Hyperpolarisation-activated cyclic nucleotide-gated channels regulate the spontaneous firing rate of olfactory receptor neurons and affect glomerular formation in mice


N. Nakashima: Department of Physiology, Faculty of Medicine, Kyoto University, Yoshida-Konoe, Sakyo-ku, Kyoto 606-8501, Japan.  Email: nakashima@nbiol.med.kyoto-u.ac.jp

Key points

  • • Olfactory receptor neurons (ORNs) utilise the spontaneous firing activity for axonal targeting.
  • • Cyclic AMP-gated HCN channels depolarised the membranes of ORNs and enhanced their spontaneous firing activity.
  • • Standing activation of the β-adrenergic receptors maintained the basal cAMP level and resulted in the opening of HCN channels.
  • • The over-expression of HCN4 resulted in a decrease in the number of glomeruli in the olfactory bulb, which was rescued by suppressing HCN4 over-expression.
  • • HCN channels, together with the standing activation of G-protein-coupled receptors, maintained the spontaneous firing activity of ORNs and were essential to glomerular formation in the olfactory bulb.

Abstract  Olfactory receptor neurons (ORNs), which undergo lifelong neurogenesis, have been studied extensively to understand how neurons form precise topographical networks. Neural projections from ORNs are principally guided by the genetic code, which directs projections from ORNs that express a specific odorant receptor to the corresponding glomerulus in the olfactory bulb. In addition, ORNs utilise spontaneous firing activity to establish and maintain the neural map. However, neither the process of generating this spontaneous activity nor its role as a guidance cue in the olfactory bulb is clearly understood. Utilising extracellular unit-recordings in mouse olfactory epithelium slices, we demonstrated that the hyperpolarisation-activated cyclic nucleotide-gated (HCN) channels in the somas of ORNs depolarise their membranes and boost their spontaneous firing rates by sensing basal cAMP levels; the odorant-sensitive cyclic nucleotide-gated (CNG) channels in cilia do not. The basal cAMP levels were maintained via the standing activation of β-adrenergic receptors. Using a Tet-off system to over-express HCN4 channels resulted in the enhancement of spontaneous ORN activity and dramatically reduced both the size and number of glomeruli in the olfactory bulb. This phenotype was rescued by the administration of doxycycline. These findings suggest that cAMP plays different roles in cilia and soma and that basal cAMP levels in the soma are directly converted via HCN channels into a spontaneous firing frequency that acts as an intrinsic guidance cue for the formation of olfactory networks.


adenylate cyclase III


β-adrenergic receptor

CNG channels

cyclic nucleotide-gated channels




glyceraldehyde-3-phosphate dehydrogenase

HCN channels

hyperpolarisation- activated cyclic nucleotide-gated channels

HCN4-Tet mice

mice with HCN4 over-expression


inward rectifying potassium channel




olfactory marker protein


odorant receptor


olfactory receptor neurons


normal Ringer solution


Ca2+- and Mg2+-free Ringer solution






l-α-phosphatidylinositol 4,5-diphosphate


phosphatidylinositol 3,4,5-trisphosphate


phosphatidylinositol phosphates


Neural networks in the brain are well designed to perform a variety of functions, and the electrical activity of the neurons frequently affects the network architecture (Goodman & Shatz, 1993; Zhang & Poo, 2001; Serizawa et al. 2006). Spontaneous neural activity has been observed in several nervous systems and is considered essential for the formation of the precise neural networks in the visual (Torborg & Feller, 2005), auditory (Jackson & Parks, 1982; Russell & Moore, 1995; Tritsch et al. 2007; Kandler et al. 2009) and olfactory systems (Yu et al. 2004a). Olfactory receptor neurons (ORNs), which are widely distributed throughout the olfactory epithelium, are unique in that they undergo lifelong neurogenesis (Graziadei & Graziadei, 1979). In addition, the terminals of the axonal projections of ORNs that express a specific odorant receptor converge onto the same set of glomeruli in the olfactory bulb (Vassar et al. 1994; Mombaerts et al. 1996). Thus, ORNs may provide important clues for understand-ing how neurons form and maintain precise networks.

The over-expression of a gene that encodes inward rectifying potassium channels (Kir2.1) suppresses the spontaneous firing activity in ORNs and results in delayed axonal growth and abnormal axonal convergence onto multiple olfactory glomeruli (Yu et al. 2004a). Spontaneous firing may affect the release of neurotransmitters from the ORN terminals, but inhibiting synaptic vesicle release by means of the genetic expression of tetanus toxin light chains throughout a population of ORNs did not result in major changes in axonal targeting (Yu et al. 2004a). Interestingly, suppressing spontaneous firing activity in only a selected population of ORNs results in a much more severe disturbance of the axonal projection pattern than suppression throughout the entire ORN population (Yu et al. 2004a). This result implies that the spontaneous firing frequency of a given ORN relative to those of the surrounding ORNs might be a crucial guidance cue; however, the regulatory mechanism of the spontaneous firing activity of the ORNs is poorly understood.

Recent genetic manipulation studies suggest that the intracellular cAMP concentrations in various ORNs may also act as guidance cues that help direct axonal projections to the appropriate glomeruli (Sakano, 2010). Knocking out adenylate cyclase III (ACIII) disturbs the pattern of normal axonal convergence onto a predicted group of olfactory glomeruli, as does the constitutive activation of the Gs α subunit of G-proteins (Gs-proteins), which are associated with cAMP signalling (Imai et al. 2006; Chesler et al. 2007; Col et al. 2007; Zou et al. 2007). Thus, we hypothesised that cAMP regulates the spontaneous firing activity observed in ORNs.

Cyclic AMP may contribute to electrical activity via direct interactions with two distinct species of ion channels in ORNs: CNG channels and HCN channels (Lynch & Barry, 1991a; Vargas & Lucero, 1999; Kleene, 2008; Biel et al. 2009). CNG channels play a principal role in generating receptor potentials in response to odorants by sensing odorant-induced cAMP production in the cilia (Kleene, 2008; Takeuchi & Kurahashi, 2008). However, ORNs in mice that lack CNG channels still exhibit spontaneous firing activity and show no major defects in axonal targeting (Brunet et al. 1996; Lin et al. 2000). In contrast, Mobley et al. (2010) histologically demonstrated that mice that lacked the HCN1 channel subunit exhibited dispersed axonal targeting in the olfactory bulb and that inhibiting HCN channels in a dissociated ORN culture delayed the extension of newly projecting neurites. Despite such insightful anatomical observations, the physiological roles that HCN channels play in glomerular formation remain unclear. We hypothesised that if the regulation of spontaneous firing does occur, it might be mediated via HCN channels. To test this hypothesis, we performed extracellular recordings from ORNs in slices of the olfactory epithelium.


Treatment of animals

We treated all of our experimental animals in accordance with the Kyoto University guidelines. We used C57/BL6 mice as wild-type animals. Animals used for electrophysiological and molecular biological studies were killed by rapid decapitation with sharp blades. Animals used for histological studies were deeply anesthetised by inhalation of isoflurane (1–5% v/v in air) in a secure plastic container prior to the perfusion fixation and decapitated after fixation.


The olfactory epithelia of postnatal day (P)0–4 mice of either sex or of 7-week-old male mice were frontally sliced using a slicer (PRO10, Dosaka, Kyoto, Japan). Each slice had a thickness of 300 μm, and the tissues were sliced in an ice-cold dissecting solution containing the following compounds (all concentrations are given in mm, and the pH of the solution was 7.4 in this and subsequent media): NaCl 130, KCl 4.5, CaCl2 2, Pipes-Na 5 and glucose 34. The slices were then pre-incubated in O2-aerated external Ringer solution (see below for the composition of this solution) for 30 min at room temperature prior to the experiments. The ORNs were morphologically and electrophysiologically (Cm= 3.8 ± 0.2 pF; Rinput= 3.0 ± 0.5 GΩ, n= 12) identified under a microscope with infrared Nomarski optics (BMX Olympus Optical, Tokyo, Japan). All recordings were obtained at a temperature of 36°C and were made using either a patch-clamp amplifier (Axoclamp 200B, Axon Instruments, Union City, CA, USA) or a current-clamp amplifier (Axoclamp 900A, Axon CNS, Molecular Devices, CA, USA). The external Ringer solution (R[+]) contained the following compounds: NaCl 155, CaCl2 2.5, MgCl2 1, glucose 17, Hepes 10 and KOH 5. CaCl2 and MgCl2 were omitted from R[+] to form the nominal divalent cation-free Ringer solution (R[–]) used in some experiments. High-K+ Ringer solution was prepared by replacing the NaCl in R[–] with KCl via isosmotic substitution. The pH values of the Ringer solutions used in the pH change experiments (Fig. 2B and C) were obtained using a mixture containing 5 mm each of piperazine (5.2), Hepes (7.6) and Tris (8.3) (Ohmori & Yoshii, 1977). The number in parentheses indicates the acidic dissociation constant (pKa) for the compounds at 36°C. During the experiments, the external solutions were continuously exchanged using a peristaltic pump (P-3, Pharmacia Fine Chemicals, USA). All of the whole-cell patch-clamp experiments were performed with a KCl-based internal solution, and the pipettes had resistances of 5–8 MΩ; the composition of the internal solution was as follows: KCl 135, NaCl 5, MgCl2 3, EDTA 5 and Hepes 10. The electrode used for multiple-unit recordings had a resistance of ∼0.2 MΩ when it was filled with the external solution. The extracellular recordings were sampled at a rate of 10 kHz, low-pass filtered at 5 kHz and high-pass filtered at 3 Hz. Currents through the HCN channels were sampled at 10 kHz and were digitally filtered at 500 Hz. The liquid junction potential (3 mV) was corrected offline.


Stock solutions of ZD 7288 (100 mm), 8-Br-cAMP (200 mm) (Tocris, USA), propranolol (100 mm), metopralol (50 mm) and butaxamine (100 mm) (Sigma, USA) were prepared in distilled water. Stock solutions of rolipram (100 mm), SQ22536 (120 mm), wortmannin (10 mm), U73122 (100 mm), ICI 118,551 (30 mm), SR59230A (100 mm) and amiloride (1.0 m) (Sigma, USA) were prepared in dimethylsulfoxide (DMSO) (Wako, Japan). To minimise changes in the osmotic pressure and ionic compositions of the bath solutions, the drugs were added to the solutions just prior to use. Desired drug concentrations were obtained via 100-fold dilutions. Ringer solutions with particularly high concentrations of SQ22536 and SR59230A were prepared by directly dissolving the reagents in the solution just prior to use. Native l-α-phosphatidylinositol 4,5-diphosphate sodium salt derived from bovine brain tissues (PIP2; P9763 from Sigma, USA) was dissolved at a final concentration of 1 μm in the intracellular solution, and the mixture was sonicated for 10 min in an ice-cooled water bath before use. In general, the final DMSO concentration did not exceed 0.1%; at this concentration, the DMSO did not have any significant pharmacological activity. Iberiotoxin and charybdotoxin (Peptide Institute Inc., Osaka, Japan) were dissolved at a concentration of 1 μm in a stock solution containing 50 mm NaCl and 50 mm Hepes-Na at a pH of 7.0.

Data analysis

Stable spontaneous firing activity was recorded from ORNs for a period of 60 min. Multi-unit spike sorting was performed using Axograph (Fig. 1C). The normalised spontaneous firing rate was defined in terms of the number of spikes that occurred in 60 s; this number was then normalised to the firing rate that had been observed 1 min prior to the onset of drug administration (t= 0). The firing rates used in a few specific conditions are described in the legends for Fig. 2A, B and E. The first 3 min of each recording was discarded, and the next 2 min were monitored to confirm the stability of the firing rates before the application of any drug(s) (t=−2 to 0 min, e.g. Fig. 2D). In the dose–response analysis, normalised inhibition was defined as (1 –normalised spontaneous rate). It was measured 10 or 15 min after perfusion, and the 50% inhibitory concentration (IC50) was determined by fitting the inhibition data to a standard Hill equation. The current–voltage (I–V) relationship of the tail currents (I/Imax) was determined using a Boltzmann equation; fitting the data to this equation yielded both the half-activation potential (V1/2) and a slope factor.

Figure 1.

Multi-unit recordings efficiently detect the ensemble activity of ORNs 
A and B, representative traces of spontaneous firing activity in single-unit (A) and multi-unit (B) recordings. C, spikes from multi-unit recordings can be sorted into several clusters by amplitude, 20–80% rise time and width (inset; this case has 6–7 clusters labelled i–vi). Amplitude variation may represent action potentials from distinct ORNs. Note that spikes that originate from different cells but that have the same amplitude cannot be clearly distinguished when the amplitudes are very small (i and i′). Inset: representative spike clusters. D and E, distribution of spontaneous spikes per second in single-unit (D, 4.2 ± 0.6 Hz, n= 171) and multi-unit recordings (E, 28.3 ± 1.6 Hz, n= 97). Each dot indicates the firing frequency of a given single-unit (D) or multi-unit recording (E). Horizontal lines indicate the means of the distributions presented in D and E. Note that the position on the abscissa has no meaning.

Figure 2.

HCN channels control spontaneous activity 
A, responses to odorants in Ringer solution with (R[+]) and without (R[–]) divalent cations (n= 7 pairs, P= 0.06, 0.001 and 0.0002 for 10–1000 nm odorant concentrations). Basal rates represent spontaneous firing activity. Inset: representative odorant responses. The vertical dotted line indicates the time at which odorant application occurred. The measured responses are represented by shaded areas. B, spontaneous firing frequency modulation by sequential perfusions of Ringer [+] or Ringer [–] solutions buffered at different external pH values (6.0, 7.0, 7.4 and 8.0). The dotted lines indicate solution exchanges. The alternating white and shaded areas indicate the period of perfusion for each solution. C, the frequencies reported in B were normalised to 1.0 based on frequency during the last 2 min of the application of Ringer [+] at a pH of 7.4 (grey square). Measurements were averaged over the 2 min that immediately preceded solution exchange and have been plotted against the pH of the external solution (pHext). (n= 5). D, spontaneous firing activity was not affected by the presence of amiloride (1 mm) in Ringer [–] (n= 8, P= 0.27). E, odorant response blocking following exposure to the HCN channel blocker ZD 7288 (ZD; filled symbols) in R[–] (open symbols) (30 μm, n= 5 pairs, open and filled squares, P= 0.09, 0.04 and 0.04 for 10–1000 nm odorant concentrations; 300 μm, n= 4 pairs, open and filled circles, P= 0.003, 0.016 and 0.0004 for 10–1000 nm odorant concentrations). Dotted lines indicate paired results. Spike counts were normalised to the spontaneous firing rates (basal) in ORNs in R[+] (A) and in R[–] (E) so that data from multi-unit recordings could be used for comparison. F, field stimulation-induced modulation of spontaneous firing activity. Representative traces from single-unit cell-attached recordings (top) and an associated raster plot (bottom). G, field stimulation-induced modulation of firing frequencies in the presence of ZD 7288 (ZD; filled symbols) and in its absence (Control; open symbols). Circles: 30 μm ZD 7288; n= 6 pairs, P= 0.0001, 0.0053, 0.0023 and 0.023 for voltages of +20, 0, −20 and −40 mV, respectively. Squares: 300 μm ZD 7288; n= 3 pairs, P= 0.0004, 0.00037, 0.00035 and 0.00045 at +20, 0, −20 and −40 mV, respectively. Dotted lines indicate paired results. Inset: representative trace from a depolarised membrane patch in the presence of 300 μm ZD 7288. Plotted data in this figure and in subsequent figures represent means ± SEMs. H, the spontaneous and odorant-evoked activity patterns recorded via multi-unit recordings using divalent-cation-free Ringer solution R[–]. ZD 7288 suppressed the firing activity, and the administration of 25 mm[K+]ext resulted in the recovery of these activity patterns. Shaded areas (a–c) in the top trace are magnified in the bottom panel. Od: odorant stimulation for 10 s.

Odorant stimulation

We selected the odorants by consulting a previous report by Bozza & Kauer (1998). Odorant mixtures were freshly prepared via 10-fold serial dilutions of 1000 μm amylacetate, cineole, octanol, R-(+)-limonene, acetophenone (Sigma, USA) and lyral (TCI, Japan) in the bath solutions. Constant local pressures were applied to the cilia for 4 or 10 s using a puffer pipette that had been placed nearby using a PV830 pneumatic pump (WPI, USA), and the estimated flow volume was approximately 2 μl s−1. The odorant responses were determined from the number of spikes measured during 3 s intervals; any spikes observed during the first second of a recording were excluded from the count until the interspike interval (ISI) became shorter than the averaged basal ISI.

Real-time quantitative PCR

Complementary DNA (cDNA) was synthesised from mRNA that had been isolated from the olfactory epithelia (of P4 mice) using SuperscriptTM II (Invitrogen, USA). Reactions using cDNA that had been synthesised in the same manner from eye tissues and whole brain total RNA were used as positive controls. Reactions in the absence of reverse transcriptase were used as negative controls. Real-time PCR was performed using SYBR Green Real-time PCR Master Mix (TOYOBO, JAPAN) and a LightCycler DX400 (Roche, USA). The following primer sequences were used: the HCN1 (NM_010408) forward primer was 5′-CGC CTT TCA AGG TTA ATC AG-3′; the HCN1 reverse primer was 5′-CAA TGA GGT TGA AGA TCC TC-3′; the HCN2 (NM_008226) forward primer was 5′-AGC AGG AAC GCG TGA AGT CG-3′; the HCN2 reverse primer was 5′-GCA TGG TGA AGT CCC AGT AG-3′; the HCN3 (NM_008227) forward primer was 5′-AAA TCG AGC AGG AGA GGG TG-3′; the HCN3 reverse primer was 5′-AGC AGC AGC ATG ATG AGA TC-3′; the HCN4 (NM_001081192) forward primer was 5′-AGT TTC ATG GAT GCC ACA GG-3′; the HCN4 reverse primer was 5′-TAA TAC GAC TCA CTA TAG G-3′; the glyceraldehyde-3-phosphate dehydrogenase (GAPDH; NM_008084) forward primer was 5′-CCT GCA CCA CCA ACT GCT TA-3′; and the GAPDH reverse primer was 5′-TGA GCC CTT CCA CAA TGC CAA A-3′. Appropriate annealing temperatures that yielded single bands were determined and used. All of the expression levels were normalised to the GAPDH expression level, which was defined as 100% expression.

Fluorescent immunohistochemistry

The olfactory epithelia from several of the mice used in our experiments (P4 or 4–7 weeks old) were fixed via incubation with 4% HCHO in PBS for 4 h, cryo-protected by overnight incubation in a 30% w/v sucrose solution that contained PBS, mounted in OCT Embedding Compound (Sakura Finetek, Japan) and sectioned into slices that were 20 μm thick using a cryostat (CM3050S, Leica Microsystems, Wetzlar, Germany). The resulting sections were then incubated at room temperature 25–30°C overnight in an appropriate blocking solution containing the primary antibodies. The sections were then washed with PBS containing 0.3% Triton X-100 (PBS-X) and incubated with the appropriate secondary antibody (Alexa-Fluor 488 or 594 conjugated, 1:200; Molecular Probes, USA; in blocking solution) for 1.5–2 h. After the secondary incubations, the sections were washed in PBS-X, mounted onto MAS-coated glass slides (Matsunami Glass, Japan), coverslipped using the Vectashield antifade reagent (Vector Labs, USA) and tightly sealed. The following primary antibodies were used in the present study: anti-HCN1 (1:1000, Dr Shigemoto, National Institute for Physiological Sciences, Japan), anti-HCN2, anti-HCN3, anti-HCN4 (1:500, Alomone Labs, Jerusalem, Israel) and anti-olfactory marker protein (OMP, 1:400, Wako, Japan). The fluorescence signals were detected using the FV1000 scanning laser confocal microscope (Olympus, Japan). The HCN4 expression level was measured as the fluorescence intensity detected in the somas of the ORNs, from which the background fluorescence of non-HCN4-expressing cells was subtracted (Fig. 4D). The olfactory bulbs were fixed, cryoprotected, sectioned into slices that were 30 μm thick and labelled with an anti-OMP antibody in a manner similar to that used for the olfactory epithelium samples. The fluorescence signals were detected using a fluorescence microscopy unit and were analysed with the appropriate software (BIOREVO BZ-9000, Keyence, Japan). The validity of the antibodies for HCN1, 2, 3 and 4 had been confirmed in retinal tissue samples (Müller et al. 2003; Supplemental Fig. S1, available online only). Omitting the primary antibodies from the incubation solution resulted in a lack of significant fluorescence signals.

Generation of knock-in mice

We generated knock-in mice in accordance with the procedure described by Bond et al. (2000). Briefly, a genetic switch using the tetracycline transactivator/tetracycline operator (tTA/teto5-CMV) system was inserted into the 5′-upstream region of the HCN4 gene by homologous recombination (Fig. 4D). HCN4-Tet mice were genotyped by tail genome PCR using Advantage2 DNA polymerase (Clontech, USA), which yielded a single band at 456 base pairs (bp) (Fig. 4B). The following specific primers and PCR cycles were used: the forward primer was 5′-CGA ATA AGA AGG CTG GCT CTG CAC C-3′; the reverse primer was 5′-GAG CAG CCT ACA TTG TAT TGG CAT G-3′; and the samples were subjected to 35 cycles in which they were heated to 98°C for 30 s, cooled to 67°C for 30 s, and heated again to 72°C for 60 s. Most of the HCN4-Tet mice were born heterozygous. Doxycycline was orally administered in the animals’ food pellets for a period of 4 weeks beginning when the mice were 3 weeks old.

Statistical analysis

The data are shown as the means ± SEMs and were analysed using appropriate Student's paired or unpaired t tests. P values of P < 0.05 (*) and P < 0.01 (**) were considered significant.


We first optimised the recording conditions to detect the electrical activity of ORNs. Because ORNs that express a given odorant receptor only respond to a limited number of odorants (Duchamp-Viret et al. 1999; Kleene, 2008), single-unit recordings (Fig. 1A) often failed to detect responses to various odorants. Therefore, the present study generally used multi-unit recordings to detect and measure the ensemble activity of multiple ORNs in an efficient manner. Multi-unit spontaneous spikes (Fig. 1B) were sorted into distinguishable clusters (Fig. 1C), and we considered the ensemble activity to be representative of the summated firings of several ORNs. The frequencies of the spontaneous activity observed in both single-unit (4.2 ± 0.6 Hz, n= 171) (Fig. 1D) and multi-unit recordings (28.3 ± 1.6 Hz, n= 97) varied between different samples (Fig. 1E); thus, unless otherwise indicated, the electrical activities reported in the following experiments were normalised to the initial spontaneous firing frequencies to elucidate the common firing properties of ORNs.

HCN channels control spontaneous firing activity; CNG channels do not

We assessed the contributions of CNG and HCN channels to the odorant-evoked and spontaneous firing activities of ORNs. Rodent ORN responses to odorants measured in vitro are reported to be far less sensitive than those measured in vivo; micromolar concentrations of odorants were often required in vitro (Oka et al. 2006; Tan et al. 2010). In contrast, even nanomolar concentrations of odorants elicited ORN responses in slice preparations when we used Ca2+- and Mg2+-free Ringer solution (R[–]; Fig 2A), which is consistent with the reported odorant concentrations required for rodent ORN activation in vivo (Duchamp-Viret et al. 1999; Bhandawat et al. 2005; Oka et al. 2006; Tan et al. 2010). The responses were strongly suppressed when Ca2+ (2.5 mm) and Mg2+ (1 mm) were added to the external solution (R[+]; Fig. 2A), which is probably because these ions block CNG channels (Frings et al. 1992; Zufall & Firestein, 1993; Bhandawat et al. 2005). The normalised spontaneous firing rate was also suppressed by the presence of divalent cations, from 2.2 (R[–]) to 1.0 (R[+]), which corresponds to an actual firing rate decrease from 16.4 Hz (R[–]) to 7.3 Hz (R[+]) in absolute values (Basal; Fig. 2A). Ca2+, Mg2+ or lanthanum ions (La3+) are known to screen surface charges on a membrane and stabilise the firing activity of a neuron by shifting the gating properties of its ionic channels along the voltage axis (Takata et al. 1966; Vogel, 1974; Ohmori & Yoshii, 1977). In the present study, these polyvalent cations also suppressed the normalised spontaneous firing rates from 2.2 (R[–]) to 1.2 (after the addition of 2.5 mm Ca2+), from 2.2 (R[–]) to 1.5 (after the addition of 2 mm Mg2+), from 2.2 (R[–]) to 1.1 (after the addition of 0.01 mm La3+) and to 0.5 (after the addition of 0.1 mm La3+). The basal firing rate was also regulated by the external pH (pHext= 6.0–8.0; Fig. 2B and C). The effectiveness of each of the cations tested was comparable to its reported effectiveness in a previous study (Ohmori & Yoshii, 1977). The effects that exchanging the solutions had on the firing rate of the ORN were quick and stable, and they were reversible at the end of the experiments (Fig. 2B), which implies that both protons and polyvalent cations directly modified the machinery responsible for spike generation. Calcium-dependent conductance might also be involved in suppressing spontaneous firing activity in mouse ORNs because they have large- and intermediate-conductance calcium-dependent potassium (BK and IK) channels (Maue & Dionne, 1987). However, inhibiting BK/IK channel activity with iberiotoxin (100 nm) or charybdotoxin (100 nm) (Brugnara et al. 1995; Ishii et al. 1997) did not significantly change the overall spontaneous ORN firing rate (data not shown). We still might not have fully excluded the contributions of possible changes in other calcium-dependent mechanisms that stabilise ORN excitability. These results, however, do indicate that the suppression of the spontaneous firing rate in the presence of polyvalent cations can be largely attributed to membrane-stabilising effects that can regulate the activation of the voltage-gated ion channels responsible for spike discharge (Hille, 1968; Ohmori & Yoshii, 1977).

In contrast, blocking CNG channels by adding amiloride (1 mm) (Frings et al. 1992) to R[–] had no significant effect on the spontaneous firing activity of ORNs (Fig. 2D), which suggests that amiloride- sensitive proteins, including CNG channels, do not significantly affect spontaneous firing rates. Taken together, these results indicate that the primary mediators of odorant-evoked responses, CNG channels, are not the principal regulators of spontaneous firing activity, which is in agreement with a previous report demonstrating that spontaneous firing activity occurred in ORNs from CNG-knockout mice (Brunet et al. 1996).

When the HCN channel blocker ZD 7288 (30 and 300 μm) was present in the R[–], both odorant-evoked and spontaneous neural activity was suppressed (Fig. 2E), and the normalised spontaneous firing rate decreased from 1.0 (Basal; R[–]) to 0.14 (following the addition of ZD 7288) (Fig. 2E). This decrease corresponded to a decrease in absolute value from 12.4 Hz (R[–]) to 0.8 Hz (ZD 7288). The spontaneous firing activity that was blocked by exposure to ZD 7288 (30 and 300 μm) was restored by field stimulation that depolarised the membrane in cell-attached single-unit recordings (Fig. 2F and G). Depolarisation caused by increasing the external K+ concentration ([K+]ext= 25 mm in R[–]) also restored both the spontaneous and odorant-evoked firing patterns that had been suppressed by exposure to ZD 7288 in R[–] (Fig. 2H). These results indicate that HCN channel activity boosts the ORN firing rate by depolarising its membrane. This in turn can affect the odorant sensitivity of ORNs (Vargas & Lucero, 1999).

Expression of HCN channel subtypes in the olfactory epithelium

We investigated the expression patterns of HCN subtypes in olfactory epithelia isolated from postnatal day 4 mice using real-time PCR and immunohistochemistry. The relative expression levels of HCN2 and HCN4 mRNA were significantly higher than the expression levels of HCN1 and HCN3 mRNA (0.4, 0.9, 0.1 and 1.0% of GAPDH levels for HCN1, 2, 3 and 4, respectively) (Fig. 3A). Immunoreactivity for HCN2 and HCN4 was predominantly detected in the somas and dendrites (Fig. 3B), so we investigated the electrical properties of transgenic mouse ORNs with HCN4 over-expression. The HCN4 subtype is one of the predominant HCN subtypes in ORNs, and it has gating properties that are more sensitive to intracellular concentrations of cAMP than any of the other three subtypes (Biel et al. 2009). These transgenic mice were created by using homologous recombination techniques to insert a tetracycline-responsive gene switch into the HCN4 gene (HCN4-Tet mice) (Fig. 4A and B. See also Bond et al. 2000). In recombinant HCN4-Tet mice, HCN4 is over-expressed in the absence of dietary doxycycline (dox), and dox administration knocks down the HCN4 expression.

Figure 3.

HCN2 and HCN4 are predominant in ORNs 
A, quantitative real-time PCR for various HCN subtypes (H1–4, n= 4). The expression levels have been normalised to the level of GAPDH expression, which was assumed to be 100% (bottom). End products were confirmed as single-band products (top). B, immunohistochemical localisation of HCN1–4 in ORNs. k, knobs; d, dendrites; s, somas; ax, axon bundles.

Figure 4.

HCN4 channel over-expression boosts the rate of spontaneous ORN firing activity 
A, schematic diagram of the generation of HCN4-Tet (Tet) mice. Abbreviations are as follows: SV40 pA, simian vacuolating virus 40 polyadenylation signal; tTA, tetracycline transactivator; TK promoter, herpes simplex virus thymidine kinase promoter; neo, neomycin resistance gene; TK pA, thymidine kinase polyadenylation signal; URA, yeast URA3 gene; hGH, human growth hormone gene; teto5-CMV, tet operator fused to cytomegalovirus promoter. The filled triangles flanking the TK promoter through the URA are loxP sites. B, genotyping of Tet mice using tail genome PCR. C, distribution of the spontaneous firing frequencies of ORNs in single-unit recordings of neurons from Tet (n= 53) and Lt (n= 50, P < 10−23) mice. D, HCN4 expression levels in the ORNs of Tet mice with or without dox treatment. The expression levels are indicated as HCN4-immunoreactivity (IR) and are given in arbitrary units (a.u.). HCN4-IR expression was not completely abolished because most of the HCN4-Tet mice were heterozygous. E, spontaneous firing frequencies obtained from single-unit recordings of ORNs from Tet-mice that had been treated with dox (n= 10) and from Tet-mice that had not been treated with dox (n= 14, P < 10−10). F, 4 weeks of dox administration did not change the spontaneous firing frequencies of ORNs in littermate (Lt) control mice (n= 17 and 13 animals with and without dox administration, respectively; P= 0.749). Horizontal bars in C, E and F indicate the means of the distributions.

HCN channels boost spontaneous firing activity

The single-unit spontaneous firing rate was significantly higher in HCN4-Tet mice in the absence of dox administration (12.3 Hz on average) than in littermate (Lt) control mice (3.0 Hz in average) (Fig. 4C). After a 4 week period of administering dox to the HCN4-Tet mice, the level of HCN4 translation in the somas of ORNs was reduced by approximately 70% (689 ± 30, n= 127, dox (+) to 198 ± 8, n= 125, dox (–)); this level was comparable to the HCN4 translation level observed in Lt mice (221 ± 11, n= 134; all immunoreactivities are given in arbitrary units) (Fig. 4D). The spontaneous firing frequency of ORNs from dox-treated HCN4-Tet mice was lower (1.8 Hz on average) than that of ORNs from untreated HCN4-Tet mice (10.5 Hz on average) (Fig. 4E), but dox administration did not significantly affect the spontaneous firing frequency of ORNs from Lt mice (average frequencies were 1.2 and 1.1 Hz for dox (–) and dox (+), respectively; Fig. 4F). These findings support the conclusion that HCN channels boost the level of spontaneous ORN firing activity.

Endogenous cAMP and phosphatidylinositol phosphates enhance spontaneous firing activity

The results presented above suggest that HCN channels contribute to the generation of spontaneous firing activity by depolarising the membranes of ORNs. Therefore, we investigated the degree to which HCN channels affect the spontaneous firing activity of ORNs at the resting membrane potentials of these cells in wild-type mice. Multi-unit recordings revealed that the application of ZD 7288 decreased the spontaneous firing rate in a dose-dependent manner, and the HCN block was nearly complete and irreversible at ZD 7288 concentrations that were higher than 100 μm (IC50= 20 μm; Fig. 5A). Another HCN channel blocker, Cs+ (10 mm), decreased the spontaneous firing activity of ORNs from wild-type mice in a similar way (Fig. 5B). The effect of Cs+ was reversible and appeared more rapidly than that of ZD 7288 because Cs+ blocks HCN channels from the outside of the membrane (Biel et al. 2009). In contrast, ZD 7288 blocks an intracellular domain of the HCN channel, so it must permeate the membrane before blocking the channels (DiFrancesco, 1982; Ludwig et al. 1998; Ishii et al. 1999; Shin et al. 2001; Cheng et al. 2007).

Figure 5.

HCN channels are active at rest 
A, dose-dependent suppression of spontaneous firing activity by ZD 7288 (10, 100, 1000 μm; n= 5, 10, 7). B, modulation of spontaneous firing activity following exposure to 10 mm Cs+ (n= 5, P < 10−13). C, representative trace of spontaneous firing activity from a single-unit recording. Suppression of spontaneous firing activity by exposure to 100 μm ZD 7288. D, changes in spontaneous firing frequency in single-unit recordings that result from the application of ZD 7288 at concentrations of 10 μm (n= 7, light grey), 100 μm (n= 7, dark grey) and 1000 μm (n= 6, black). E, normalised changes in spontaneous firing activity based on the data in D. The thick lines in D and E represent means, and the vertical bars represent the associated SEMs. FI, modulation of spontaneous firing activity following exposure to PDE4 inhibitor (F, rolipram, 10 μm, n= 5, P= 0.0016), adenylate cyclase inhibitor (G, SQ22536, 100, 300, 1000 μm; n= 4, 7, 3), PI-kinase inhibitor (H, wortmannin, 10, 100, 1000 nm; n= 5, 5, 4) and 300 μm U73122 (I, n= 6, P= 0.76). Ctrl: control experiments.

The effect of ZD 7288 was also confirmed using single-unit recordings (Fig. 5C); these recordings showed that the administration of ZD 7288 decreased the frequency of spontaneous firing activity in a dose-dependent manner (Fig. 5D). The normalised values of the ZD 7288-induced suppression of spontaneous firing obtained from single-unit recordings were approximately 40, 80 and 95% at ZD 7288 concentrations of 10, 100 and 1000 μm, respectively (Fig. 5E), and these values largely agreed with the results obtained in multi-unit recordings (Fig. 5A). These results, which were obtained via single-unit recordings, also support the assumption that the ensemble activity in multi-unit recordings most likely corresponds to the sum of the firing activities of individual ORNs. Together, these results indicate that HCN channels effectively regulate the spontaneous firing activity of resting ORNs.

HCN channel gating is directly enhanced by cAMP and by phosphatidylinositol-phosphates (PIPs; namely PIP2 and phosphatidylinositol 3,4,5-trisphosphate (PIP3)) (Pian et al. 2006; Zolles et al. 2006; Biel et al. 2009). Elevating the basal cAMP level by inhibiting phosphodiesterase (PDE) 4 activity (rolipram, 10 μm) slightly but significantly increased spontaneous firing activity in an irreversible manner (Fig. 5F), whereas reducing the basal cAMP level by inhibiting adenylate cyclase activity (SQ22536; 100–1000 μm) decreased spontaneous firing activity and almost completely and irreversibly suppressed the firing activity when the adenylate cyclase inhibitor was administered at concentrations above 300 μm (IC50= 60 μm, Fig. 5G); adenylate cyclase was constitutively active. Similarly, depleting the concentration of PIPs by inhibiting phosphatidylinositol (PI) kinases (wortmannin; 10–1000 nm) decreased spontaneous firing activity irreversibly, although the inhibition was incomplete (Fig. 5H). PIP2 accumulation that resulted from phospholipase C inhibition (U73122; 100 μm) did not change the firing activity of the cell (Fig. 5I). These results indicate that the endogenous levels of cAMP and PIPs in ORNs are sufficient to maintain the spontaneous ORN firing activity by means of their effects on HCN channels. Notably, these electrophysiological properties are consistent with both histological and real-time PCR results; both of the predominant HCN channel subtypes in ORNs, HCN2 and HCN4, are highly sensitive to cAMP and PIP2 (Pian et al. 2006; Zolles et al. 2006; Biel et al. 2009).

HCN channels can open at the resting membrane potential of ORNs

Previous reports have concluded that few HCN channels in the membranes of ORNs are open near the resting potential. The HCN channel half-activation voltage (V1/2) was between −110 and −130 mV, activation began at a potential of −90 to −100 mV (Lynch & Barry, 1991a; Vargas & Lucero, 1999; Mobley et al. 2010), and the estimated resting membrane potential was −70 mV (Lagostena & Menini, 2003). Thus far, however, our results have suggested that sufficient numbers of HCN channels are open at the resting membrane potential of ORNs. We therefore performed whole-cell voltage-clamp recordings to investigate HCN channel gating modulation. We found that currents through HCN channels were activated by membrane hyperpolarisation, inwardly rectifying with the initial instantaneous currents and not inactivating (Fig. 6A). The recorded current was at half-maximum at a voltage of −115.7 ± 3.5 mV and had a slope of −8.6 ± 0.8 mV (n= 7). In general, activation was slow, and the time constants derived from single exponential fits ranged from 100 to 250 ms (Fig. 6A). Because both PIP2 and cAMP may wash out during whole-cell recordings, which could lead to a reduction in HCN channel activity (Pian et al. 2006; Zolles et al. 2006), we included native PIP2 (1 μm) in the pipette for some recordings. This resulted in a positive V1/2 shift of nearly 18 mV (V1/2=−98.2 ± 4.6 mV, slope =−8.8 ± 0.8 mV, n= 10), and a further bath application of the membrane-permeable cAMP analogue 8-Br-cAMP resulted in an even more positive V1/2 shift (an additional 12 mV) and faster activation kinetics (1 mm; V1/2=−86.0 ± 1.5 mV, −8.0 ± 1.9 mV, n= 5) (Fig. 6B and C). These results indicate that both PIP2 and cAMP work to shift the activation voltage of HCN channels to a more positive potential, which can induce the opening of HCN channels of an ORN at its resting potential. HCN channels conduct Na+ in addition to K+ and the reversal potential lies between −20 and −30 mV at physiological ionic conditions (Lynch & Barry, 1991a; Biel et al. 2009). Thus, the HCN currents that occur at the resting membrane potential are generally inward and depolarise the membrane, which boosts the spontaneous firing activity.

Figure 6.

HCN channels are gate-shifted by cAMP and PIP2
A and B, representative traces of hyperpolarisation-activated currents recorded when the pipettes contained a KCl-based internal solution (A, Control) and traces recorded after adding PIP2 to the pipette solution and 8-Br-cAMP to the bath (B, +PIP2/cAMP). C, voltage dependence of HCN channel activation in the control condition (filled triangle), the PIP2 condition (open triangle) and the condition in which cells were exposed to both PIP2 and 8-Br-cAMP (filled circle).

The β-adrenergic receptors are basally activated

How might basal cAMP expression be maintained at levels sufficient for the activation of HCN channels in ORNs? One possibility is that the basal activity of some Gs protein-coupled receptors results in constitutive activation of adenylate cyclase. Adrenaline controls the ORN excitability by controlling cAMP concentrations (Kawai et al. 1999) and sympathetic stimulation enhances the odorant responses (Beidler, 1961). These reports led us to investigate whether the basal activation of β-adrenergic receptors (ADRBs; three subtypes are known) influences spontaneous firing activity by constitutively activating adenylate cyclase and thereby maintaining elevated cAMP levels. Propranolol, which blocks ADRBs, decreased the spontaneous firing activity by nearly 90% when applied at a concentration of 300 μm (IC50= 60 μm) (Fig. 7A). Similarly, two β2-selective blockers, butaxamine (IC50= 50 μm) and ICI 118,551 (IC50= 4 μm), inhibited spontaneous firing in a dose-dependent manner (Fig. 7B) and the inhibitory effects of β2-blocker administration were reversible. In contrast, blocking either ADRB1 (Metoprolol) or ADRB3 (SR59230A) had little influence on spontaneous ORN firing rates; the effects were only observed when the blocking compounds were administered at very high concentrations (1000 μm) (Fig. 7B). These results indicate that ADRBs constitutively influence the spontaneous firing activity of ORNs; ADRB2 shows more contribution to these effects than ADRB1 or ADRB3.

Figure 7.

The β-adrenergic receptors maintain basal cAMP levels 
A, dose-dependent suppression of spontaneous activity after administration of a non-selective β blocker (propranolol; 1–300 μm; n= 4 each). Ctrl: control in Ringer [+]. B, dose-dependent spontaneous activity inhibition by selective β blockers: ICI 118,551 (ICI; open circle, n= 5), butaxamine (Buta; open diamond, n= 6), propranolol (Prop; filled circle, n= 4), metoprolol (Met; open square, n= 3), SR59230A (SR; open triangle, n= 4). β(1–3) indicates specificity of each inhibitor to the β-adrenergic receptor subtypes.

HCN4 over-expression reduces the size and number of glomeruli

The importance of spontaneous neuronal activity is usually demonstrated by eliminating this electrical activity (Yu et al. 2004a; Davis 2006; Kandler et al. 2009), but it is worth asking what would occur if the activity were heightened instead. We next investigated how the enhanced spontaneous firing activity observed in ORNs affects the development of the olfactory projection map in mice with HCN4 over-expression (HCN4-Tet mice in Fig. 4C). The olfactory map is formed by glomeruli in the olfactory bulb (Yu et al. 2004a; Sakano, 2010), which is located rostrally in the forebrain with an eminence in the caudal portion (‘Em’ in Fig. 8A, top; photo). Macroscopic observation instantly revealed that 4-week-old HCN4-Tet mice had smaller olfactory bulbs compared with those of age-matched Lt mice; specifically, the olfactory bulbs of the HCN4-Tet mice were shorter along the rostro-caudal axis. The overall size of the bulb in each animal was quantified by measuring the total cross-sectional area of a series of frontal slices taken along the rostro-caudal axis. The area was defined by encircling each slice of the olfactory bulb beneath the olfactory nerve layer from the rostral end (defined as 0 mm) to the caudal end; slice areas were measured every 150 μm. This procedure was implemented to avoid overestimating the size of the olfactory bulb by including the remnants of the olfactory nerve (see also Fig. 8B and C). The areas around the eminence of the bulb over the transition to the olfactory tract (1.2–3.0 mm) were significantly smaller in HCN4-Tet mice (Fig. 8A, bottom). The area that corresponded to the dorsal eminence of the olfactory bulb (‘Em’ in Fig. 8A, top) was as large as 3.99 ± 0.09 mm2 at 1.95 mm in Lt mice, but only reached a size of 2.87 ± 0.1 mm2 at 1.65 mm in HCN4-Tet mice (Fig. 8A, bottom, n= 3 each). Olfactory glomeruli, defined as mature ORN marker protein (OMP)-immunoreactive globular convergences of axon terminals with distinguishable contours located in the olfactory bulb (Monti-Graziadei et al. 1977; Danciger et al. 1989), formed a thick, dense glomerular layer around the perimeter of olfactory bulb slices in Lt mice (Fig. 8B beneath the dotted line, frontal section at 1.65 mm from the rostral end). In HCN4-Tet mice, however, not only was the corresponding slice smaller, but the glomeruli also formed only a rather thin and sparse layer (Fig. 8C). We then scrutinised the distribution of the glomeruli relative to the normalised perimeter of the olfactory bulb slice. The perimeter was normalised such that the ventral peak of the bulb was assigned a value of 0, and subsequent points encircled the bulb laterally back to the same point, which was then assigned a value of 1.0, which is shown in Fig. 8C. The glomeruli in the normal Lt mice gradually increased in number as the middle of the bulb was approached, which corresponds to the eminence noted in Fig. 8A. The number of glomeruli then decreased caudally and became nearly evenly distributed around the perimeter of the bulb with slightly denser regions in the medial and lateral areas and with the exception of a caudo-lateral region that was devoid of glomeruli altogether (this region corresponded to the point of transition to the olfactory tract) (Fig. 8D). In HCN4-Tet mice, the pattern of glomerular distribution resembled the pattern observed in Lt mice, including the increase in the number of glomeruli near the eminence of the olfactory bulb, the corresponding caudal decrease and the dense localisation of the glomeruli in the medial and lateral regions of the olfactory bulb. However, there were obviously fewer glomeruli in the olfactory bulbs of HCN4-Tet mice, particularly in the dorsal region (Fig. 8E). We therefore counted the numbers of glomeruli in the appropriate location surrounding the olfactory bulb (0–1.0) in series slices taken at 150 μm increments, starting from the rostral end of the olfactory bulb. We classified the glomeruli by designating a dorsal region of the olfactory bulb as the normalised perimeter that corresponded to values of 0.25–0.75 (Fig. 8F), and the ventral region was represented by the remainder of the perimeter values (0–0.25 and 0.75–1.0, Fig. 8G). The reduction in the number of glomeruli was much more apparent and drastic in the dorsal region of the olfactory bulb. This reduction was almost one-half; the maximum number of glomeruli in bulb slices from Lt mice was 42.5 ± 1.8, and it was only 23.1 ± 0.5 in HCN4-Tet mice (n= 6 bulbs from 3 mice in each group; Fig. 8F). In addition, a significant reduction in the number of glomeruli in the ventral region was also observed; the maximum number of ventral glomeruli in a sliced bulb was 45.5 ± 1.5 in Lt mice and 33.5 ± 3.4 in HCN4-Tet mice (n= 6 bulbs from 3 mice in each group; Fig. 8G). The sizes of the glomeruli were measured by outlining the perimeters of areas of OMP-positive axonal convergence in serial slices from 1.35 to 1.8 mm. We found that the glomeruli in HCN4-Tet mice were relatively small in comparison to the glomeruli of Lt mice (Tet; 2677 ± 1344 μm2, n= 674, Lt; 3698 ± 2067 μm2, n= 435; Fig. 8H; see Discussion). Finally, we suppressed the over-expression of HCN4 using dox to confirm the involvement of HCN4 in glomerular formation. In the HCN4-Tet mice, dox treatment from embryonic day (E)1 through to P28 rescued the decrease in the number of glomeruli; the total numbers of glomeruli in the bulbs of dox-treated HCN4-Tet mice were approximately equal to those of Lt mice (n= 6 bulbs from 3 mice in each group; Fig. 8I). Moreover, the loss of glomeruli in the dorsal region of the bulb was also markedly rescued (Fig. 8J and K; Lt and HCN4-Tet mice, respectively). These data demonstrate that HCN4 channel hyperactivity results in the deterioration of olfactory map formation, and thereby implies that the rate of spontaneous ORN activity might correlate with the number of target glomeruli in the bulb.

Figure 8.

HCN4 over-expression resulted in reductions in olfactory glomeruli 
A, dorsal (top) and cross-sectional views (bottom; frontal slices, n= 3) of the olfactory bulb (OB) from 4-week-old HCN4-Tet (Tet) and littermate (Lt) mice. The eminent part (Em) of the olfactory bulb indicated in the photo (shown with the rostral end down) largely corresponds to the graded blue square on the graph below. Note that the caudal part of the olfactory bulb extends further underneath the frontal lobe of the cerebral cortex (CC) and is not visible in the photo. B and C, olfactory bulb slices taken 1.65 mm from the rostral end (11th slice in A) from Lt (B) and Tet (C) mice and stained with an anti-OMP antibody. The perimeters are scaled 0 to 1.0 relative to the ventral edge. V, L, D and M stand for the ventral, lateral, dorsal and medial sides, respectively, in both B and C. The olfactory bulb is encircled by a dotted line. Strong immunoreactivities observed outside the olfactory bulb represent remnants of the olfactory nerve (ON). D and E, representative distribution of glomeruli near the olfactory bulb in Lt and Tet mice. Each dot represents the location of a glomerulus and was plotted along the normalised perimeter shown in C. F, the number of glomeruli in the dorsal region (0.25–0.75 in D and E) of the olfactory bulbs from Lt and Tet mice. G, the number of glomeruli in the ventral region (0–0.25 and 0.75–1.0 in D and E) of the olfactory bulb in Lt and Tet mice. H, distribution of glomeruli sizes. The bin width is 400 μm2. I, the total number of glomeruli in the ventral and dorsal regions of olfactory bulbs from dox-treated Lt and Tet mice. J and K, anti-OMP antibody-stained olfactory bulbs from Lt and Tet mice after dox treatment (in J and K, respectively). ** in A, F and G indicates statistical significance with P= 0.0005–0.008 between the underscored pairs of Lt and Tet mice.


We have shown that HCN channels regulate spontaneous ORN firing activity by depolarising the membranes of these cells. This result can be attributed to the fact that due to the basal level of ADRB2 activation, the basal cAMP levels of ORNs are sufficient to adjust the activation voltage of the HCN channels in such a way that the channels are active at the resting membrane potential.

Characterisation of HCN channels in ORNs

We used ZD 7288 to elucidate the role of HCN channels. Our results (ZD 7288; IC50= 20 μm) are similar to those of previous studies that used endogenously and heterologously expressed HCN channels (Gasparini & DiFrancesco, 1997; Shin et al. 2001; Cheng et al. 2007). However, the reported IC50 values for ZD 7288 range from 1 to 80 μm, depending on the study (Gasparini & DiFrancesco, 1997; Shin et al. 2001; Cheng et al. 2007; Biel et al. 2009). Moreover, it has been reported that HCN1 is less sensitive to ZD 7288 (IC50= 40–80 μm) than HCN2 (IC50= 13 μm) (Shin et al. 2001; Cheng et al. 2007). High concentrations of ZD 7288 also block T-type calcium channels (Felix et al. 2003; Sánchez-Alonso et al. 2008) or induce presynaptic depression (Chevaleyre & Castillo, 2002; Agmon & Wells, 2003). However, ORNs supposedly receive no extrinsic innervation within the olfactory epithelium (Getchell, 1986) and have no detectable T-type calcium currents (Trombley & Westbrook, 1991), so it is not surprising that a T-type calcium channel blocker (mibefradil, 10 and 50 μm) had little effect on the spontaneous firing activity of these neurons (Supplemental Fig. S2). Although we cannot exclude the possibility of other non-specific effects of ZD 7288 altogether, our results collectively indicate that HCN channels are largely responsible for the cAMP-dependent regulation of the spontaneous ORN firing rate.

Because ORNs have a high input resistance (3.0 ± 0.5 GΩ, n= 12; see also Lagostena & Menini, 2003), it is reasonable to think that the slow and small currents that pass through the HCN channels are suitable for controlling the membrane potentials of these neurons. The present study illustrates the predominant expression of HCN channel subtypes with relatively slow activation kinetics (HCN2 and HCN4) at P4, but one recent study showed that HCN1 and HCN4 are predominantly expressed during earlier mouse development (E13 through to P4; Mobley et al. 2010). Mobley et al. (2010) also provided evidence that the level of HCN2 expression increases after birth, so ORNs might utilise different subtypes of HCN channels for establishing and maintaining the olfactory map. At present, the roles played by HCN1 during early developmental stages are unclear. HCN channels are associated with small voltage-independent currents (Proenza & Yellen, 2006) and can form heteromers that allow them to have variable cAMP sensitivities (Chen et al. 2001; Ulens & Tytgat, 2001; Altomare et al. 2003). These properties of HCN channels (among others) might affect the excitability of ORNs by modulating the membrane potentials of these neurons.

Contribution of basal cAMP and PIPs to electrical activities

In the present work, we have attributed the contribution of second messenger pathways to the resting membrane potential via the basal activation of HCN channels. Interestingly, PIP3 suppresses the gating of CNG channels (Brady et al. 2006), whereas PIP2 promotes the gating of HCN channels (Fig. 6B and C; Pian et al. 2006; Zolles et al. 2006). Because of this, a reservoir of polyanionic PIPs may also have a role in the regulation of the resting membrane excitability of ORNs (Hegg et al. 2003; Suh & Hille, 2007). The physiological and regulatory roles played by PIPs in ORNs remain to be elucidated (Barry, 2003), which may require intensive future investigations.

How, then, are ADRB2s basally active in ORNs and therefore responsible for the maintenance of basal cAMP levels? The slice preparations were pre-incubated in an external Ringer solution bath for 30 min at room temperature, and the external solutions were continuously exchanged during the experiments, so there could be little remaining adrenaline or noradrenaline in the blood supply near the ORNs. Although the alternative possibility that damaged sympathetic terminals might continue to release the aforementioned catecholamines cannot be ruled out, the most parsimonious explanation is that there is some level of basal ADRB2 activity that occurs even in the absence of ligands, similar to what has been demonstrated in cardiac cells (Zhou et al. 2000; Chakir et al. 2003). Indeed, we confirmed that an approximately 20 mV positive shift in the activation curve of homomeric HCN2 channels that were heterologously expressed in HEK 293 cells occurred in the presence of ADRB2s (Supplemental Fig. S3). This shift is compatible with the cAMP sensitivity of HCN2 and is slightly larger than other reported values (approximately 15 mV, Chen et al. 2001; Ulens & Tytgat, 2001; Biel et al. 2009). ADRB2s might contribute to the spontaneous ORN firing activity via several mechanisms in addition to its role in maintaining basal cAMP levels; for example, ADRB2s might interact directly with HCN2 channels or might have a role in the regulation of PIP2 metabolism that can modulate HCN2 channel gating (Pian et al. 2006; Zolles et al. 2006; Biel et al. 2009). Basal activation of odorant receptors (ORs) has also been demonstrated using noise analysis of CNG channel gating and has been associated with basal cAMP production in cilia (Reisert, 2010). Considering that these ORs are also present in the somas of ORNs (Barnea et al. 2004; Strotmann et al. 2004), ORs might interact with Gs-proteins or directly interact with ADRB2s (Hague et al. 2004). Interestingly, the loss of an odorant receptor that is situated upstream of CNG channels in the olfactory signalling pathway also disturbs glomerular formation, and this disturbance can be rescued by the expression of ADRB2 (Mombaerts et al. 1996; Wang et al. 1998; Feinstein et al. 2004). This result suggests that ADRB2s and ORs might have complementary or overlapping functions in ORNs. Furthermore, the fact that genetic knockout of Golf, a G-protein involved in olfactory signal transduction in the cilia, causes no obvious disturbance in the formation of glomeruli (Belluscio et al. 1998) also supports the notion that additional cAMP-signalling pathways outside the cilia of ORNs exist. Histological studies indicate that the expression patterns of cAMP-associated genes are compartmentalised within ORNs; PDE1C, CNG channels and ACIII are expressed in the cilia, whereas PDE4, HCN channels are expressed in and near the soma (Nakamura & Gold, 1987; Brunet et al. 1996; Michalakis et al. 2006; Chesler et al. 2007; Col et al. 2007; Song et al. 2008; Cygnar & Zhao, 2009; Fig. 2B). We conducted a preliminary immunohistochemical investigation on ADRBs, which also implied that ADRB2 might operate outside the cilia (Supplemental Figs S4 and S5). These data suggest that the observed increase in the spontaneous firing activity of an ORN in the presence of a PDE4-selective inhibitor (rolipram; Fig. 5F) implies that the cAMP level in the soma is sufficiently elevated to regulate the spontaneous firing activity. However, the diffusion of odorant-induced cAMP is considered to be spatially restricted to a very thin structure within the cilia (Takeuchi & Kurahashi, 2008). The distinct expression patterns of cAMP-regulating genes, such as those associated with adenylate cyclases and PDEs may also lead to the spatial compartmentalisation of cAMP concentrations (Houslay & Adams, 2003; Baillie, 2009). This implies that cAMP can play two different roles in ORNs: in the cilia, cAMP can serve as a second messenger in odorant response pathways, and in the soma, it can serve as a basal regulator of spontaneous firing activity. The segregation of CNG channels and HCN channels in both location and function might partly explain why genetic knockout of the HCN1 channel significantly affects glomerular map formation in the olfactory bulb (Mobley et al. 2010), whereas knockout of the A2 subunit of the CNG channel does not (Brunet et al. 1996; Lin et al. 2000).

Spontaneous activities for glomerular formation

During development, the general locations of ORN axon convergence (namely, glomeruli) appear to be genetically determined by the graded distributions of and repulsive interactions between receptors and ligands within the olfactory bulb (Takeuchi et al. 2010). The cAMP levels in ORNs regulate target locations along the anterior–posterior axis of the olfactory bulb; higher levels of cAMP are associated with more posterior glomerulus locations (Mori & Sakano, 2011). In addition, electrical activity affects the size, number and placement of glomeruli. One study found that the generalised suppression of the spontaneous activity of all ORNs via the forced expression of Kir2.1 resulted in a reduction in the total number of glomeruli, whereas the partial suppression of activity in a selected subpopulation of ORNs resulted in the dispersion of the axonal terminals associated with those ORNs from their target glomeruli (Yu et al. 2004a). Knocking out HCN1, which may slightly reduce the basal excitability of ORNs, resulted in larger glomeruli, but the number of them did not change (Mobley et al. 2010). In the present study, HCN4 over-expression increased the spontaneous firing activity of ORNs, which resulted in glomeruli that were smaller and, surprisingly, fewer in number (Figs 6C and 8DH). As expected, the use of dox to switch off the over-expression of HCN4 in Tet-off animals rescued the apparent loss of glomeruli in the olfactory bulb (Fig. 8IK). Glomerular formation was not disturbed by dox treatment, which is most likely because the HCN4-Tet mice were heterozygous, so dox treatment did not completely abolish HCN4 expression (Fig. 4D). Instead, dox administration reduced the expression of HCN4 to a normal level, which is evidenced by the apparently normal spontaneous firing activity shown in Fig. 4E.

Although there is currently no obvious physiological mechanism that explains how spontaneous electrical activity relates to specific parameters that determine the localisation of glomeruli, it appears that HCN4 plays a critical role in determining the number of glomeruli; the activity-dependent regulation of repulsive/adhesive molecules between the neighbouring glomeuli (Serizawa et al. 2006) may be also affected. Accelerated spontaneous firing rates may reduce glomerular size by inducing the pruning of glomerular structures (Zhang & Poo, 2001), but this behaviour would not be a simple linear parametric cue that determines either the number or the locations of the glomeruli. Instead, the range of spontaneous firing rates might be set to maximise the network complexity. Moreover, the constitutive activation levels of ADRB2s and adenylate cyclase may be both necessary and sufficient to establish this level of spontaneous firing activity in the basal condition. In genetic studies in which ACIII (Zou et al. 2007), Kir2.1 (Yu et al. 2004a), HCN1 (Mobley et al. 2010) and HCN4 (Fig. 8DG) have been manipulated, the dorsal region of the olfactory bulb was more severely affected than the ventral region, and this region was found to be functionally involved in innate aversive actions (Kobayakawa et al. 2007). A developmental study also showed that axonal projections from dorsally located ORNs that projected to the dorsal region of the olfactory bulb began to develop earlier than ventral region projections (Takeuchi et al. 2010). Thus, the susceptibility to guidance cues might differ along the dorso-ventral axis.

In contrast, HCN4 over-expression may damage ORNs and reduce the number of glomeruli. The HCN4 channel conducts a slight Ca2+ current (Yu et al. 2004b), and the increase in the spontaneous firing rate may result in a larger influx of Ca2+ through high-threshold (L-type) calcium channels (Maue & Dionne, 1987; Trombley & Westbrook, 1991) that eventually leads to calcium-induced apoptosis (Orrenius et al. 2003). However, we frequently observed high spontaneous firing rates in ORNs from the epithelia of HCN4-Tet mice (Fig. 4C) that had an almost normal density of ORNs (data not shown). Therefore, understanding the developmental fate of HCN4 over-expressing ORNs requires further study.

At present, it appears that ORNs have a crucial and intrinsic function of guiding olfactory axonal targeting (Yu et al. 2004a; Sakano, 2010; Mori & Sakano, 2011). However, it is also worth noting that increasing the excitability of the postsynaptic mitral cells by deleting the Kv1.3 Shaker channel resulted in a disturbance of the axonal targeting of presynaptic ORNs (Biju et al. 2008). It should be noted that HCN channels are also expressed in both the mitral cells (Angelo & Margrie, 2011) and the juxtaglomerular cells (Holderith et al. 2003; Fried et al. 2010) of the olfactory bulb, so the morphological changes observed in HCN1 knockout mice (Mobley et al. 2010) or in mice in which HCN4 was over-expressed (e.g. the present study) may result from the malfunctions of any of those cells. The olfactory glomeruli are not simple synaptic junctions between ORNs and mitral/tufted cells; rather, they are complex computational units that involve connections between several cell types (Kosaka & Kosaka, 2005). The interactions among the cells that constitute a glomerulus and between these cells and their matrix should be investigated as a whole to facilitate the improvement of our understanding of how odour information is processed. These interactions should also be investigated with the goals of understanding how electrical activity modifies the olfactory glomeruli and their synapses in concert with genetic instructions and how it stabilises neural functions despite the ever-changing statuses of neurons (Davis, 2006). Further studies utilising conditional gene manipulation techniques may help elucidate the process by which spontaneous firing activity is processed within the glomerular circuitry in the olfactory bulb.

Given the information described above, we propose that cAMP levels within the soma may generally be regulated by ADRB2s and may then be further refined in an OR-specific manner by the specific OR expression pattern in any given ORN. Thus, the integration of these two effects can modulate the resting level of HCN channel activity and thereby establish the spontaneous firing frequency of a given ORN in an OR-specific manner. In this scenario, GPCRs could serve as both dynamic signal mediators and as static regulators of the basal status of the cell even in the absence of specific ligands. In addition, the spontaneous firing frequency itself can be a parametric guidance cue, and coupled with the distributions of various adhesive/repulsive molecules and patterns of spontaneous frequency, the frequency itself may be important for the formation and maintenance of an evenly distributed and diverse olfactory map that is fundamental to the accurate encoding of olfactory information.

Phylogenic implications on the olfactory map diversity

Finally, HCN currents have not been identified in the ORNs of several amphibian species (Firestein & Weblin, 1987; Schild, 1989; Delgado & Labarca, 1993; Kawai, 2002). ORNs vary uniquely among species in terms of both their size and their electrical properties (Firestein & Weblin, 1987; Lynch & Barry, 1991b; Imanaka & Takeuchi, 2001). Furthermore, vertebrate ORNs express either of two phylogenetically distinct classes of ORs (class I and II) largely depending on the habitats of the animals (water vs. land; Glusman et al. 2000; Niimura & Nei, 2006; Kobayakawa et al. 2007). Mice, however, express both classes of ORs and the expression patterns of the two OR classes are related to the dorsal and ventral pathways to the olfactory bulb (Kobayakawa et al. 2007). Thus, the mammalian olfactory system may illustrate the evolutionary preservation of two distinct principles that affect the formation and maintenance of the dorsal and/or ventral neural networks (Yu et al. 2004a; Zou et al. 2007; Mobley et al. 2010; Takeuchi et al. 2010; Fig. 8DG). Comparative studies of olfactory patterning in different species might be further necessary to ultimately understand the processing of olfactory information in the human brain.


Author contributions

N.N. and T.M.I. designed and performed all of the experiments, and they collected, analysed and interpreted all of the data. Y.B. and R.K. performed homologous recombination using ES cells to develop the HCN4-Tet mice. N.N., T.M.I. and H.O. wrote the manuscript. All authors discussed the results.


We thank Dr John P. Adelman (Oregon Health and Science University) for providing us with the material vector and for critically reading and editing the manuscript. We thank Dr Takuya Notomi and Dr Ryuichi Shigemoto (National Institute for Physiological Sciences) for providing us with antibodies against the HCN channels. We thank Dr Akira Kakizuka and Dr Seiji Hori (Kyoto University) for the use of the real-time PCR LightCycler and for the helpful technical discussion. We thank Dr Takashi Kurahashi, Dr Hiroko Takeuchi (Osaka University), Dr Hiroshi Kuba and Dr Rei Yamada (Nagoya University) for helpful discussion and for their critical comments. We also thank Mrs Midori Sakiyama for her assistance in maintaining the transgenic animals. This work was supported by the Japan Society for the Promotion of Science (JSPS) Grant-in-Aid for JSPS Fellows and by the Global COE Program Center for Frontier Medicine of the Japanese Ministry of Education, Culture, Sports, Science, and Technology (MEXT) to N.N.; it was also supported by Grants-in-Aid for Scientific Research from MEXT to T.M.I. (21590232) and H.O. (20220008).