct containing an intact C-terminal tripeptide. However, increasing amounts of cAMP reduced the binding of TRIP8b to an HCN1 construct lacking the C-terminal tripeptide in a concentration-dependent manner, with an IC 50 of 40 nM (n ϭ 3). This inhibition of TRIP8b binding to the HCN1 CNBD by cAMP indicates that TRIP8b and cAMP might compete at overlapping binding sites to regulate HCN channel gating.
The crystal structure of mouse HCN2 has demonstrated that cAMP interacts directly with the highly conserved Arg-591 (Arg-538 in HCN1) within the CNBD, leading to allosteric changes coupled to gating (5,19,20). Mutation of HCN1 Arg-538 (or HCN2 Arg-591) to glutamate (HCN1[R538E]) eliminates the effects of cAMP on channel function without altering the ability of channels to be regulated by voltage (5). If TRIP8b and cAMP compete for binding to the same site in the CNBD, we reasoned that R538E mutation might prevent TRIP8b binding. We performed a directed yeast-two hybrid assay between TRIP8b(1a-2-4) and an HCN1 CNBD construct harboring the R538E mutation (HCN1-(386 -591)[R538E]) and found the R538E mutation indeed blocked binding of TRIP8b to the CNBD (Table 4).
Because the overall structure of the CNBD, including the importance of a critical Arg for cAMP binding, is shared by other cyclic nucleotide-binding proteins (supplemental Fig.  3 and Flynn et al. (21)), we wondered if TRIP8b might bind to other cyclic nucleotide-gated channels. In directed yeast two-hybrid studies, although TRIP8b interacted with the CNBD of all HCN subunits (Fig. 1A), it did not interact with the CNBD of olfactory or retinal cyclic nucleotide gated channels CNGA1-4, CNGB1, or CNGB3, nor did it bind the CNBD of the ether-a-go-go-like K ϩ channel subunit, KCNH1 (Table 4). TRIP8b also failed to interact with the CNBD of the sea urchin HCN channel, SpIH. These results suggest that the presence of the critical Arg is required but not sufficient to define an overlapping binding site of TRIP8b and cAMP in the CNBD of cyclic-nucleotide regulated K ϩ channels. This implies that TRIP8b-mediated inhibition of cAMP effects may not be a general phenomenon in this group but may rather be specific for vertebrate HCN channels.
cAMP Regulates TRIP8b-mediated Effects on HCN1 Trafficking-Previous studies demonstrated that increasing cAMP levels reverses TRIP8b inhibition of HCN2 gating and kinetics (9). Because we found that TRIP8b interaction with the HCN CNBD affected both channel activation and trafficking, we tested if dissociation of the TRIP8b/HCN1 interaction by cAMP might impair the effects of TRIP8b on surface expression of the HCN1 subunit. We performed flow cytometry studies in untreated HEK293 cells and HEK293 cells treated with forskolin to increase intracellular cAMP levels, evaluating surface HCN1 expression in cells transfected alone or together with TRIP8b constructs. As shown in Fig. 6D, forskolin treatment had little effect on HCN1 surface expression when the channel was expressed alone (30% reduction, n ϭ 6) or when coex-pressed with wild type TRIP8b(1a-2-4) (20% reduction, n ϭ 6). However, when coexpressed with a TRIP8b molecule that increases HCN1 surface expression only via interaction with the CNBD, TRIP8b(1a-2-4)[TPR3-N13A], forskolin reduced HCN1 surface expression by 80% (n ϭ 6, p Ͻ 0.001 versus HCN1). These results are intriguing because they indicate that cAMP and TRIP8b directly compete for a CNBD site that governs channel surface expression.

DISCUSSION
The principal findings of these studies are 1) TRIP8b functions as an auxiliary subunit of HCN channels by interacting with HCN1-4 subunits in a complex manner that involves at least two distinct domains to modulate selectively gating and surface expression, 2) whereas HCN surface expression is governed by TRIP8b binding to both the HCN CNBD and the C-terminal tripeptide binding domains, modulation of gating is dependent on the interaction with the CNBD site alone, and 3) direct binding of TRIP8b and cAMP to the HCN CNBD domain leads to competitive modulation of HCN channel surface expression.
The original identification of TRIP8b as an HCN channelinteracting partner identified the TPR/tripeptide interaction site and also revealed the importance of this interaction for  Table 3 for values and statistical significance. mediating the trafficking effects of a TRIP8b isoform, TRIP8b(1b-2), that resulted in a down-regulation of HCN channel function (22). Here we find that a TRIP8b isoform that enhances HCN function, TRIP8b(1a-2-4), also employs the TPR/tripeptide interaction in regulating channel trafficking. Importantly, the current study identifies a second binding site involving the TRIP8b residues N-terminal to the TPR domains and the HCN CNBD that is also important for TRIP8b-mediated regulation of HCN1 trafficking. Thus, only when both interactions were eliminated did we block TRIP8b-mediated up-regulation of HCN surface expression.
In addition to contributing to TRIP8b-mediated effects on channel trafficking, we find that this second interaction site alone regulates TRIP8b effects on HCN1 gating. Prior studies demonstrated that TRIP8b acts as a competitive antagonist of cAMP function on HCN2 gating (9). We show here that the mechanism for TRIP8b antagonism of cAMP effects is via direct binding to the CNBD. We found that the interaction between TRIP8b and the CNBD was competitively blocked by increasing concentrations of cAMP and that the concentration at which cAMP inhibited 50% of TRIP8b binding to the HCN1 CNBD (40 nM) was similar to that reported for shifting the t1 ⁄ 2 of I h mediated by HCN1 in oocytes (60 nM) (5). These data are consistent with the possibility that TRIP8b binds to and stabilizes the non-cAMP bound CNBD conformation, whereas conformational changes induced by cAMP (19) eliminate TRIP8b binding. Alternatively, TRIP8b and cAMP might share overlapping binding sites. Along these lines, mutation of a conserved Arg residue in the HCN1 or HCN2 CNBD demonstrated that this residue was necessary for both cAMP and TRIP8b binding. Crystallization of the CNBD in the presence of TRIP8b should help resolve the structural mechanisms by which cAMP and TRIP8b compete at the CNBD.
Although we found TRIP8b interacted with the CNBDs of HCN1-HCN4, it is important to note that TRIP8b did not bind the CNBD of CNG channel subunits, suggesting that TRIP8b is not a general antagonist of cAMP effects on CNBD-containing proteins. It is also notable that TRIP8b did not bind the CNBD of the sea urchin HCN subunit, SpIH. SpIH and other invertebrate HCN channel proteins (i.e. fruit fly, spiny lobster, honeybee, Egyptian mosquito) show tremendous diversity in their C termini and all lack the tripeptide SNL conserved across most vertebrate HCN channels. These observations suggest that an important role for TRIP8b in HCN channel function evolved more recently and is consistent with the gene encoding TRIP8b (Pex5l) being present only in chordates. Nonetheless, further studies will be required to determine whether TRIP8b is spe-cific for antagonizing cAMP effects only on vertebrate HCN channels.
Intriguingly, we find that the interaction between TRIP8b and the HCN1 CNBD contributed to the trafficking effects of TRIP8b on the HCN channel. Also surprising was the observation that forskolin was able to regulate the trafficking effects of TRIP8b in a TRIP8b construct that interacted only with the HCN CNBD domain. It remains unclear whether cAMP might play a role in regulating trafficking of HCN channels in the brain, where neuronal activity that typically augments cellular cAMP levels promotes HCN1 membrane expression (23). TRlP8b may facilitate HCN channel internalization via interactions with adapter proteins such as AP-2 (11), and it is interesting to speculate that cAMP in neurons might contribute to dissociation of TRlP8b from HCN channels and increase their surface expression. Future experiments are planned to evaluate further the role cAMP might play in regulating TRlP8b-mediated trafficking effects in vivo.
In summary, the results of our study give insight into the functional importance of two distinct interactions between HCN1-4 subunits and TRIP8b. Whereas both interactions contribute to TRIP8b regulation of current density and intracellular trafficking, unique interaction with the HCN CNBD allows TRIP8b to regulate gating. Our data further suggest that regulation of HCN channel function by TRIP8b is closely linked to cAMP metabolism, as cAMP antagonizes the effects of TRIP8b on HCN function by directly disrupting the interaction between TRIP8b and the HCN CNBD.