GABAB receptor/HCN channel complexes in VTA dopamine neurons limit synaptic inhibition and prevent anxiety-like behavior

Aversive stimuli inhibiting dopamine neurons in the ventral tegmental area (DAVTA neurons) induce anxiety-like behaviors. The inhibition of DAVTA neurons is prolonged by GABAB receptor (GBR)-activated K+-currents, which exhibit a rapid desensitization of unknown physiological relevance. We now report that GBRs associate via auxiliary KCTD16 subunits with HCN channels, which facilitates activation of hyperpolarization- activated currents (Ih) by GBR-activated K+ currents. Activation of Ih underlies rapid K+ current desensitization in DAVTA neurons and limits GBR-mediated inhibition. Disruption of the GBR/HCN complex in KCTD16-/- mice or blockade of Ih prolongs optogenetically driven inhibition of DAVTA neuron firing. KCTD16-/- mice exhibit an increased anxiety-like behavior in response to stressful stimuli, which is reproduced by in vivo CRISPR/Cas9-mediated KCTD16 ablation in DAVTA neurons or intra-VTA infusion of HCN antagonist to wild-type mice. Our data reveal that GBR-induced Ih protect DAVTA neurons from prolonged GBR- mediated inhibition in response to stressors, which moderates anxiety-like behaviors.


Introduction 16
Dopamine neurons of the ventral tegmental area (DA VTA ) are implicated in the processing 17 of reward and emotions (Chaudhury et al., 2013;Coque et al., 2011;Tovote, Fadok, & 18 Luthi, 2015;Tye et al., 2013;Zweifel et al., 2011). Aberrant DA VTA neuron activity is 19 associated with psychiatric disorders, including addiction, depression and anxiety 20 (Bellone, Loureiro, & Luscher, 2021;Jennings et al., 2013;Nestler & Carlezon, 2006). In 21 the latter case, environmental stimuli are perceived as threatening or overly aversive. 22 Aversive stimuli that attenuate DA VTA neuron activation induce avoidance and cause a 23 persistent, generalized anxiety-like phenotype in mice (Tan et al., 2012;Ungless, Magill, GBRs co-purified HCN2 in the presence of KCTD16 but none of the other KCTDs from 23 transfected HEK293 cells ( Figure 1B). KCTD16 also co-purified HCN2 in the absence of GBRs, indicating that HCN2 binds via KCTD16 to GBRs (Figure 1 -figure supplement 1 1A). Indeed, GBRs did not co-purify HCN2 when mutation of Y902A in GB2 (Schwenk et 2 al., 2016;Sereikaite et al., 2019) prevented binding of KCTD16 to the receptor ( Figure  3 1C). Four different subunits assemble homo-and hetero-tetrameric HCN channels (Biel 4 et al., 2009). KCTD16 selectively binds to HCN2 and HCN3 subunits (Figure 1 -figure  5 supplement 1B). We determined that the H2 domain of KCTD16 (Gassmann & Bettler,6 2012) binds to the N-terminal intracellular domain of HCN2 (aa1-215, Figure 1D-F). In 7 summary, biochemical data show that KCTD16 recruits HCN channels containing HCN2 8 or HCN3 to GBRs. 9 Superimposed I h mimics I Kir3 desensitization 11 DA VTA neurons, which express GBRs, HCN channels and KCTD16 constitute a suitable 12 cellular system to study putative functional interactions between GBRs and HCN channels 13 we typically identified DA VTA neurons in horizontal midbrain slices by their spontaneous 16 firing activity, the presence of I h and, occasionally, by staining for tyrosine hydroxylase 17 (TH) (Figure 2A and C) (Cruz et al., 2004;Ungless & Grace, 2012). When clamped at -60 18 mV, DA VTA neurons exhibited a pronounced I bac desensitization (46.0 ± 4.0%, mean ± 19 s.e.m. n = 9 neurons; Fig. 2b, d), as reported earlier (Arora et al., 2011;Cruz et al., 2004;20 Labouebe et al., 2007;Lalive et al., 2014;Padgett et al., 2012). During I bac desensitization, 21 the input resistance Ri surprisingly only recovered by 9.4 ± 3.1%, while the Ri strongly 22 increased after GBR inhibition with the antagonist CGP54626 ( Figure 2B and D). The 23 small decrease in I bac conductivity during continuous baclofen application is at odds with 24 the pronounced I bac desensitization. Because GABA VTA neurons, which do not express HCN channels, lack the pronounced I bac desensitization observed in DA VTA neurons (Cruz 1 et al., 2004), we tested whether pharmacological blockade of I h influences the time course 2 of I bac in DA VTA neurons. Indeed, the I h blocker zatebradine (Bucchi, Baruscotti, & 3 DiFrancesco, 2002) significantly attenuated I bac desensitization in WT DA VTA neurons (p < 4 0.001, Figure 2B and D), showing that I h contribute to I bac desensitization and that I Kir3 5 exhibit per se little desensitization. KCTD16 -/-DA VTA neurons also exhibited a significantly 6 reduced I bac desensitization compared to WT neurons (p < 0.01, Figure 2B and D), 7 suggesting that disruption of the GBR/HCN complex reduces desensitization. KCTD12 -/-8 neurons exhibited similar I bac desensitization as WT neurons, showing that KCTD12 does 9 not contribute to I bac desensitization in DA VTA neurons. These findings support that 10 activation of I h counteracts I Kir3 and that superimposed I h and I Kir3 underlie the apparent I bac 11 desensitization. The results additionally indicate that GBRs are more efficient in triggering 12 I h in WT than in KCTD16 -/neurons, showing that HCN/GBR complex formation promotes 13 I h activation. 14 15

GBR-induced hyperpolarization activates dendritic I h 16
For I bac recordings, DA VTA neurons were clamped at -60 mV, which normally should 17 prevent I h activation by hyperpolarization. We therefore tested whether incomplete space 18 clamp is responsible for hyperpolarization of DA VTA neurons by GBR-activated I Kir3 and for 19 activation of I h (Williams & Mitchell, 2008). Indeed, somatic voltage clamp did not prevent decreasing influence of the somatic voltage clamp. The I bac exhibited a strong attenuation dendritic hyperpolarization due to incomplete voltage clamp activates dendritic I h , which 1 opposes the I Kir3 . Of note, due to the influence of the somatic voltage clamp, I bac -induced 2 hyperpolarization in the proximal dendrites was slow with maximal hyperpolarization only 3 observed after approximately a minute (Figure 2 -figure supplement 1B). 4 Blocking I h with zatebradine in somatic current-clamp recordings resulted in a 5 significantly larger baclofen-induced maximal hyperpolarization (control: -75.2 ± 0.9 mV, 6 n = 10 neurons; zatebradine: -85.6 ± 1.4 mV, n = 6 neurons; mean ± s.e.m. p < 0.001, 7 unpaired Welch's t-test) ( Figure 3A-C). The conductance at maximal hyperpolarization 8 (g max ) was significantly reduced in the presence of zatebradine, revealing that g HCN 9 contributes to the conductance induced by GBR activation (control: 15.6 ± 1.7 nS, 10 zatebradine: 5.2 ± 1.6 nS; p < 0.001, unpaired Welch's t-test). Zatebradine also 11 significantly reduced the conductance at steady state (g steady ) ( Figure 3D), showing that 12

HCN channels largely remain open during GBR activation by baclofen. 13
The activation kinetics of HCN channels reported in the literature are 14 approximately 100 ms (Biel et al., 2009;Kaupp & Seifert, 2001). Therefore, voltage-15 dependent activation of I h appears to be too fast to account for the attenuation of the I bac 16 in DA VTA neuron (τ = 156.4 ± 19.2 sec, mean ± s.e.m. n = 9 neurons, calculated for the 5 17 minutes following I max ; Figure 2B). We therefore determined I h activation kinetics in DA VTA 18 neurons by injecting rectangular voltage commands from -60 (the holding potential in 19 voltage-clamp experiments) to -75 mV (the maximal hyperpolarization induced by baclofen 20 in DA VTA neuron) ( Figure 3E). Fitting a double exponential function to the I h yielded time 21 constants in the order of tens of seconds (τ 1 = 9.5 ± 2.1 and τ 2 = 58.0 ± 7.6 sec, mean ± 22 s.e.m. n = 8 neurons, Figure 3F). The non-instantaneous hyperpolarization induced by 23 diffusion of baclofen into the brain slice and the somatic voltage-clamp opposing 24 hyperpolarization will further slow I h activation kinetics. Therefore, the kinetics of I h activation in DA VTA neurons under our experimental conditions are compatible with the 1 kinetics of I bac attenuation. supplement 1B). Altogether, these data indicate that GBRs do not allosterically influence 12 HCN channel activity. Therefore, exclusively hyperpolarization-dependent activation of 13 HCN channels accounts for the observed attenuation of GBR-mediated inhibition. 14 Dissociation of HCN channels from GBR-complexes in the absence of KCTD16 impairs 15 this regulatory mechanism, indicating that it relies on the GBR-dependent subcellular 16 localization of HCN channels in the dendrites. ChR2 under the VGAT promoter (ChR2 VGAT ). Illumination of ChR2 VGAT horizontal midbrain 22 slices with blue light via the microscope objective induced intense firing activity in GABA VTA 23 neurons ( Figure 4A) and evoked IPSPs in DA VTA neurons ( Figure 4B). In WT::ChR2 VGAT mice, these IPSPs were followed by a rebound depolarization ( Figure 4B) that has been 1 attributed to I h activation (Biel et al., 2009). Pharmacological blockade of HCN channels 2 with zatebradine prevented rebound depolarization and prolonged the duration of IPSPs 3 (p < 0.01, Figure 4B and C). Similarly, DA VTA neurons from KCTD16 -/-::ChR2 VGAT mice 4 exhibited longer IPSPs compared to DA VTA neurons from WT::ChR2 VGAT mice (p < 0.05, 5 Figure 4B and C) and rebound depolarization was observed less frequently (in 2 out of 8 6 neurons). IPSP amplitudes were similar in all groups ( Figure 4B and C). 7 To study whether KCTD16 regulates DA VTA neuron firing in vivo, we performed 8 optogenetic experiments in anesthetized ChR2 VGAT mice ( Figure 5A). As previously 9 observed (Tan et al., 2012), optogenetic activation of GABA VTA neurons ( Figure 5B) 10 resulted in a time-locked decrease of DA VTA neuron firing ( Figure 5C). The inhibition of 11 DA VTA neurons was more pronounced in KCTD16 -/-::ChR2 VGAT mice compared to 12 WT::ChR2 VGAT mice, as indicated by a reduced firing rate (p < 0.05, Figure 5D) and a 13 longer latency to fire an action potential after the onset of inhibition (p < 0.05, Figure 5E). 14 GABA VTA and DA VTA neurons were identified based on validated in vivo 15 electrophysiological properties, such as action potential width and the basal firing 16 frequency (Ungless & Grace, 2012), which were similar between genotypes ( Figure 5F  17 and G). In vivo recordings thus corroborate that GBR/HCN complexes serve to limit 18 synaptic inhibition of DA VTA neurons. 19 20

KCTD16 -/mice exhibit increased anxiety-like behavior 21
A decrease in DA VTA neuron excitability was shown to result in an anxiety-like behavior in 22 mice (Zweifel et al., 2011). The increased inhibition of DA VTA neuron firing observed in 23 KCTD16 -/mice in response to GABA VTA activation may therefore increase anxiety-like behaviors in stressful situations. We used the light/dark chamber test to analyze whether 1 KCTD16 -/mice exposed to a mild naturalistic stressor exhibit an exacerbated anxiety-like 2 behavior. This test relies on the innate aversion of mice to brightly illuminated areas. 3 However, when presented in a novel environment, mice also have a tendency to explore. 4 This conflicting situation leads to anxiety-like behaviors (Takao & Miyakawa, 2006). In the 5 light/dark chamber test, KCTD16 -/mice showed increased anxiety-like behavior, as 6 indicated by fewer entries (p < 0.05) and less time spent in the light chamber (p < 0.01) 7 compared to WT littermate mice ( Figure 6A and C). KCTD16 -/mice also covered less 8 distance than WT mice (p < 0.01, Figure 6B). This was not due to a motor impairment, as 9 both genotypes exhibited similar locomotor behavior in an open arena, when spontaneous 10 locomotion was decomposed into bouts (number, length, speed and duration) to test for 11 possible motor differences (Figure 6 -figure supplement 1). 12 To confirm exaggerated anxiety in response to stressors, we tested mice in a mild 13 version of the classical fear-conditioning paradigm. The test consisted of one training 14 session with five auditory stimuli (tones) paired with an aversive foot shock. 24 hours later, 15 in a novel context (different from the training context), freezing behavior in response to the 16 novel context and to the tones in the absence of foot shock was assessed ( Figure 6D). At 17 baseline, both genotypes showed a similar low level of freezing in the test chamber ( Figure  18 6E). However, after the first tone and throughout the entire session, KCTD16 -/mice 19 displayed significantly more freezing episodes (p < 0.01) and spent more time freezing (p 20 < 0.05) than WT littermate mice ( Figure 6F). These data indicate an increased anxiety-like 21 behavior in response to the novel context and the predictive tone. 22 In further experiments we used the elevated zero and plus mazes as non-invasive 23 tests for anxiety. KCTD16 -/and WT littermate mice spent significantly more time (p < 24 0.001, Figure 6G explored the open arms significantly less than WT mice, as shown by a decreased dwell 2 time (p < 0.05, Figure 6I and M), decreased number of entries (p < 0.05, Figure 6J  These data corroborate that KCTD16 -/mice exhibit increased anxiety-like behaviors in 5 stress situations. 6 7 CRISPR/Cas9-mediated KCTD16 ablation in DA VTA neurons promotes anxiety-like 8 behaviors 9 AAV-mediated CRISPR/Cas9 technology has recently been established for cell-type 10 specific gene editing in the VTA (DeBaker, Marron Fernandez de Velasco, McCall, Lee, & 11 Wickman, 2021). We applied this approach to ablate KCTD16 specifically in DA VTA 12 neurons of adult mice ( Figure 7A  (AAV8-sgKCTD16-mCherry). A 1:1 mixture of both AAV8-sgKCTD16-mCherry was then 18 injected into the VTA of Cre-dependent LSL-Cas9/EGFP knock-in mice heterozygous for 19 the DA neuron-directed driver DAT-Cre (LSL-Cas9/EGFP::DAT-Cre) to produce mice that 20 are deficient for KCTD16 specifically in DA VTA neurons (DA VTA -KCTD16 -/mice). As a 21 control, the same AAV8 preparation was injected into LSL-Cas9/EGFP knock-in mice 22 lacking the DAT-Cre driver ( Figure 7A). 6 weeks after AAV8 injection, we tested mice for 23 their anxiety levels in the elevated zero and elevated plus mazes ( Figure 7C-H). DA VTAcontrol mice (p < 0.01 and p < 0.05, Figure 7D Figure 7I). Importantly, more than 80% of the editing events resulted in out-of-frame 7 mutations. These data confirm that selective ablation of KCTD16 in DA VTA cells is sufficient 8 to exacerbate anxiety-like behaviors in stressful conditions. 9 10

Intra-VTA inhibition of HCN channels induces an anxiety-like phenotype 11
Our data support that the increased anxiety-like behavior of KCTD16 -/mice in stress 12 situations relates to a reduced activation of HCN channels in DA VTA neurons. We therefore 13 tested whether zatebradine exhibits anxiogenic properties when infused into the VTA. One 14 group of WT mice was first infused with saline and 24 hrs later with zatebradine. The mice 15 were tested 30 min after each infusion for their anxiety levels in the elevated zero maze 16 and subsequently in the elevated plus maze ( Figure 8A). A second group of mice first 17 received zatebradine and 24 hrs later saline before testing in the mazes ( Figure 8A). 18 Because the behavioral data in the presence of saline or zatebradine of the two groups of 19 mice were similar, we pooled the data for analysis. After infusion with zatebradine, all mice 20 spent less time in the open arms of the mazes (p < 0.05, Figure 8C-F), thus mimicking the 21 anxiety-like behavior of KCTD16 -/mice in the mazes. The infusion site was validated by 22 infusion of green fluorescent beads and staining for TH in brain slices ( Figure 8B). 23 Pharmacological data therefore corroborate that lack of HCN channel activity in DA VTA influences on neuronal activity, the gating of Kir3 channels by GBRs must be under 6 stringent temporal control. Endocytosis of GBRs, which typically terminates second 7 messenger signaling, is too slow to terminate Kir3 activity in a timely manner (Gassmann 8 & Bettler, 2012). GBRs therefore use RGS and KCTD12 proteins, which directly interfere 9 with G protein activity, to rapidly terminate signaling to Kir3 channels (Mutneja, Berton, 10 Suen, Luscher, & Slesinger, 2005;Raveh, Turecek, & Bettler, 2015;Schwenk et al., 2010;11 Turecek et al., 2014;Xie et al., 2010;Zheng et al., 2019). Here, we describe a novel 12 mechanism that allows for rapid inactivation of GBR-activated I Kir3 . This mechanism relies 13 on association of HCN channels with GBRs, which facilitates activation of HCN channels 14 through hyperpolarization by I Kir3 . In contrast to rapid termination of I Kir3 by RGS and 15 KCTD12 acting at the G protein, this novel mode of l Kir3 inactivation relies on activation of 16 I h that counteract l Kir3 . GBR-dependent I h activation provides a straightforward explanation 17 for the more pronounced I bac desensitization observed in DA VTA neurons compared to 18 neighboring GABA VTA neurons, as the latter lack I h (Cruz et al., 2004).  Likewise, optogenetic activation of GABA VTA neurons in vivo produces a more pronounced 23 and prolonged inhibition of DA VTA neuron firing in KCTD16 -/mice than in WT mice, further inhibition. Our data are consistent with GBRs capturing HCN channels at postsynaptic 1 sites where they are gated by hyperpolarizing IPSPs. The activation of HCN channels is 2 not dependent on a direct modulation or allosteric regulation by GBRs and simply relies 3 on the spatial co-localization with hyperpolarizing conductances. HCN channels have a 4 propensity to traffic to the distal dendrites (Lorincz, Notomi, Tamas, Shigemoto, & Nusser, 5 2002). Translocation of HCN channels to distal dendrites therefore likely explains why 6 HCN channels fail to regulate GBR-mediated inhibition in KCTD16 -/mice. I h activation 7 explains the small contribution of GBR-activated I Kir3 to IPSPs in DA VTA neurons during 8 inhibitory activity, for example during reward omission or the presentation of aversive 9 stimuli (Bromberg-Martin, Matsumoto, & Hikosaka, 2010;Cohen, Haesler, Vong, Lowell, 10 & Uchida, 2012;Schultz, 2013;Tan et al., 2012). GBR-dependent I h activation may not be 11 limited to DA VTA neurons and is expected to play a role in shortening the duration of IPSPs 12 in other neurons as well. It is assumed that GBR-mediated slow IPSPs are mostly 13 generated during periods of intense GABAergic activity, when GABA concentrations are 14 high enough to reach extrasynaptic GBRs (Gassmann & Bettler, 2012). Our data imply 15 that postsynaptic GBRs may also be activated during less intense GABAergic activity, but 16 that I h occlude GBR-mediated IPSPs. increased in KCTD16 -/mice. Since a reduced excitability of DA VTA neurons is anxiogenic 24 (Zweifel et al., 2011), we expected that increased DA VTA neuron inhibition in response to stressors underlies the increased anxiety-like behavior of KCTD16 -/mice. Indeed, cell-1 type specific ablation of KCTD16 in DA VTA neurons is sufficient to promote anxiety 2 phenotypes, supporting that the GBR/HCN interaction in DA VTA neurons is behaviorally 3 relevant. This is corroborated by intra-VTA blockade of I h in WT mice, which is expected 4 to increase DA VTA neuron inhibition in response to stressors and replicates the anxiety 5 phenotype of KCTD16 -/mice. 6 DA VTA neurons can be separated into anatomically and functionally distinct 7 subpopulations that interact with specific neural circuits associated with anxiety and 8 reward-seeking behaviors (Morales & Margolis, 2017). The GBR/HCN regulatory 9 mechanism described here will be operational in "conventional" DA VTA neurons 10 characterized by large I h current densities but not in DA VTA neurons lacking I h currents 11 Brains were washed in ice-cold PBS (137 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 1.8 9 mM KH 2 PO 4 ) and placed at 100 mg/ml in ice-cold homogenization media containing: 10 320 mM sucrose, 4 mM HEPES, pH 7.5, 1 mM EDTA, 1 mM EGTA and protease 11 inhibitors (Roche Diagnostics). For homogenization, we used a glass-teflon homogenizer 12 (mouse brain) or a microcentrifuge tube pestle grinder (VTA brain slice) with 30 passes 13 on ice. The homogenized material was cleared by centrifugation at 10 3 x g (4°C, 15 min), 14 the membrane-enriched fraction isolated by ultracentrifugation at 4.8 x 10 4 x g (4°C, 15 45 min) and solubilized at 2 mg protein/ml for 3 hrs at 4°C in NETN-based solubilization 16 buffer containing: 100 mm NaCl, 1 mm EDTA, 0.5% Nonidet P-40, 20 mm Tris/HCl (pH 17 7.4) and protease inhibitors. The solubilized fraction was cleared by ultracentrifugation at 18 10 5 x g (4°C, 45 min) and directly used for Western blots (input 100 μg of total membrane 19 protein) or precleared for 1 hr using 30 μl dry volume of a 1:1 mixture of protein-A and 20 protein-G agarose (Roche Diagnostics). Solubilized membranes (3 mg of total membrane 21 protein) of WT and KCTD16 -/mouse brains were incubated overnight at 4°C with 2 μl of 22 affinity-purified guinea pig anti-KCTD16 antibody, followed by 40 min of incubation with 10 μl of a 1:1 mixture of protein-A/-G agarose. After repeated washing, bound proteins were 1 eluted with Laemmli buffer and resolved using standard SDS-PAGE. 2 HEK293 cells were transfected at 80-90% confluency using PEI transfection reagent 3 (Sigma) with 2 mg/ml PEI per mg of DNA. The total amount of DNA in the transfections 4 was kept constant by supplementing with empty pCI plasmid (Promega). Plasmids 5 encodings HCN1-4, Flag-, Myc-and HA-tagged GB1 and GB2 as well as Flag-and Myc-6 tagged KCTDs were as described (Schwenk et al., 2010;Seddik et al., 2012;Zolles et al., 7 2009). All mutant constructs were generated using overlap extension PCR. To generate 8 KCTD16ΔH2 a stop codon was introduced after E 279 . To generate KCTD16T1 a stop 9 codon was introduced after L 118 . To generate KCTD16ΔT1 the region between M 1 and V 119 10 was excised. To generate KCTD16H1 a stop codon was introduced in KCTD16ΔT1 after 11 E 279 ; in KCTD16H2 the region between M 1 and P 280 was excised. To generate HCN2 aa1 12 -215 a stop codon was introduced after D 215 . HEK293 cells were harvested 48 hrs after 13 transfection in NETN lysis buffer (with protease inhibitors added) by rotating them for 10 14 min at 4°C. Lysates were cleared by centrifugation at 10 3 x g for 20 min at 4°C and directly 15 used for Western blot analysis (input lanes) or precleared as described above and 16 immunoprecipitated for 3 hrs at 4°C with the following antibodies: mouse anti-Myc (9E10, 17 Santa Cruz Biotechnology, 1 μg) or mouse anti-Flag (M2, Sigma, 1 μg), followed by 40 18 min of incubation with 10 μl of a 1:1 mixture of protein-A/-G agarose. Lysates and 19 immunoprecipitates were resolved using SDS-PAGE and probed with the primary 20 antibodies rabbit anti-KCTD16 (RRID: AB_2631050, 1:2500), rabbit anti-GB1 (AB26, rat 21 Alomone labs, 1:1000), guinea pig anti-HCN4 (AGP-004, Alomone labs, 1:1000), mouse 2 anti-β-tubulin III (T8660, Sigma, 1:1000), rabbit anti-TH (ab112, Abcam, 1:1000) and 3 peroxidase-coupled secondary antibodies donkey anti-guinea pig (A7289, Sigma, 4 1:10000), donkey anti-rabbit (NA934, GE Healthcare, 1:10000) and sheep anti-mouse 5 (NA931, GE Healthcare, 1:10000). The guinea-pig anti-KCTD16 antibody was raised 6 against a synthetic peptide derived from mouse KCTD16 (aa7-23) (Metz et al, 2011). The 7 Bradford assay was used to ensure equal protein loading. Antibody incubation was in 5% 8 nonfat dry milk in PBS containing 0.1% Tween-20. The chemiluminescence detection kit 9 SuperSignal West (Thermo Scientific) was used for visualization of proteins on Western 10 blots. 11 12

Behavioral tests 20
To assess spontaneous locomotion in an open field, mice were free to explore a 21 rectangular (30 x 45 cm) arena for a total of 6 min. Automated software (ANY-maze, 22 Stoelting Co) was used to track movements. Data were post hoc analyzed using MATLAB, 23 as described (Ruder, Takeoka, & Arber, 2016). In the light/dark chamber test the were allowed to explore and move freely between the chambers for a total of 15 min. The 1 apparatus was elevated 60 cm from the table and mice video tracked using ANY-Maze 2 software. The dark chamber was lit with red LEDs to allow video tracking. The reported 3 parameters were automatically calculated by the software. In the elevated zero maze mice 4 were free to explore for a total of 6 min an O-shaped maze (5 cm width, 25 cm radius) 5 divided into two open arms and two closed arms (15 cm height). The maze was elevated 6 60 cm from the table. The bottom of the floor of each arm was coated with a semi-7 transparent paint that allowed video tracking of the mice. In the elevated plus maze mice 8 were placed for 6 min in a plus-shaped maze consisting of two open arms (7.5 width x 30 9 cm length) and two enclosed arms (7.5 cm width x 30 cm length x 30 cm height) extending 10 from a center platform (7.5 cm width x 7.5 cm length). The maze was elevated 60 cm from 11 the table. ANY-Maze software was used to track movements. For pharmacological 12 inhibtion of HCN channels in vivo, four WT mice were infused with 500 µl of saline and 24 13 hrs later with zatebradine (10 μM) via a cannula implanted above the VTA (Plastics One, 14 5 mm length guide with a 26 gauge, implanted from bregma at AP-3.3, ML-0.8 with a 10° 15 angle, DV-4.1). 30 min after saline or zatebradine infusion mice were tested in the mazes. 16 To avoid that the sequence of the infusions impacts behavioral results, we reversed the 17 order of the infusions such that four additional mice first received zatebradine and 24 hrs 18 The experiment was a simplified version of the classical fear-conditioning task and was 24 divided in two steps with a conditioning session followed by a test session 24 hrs later (10 and diagonal stripes whereas the test context (17 cm x 26 cm x 17 cm) exhibited large 1 black dots. During the conditioning session, mice received 5 auditory stimuli (25 sec, 1 Hz 2 pure tone), each paired with a foot shock (2 s, 0.6 mA). On the test session freezing was 3 measured in response to the predictive auditory stimuli in the absence of foot shocks. 4 Freezing was continuously assessed with ANY-Maze. were randomly allocated to the experimental groups unless they were predefined by the 15 genotype. Masking was not performed for any experiment. Data are presented as mean ± 16 standard error of the mean (s.e.m.) unless indicated otherwise. To determine whether to 17 use parametric or non-parametric statistical tests, Shapiro-Wilk test for normality of 18 residuals was applied. The statistical tests adopted are described in the figure legends. 19 All p values reported are two-tailed, and a p value of < 0.05 was considered statistically 20 significant. 21 22

Data availability 1
All data needed to evaluate the conclusions in this paper are present in the main 2 manuscript and/or the Supplementary Materials. Numerical data that are represented in 3 graphs are also provided as source data excel files.  s.e.m., n indicates the number of mice tested. Source data for all plots are provided in 10