Structure-guided design of a cell penetrating peptide preventing cAMP modulation of HCN channels

The auxiliary subunit TRIP8b prevents cAMP activation of HCN channels by antagonizing its binding to their cyclic-nucleotide binding domain (CNBD). By determining an NMR-derived structure of the complex formed by the HCN2 channel CNBD and a minimal TRIP8b fragment, TRIPnano, we show here a bipartite interaction between the peptide and CNBD which prevents cAMP binding in two ways: through direct competition for binding at the distal C-helix of the CNBD; and through an allosteric reduction in cAMP affinity induced by TRIP8b binding to the CNBD N-bundle loop. TRIPnano abolishes cAMP binding in all three isoforms, HCN1, HCN2 and HCN4 and can be used to prevent cAMP stimulation in native f-channels. Application of TRIP8bnano, or its delivery via a cell-penetrating sequence, in sinoatrial node myocytes, selectively inhibits beta-adrenergic stimulation of the native If current and mimics the physiological concentrations of acetylcholine leading to a 30% reduction in the spontaneus rate of action potential firing.


Introduction
Hyperpolarization-activated cyclic nucleotide-regulated (HCN1-4) channels are the molecular correlate of the If/Ih current, which plays a key role in controlling several higher order electrophysiological functions, including dendritic integration and intrinsic rhythmicity both in cardiac and neuronal cells (Robinson & Siegelbaum, 2003). Unique among the voltage-gated ion channel superfamily, HCN channels are modulated by the direct binding of cAMP to their C-terminal CNBD. cAMP binding enhances channel opening upon hyperpolarization via conformational changes in the CNBD that are propagated, through the C-linker domain, to the pore (Wainger et al, 2001;Zagotta et al, 2003).
In addition to cAMP, HCN channels are regulated by TRIP8b, their brain-specific auxiliary (β) subunit, which modulates channel trafficking and gating (Santoro et al, 2009;Zolles et al, 2009). TRIP8b binds HCN channels in two distinct sites: the tetratricopeptide repeat (TPR) domain, which binds the last three amino acids (SNL) of HCN channels; and the TRIP8bcore domain, which interacts with the HCN channel CNBD domain . Although TRIP8b is subject to alternative splicing giving rise to nine isoforms with different effects on HCN trafficking (Santoro et al, 2009;Piskorowski et al, 2011;Lewis et al, 2009), all isoforms act in the same way on the electrical activity of HCN channels: they bind to the cAMP-free state of the CNBD and thus antagonize the effect of the ligand on the voltage dependency of the channel (Hu et al, 2013).
In previous work, we contributed to elucidating the dual mechanism of action of cAMP and TRIP8b on HCN channels by using a simplified system that included the HCN2 CNBD and the TRIP8bcore protein fragments (Saponaro et al, 2014). By means of solution NMR spectroscopy we described in atomic detail the cAMP-induced conformational changes occurring in the HCN channel CNBD. These include a major rearrangement of the Cterminal helix of the CNBD (C-helix), which undergoes lateral translation and folding upon cAMP binding; and an upward movement of the N-terminal helical bundle, a helix-turn-helix motif immediately upstream of the CNBD -roll, which presumably initiates the cAMPinduced facilitation of pore opening. These movements are ostensibly inhibited by TRIP8b binding to both the C-helix and N-terminal helical bundle, thus explaining the allosteric inhibition exerted by TRIP8b on the HCN channel response to cAMP (Hu et al, 2013). The above TRIP8b interaction sites were validated by deletion analysis, confirming that both the N-terminal helical bundle and the C-helix are necessary but not sufficient for TRIP8bcore binding to the CNBD (Saponaro et al, 2014). Despite these findings, several reports have suggested a direct competition mechanism between the two ligands for a common binding region that comprises the C-helix and the phosphate binding cassette, PBC, in the HCN channel CNBD (Han et al, 2011;Deberg et al, 2015;Bankston et al, 2017).
A comprehensive structural model of the complex between TRIP8b and the HCN channel CNBD, which may explain the interaction in atomic detail, is therefore crucial for solving the apparent discrepancies between the two proposed models.
We present here a NMR-based 3D model structure of the complex formed by a minimal TRIP8b fragment and the CNBD of the human HCN2 channel isoform, which provides molecular support for both direct and indirect (allosteric) competition modes exerted by TRIP8b on cAMP binding. TRIP8bnano prevented cAMP stimulation in all HCN isoform tested (HCN1, 2 and 4) and was further employed to manipulate the native If current in cardiac pacemaker cells. This result opens the possibility of controlling in vivo the cAMP-dependent facilitation of HCN channel opening, which represents the basis for the autonomic regulation of cardiac activity and spontaneous neuronal firing (Robinson & Siegelbaum, 2003).

Results
We have previously shown that TRIP8bcore (residues 223 -303 of mouse TRIP8b splice variant 1a4, hereafter TRIP8b) interacts with two elements of the isolated CNBD protein fragment from HCN channels (residues 521 -672 of human HCN2, hereafter CNBD): the Chelix and the N-bundle loop, a sequence connecting helix E' of the C-linker with helix A of the CNBD (Saponaro et al, 2014). Biochemical assays confirmed that each of these two elements i.e. the N-bundle loop and C-helix is necessary but not sufficient for binding (Saponaro et al, 2014).
To understand the interaction in atomic detail, we used solution NMR spectroscopy to characterize the structural properties of the CNBD-TRIP8bcore complex. However, the NMR spectra of TRIP8bcore showed very few signals. In order to improve the quality of the NMR spectra, we reduced the length of the TRIP8b fragment by progressively removing residues at the N-and C-termini with no predicted secondary structure. The truncated peptides were then tested for CNBD binding activity by isothermal titration calorimetry (ITC). We thus identified a 40 amino acid peptide (TRIP8bnano, comprising residues 235 -275 of TRIP8b, Fig 1A) with a binding KD of 1.5 ± 0.1 µM, a value very similar to the KD of 1.2 ± 0.1 µM obtained with TRIP8bcore ( Fig 1B). TRIP8bnano was therefore employed for all subsequent NMR experiments, resulting in a remarkable improvement in the spectral quality and sample stability.

Structural characterization of TRIP8bnano bound to CNBD
The comparison of the 1 H-15 N HSQC spectra of TRIP8bnano with and without CNBD bound shows that the peptide folds upon interaction with the CNBD. Thus, 1 H-15 N HSQC spectrum of TRIP8bnano without CNBD shows a limited 1 H resonance dispersion characteristic of unstructured proteins (Dyson & Wright, 2004), while a larger number of well-dispersed amide signals appear in the spectrum of the CNBD-bound form ( Fig 1C). Importantly, we were now able to assign the backbone chemical shift resonances of TRIP8bnano bound to the CNBD. The φ and ψ dihedral angles obtained from the NMR assignment indicate that the peptide displays two α-helices (stretch L238-E250 named helix N and stretch T253-R269 named helix C) when bound to CNBD. The helices are separated by two amino acids; three and six residues at the N-and C-termini, respectively, are unstructured ( Fig 1D).

Structural characterization of CNBD bound to TRIP8bnano
NMR-analysis of the CNBD fragment bound to TRIP8bnano revealed that the interaction with the peptide does not affect the overall fold of the protein. Thus, the CNBD adopts the typical fold of the cAMP-free state, in line with previous evidence that this is the form bound by TRIP8b (Deberg et al, 2015;Saponaro et al, 2014). More specifically, the secondary structure elements of the cAMP-free CNBD are all conserved in the TRIP8bnano-bound CNBD (Fig 2). This finding generally agrees with a recent double electron-electron resonance (DEER) analysis of the CNBD-TRIP8b interaction, which showed that TRIP8b binds to a conformation largely similar to the cAMP-free state (Deberg et al, 2015). Despite the overall agreement with the DEER study, the NMR data also reveal a new and unexpected feature of TRIP8b binding to the CNBD. Our results surprisingly show that TRIP8bnano induces, upon binding to the CNBD, a well-defined secondary structure of the distal region of the C-helix (Fig 2). This means that the distal region of the C-helix (residues 657-662), which is unstructured in the free form of the CNBD (Saponaro et al, 2014;Lee & MacKinnon, 2017), extends into a helical structure upon ligand binding irrespectively of whether the ligand is cAMP (Saponaro et al, 2014;Puljung & Zagotta, 2013;Lee & MacKinnon, 2017) or TRIP8b (Fig 2). In contrast, and very differently from cAMP, which directly contacts the P-helix in the Phosphate Binding cassette (PBC) and causes its folding (Saponaro et al, 2014;Lee & MacKinnon, 2017), the NMR data show that TRIP8bnano binding to the CNBD does not induce P-helix formation (Fig 2).

Modelling the CNBD-TRIP8bnano complex
Despite the significant improvement in sample stability and NMR spectra quality achieved upon TRIP8bnano binding, we were still unable to assign the side chains of both proteins in the complex and thus could not solve the solution structure of the complex by the canonical NMR procedure. We therefore built a model of the CNBD-TRIP8bnano complex by docking the two NMR-derived structures using the Haddock program (a detailed description of how the respective structures were generated is provided in Materials and Methods and Appendix Table S1).
In order to define the active residues (ambiguous interaction restraints) on the CNBD we used the chemical shift perturbation values as described in Appendix Fig S1. For TRIP8bnano, we defined as active a stretch of residues, E239-E243, previously identified as critical for the interaction (Santoro et al, 2009. Output clusters of this first molecular docking calculation (settings can be found in Material and Methods) were further screened for TRIP8bnano orientations in agreement with a previous DEER analysis, which placed TRIP8b residue A248 closer to the proximal portion and TRIP8b residue A261 closer to the distal portion of the CNBD C-helix (Deberg et al, 2015). Remarkably, in all clusters thus selected, residues E264 or E265 in TRIP8b were found to interact with residues K665 or K666 of the CNBD (Appendix Fig S2). This finding was notable, because we previously identified K665/K666 as being critical for TRIP8b interaction in a biochemical binding assay (Saponaro et al, 2014).
We thus proceeded to individually mutate each of these four positions, and test their effect on binding affinity through ITC. As expected, reverse charge mutations K665E or K666E (CNBD) as well as E264K or E265K (TRIP8bnano) each strongly reduced the CNBD/TRIP8bnano binding affinity (Appendix Fig S3).
Based on these observations, we performed a second molecular docking calculation, including E264 and E265 as additional active residues for TRIP8bnano. This procedure resulted in the model shown in Fig 3, which represents the top-ranking cluster for energetic and scoring function (Appendix Table S2) and was fully validated by mutagenesis analysis as described below. Scrutiny of the model shows that TRIP8bnano binds to both the C-helix and the N-bundle loop ( Fig 3A). Binding to the C-helix is mainly guided by electrostatic interactions between the negative charges on TRIP8bnano, and the positive charges on the CNBD ( Fig 3A). As shown in Fig 3B, the model highlights a double saline bridge (K665 and K666 of CNBD with E265 and E264 of TRIP8bnano) in line with the ITC results described above (Appendix Fig S3). Of note, the contribution of residue R662 to the binding is also consistent with previous experiments showing residual TRIP8b interaction in a CNBD deletion mutant ending at position 663 (Saponaro et al, 2014). Our modelling data suggest that, upon folding of the distal portion of the C-helix, the side chains of residues R662 and R665 face to the inside when contacting cAMP but face to the outside when binding TRIP8b (Appendix Fig S4).
In addition to clarifying the role of residues in the distal portion of the CNBD C-helix, the model also highlights a second important cluster of electrostatic interactions with R650 in the proximal portion of the CNBD C-helix contacting E240 and E241 in helix N of TRIP8bnano ( Fig   3C). To confirm the contribution of these residues, we reversed charges and tested each residue mutation for binding in ITC. The results in Appendix Fig S3 show that R650E caused a more than six-fold reduction in binding affinity for TRIP8bnano, with smaller but significant effects seen also for E240R and E241R.
A third important contact highlighted by the model is the interaction between N547 in the Nbundle loop of the CNBD and D252 in the link between helix N and helix C of TRIP8bnano ( Fig   3D). We tested this potential interaction by disrupting the expected hydrogen bond between N547 and the carboxyl group of the negative residue (D252) in TRIP8bnano. The asparagine in CNBD was mutated into aspartate (N547D) to generate an electrostatic repulsion for D252, and the carboxyl group in D252 of TRIP8bnano was removed by mutation into asparagine (D252N). As predicted, N547D greatly reduced binding to TRIP8b in ITC assays (Appendix Fig   S3 ), with a smaller but significant effect observed also for D252N (Appendix Fig S3). These results confirm and extend our previous finding that the N-bundle loop contributes in a substantial manner to the binding of TRIP8b (Saponaro et al, 2014).

TRIP8bnano as a tool for the direct regulation of native HCN currents
Next, we asked whether the relatively short TRIP8bnano peptide could be used to block the response of HCN channels to cAMP by delivering the peptide to full length channels heterologusly expressed in cells. To this end, we dialyzed TRIP8bnano into the cytosol of HEK 293T cells transfected either with HCN1, HCN2, or HCN4 channels. The peptide was added (10 µM) in the recording pipette together with a non-saturating concentration of cAMP for each isoform (5 µM for HCN2, 1 µM for HCN4) expected to produce a ~10 mV rightward shift in the mid-activation potential (V1/2) in the wildtype channels (Fig 4). No cAMP was added in the case of HCN1, as this isoform is already fully shifted by low endogenous cAMP levels ( Thus, we reckoned it may be employed as a regulatory tool for native If/Ih currents. As proof of principle, we tested whether TRIP8bnano can modulate the frequency of action potential firing in sinoatrial node (SAN) cells. In these cells, If contributes substantially to the diastolic depolarization phase of the action potential. Moreover, the autonomic nervous system modulates the frequency of action potential firing by changing intracellular cAMP levels, which in turn acts on HCN channel open probability (DiFrancesco, 1993). The native If current from cardiomyocytes acutely isolated from rabbit sinoatrial node (SAN) was recorded with and without 10 µM TRIP8bnano in the pipette solution ( Fig 5A). Fig 5B shows that the activation curve recorded in presence of TRIP8bnano is significantly hyperpolarized compared to the control. This indicates that the peptide is displacing the binding of endogenous cAMP to native HCN channels. Moreover, when the experiment was repeated in the presence of 1 µM cAMP, TRIP8bnano prevented the typical cAMP-dependent potentiation of the native If current (Fig 5B). In light of these results, we tested whether TRIP8bnano is also able to modulate cardiac automaticity by antagonizing basal cAMP. The data in Fig 5C show that TRIP8bnano indeed significantly decreased the rate of action potential firing in single SAN cells. Strikingly, the observed 30% decrease in action potential rate corresponds to the effect induced by physiological concentrations of acetylcholine (DiFrancesco et al, 1989).
To conclusively prove that the inhibition of the native If current was specifically due to TRIP8bnano rather than caused by the dilution of the cellular content followed by whole cell configuration, we created a TAT version of TRIP8bnano (hereafter TAT-TRIP8bnano). Indeed, TAT sequence allows the entry of biomolecules into a cell via endocytosis, thus leaving unaltered the cytosolic content (Guidotti et al, 2017).
We thus tested whether both TRIP8bnano and TAT-TRIP8bnano were able to selectively inhibit the beta-adrenergic stimulation of If current, while leaving unaltered the potentiation of Ltype Ca 2+ current (ICa,L). To this end, we recorded either the native If or ICa,L current from cardiomyocytes acutely isolated from mouse sinoatrial node (SAN) in the presence and in the absence of 10 µM TRIP8bnano or TAT-TRIP8bnano, before and after stimulation with 100 nM isoproterenol, a β-adrenergic receptor agonist (Fig 6). Strikingly, TRIP8bnano prevented the isoproterenol-induced increase of If current density, both when the peptide was added in the recording pipette solution (Fig 6A and 6B), and when it was used in the TAT version ( Fig 6A and 6C). The specificity of TRIP8bnano for If current was confirmed by the absence inhibition of the ICa,L current ( fig 6D). Indeed, we failed to record a significant difference in the isoproterenol-stimulated increase of the ICa,L current density between the control condition and 10µM TRIP8bnano ( fig 6E) or TAT-TRIP8bnano ( fig 6F) conditions.

TRIP8b-CNBD complex
In the present study, we describe the interaction between TRIP8b and the HCN channel CNBD at atomic level, based on a NMR-derived structural model of their complex. The data show that the minimal binding unit of TRIP8b, TRIP8bnano, folds in two helices upon binding, suggesting that this region of the regulatory subunit has an intrinsically disordered behavior when not in the complex. The model structurally validates previous indirect evidence, which suggested that TRIP8b binds to two different elements of the CNBD, the N-bundle loop and the C-helix. As a consequence of the interaction with TRIP8bnano, the C-helix in the CNBD increases in length, a behavior already observed in the case of cAMP binding. The model also identifies residues R662 and K665 in the CNBD as interaction partners with TRIP8bnano, two cationic residues also involved in cAMP binding (Zhou & Siegelbaum, 2007). The finding that TRIP8b and cAMP share binding sites on the C-helix provides a solid molecular explanation for functional data which underscore a competition between the two regulators (Han et al, 2011;Deberg et al, 2015;Bankston et al, 2017). However, it has been previously suggested that a direct competition model does not fully explain the mutually antagonistic effect of the two ligands (Hu et al, 2013). Specifically, the fact that the inhibitory effect of TRIP8b on channel activity persists even at saturating cAMP concentrations suggests an allosteric component in the regulation mechanism. Our structural model provides the missing molecular evidence for this allosteric component. TRIP8bnano binds to the solventexposed elements of CNBD (N-bundle loop and C-helix) and does not interact directly with the buried PBC, which remains unfolded, as demonstrated by the observation that the P-helix does not form upon TRIP8b binding. This rules out the possibility that TRIP8bnano controls the affinity for cAMP by directly binding to the PBC. Rather, it confirms that the PBC in the complex is indirectly kept in the low affinity state for cAMP binding (unfolded form) through allosteric long-range interactions.

TRIP8bnano as a tool for modulating native If currents
TRIP8bnano is a minimal protein fragment, which binds the HCN channel CNBD with high affinity and fully abolishes the cAMP effect in all tested isoforms (HCN1, 2 and 4). Given the small size of the peptide (<5kD), TRIP8bnano may be easily adapted for in vivo delivery, and thus constitutes a promising tool for the study or modulation of native HCN channels in systems where the regulatory protein is not expressed, or is expressed at low levels. To this end, we successfully fused to TRIP8bnano an internalization sequence that delivered the peptide into sinoatrial node myocytes via the physiological pathway of the endocytosis without affecting TRIP8bnano function. This modification may significantly expand the use of TRIP8bnano as a tool for non-invasive and in vivo functional assays.
Our structural model explains why a previous attempt at identifying the minimal domain required for TRIP8b activity resulted in a peptide with strongly reduced CNDB affinity, as the fragment selected in the study by Lyman et al (Lyman et al, 2017) is lacking an important contact residue (E240). In the present study, we successfully used TRIP8bnano to selectively control native If currents and pacemaking in sinoatrial node cardiomyocytes. Unlike channel blockers, which inhibit ionic currents, the peptide only interferes with the cAMP-based regulation of HCN channels, while leaving basal HCN functions unaltered. In addition, and in contrast to even the most selective blockers, it is entirely specific for HCN channels. The ability to selectively control a specific molecular mechanism to modulate channel activity represents a novel approach, which yields the promise of a more targeted therapeutic intervention compared to pore blockers.
Mutations were generated by site-directed mutagenesis (QuikChange site-directed mutagenesis kit; Agilent Technologies) and confirmed by sequencing.

Preparation of proteins
The HCN2 CNBD WT and mutant proteins, as well as the TRIP8bcore and TRIP8bnano proteins (WT and mutants) were produced and purified following the procedure previously described (Saponaro et al, 2014).

TRIP8bnano and vice versa
NMR experiments were acquired on Bruker Avance III 950, 700 and 500 MHz NMR spectrometers equipped with a TXI-cryoprobe at 298 K. The acquired triple resonance NMR experiments for the assignment of backbone resonances of cAMP-free HCN2 CNBD (CNBD hereafter) in complex with TRIP8bnano and vice versa are summarized in Table S1. 15 N, 13 C', 13 Cα, 13 Cβ, and Hα chemical shifts were used to derive ϕ and ψ dihedral angles by TALOS+ program (Cornilescu et al, 1999) for both CNBD and TRIP8bnano. For TRIP8bnano, CYANA-2.1 structure calculation (Guntert & Buchner, 2015) was performed using 68 ϕ and ψ dihedral angles and 40 backbone hydrogen bonds as input. For CNBD, CYANA-2.1 structure calculation was performed using 108 ϕ and ψ dihedral angles, combined with the NOEs obtained in our previous determination of the cAMP-free form of the CNBD (Saponaro et al, 2014) for those regions not affected by the interaction with TRIP8bnano. The 10 conformers of TRIP8bnano and CNBD with the lowest residual target function values were subjected to restrained energy minimization with AMBER 12.0  (http://pyenmr.cerm.unifi.it/access/index/amps-nmr) and used as input in docking calculations.

Docking calculations
Docking calculations were performed with HADDOCK2.2 implemented in the WeNMR/West-Life GRID-enabled web portal (www.wenmr.eu). The docking calculations are driven by ambiguous interaction restraints (AIRs) between all residues involved in the intermolecular interactions (Dominguez et al, 2003). Active residues of the CNBD were defined as the surface exposed residues (at least 50% of solvent accessibility), which show chemical shift perturbation upon TRIP8bnano binding.
The assignment of the CNBD bound to TRIP8bnano allowed to highlight the residues of CNBD whose backbone featured appreciable Combined Chemical Shift Perturbation (CSP) ( Fig   S1).
Passive residues of CNBD were defined as the residues close in space to active residues and with at least 50% solvent accessibility.
In the case of TRIP8bnano, the conserved stretch E239-E243, located in helix N, was defined as active region in a first docking calculation, while all the other solvent accessible residues of the peptide were defined as passive. This docking calculation generated several clusters.
A post-docking filter step allowed us to select those clusters having an orientation of TRIP8bnano bound to CNBD in agreement with a DEER study on the CNBD -TRIP8bnano interaction (Deberg et al, 2015). The selected clusters grouped in two classes on the basis of the orientation of helix N of TRIP8bnano (N) relative to CNBD (Fig S2). A second docking calculation was subsequently performed introducing also residues E264-E265, located in helix C of TRIP8bnano as active residues. The active residues for CNBD were the same used for the first calculation. For this second HADDOCK calculation 14 clusters were obtained and ranked according to their HADDOCK score. Among them only four clusters showed both an orientation of TRIP8bnano bound to CNBD in agreement with the DEER study (Deberg et al, 2015) and the involvement of E239-E243 stretch of TRIP8bnano in the binding to CNBD. These clusters were manually analyzed and subjected to a per-cluster re-analysis following the protocol reported in http://www.bonvinlab.org/software/haddock2.2/analysis/#reanal. From this analysis, it resulted that the top-ranking cluster, i.e. the one with the best energetic and scoring functions, has a conformation in agreement with mutagenesis experiments (Fig S3).
Energy parameters (van der Waals energy, electrostatic energy, desolvation energy, and the penalty energy due to violation of restraints) for this complex model are reported in Table   S2.
Both docking calculations were performed using 10 NMR conformers of both the CNBD and the TRIP8bnano structures calculated as described above. In the TRIP8bnano structures the unfolded N-and C-terminal regions were removed, while in the CNBD structures only the unfolded N-terminal region was removed. This is because the C-terminal region of the CNBD is known to comprise residues involved in TRIP8bnano binding (Saponaro et al, 2014).
Flexible regions of the proteins were defined based on the active and passive residues plus two preceding and following residues. The residue solvent accessibility was calculated with the program NACCESS (Hu et al, 2013). In the initial rigid body docking calculation phase, TRIP8bnano was added (10 µM) to the pipette solution. cAMP was added at different concentration to the pipette solution depending on the HCN isoform used: 0 µM for HCN1, 5 µM for HCN2 and 1 µM for HCN4.

Isolation and electrophysiology of rabbit sinoatrial node cells
Animal If activation curves were obtained using a two-step protocol in which test voltage steps (from -30 to -120 mV, 15 mV interval) were applied from a holding potential of -30 mV and were followed by a step to -125 mV. Test steps had variable durations so as to reach steadystate activation at all voltages.
In current-clamp studies, spontaneous action potentials were recorded from single cells superfused with Tyrode solution, and rate was measured from the interval between successive action potential. When indicated cAMP (1 µM) and/or nanoTRIP8b (10 µM) were added to the pipette solution.

Isolation and electrophysiology of mouse sinoatrial node cells
Mice were killed by cervical dislocation under general anesthesia consisting of 0.01 mg/g xylazine (2% Rompun; Bayer AG), 0.1 mg/g ketamine (Imalgène; Merial) and 0.04mg/g of Na-pentobarbital (Euthanasol VET, Laboratoire TVM, Lempdes, France), and beating hearts were quickly removed. The SAN region was excised in warmed ( Electrodes had a resistance of about 3 MΩ. Seal resistances were in the range of 2-5 GΩ.
10µM TRIPb8nano was added to pipette solution. 10µM TAT-TRIPb8nano was added in cell storage solution for at least 30 min before patch clamp recording.

Data analysis
Data were acquired at 1 kHz using an Axopatch 200B    region of the C-helix (residues 657-662), which is unfolded in the free form of the CNBD (Saponaro et al, 2014) and folds upon TRIP8bnano binding, is coloured in red. The unfolded regions at the N-and C-termini of the construct (residues 521-532 and 663 -672 respectively) are omitted for clarity.     Appendix Table S1. Acquisition parameters for NMR experiments performed on cAMP-free human HCN2 CNBD in complex with TRIP8bnano and vice-versa.