Hyperpolarization-Activated Currents and Subthreshold Resonance in Granule Cells of the Olfactory Bulb

Abstract An important contribution to neural circuit oscillatory dynamics is the ongoing activation and inactivation of hyperpolarization-activated currents (Ih). Network synchrony dynamics play an important role in the initial processing of odor signals by the main olfactory bulb (MOB) and accessory olfactory bulb (AOB). In the mouse olfactory bulb, we show that Ih is present in granule cells (GCs), the most prominent inhibitory neuron in the olfactory bulb, and that Ih underlies subthreshold resonance in GCs. In accord with the properties of Ih, the currents exhibited sensitivity to changes in extracellular K+ concentration and ZD7288 (4-ethylphenylamino-1,2-dimethyl-6-methylaminopyrimidin chloride), a blocker of Ih. ZD7288 also caused GCs to hyperpolarize and increase their input resistance, suggesting that Ih is active at rest in GCs. The inclusion of cAMP in the intracellular solution shifted the activation of Ih to less negative potentials in the MOB, but not in the AOB, suggesting that channels with different subunit composition mediate Ih in these regions. Furthermore, we show that mature GCs exhibit Ih-dependent subthreshold resonance in the theta frequency range (4–12 Hz). Another inhibitory subtype in the MOB, the periglomerular cells, exhibited Ih-dependent subthreshold resonance in the delta range (1–4 Hz), while principal neurons, the mitral cells, do not exhibit Ih-dependent subthreshold resonance. Importantly, Ih size, as well as the strength and frequency of resonance in GCs, exhibited a postnatal developmental progression, suggesting that this development of Ih in GCs may differentially contribute to their integration of sensory input and contribution to oscillatory circuit dynamics.


Introduction
Hyperpolarization-activated cyclic nucleotide-gated (HCN) channels play an essential role in neuronal function and network oscillatory dynamics, including the control of membrane excitability and integration of synaptic inputs, the generation of membrane potential oscillations, and subthreshold resonance of neurons (McCormick and Pape, 1990;Magee, 1998;Fan et al., 2005;George et al., 2009;Hu et al., 2009). Moreover, the voltage sensitivity and kinetic properties of the hyperpolarization-activated current (I h ), is dependent on the subunit composition of the channels (HCN1-HCN4), which confers differing sensitivities to cyclic nucleotides, providing a rich heterogeneity to the contribution of these channels to network function (Kaupp and Seifert, 2001;Wainger et al., 2001). Thus, changes in intracellular cAMP concentration in response to neuronal metabolic state or to the action of neuromodulators can affect the I h and, in turn, alter membrane potential oscillations of pacemaker cells in the brain and in the heart (Pape and McCormick, 1989;Valsecchi et al., 2013).
In olfaction, initial processing of odor signals occurs through two parallel, albeit complementary, pathways in the olfactory bulb (OB). In general, the main OB (MOB) receives input from the nasal epithelium and is largely tuned to environmental odors, whereas the accessory OB (AOB) receives input from the vomeronasal organ (VNO) and is largely responsible for the detection of semiochemicals (Shepherd, 1972;Keverne, 1999;Lledo et al., 2005). In the MOB, network synchrony is a substrate to important aspects of olfactory coding, such as sparsening and feature binding of odor representations (Brody and Hopfield, 2003;Binns and Brennan, 2005;Osinski and Kay, 2016), and network synchrony may also play a role in pheromonal processing by the AOB (Binns and Brennan, 2005;Leszkowicz et al., 2012). This network synchrony arises from the activity of dendrodendritic synapses between a large number of intrinsic inhibitory neurons, including the granule cells (GCs) and periglomerular cells (PGCs), and the output neurons, the mitral cells (MCs) and tufted cells (Lagier et al., 2004;Lepousez and Lledo, 2013;Kay, 2014;Gschwend et al., 2015). In the MOB, network oscillations in the theta frequency (4 -12 Hz), entrained by the respiratory cycle, are in part mediated by PGCs (Fukunaga et al., 2014), while gamma oscillations  are mediated by GCs (Lagier et al., 2004;Fukunaga et al., 2014;Kay, 2014).
Surprisingly, despite the influence of inhibitory neurons in promoting network oscillations in the OB, the presence and contribution of I h to GC excitability has not been addressed. Here, using whole-cell patch-clamp recordings, we show that most GCs of the OB exhibit I h . Targeted recording of postnatally labeled neurons indicated that the contribution of I h to GC physiology increases throughout postnatal development, leveling between 4 and 6 weeks after birth. In the presence of ZD7288 (4-ethylphenylamino-1,2-dimethyl-6methylaminopyrimidin chloride), an I h blocker, GCs hyperpolarized and their input resistance increased, suggesting that, at rest, I h contributes to the excitability and passive properties of GCs. Importantly, mature MOB GCs and PGCs exhibited an I h -dependent subthreshold resonance with resonant frequencies in the delta and theta ranges (1-12 Hz). These results suggest that the expression of I h and subthreshold resonance in inhibitory neurons may impart unique features to odor processing in the OB, and facilitate oscillatory network activity in both the main olfactory and vomeronasal systems.

Animals
All experiments were conducted following the guidelines of the institutional animal care and use committee of the University of Maryland and the Marine Biology Laboratory. Experiments were performed on wild-type (C57BL/6) mice of either sex, ranging in age from postnatal day 15 (P15) to 60 (P60).

Electroporation
To label postnatally born neurons with the green fluorescent protein (GFP), mice (P1-P4) were anesthetized by hypothermia and 2 l (4 g/l) of pCAG-GFP plasmid (Addgene) was injected into the lateral ventricle. Immediately after the injection, twizzer-type electrodes (BTX) were placed on the sides of the head for electroporation (5 ϫ 100 V pulses of 50 ms) using an ECM 830 Electro Square Electroporation System (BTX). Recordings of GCs were conducted at least 1 week postelectroporation, and onward, in cells identified by the expression of GFP under a fluorescent microscope.

Slice preparation
Experiments were performed in OB slices using methods previously described (Smith et al, 2015). Briefly, brain slices were prepared in an oxygenated ice-cold artificial CSF (ACSF) containing low Ca 2ϩ (0.5 mM) and high Mg 2ϩ (6 mM). Sagittal sections of the OB (250 m) were then transferred to an incubation chamber containing normal ACSF (see below) and were left to recuperate for at least 30 min at 35°C until the recordings. In all experiments, unless otherwise stated, the extracellular solution is ACSF of the following composition (in mM): 125 NaCl, 26 NaHCO3, 1.25 NaH2PO4, 2.5 KCl, 2 CaCl2, 1 MgCl2, 1 myoinositol, 0.4 ascorbic acid, 2 Napyruvate, and 15 glucose, continuously oxygenated (95% O 2 -5% CO 2 ), which produced a pH of 7.4 and an osmolarity of 305 mOsm.

Electrophysiological recordings
Neurons were visualized using an Olympus BX51W1 microscope and were recorded using a dual EPC10 amplifier (HEKA) in voltage-clamp and current-clamp modes. In a subset of the experiments, to visualize and confirm the identity and morphology of GCs, the fluorophore Alexa Fluor 594 (red) was included in the recording pipette solution (20 M). Electrical stimulations and recordings were performed using the PatchMaster software. Experiments were performed at room temperature, or at 32 Ϯ 2°C, using the TC-342B Automatic Temperature Controller (Warner Instruments) with a perfusion speed of 2-3 ml/min. Cells were patched using standard patch pipettes (resistance, 4 -8 M⍀). The membrane potential was not corrected for junction potential. Although our experiments in whole-cell recordings precluded an accurate measurement of membrane potential (V m ), right after rupturing the seal, the V m in GCs of the MOB was Ϫ71.9 Ϯ 1.5 mV (n ϭ 26) and Ϫ64.4 Ϯ 1.5 mV (n ϭ 5) in the AOB. These measurements are in agreement with those from previous studies indicating that GCs maintain a hyperpolarized V m in the slice preparation (Cang and Isaacson, 2003); Egger et al., 2005). Therefore, unless otherwise indicated, voltage-clamp and current-clamp experiments in GCs were conducted at Ϫ60 or Ϫ70 mV.

Solutions and pharmacological agents
In whole-cell recordings, the internal solution had the following composition (in mM): 120 K-gluconate, 10 Na-Gluconate, 4 NaCl, 10 HEPES-K, 10 Na phosphocreatine, 2 Na-ATP, 4 Mg-ATP, and 0.3 GTP adjusted to pH 7.3 with KOH. This internal solution, which contains Na, was used in order to compare the physiology of GCs with previous work published from our laboratory (Smith et al., 2015).
In some experiments, 0.5 mM cAMP was added to the internal solution. The osmolarity of the internal solutions was adjusted to 290 -305 mOsm. For experiments using high extracellular K ϩ , the ACSF contained 25 mM KCl and 102.5 mM NaCl. Drugs were prepared fresh from stocks and diluted into the external solution; ZD7288 was purchased from Tocris Cookson, and it was applied at 30 M.

Data analysis
All electrophysiological recordings were analyzed in MATLAB (MathWorks). In order to isolate I h , leak currents were subtracted from the voltage-current relationships using multiples of a small voltage step (Ϫ60 to Ϫ65 mV) elicited at the end of each voltage-clamp trace. Recorded GCs exhibited heterogeneity in physiological properties, such as input resistance and capacitance, presumably due to the presence of a heterogeneous population that included mature and immature cells. To compensate for these differences, in some instances (i.e., high K ϩ or ZD288) I h was normalized to the current elicited by a voltage step from Ϫ60 to Ϫ130 mV.
Voltage dependency was calculated by fitting the tailcurrent maximum values to a Boltzmann function of the following form (Ludwig et al., 1998;Dibattista et al., 2008): where I max is the maximal current, V is the prepulse potential, V half is the half-activation potential of the current, and k s is the slope factor. The current (I) values were measured as the peak current obtained with a step to Ϫ130 mV from the prepulse potential (see Fig. 2A, inset) and were then normalized to the fitted I max value for each cell. The activation time constant () for I h was determined by fitting a single exponential function to the raw voltage trace when stepping from Ϫ60 to Ϫ130 mV.
The sag potential was calculated from the voltage deflection elicited by a negative current pulse, and corresponds to the difference between the peak minimum voltage and the V m upon reaching a steady state (after 1 s). The change in membrane potential (⌬V m ) in response to ZD7288 was measured as the difference between the baseline V m value (before the administration of the drug) and the V m value at the peak of the response, which, under our perfusion speed, occurred within ϳ6 -10 min. Subthreshold resonance was measured using a standard impedance amplitude protocol (ZAP), in which a stimulus of sinusoidal current of constant amplitude and exponentially increasing frequency (0.2-20 Hz) is injected into the cell. The ZAP protocol was obtained at different negative potentials (Ϫ60 to Ϫ90 mV), which are correspondingly indicated throughout the text. The exponential ZAP protocol we used works best on cells with lower resonant frequencies; therefore, we also confirmed our findings in mature cells using a linear ZAP protocol Narayanan and Johnston, 2007), allowing us to better sample higher frequencies (data not shown). The impedance profile was calculated as the magnitude of the ratio of the fast Fourier transforms of the voltage response and current input (Gutfreund et al., 1995;Hutcheon and Yarom, 2000;Vera et al., 2014) as follows: The impedance profile is smoothed, and the resonant frequency (f res ) is the frequency at which the maximal impedance value occurs. If the peak value is not clear, the impedance profile is fit with a quadratic function to provide an estimate. The strength of resonance (Q factor) is calculated as the ratio between the maximal impedance (|Z(f res )|) and the lowest frequency impedance values (|Z(f low )|; 0.5 Hz). The cell is considered resonant if f res Ͼ f low and Q Ͼ 1. Statistical significance was determined by a Student's t test, Mann-Whitney U test, one-way ANOVA, logistic regression, or linear regression.

I h is present in GCs of the AOB and MOB
We conducted whole-cell patch-clamp recordings in GCs of the MOB and AOB. In both regions, GCs are located in easily identified layers and exhibit characteristic morphology, including the presence of basal dendrites and a single apical dendrite, which bifurcates into several branches, populated by prominent dendritic spines ( Fig.  1A; Price and Powell, 1970;Larriva-Sahd, 2008). In voltage clamp (Fig. 1B), hyperpolarizing steps from Ϫ60 to Ϫ130 mV revealed a slowly developing inward current, with characteristics of I h , that occurred in the majority of GCs [ϳ96% in the AOB (80 of 83 GCs) and ϳ90% in the MOB (106 of 118 GCs)]. In agreement with the properties of I h , the peak amplitude of the inward current was enhanced by approximately threefold by increasing the external K ϩ concentration from 2.5 to 25 mM (Fig. 1C). Thus, the normalized mean currents in high K ϩ were 3.7 Ϯ 0.9 (n ϭ 5; Mann-Whitney U test, p Ͻ 0.01) in the AOB and 3.5 Ϯ 1.1 in the MOB (n ϭ 4; Mann-Whitney U test, p Ͻ 0.01).
Importantly, in the presence of the selective I h blocker ZD7288 (30 M; see Materials and Methods), the inward current was reduced by over 60% (Fig. 1C). The normalized mean current in the presence of ZD7288 was 0.4 Ϯ 0.1 in the AOB (n ϭ 4; Mann-Whitney U test, p Ͻ 0.03) and 0.36 Ϯ 0.16 in the MOB (n ϭ 7; Mann-Whitney U test, p Ͻ 0.001).
Different subunit composition in the tetrameric HCN channels confers distinct cyclic nucleotide sensitivity to I h , which in turn influences the voltage sensitivity and kinetics of the channels (Craven and Zagotta, 2006;Benarroch, 2013). Therefore, we compared the properties of I h by recording GCs with and without cAMP (0.5 mM) in the internal solution, determining the kinetics of I h activation and maximal current elicited in voltage clamp ( Fig. 2A). Analysis of the voltage dependency using a Boltzmann fit (see Materials and Methods) and the kinetics of activation of I h revealed significant differences between GCs in the AOB and MOB. In the absence of cAMP, the activation of I h occurred at more positive potentials in the AOB than in the MOB; the V half of the activation of I h in AOB was Ϫ94.8 Ϯ 4 mV (n ϭ 15), while in the MOB V half was Ϫ103 Ϯ 6 mV (n ϭ 10; AOB vs MOB, Student's t test, p Ͻ 0.03). Furthermore, in the absence of cAMP, the activation of I h was slower in the AOB than in the MOB ( Fig. 2B; ϭ 123 Ϯ 14 ms vs ϭ 58.8 Ϯ 12.9 ms, respectively; Student's t test, p Ͻ 0.007). Interestingly, adding cAMP to the internal solution shifted V half by ϳ10 mV in the MOB (Ϫ94 Ϯ 5 mV, n ϭ 11, Student's t test, p Ͻ 0.04), while it produced no significant effect in the AOB (Ϫ94.5 Ϯ 3 mV, n ϭ 12; Student's t test, p Ͼ 0.4). In contrast, decreased by approximately half in both regions (AOB: 55.3 Ϯ 8.7 ms; Student's t test, p Ͻ 0.001; MOB: 30 Ϯ 6.8 ms; Student's t test, p Ͻ 0.03). Together, these findings indicate that the sensitivity to cAMP is region specific and suggest that GCs may express HCN channels with different subunit composition in these regions.
The voltage dependency of I h in GCs suggests that it could be active at resting V m (see Materials and Methods). In agreement with this possibility, in current-clamp experiments bath perfusion of ZD7288 invariably produced a hyperpolarization in GCs (Fig. 3). Blocking I h reduced the membrane potential by 6 Ϯ 1.3 mV in the AOB (orange, n ϭ 6; Student's t test, p Ͻ 0.005) and by 8.8 Ϯ 3.7 mV in the MOB (blue, n ϭ 5; Student's t test, p Ͻ 0.04). This effect of ZD7288 was still present in the presence of  At the end of each step, a test pulse to Ϫ130 mV (V test ) was elicited to measure the maximal current at each potential (traces shown in the bottom box inset). The current amplitudes measured at the peak were normalized to I max and fitted to a Boltzmann function (see Materials and Methods). The value was estimated by fitting the current elicited by stepping from Ϫ60 to Ϫ130 mV (red dotted line, 25.5 ms for this trace). B, Top, Boltzmann fits for I h in GCs of the MOB (left) and AOB (right) in recordings conducted with internal solutions without cAMP (filled circles) or containing cAMP (0.5 mM, empty circles). Bottom, Summary plot comparing the cAMP sensitivity of V half and in AOB GCs (orange) and MOB GCs (blue). In the presence of cAMP, the V half was shifted to more positive potentials in MOB CGs, while in the AOB, was reduced (shaded oval).

GCs exhibit subthreshold resonance that depends on I h
In other brain regions, I h plays a critical role in generating subthreshold resonance of neurons, which in turn contributes to network oscillatory properties Narayanan and Johnston, 2007;van Brederode and Berger, 2011;Vera et al., 2014). Therefore, we examined subthreshold resonance in GCs that exhibited I h , using a sinusoidal current stimulus of increasing frequency (ZAP stimulus; see Materials and Methods) in animals older than P30. Surprisingly, we observed a significant heterogeneity in responses to the ZAP stimulus among GCs, with a mixed population of resonant and nonresonant cells (Fig. 4A). The percentage of resonant cells in the AOB was ϳ71% (15 of 21 cells), and in the MOB was ϳ52% (10 of 19 cells). In addition, among all resonant cells (AOB plus MOB), the response to the ZAP stimulus revealed heterogeneity in the f res , with cells distributed along the delta and theta range (data not shown). Importantly, this subthreshold resonance could affect the global excitability of GCs. Accordingly, when we adjusted the ZAP protocol to elicit action potentials in a subset of resonant MOB and AOB GCs, these action potentials occurred at a frequency within the range of f res for each cell, with 79% of the action potentials falling within Ϯ3 Hz relative to the resonant frequency of the cell (Fig. 4B).
The heterogeneity of subthreshold resonance that we observed prompted us to examine whether the size of I h in GCs could contribute to these differences. Across the total population of GCs studied (AOB and MOB), resonant cells had larger I h values than nonresonant cells, suggesting that I h amplitude is a predictor of resonance ( Fig. 4C; resonant cells, I max ϭ Ϫ97.5 Ϯ 10.3 pA, n ϭ 25; nonresonant cells, I max ϭ Ϫ61 Ϯ 13.0 pA, n ϭ 15; logistic regression, p Ͻ 0.02). Similarly, I h amplitude is predictive of resonance strength (I max vs Q, in a linear regression model, p Ͻ 0.02). In agreement with a larger I h value, R i was significantly lower in resonant cells than in nonresonant cells (1.14 Ϯ 0.01 vs 1.61 Ϯ 0.17 G⍀; logistic regression, p Ͻ 0.01). Notably, the subthreshold resonance was abolished in the presence of ZD7288, (Fig. 4C); the strength of resonance or Q factor (see Materials and Methods) was reduced from 1.12 Ϯ 0.02 to 1.02 Ϯ 0.01 in the presence of ZD7288 (n ϭ 8; Student's t test, p Ͻ 0.0005). Together, these results suggest that I h is a major contributor to the expression of subthreshold resonance in AOB and MOB GCs.

Developmentally timed expression of I h and subthreshold resonance in GCs in the MOB
GCs undergo a critical period and exhibit varied physiological properties over the course of development (Yamaguchi and Mori, 2005;Lepousez et al., 2013); therefore, we compared I h and subthreshold resonance expression in GCs at 2 and 6 weeks after birth in the MOB. As shown in Fig. 6A, we found that I h -mediated sag amplitude in GCs significantly increased between 2 weeks (n ϭ 5) and 6 weeks (n ϭ 5) after birth, (Ϫ1.9 Ϯ 0.4 vs Ϫ4.6 Ϯ 0.7 mV; Student's t test, p Ͻ 0.006). Similarly, the I h amplitude nearly doubled between 2 and 6 weeks, although it was not significant within our sampled population of cells (Ϫ121.1 Ϯ 31.1 vs Ϫ215.9 Ϯ 37.0 pA; Student's t test, p Ͻ 0.09). Nevertheless, and in agreement with the developmental increase in I h -mediated sag amplitude, resonance in GCs increased as a function of cell age (Fig. 6B). Between 2 and 6 weeks, resonance strength increased from 1.05 Ϯ 0.01 to 1.10 Ϯ 0.01 (Student's t test, p Ͻ 0.02). Interestingly, we also observed a developmental increase in f res . At 2 weeks, f res was 3.44 Ϯ 0.98 H, and at 6 weeks, 7.32 Ϯ 0.79 Hz (Student's t test, p Ͻ 0.02). Altogether, these results indicate that as GCs mature, their f res becomes tuned towards the physiological breathing rates of mice (1-12 Hz).

Subthreshold resonance in other OB subtypes
We next examined whether other neurons in the OB circuit exhibit subthreshold resonance. In particular, we examined another inhibitory neuron, the PGC, which con-  5 and 3, respectively). B, Left, In MOB GCs, Q is strongest at more negative potentials, while in AOB GCs Q is larger at more positive potentials. Right, At Ϫ90 mV the average peak frequency in MOB GCs is significantly higher than in the AOB GCs (see Results). tributes to network oscillations entrained by the respiratory cycle (Fukunaga et al., 2014). Therefore, we examined the presence of subthreshold resonance in PGCs sampled from animals older than P30. MOB PGCs exhibited a prominent I h , with a sag potential of 5.2 Ϯ 1.5 mV (n ϭ 9). In recordings conducted with cAMP in the pipette, the V half for I h in PGCs was more positive than in MOB GCs (Fig. 7A; Ϫ82.6 Ϯ 3.4 mV, n ϭ 9; Student's t test, p Ͻ 0.03), while was significantly slower (81.1 Ϯ 22.5; n ϭ 9; Student's t test, p Ͻ 0.01), indicating differences in I h between these inhibitory subtypes. Importantly, the majority of MOB PGCs (9 of 11 PGC) exhibited subthreshold resonance with a mean resonant frequency of 2.02 Ϯ 0.16 Hz and a Q factor of 1.15 Ϯ 0.03. Moreover, application of ZD7288 completely abolished the subthreshold resonance ( Fig. 7A; n ϭ 3), indicating that subthreshold resonance in PGCs, like in GCs, depends on the expression of I h .
In addition, we examined subthreshold resonance in MOB MCs, which have been previously shown to have I h , with its expression decreasing with age (Angelo and Margrie, 2011;Yu et al., 2015). To this extent, we recorded from two age groups of MCs, young (P11; n ϭ 8) and old (P35; n ϭ 6). In both age groups, MCs exhibited I h , and, in agreement with previous work, I h was more prominent in the younger group, with sag values of 8.9 Ϯ 1.3 mV (young) and 5.0 Ϯ 1.2 mV (old; Fig. 7B; Student's t test, p Ͻ 0.02). In these recordings, which we conducted with cAMP in the pipette, the V half for I h was Ϫ97 Ϯ 2.8 mV, and was 500 Ϯ 84 ms at P11 (MC vs PGC , Student's t test, p Ͻ 0.0001). Intriguingly, despite the expression of I h , none of the recorded MCs, in either group exhibited subthreshold resonance. Furthermore, while I h has also been observed in AOB MCs (data not shown; but see Gorin et al., 2016), we did not observe subthreshold resonance in recorded AOB MCs (n ϭ 5) nor in AOB PGCs (n ϭ 4).

Discussion
Throughout the brain, the regulation of membrane excitability by I h plays an important role in neuronal function and network dynamics (Lüthi and McCormick, 1998;Narayanan and Johnston, 2007;George et al., 2009;Benarroch, 2013). Network synchrony dynamics are an im- Mice were electroporated at P1, and recordings were conducted at different postnatal weeks. Right, Sample current-clamp traces from labeled GCs at 2 and 6 weeks; recorded cells exhibit robust action potentials. At both ages, a hyperpolarizing current step elicits an I h -mediated sag upon hyperpolarization (see inset). Calibration: 500 ms and 20 mV. The colored inset compares the sag amplitudes of the cells shown at 2 weeks (green) and 6 weeks (red). Calibration: 100 ms and 2 mV. The bar graph summarizes the change in sag amplitude between 2 and 6 weeks postlabeling. B, Left, Example plots of the impedance profile at 2 and 6 weeks after electroporation obtained using a ZAP protocol, corresponding to the cells seen above in A. At 2 weeks, Q is 1.02, while at 6 weeks Q is 1.14 and f res is 5.18 Hz (dotted line). Right, Bar graphs showing the increase in Q factors and peak frequencies between 2 and 6 weeks.
New Research portant component of the initial processing of odor signals in the OB. GCs comprise the largest population of inhibitory neurons in the OB, where they integrate local and afferent top-down signals to convey inhibition to output neurons, the MCs. Here, we show that I h is present in GCs and that its expression follows the postnatal development of GCs, increasing with age. In addition, we show that GCs and MOB PGCs exhibit an I h -dependent subthreshold resonance in the 1-12 Hz frequency range. These results suggest that the regulation of I h can play a role on circuit dynamics in the OB entrained by respiration.
GCs are a heterogeneous group of neurons, and they exhibit distinct characteristics in the AOB and MOB, including synaptic properties and regulation by neuromodulatory transmitters (Castro et al., 2007;Zimnik et al., 2013;Burton and Urban, 2015;Smith et al., 2015). Accordingly, we found that the voltage sensitivity of I h was modulated by cAMP in the MOB, but not in the AOB; nevertheless, I h activation was faster in the AOB in the presence of cAMP. Among HCN subunits, HCN1 exhibit the fastest kinetics, but the weakest cAMP sensitivity (Kaupp and Seifert, 2001;Accili et al., 2002), while HCN2 and HCN4 exhibit slower kinetics but are more cAMP sensitive (Kaupp and Seifert, 2001;Wainger et al., 2001). On the other hand, the HCN3 subunit exhibits slow kinetics but is not cAMP sensitive (Mistrik et al., 2005). Therefore, the kinetic parameters that we determined suggest that MOB GCs may express a combination HCN1, HCN2, and HCN4 subunits, while HCN3 subunits predominantly contribute to I h in AOB GCs. In agreement with this possibility, all four HCN subunits are expressed in the OB, with HCN1 present in PGCs and GCs (Holderith et al., 2003;Fried et al., 2010), whereas HCN2 to HCN4 are more strongly expressed in the inner layers of the MOB (Notomi and Shigemoto, 2004). In the AOB, HCN1, HCN2, and HCN4 are moderately expressed, while HCN3 exhibits the strongest expression (Notomi and Shigemoto, 2004). Interestingly, despite exhibiting different kinetics and voltage sensitivity, our data with ZD7288 suggest that I h in GCs of both regions is active at rest. In other neurons, I h has been shown to contribute to baseline membrane excitability as well as dendritic integration (Magee, 1998;Fan et al., 2005;Benarroch, 2013); we propose a similar , while a depolarizing current step (5 pA) elicits a burst of action potentials that is characteristic of PGCs. The resting V m in this cell is Ϫ70 mV. Calibration: 20 mV and 200 ms. Middle, Summary bar graph showing that PGCs (n ϭ 9) exhibit a more depolarized V half than GCs (n ϭ 11). Right, Impedance profile of a different PGC obtained using a ZAP protocol while holding at Ϫ70 mV. The subthreshold resonance is abolished in the presence of ZD7288 (control: black line, f res ϭ 2.24 Hz, indicated by the vertical dashed line, and Q ϭ 1.10; ZD7288, dashed line, f res ϭ f low , and Q ϭ 1). B, Left, A current-clamp response of an MC at P11; a hyperpolarizing current step (Ϫ350 pA) produces a sag in V m (arrow), while a depolarizing current step (350 pA) elicits a train of action potentials. The resting V m in this cell is Ϫ60 mV. Calibration: 20 mV and 1 s. Middle, Summary bar graph showing the decrease in sag potential in MCs between P11 and P35 (P11, n ϭ 8; P35, n ϭ 6). Right, Impedance profile of the MC shown on the left, obtained using a ZAP protocol while held at Ϫ90 mV. Despite exhibiting I h , the cell is nonresonant (f res ϭ f low , and Q ϭ 1). function of I h in GCs. In fact, as they lack axons, the integration of synaptic inputs and the output occurs solely in dendritic processes in GCs. Interestingly, GCs receive most afferent excitatory input in basal and proximal dendrites (Boyd et al., 2012;Markopoulos et al., 2012); therefore, the extent of regulation by these inputs of dendrodendritic synapses in distal regions of GC's apical dendrites could also be modulated by the tonic influence of I h .
Mature AOB and MOB GCs exhibited subthreshold resonance, but there was a stark contrast in the membrane potential at which the resonance was strongest, as well as the peak frequency. In AOB GCs, subthreshold resonance was stronger at more depolarized potentials, and GCs were resonant in the delta range (1-4 Hz), whereas in MOB GC resonance was strongest at more negative potentials and resonance frequencies were in the theta range (4 -12 Hz). These differences in GCs could underlie different circuit dynamics to accommodate their specific sensory inputs, which could be arriving at these specific frequency bands. In the MOB, inputs from the main olfactory epithelium are driven by breathing and arrive in the theta range (Smear et al., 2011;Manabe and Mori, 2013). Thus, in the theta range, subthreshold resonance for GCs may act as a bandpass filter for inputs at this frequency. In the AOB, however, the VNO is driven by the constriction and dilation of large blood vessels acting as a pump (Keverne, 1999); therefore, activity in the AOB may not be directly driven by breathing, but instead by the autonomic nervous system. Interestingly, previous studies (Dibattista et al., 2008;Cichy et al., 2015) have shown the presence of I h in VNO sensory neurons; however, at this time it is unknown whether these cells exhibit subthreshold resonance. The lower frequency in AOB GCs suggests that the information processing and timing of inputs can work on a different timescale compared with the MOB. For example, the delta range resonance that we found in the AOB matches the lower-frequency baseline activity of AOB MCs near ϳ4 Hz (Luo et al., 2003), as well as VNO activity near ϳ1 Hz (Meredith, 1994). We note that differences in resonant frequency could be due to the contribution of other currents to the subthreshold resonance, for example, the M-current (I m ), another noninactivating potassium current associated with the KCNQ family of ion channels (Hutcheon and Yarom, 2000;Hu et al., 2009); however, our data indicate that in the MOB the subthreshold resonance arises from I h and passive properties of GCs. Further studies should examine the possibilities of other contributing currents known to play a role in subthreshold resonance, such as I m , or I NaP , and corroborated with a computational model.
In several brain regions, I h underlies the expression of subthreshold resonance in principal neurons, as well as interneurons (Hutcheon and Yarom, 2000;Ulrich, 2002Ulrich, , 2014Hu et al., 2009;Zemankovics et al., 2010;Vera et al., 2014). Our data indicate that inhibitory neurons in the MOB, the GC and PGCs, exhibit subthreshold resonance and that this property is dependent on the presence of I h in both subtypes; however, in the AOB, I hmediated subthresold resonance may be present only in GCs. In fact, we found that subthreshold resonance in GCs strongly correlated with the presence of I h . This was particularly evident when we compared I h expression and the presence of resonance during postnatal development. In 2-week-old GCs, I h was smaller and the resonance strength was low, while they were significantly larger at 6 weeks. Although we did not perform a developmental analysis in MOB PGCs, we hypothesize that they exhibit a similar developmental progression of I h size and the expression of subthreshold resonance. However, this was not the case for principal neurons, the MCs. In agreement with a previous report (Yu et al., 2015), we observed a developmental decrease in I h , yet, even when I h was present, MCs did not exhibit I h -mediated subthreshold resonance. However, it has been suggested that MOB MCs exhibit gamma-frequency subthreshold resonance due to a persistent Na ϩ current (Brea et al., 2009), which was not addressed in our studies. Even in the AOB, I h has been described in MCs, but blocking it had no effect on the intrinsic oscillations of MCs (Gorin et al., 2016). In agreement with these findings, we found no I h -mediated subthreshold resonance in AOB MCs.
In summary, we provide evidence that the two predominant subtypes of inhibitory neurons in the MOB, the GCs and PGCs, exhibit subthreshold resonance mediated by I h , but the primary output neurons, the MCs, do not. Various studies indicate that top-down neuromodulatory projections and cortical feedback principally target PGCs and GCs. Therefore, it is possible that subthreshold resonance and underlying I h contribute to the integration of this top-down information. In addition, the activation of cAMP-dependent pathways by neuromodulators could lead to changes in I h kinetics, affecting the resonant frequency and oscillations at the network level. Future in vivo studies will address how these properties of inhibitory neurons may contribute to OB circuit dynamics and odor processing.