Abstract
The accessory olfactory system controls social and sexual behavior. However, key aspects of sensory signaling along the accessory olfactory pathway remain largely unknown. Here, we investigate patterns of spontaneous neuronal activity in mouse accessory olfactory bulb mitral cells, the direct neural link between vomeronasal sensory input and limbic output. Both in vitro and in vivo, we identify a subpopulation of mitral cells that exhibit slow stereotypical rhythmic discharge. In intrinsically rhythmogenic neurons, these periodic activity patterns are maintained in absence of fast synaptic drive. The physiological mechanism underlying mitral cell autorhythmicity involves cyclic activation of three interdependent ionic conductances: subthreshold persistent Na+ current, R-type Ca2+ current, and Ca2+-activated big conductance K+ current. Together, the interplay of these distinct conductances triggers infraslow intrinsic oscillations with remarkable periodicity, a default output state likely to affect sensory processing in limbic circuits.
SIGNIFICANCE STATEMENT We show for the first time that some rodent accessory olfactory bulb mitral cells—the direct link between vomeronasal sensory input and limbic output—are intrinsically rhythmogenic. Driven by ≥3 distinct interdependent ionic conductances, infraslow intrinsic oscillations show remarkable periodicity both in vitro and in vivo. As a novel default state, infraslow autorhythmicity is likely to affect limbic processing of pheromonal information.
Introduction
In rodents, the accessory olfactory system controls conspecific chemical communication during social interactions. Behaviorally relevant chemosignals are detected by neurons in the vomeronasal organ and processed along sensory pathways that include the accessory olfactory bulb (AOB), amygdala, and hypothalamus (Stowers and Logan, 2010). While two of these three primary processing centers have been intensely studied (LeDoux, 2000; Simerly, 2002), central aspects of AOB physiology remain largely unexplored and functional analogies with neurons of the main olfactory bulb (MOB) are mostly speculative (Dulac and Wagner, 2006). Principal neurons, conventionally referred to as mitral cells, are the sole projection neurons of the AOB. These neurons extend complex, branched primary dendrites that receive excitatory synaptic input from vomeronasal sensory neurons in multiple glomeruli (Larriva-Sahd, 2008). This morphology indicates that sensory computation in AOB mitral cells is fundamentally different from the processing of “glomerulus-specific” information by their MOB counterparts (Dulac and Torello, 2003). Despite important recent insights into the organizational principles of connectivity, sensory input, and integration in the AOB (Del Punta et al., 2002; Ma and Lowe, 2004; Sugai et al., 2005; Wagner et al., 2006; Castro et al., 2007; Ben-Shaul et al., 2010; Smith and Araneda, 2010; Hovis et al., 2012; Leszkowicz et al., 2012; Shpak et al., 2012; Tolokh et al., 2013; Hammen et al., 2014), a conceptual understanding of how the biophysical properties of AOB mitral cells affect their computations is lacking.
Spontaneous activity is a major determinant of a neuron's coding capacity and information transfer function (Rieke et al., 1997). Spontaneous discharge may be sparse or dense, irregular or periodic, generating a continuum of neuronal firing patterns from Poisson-like discharge (Shadlen and Newsome, 1998) to rhythmic clock-like firing (Blankenship and Feller, 2010). Patterned discharge, such as burst firing and/or oscillatory activity, is particularly relevant for neural processing: bursts often represent units of information (Izhikevich et al., 2003), whereas oscillations provide precise temporal windows of excitability for circuit computations (Mizuseki et al., 2009). Different frequencies have been reported to encode specific brain states (Buzsáki et al., 2013). At the lower end of the time scale, oscillations extend into the slow (0.1–1 Hz) and infraslow (<0.1 Hz) range (Schroeder and Lakatos, 2009). Slow stereotypical episodic “up” and “down” states (Blethyn et al., 2006) could either result from regular recurrence of excitatory and inhibitory synaptic barrages (Crunelli and Hughes, 2010) or, alternatively, represent a network-independent intrinsic neuronal property, caused by a cyclical interplay of autonomous depolarizing and hyperpolarizing conductances (Blankenship and Feller, 2010). Such intrinsic pacemaker-like discharge has been implicated both in controlling rhythmic behaviors (Peña et al., 2004; Bucher et al., 2006; Koizumi and Smith, 2008; Tazerart et al., 2008) and in driving oscillatory circuits that play important roles in sensory perception, attention, memory formation, and decision making (Gutierrez et al., 2013). In the MOB, sensory input evokes oscillatory network activity across a wide frequency range (Kay et al., 2009). Autonomous, episodic burst firing of external tufted (ET) cells within a small spectral window (∼0.2–10 Hz; Hayar et al., 2004) has been attributed a role in setting sniff cycle-dependent glomerular synchrony (Hayar et al., 2005; Shao et al., 2009) and, consequently, distinct phase-locking of MOB principle neurons (Smear et al., 2011; Fukunaga et al., 2012). Whether aspects of sensory coding in the AOB are also affected by single cell/network rhythmicity, however, is unclear.
Here, we identify patterns of coordinated spontaneous activity among AOB mitral cells. We demonstrate that a subpopulation of these neurons reveals slow stereotypical rhythmic discharge both in vitro and in vivo. Among these neurons, a group of mitral cells is intrinsically rhythmogenic. Their autorhythmicity is driven by a precisely orchestrated ensemble of interdependent Na+, Ca2+, and K+ conductances. The cyclic (in)activation of these three distinct conductances provides a mechanistic basis for infraslow intrinsic oscillations of remarkable periodicity in a subset of AOB projection neurons.
Materials and Methods
Animals.
All animal procedures were approved by local authorities and were in compliance with European Union legislation (Directive 86/609/EEC) and recommendations by the Federation of European Laboratory Animal Science Associations. C57BL/6 mice (Charles River Laboratories) were housed in groups of both sexes [room temperature (RT); 12 h light/dark cycle; food and water available ad libitum]. For in vivo recordings, sexually naive male BALB/c mice were used (Harlan Laboratories). All in vivo experiments were performed in compliance with the Hebrew University Animal Care and Use Committee.
Chemicals and solutions.
The following solutions, labeled S1–S9, were used: S1, HEPES buffered extracellular solution containing (in mm) 145 NaCl, 5 KCl, 1 CaCl2, 1 MgCl2, 10 HEPES, pH 7.3 (adjusted with NaOH), 300 mOsm (adjusted with glucose); S2, oxygenated (95% O2, 5% CO2) artificial CSF (aCSF) containing (in mm) 124 NaCl, 26 NaHCO3, 3 KCl, 1.25 NaH2PO4, 1.3 MgSO4, 1.3 CaCl2, 10 glucose, pH 7.3, 300 mOsm (adjusted with glucose); S3, oxygenated (95% O2, 5% CO2) cutting solution containing (in mm) 220 sucrose, 26 NaHCO3, 3 KCl, 1.25 NaH2PO4, 2.6 MgSO4, 10 glucose, pH 7.3, 300 mOsm (adjusted with glucose); S4, extracellular Ba2+ solution containing (in mm) 100 NaCl, 25 tetraethylammonium chloride (TEA-Cl), 10 4-aminopyridine, 10 BaCl2, 1 MgCl2, 10 HEPES, 0.001 tetrodotoxin (TTX), pH 7.3 (adjusted with HCl), 300 mOsm; S5, standard pipette solution containing (in mm) 10.5 KCl, 125 KOH, 125 gluconic acid, 2 MgCl2, 1 EGTA, 0.3 CaCl2, 10 HEPES, 2 Mg-ATP, 1 Na-GTP (free Mg2+, 2 mm; free Ca2+, 130 nm), pH 7.1 (adjusted with KOH), 290 mOsm; S6, symmetrical chloride pipette solution containing (in mm) 143 KCl, 2 KOH, 1 EGTA, 0.3 CaCl2, 10 HEPES, 2 Mg-ATP, 1 Na-GTP (free Ca2+, 130 nm), pH 7.1 (adjusted with KOH), 290 mOsm; S7, Cs+-based pipette solution containing (in mm) 125 gluconic acid, 105 CsOH, 20 NaOH,10.5 CsCl, 2 MgCl2, 1 EGTA, 0.3 CaCl2, 10 HEPES, 2 Mg-ATP, 1 Na-GTP (free Mg2+, 2 mm; free Ca2+, 130 nm), pH 7.1 (adjusted with CsOH), 290 mOsm; S8, standard blocking solution containing 5% normal bovine serum (Dianova), 0.8% Triton X-100, and 0.05% NaN3 in Ca2+/Mg2+-free PBS (PBS−/−; 100 mm); S9, standard staining solution containing 3% bovine serum albumin (IgG-free, protease-free), 0.05% NaN3, and Alexa Fluor 488 or 633 streptavidin conjugate (1:800; Life Technologies) in PBS−/− (100 mm).
Free Ca2+ concentrations were calculated using WEBMAXC (http://web.stanford.edu/∼cpatton/webmaxcE.htm). If not stated otherwise, chemicals were purchased from Sigma-Aldrich. Alexa Fluor hydrazide was purchased from Life Technologies; ω-agatoxin IVA and SNX 482 were purchased from Tocris Bioscience; paxilline was purchased from Enzo Life Science; 2-(3-carboxypropyl)-3-amino-6-(4 methoxyphenyl)pyridazinium bromide (gabazine), D-(-)-2-amino-5-phosphonopentanoic acid (AP5), 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide (NBQX), ω-conotoxin GVIA, and 4-ethylphenylamino-1,2-dimethyl-6-methylaminopyrimidinium chloride (ZD7288) were purchased from Abcam. Final solvent concentrations were ≤0.1%. Solutions and pharmacological agents were applied either by bath or from air pressure-driven reservoirs via an eight-in-one multibarrel “perfusion pencil” (Science Products). Changes in focal superfusion (Veitinger et al., 2011) were software-controlled and, if required, synchronized with data acquisition by transistor–transistor logic input to 12 V DC solenoid valves using a TIB 14S digital output trigger interface (HEKA Elektronik).
Slice preparation.
Mice were killed by brief exposure to a CO2 atmosphere and decapitation. The left and right olfactory bulbs were rapidly removed, separated with a razor blade, embedded in 4% low-gelling temperature agarose (VWR International), and placed in ice-cold oxygenated cutting solution (S3). Parasagittal slices (250 μm) were cut with a VT1000S vibratome (Leica Biosystems). Two slices per bulb, each including the AOB, were transferred to a submerged, oxygenated storage container and allowed to recover for ≥1 h in aCSF (S2). Slices were then stored at RT until use.
Histology and immunochemistry.
For immunohistochemical labeling of AOB sections, olfactory bulbs were fixed in 4% (w/v) paraformaldehyde (PFA) in PBS−/− [10 mm, pH 7.4 (3 h, 4°C)] and subsequently cryoprotected in PBS−/− containing 30% sucrose (≥24 h, 4°C). Samples were embedded in Tissue Freezing Medium (Leica Biosystems), sectioned at 20 μm on a Leica CM1950 cryostat (Leica Biosystems), and mounted on Superfrost Plus slides (Menzel). Next, sections were incubated in 4% PFA (15 min, 4°C), washed several times in PBS−/−, and then incubated (1 h, RT) in PBS−/−-based blocking solution containing 0.8% Triton X-100, 0.05% NaN3, and normal bovine serum. Primary antibodies were anti-Slo1 (KCa1.1, 1:500; Alomone Labs), anti-Sloβ4 (1:100; Alomone Labs), and anti-CaV2.3 (1:100; Alomone Labs). Cryosections were incubated with primary antibodies (19 h, 4°C) in a dark humidified chamber. After washing in PBS−/− (3×, 10 min), sections were incubated in PBS−/− (1 h, RT) containing Alexa Fluor-conjugated secondary antibodies (1:500; Life Technologies). Excess antibodies were removed by washing (5×, 10 min). Fluorescent images were taken using an upright fixed-stage scanning confocal microscope (Leica TCS SP5 DM6000 CFS, Leica Microsystems) equipped with a 20× 1.0 numerical aperture water-immersion objective (HCX APO L, Leica Microsystems). To control for nonspecific staining and to demonstrate antibody specificity, we performed (1) antigen preadsorption controls and (2) experiments in which the primary antibodies were omitted in parallel with each procedure. Nuclei were counterstained using 1, 5-bis{[2-(di-methylamino)ethyl]amino}-4, 8-dihydroxyanthracene-9, 10-dione (DRAQ5, 1:200; ThermoFisher Scientific). Digital images were uniformly adjusted for brightness and contrast using Adobe Photoshop CS6 (Adobe Systems).
For immunoblotting, both AOB and control tissue were homogenized in lysis buffer (100 μl, 0.1% Triton X-100, 4°C) in the presence of Complete Mini protease inhibitor mixture tablets (Roche). The homogenate was sonicated and centrifuged for 10 min (4°C, 1000 × g). The supernatant was resuspended in lysis buffer and protein concentration was determined (BioPhotometer Plus, Eppendorf). Thirty microliters of Laemmli buffer (20% glycerol, 4% SDS, 125 mm Tris-HCl, and 0.02% bromphenol blue, pH 6.8) was added and equal amounts of protein were fractionated by SDS-PAGE. Separated proteins were transferred using a Bio-Rad Criterion Blotter wet-blotting system. Membranes were washed, stained with Ponceau S to control for protein transfer, and again washed with TBST (61 mm Tris-HCl, 88 mm NaCl, 0.1% Tween 20, pH 7.5). Blocking was performed in 5% non-fat dry milk/TBST overnight. Blots were then incubated in 2.5% non-fat dry milk/TBST (1 h, 4°C) with anti-Slo1 (1:200; Alomone Labs) and anti-Sloβ4 (1:50; Alomone Labs) antibodies. Membranes were then washed (4 × 15 min) in TBST and incubated with horseradish peroxidase-conjugated goat anti-rabbit IgG (Bio-Rad; 1 h, RT, 1:5000) in 2.5% non-fat dry milk/TBST. Blots were again washed in TBST (4 × 15 min) and antibody binding was detected using 2.5 ml of Lumi-Light solution (Roche).
For post hoc visualization, 3D reconstruction, and morphometric analysis of AOB mitral cells diffusion-loaded (≥20 min) with biocytin during whole-cell patch-clamp recordings, slices were kept for another 15 min in the recording chamber to wash out excess biocytin (Marx et al., 2012). AOB slices were then fixed in 4% PFA in PBS−/− [0.1 m; pH 7.4 (4°C, ≥12 h)]. Next, slices were washed (4 × 10 min) in PBS−/− and incubated in blocking/permeabilization solution (S8; 4°C, 90 min). Subsequently, biocytin-filled neurons were stained in a dark humidified chamber (S9; 60 min, RT) with fluorophore-conjugated streptavidin. After washing in PBS−/− (4 × 10 min), slices were mounted on slides and coverslipped. Confocal fluorescent z-stack images were taken using an upright fixed-stage scanning confocal microscope (TCS SP5 DM6000CFS, Leica Microsystems) equipped with a 20×/1.0 numerical aperture water-immersion objective (HCX APO L, Leica Microsystems). Alexa Fluor 488 or 633 streptavidin conjugate were excited using the 488 nm line of an argon laser or a 633 nm HeNe laser, respectively. Rendering of 3D data and morphometric measurements were performed using Imaris 8.0 software (Bitplane).
In vivo recordings.
Electrophysiological recordings of AOB neurons were performed as previously described in detail (Ben-Shaul et al., 2010; Cichy et al., 2015). Mice were anesthetized with 100 mg/kg ketamine and 10 mg/kg xylazine and a tracheotomy was performed using a polyethylene tube (inner diameter, 0.76 mm; outer diameter, 1.22 mm). Though not used during the recordings presented here, a cuff electrode was placed on the sympathetic nerve trunk. Incisions were closed and the mouse was placed in a custom-built stereotaxic apparatus where anesthesia was maintained throughout the entire experiment (0.5–1% isoflurane in O2). A craniotomy was made immediately rostral to the rhinal sinus, the dura was removed around the penetration site, and electrophysiological probes were advanced into the AOB using an electronic micromanipulator (MP-285, Sutter Instruments). All recordings were made with 32-channel multisite electrodes (A4x8-5mm-100-200-5 177, NeuroNexus Technologies). Spontaneous neuronal activity was recorded without vomeronasal stimulation for 3–10 min. Before each recording, electrodes were dipped in fluorescent dye (DiI; Invitrogen) and targeting of the AOB external cellular layer was confirmed from electrode tracts post mortem. During the experiment, large well isolated spikes were found exclusively when electrode tracks traversed the AOB external cellular layer. By contrast, negligible neural activity was observed from electrode tracks located in the granule cell layer. Recordings from MOB units were identified by both electrode tracks and signal–response patterns, and were thus removed from the analysis. Using an RZ2 processor, PZ2 preamplifier, and two RA16CH head-stage amplifiers (Tucker-Davis Technologies), neuronal activity was sampled at 25 kHz and bandpass filtered at 0.3–5 kHz. Candidate spike events were derived from continuous electrode data using the Matlab thselect function with a minimaxi criterion. The threshold was set conservatively to avoid loss of candidate spikes. Because each shank of the probe contains eight recording sites, each spike event is defined by eight simultaneously measured waveforms (spanning 3.5 ms; detected at any site). Then, for each recording site, we calculated the first two principal components of all recorded waveforms. Each spike event was then characterized according to its projections on the first two principal components of each of the eight channels per shank. Thus, each spike waveform was defined by 16 numbers. These numbers were then fed to the automatic spike-sorting algorithm KlustaKwik (Harris et al., 2000), which assigned each event to one of several clusters. Clusters were then manually examined and adjusted using the Klusters program (Hazan et al., 2006). This latter control procedure is required since the KlustaKwik function tends to overseparate clusters. Criteria for cluster evaluation included the spike waveforms themselves, their projections on the principal component space, and the interspike interval (ISI) histogram. Spike clusters were defined as single units if (1) they had a distinct spike shape, (2) they were fully separated from both the origin (noise) and other clusters in ≥1 principal component projection, and (3) their ISI histogram demonstrated a clear trough of ≥10 ms duration around time 0. If a cluster was composed of >1 single unit, it was designated as multiunit activity and excluded from the analysis. All spike-sorting procedures (except KlustaKwik and Klusters) were implemented using custom-written Matlab code.
In vitro electrophysiology.
Olfactory bulb slices were transferred to a recording chamber (Luigs & Neumann), positioned with stainless steel anchors, and visualized using an upright fixed-stage video-microscope (DM6000FS, Leica Microsystems) equipped for infrared-optimized differential interference contrast. Slices were continuously superfused with oxygenated S2 (∼3 ml/min, gravity flow, 25°C). Bath temperature was measured and controlled using the temperature controller TC07 and the PTC Minibath Chamber IV Plus Upgrade kit (Luigs & Neumann). Neurons were visualized using a 5× (N Plan 5x/0.12) and 25× (HCX IRAPO L25x/0.95W) objective, a three-position magnification changer (0.35×, 1.25×, and 4.0×) and a cooled CCD camera (DFC360FX, Leica Microsystems). Patch pipettes (5–8 MΩ) were pulled from borosilicate glass capillaries (outer diameter, 1.50 mm; inner diameter, 0.86 mm; Science Products) on a PC-10 micropipette puller (Narishige Instruments), fire-polished (MF-830 Microforge, Narishige Instruments), and filled with pipette solution (S5–S7, depending on experimental design). Alexa Fluor 488 hydrazide (20 μm) and, in some recordings, biocytin [0.3% (w/v)] was routinely added to the pipette solution to enable on-line evaluation of cell morphology and post hoc 3D reconstruction of recorded neurons, respectively. Neither chemical showed an evident effect on mitral cell electrophysiology. An agar bridge (150 mm KCl) connected the reference electrode and bath solution. An EPC-10 USB amplifier controlled by Patchmaster 2.67 software (HEKA Elektronik) was used for data acquisition. We monitored and compensated pipette and membrane capacitance (Cmem) as well as series resistance. Only neurons exhibiting relatively low (<30 MΩ) and stable access resistances were used for analysis. Liquid junction potentials were calculated using JPCalcW software (Barry, 1994) and corrected on-line. Signals were low-pass filtered [analog three-pole and four-pole Bessel filters (−3 dB); adjusted to one-quarter to one-fifth of the sampling rate (10 kHz)]. If not stated otherwise, holding potential (Vhold) was −70 mV. All electrophysiological data were recorded at RT. Mitral cells were identified according to their location [residing in the external cellular layer between the AOB glomerular layer and the lateral olfactory tract (Larriva-Sahd, 2008)], soma size (large somata; average Cmem, 14.5 ± 0.4 pF), and dendritic morphology (multiple apical/primary dendrites that terminate as tufts in the glomerular layer). “Loose-patch” recordings were performed from intact mitral cell somata to prevent dialysis of intracellular components. Action potential-driven capacitive currents were recorded in loose-seal cell-attached configuration (seal resistance, 30–150 MΩ; pipettes filled with S1). Spikes were analyzed using Igor Pro functions (SpAcAn, written by Guillaume Dugué and Charly Rousseau) for detection and analysis of spontaneous events by a threshold detection algorithm. Passive membrane properties [i.e., input resistance (Rinput), Cmem, and membrane time constant (τmem)] were obtained immediately after membrane rupture. Treated, to a first approximation, as a “biological constant” with a value of ∼1 μF/cm2 (Gentet et al., 2000), Cmem was determined using a square-pulse (5 mV, 10 ms) routine. Rinput at the mitral cell soma was determined by measuring the steady-state voltage response to a hyperpolarizing current step of −20 pA. Linear passive voltage responses were also used to estimate τmem from monoexponential fits to the voltage responses (from onset to steady state). General action potential parameters were calculated from averaged spike waveforms. Spike amplitude was measured as the threshold-to-peak distance, spike duration was calculated as the full duration at half-maximum (FDHM), and spike-generating kinetics were measured as the time-to-peak. All electrophysiological in vitro experiments used slices from young adults of either sex. We did not observe obvious gender-dependent differences.
Modeling.
The model design and specification was set up in neuroConstruct software (Gleeson et al., 2007). neuroConstruct facilitates the building and validation of models conforming to the NeuroML specification (www.neuroml.org). Model scripts were generated at multiple stages during the development process by neuroConstruct to run simulations using the Neuron simulator software (Hines and Carnevale, 1997, 2001). Simulator software contains solvers that run all the differential equations that are spatially distributed within the morphological framework of the model at a specified time step. It also allows the output of these differential equations to be recorded at specified locations and for specified parameters.
For model development, the digitized 3D reconstruction of a representative intrinsically oscillating AOB mitral cell was chosen. Morphology parameters were saved as a .hoc file and imported into neuroConstruct. Compartments were separated into groups (soma, dendrites, axon) for specifying channel allocation.
The model's specific passive parameters were tuned to target mean experimental measurements/estimates, such as somatic Rinput, τmem, and Cmem. Based on simulated somatic current injections (70 pA, 400 ms), Rinput was calculated as follows: ΔVmem/Iinject, with Vmem defined as cell membrane potential and Iinject defined as injection current. τmem Was estimated by monoexponential fits using the Octave expfit function (http://octave.sourceforge.net/). Then, a simulated voltage-clamp test pulse (5 ms, 5 mV, 20 MΩ) was applied to the soma and Cmem was estimated by fitting an exponential to the capacitive transient (Octave expfit). Iterations were repeated until passive model parameters replicated experimental values. The passive parameters were as follows: Cmem = 0.79 μF cm−2; Rmem = 59,524 Ω μm−2; and Rinput = 560 Ωcm.
New models for outward-rectifying K+-channel current (IKdr), transient K+-channel current (IKA), persistent Na+ current (INaP), and voltage-gated Ca2+ current (ICaV) were developed from current–voltage (I–V) plots, steady-state plots, and time-course datasets for activation, inactivation, and deactivation measured in voltage-clamp experiments. These models were used in combination with previously published models for transient Na+ current (INaT; Migliore et al., 2005) and big conductance KCa (BK) current (IBK; Maex and De Schutter, 1998). Ionic reversal potentials were (in millivolts) as follows: Eleak = −74.1, ENa = 67, EK = −86.5, and ECa = 80. All equations and channel densities (Gmax) for each model are found at https://github.com/Simon-at-Ely/Channel_Kinetics/blob/master/AccessoryOlfactoryBulb/MitralCell/ipython/MultipleChannelModel/iAMT_model_specification.ipynb (any tuning adjustments have been highlighted in red or blue).
Data analysis.
All in vitro data were obtained from independent experiments performed on ≥3 d using ≥3 different animals. Individual numbers of cells/experiments (n) are denoted in the figure and/or captions. If not stated otherwise, results are presented as means ± SEM. Statistical analyses were performed using paired or unpaired t tests, one-way ANOVA with Tukey's HSD post hoc test, Mann–Whitney U tests, Wilcoxon signed-rank tests, or Kruskal–Wallis tests (as dictated by data distribution and experimental design). Tests and corresponding p values that report statistical significance (≤0.05) are individually specified in the figure legends. Data were analyzed off-line using FitMaster 2.67 (HEKA Elektronik), Imaris 8.0 (Bitplane), IGOR Pro 6.3 (WaveMetrics), Matlab (Mathworks), and Excel 2013 (15.0.4779.1001, Microsoft) software. Activation curves were fitted by the Boltzmann equation to calculate the Vmem of half-maximal activation (V1/2). Time constants (τ) were calculated by fitting individual traces to monoexponential functions I(t) = I1 [exp (−t/τ)] + I0. Linear correlation of two parameters was analyzed by calculating the Pearson correlation coefficient using IGOR Pro's linear correlation procedure. Data fitting for model development was carried out using IPython Notebook (ipython.org/notebook.html) and a least-square fitting method.
To account for the variability in single-unit firing rates obtained from in vivo recordings, unit classification required statistical analysis. Bursting activity was defined statistically as follows. First, for each unit, an expected ISI distribution assuming random independent (Poisson) spiking was derived. Poisson distribution is defined by a single parameter λ, which corresponds to that unit's observed average firing rate, calculated over the entire recording period. From this distribution, the median ISI value for each unit (i.e., intervals smaller or larger than this interval were equally likely) was derived. A burst was defined as a sequence of ≥4 consecutive spikes separated by intervals smaller than the median ISI (Fig. 1Ei–Hi, red horizontal bars). A unit was classified as “nonbursting” if <50% of all spikes occurred in bursts. By contrast, a unit was classified as “bursting” if (1) >50% of all spikes occurred in bursts, and (2) that unit's ISI distribution was significantly different (p < 0.05; two-sample Kolmogorov–Smirnov test) from the expected Poisson distribution under the null hypothesis of random firing at the same mean firing rate. For additional classification of (ir)regularity, spike time autocorrelation histograms (ACHs) were calculated for each unit across multiple time periods: 5, 10, 20, 30, 60, 90, and 120 s. Each period was divided into 200 equal bins (5 s ACH ≙ 25 ms bins, 120 s ACH ≙ 600 ms bins). Based on each unit's autocorrelograms, we determined ACH-specific trough indices. The trough index (TI) was defined as follows: TI =
In addition to each unit's classification as either nonbursting, irregular bursting, or regular bursting (i.e., oscillating), oscillatory firing patterns were further characterized by various statistics. These included the number and duration of bursts as well as the number and frequency of spikes within bursts.
Results
Slow oscillatory bursting of AOB neurons in vivo
Brain circuit computations are not only determined by sensory input, but also by intrinsically generated spatiotemporally structured patterns of spontaneous activity (Romano et al., 2015), both at the single-neuron and network levels. As information about spontaneous AOB activity is lacking, we performed in vivo recordings from the AOB mitral cell layer of anesthetized mice (Fig. 1A–C). We continuously monitored spontaneous multiunit activity of AOB neurons and validated putative single units by feature-based clustering (Fig. 1C). Inspection of spike trains over time revealed large firing-pattern heterogeneity in AOB neurons, ranging from apparently random (Fig. 1Ci) to more regular firing (Fig. 1Ciii).
Based on these initial findings, we next quantified and classified the observed patterns (Fig. 1D–H). Consideration of multiple measures, including deviation from Poissonian spiking, burst-firing parameters, autocorrelation analysis, and each unit's TI (see Materials and Methods), revealed two different discharge categories: 339 of 466 (72.8%) units were classified as irregular nonbursting (Fig. 1D), whereas 127 of 466 (27.2%) units were classified as bursting (Fig. 1E–H). Among the bursting subpopulation of AOB neurons, 49 (10.5%) displayed irregular burst firing characterized by an exponential decay of the ISI histogram and the lack of pronounced autocorrelogram “side” peaks. These units were classified as irregular bursting neurons (Fig. 1E). By contrast, 78 (16.7%) bursting neurons showed highly rhythmic oscillatory discharge with prominent ISI and ACH peaks, corresponding to large TI values (see Materials and Methods). These neurons were classified as oscillating (Fig. 1F–H). Within this subpopulation of oscillating AOB neurons, individual discharge patterns were heterogeneous and, accordingly, interburst intervals (IBIs), burst durations, and firing rates were highly variable (Fig. 1I). IBIs ranged from 1.4 to 112.2 s and bursts lasted for ≤15.0 s. The non-normal distribution of these values indicates the absence of a single characteristic rhythm governing AOB unit oscillations. In general, the activity of oscillating neurons was markedly increased compared with previously observed low baseline rates (1–2 Hz) of irregularly firing neurons (Luo et al., 2003; Ben-Shaul et al., 2010). To identify whether the characteristics of oscillating units define a distinct subpopulation of AOB neurons or they instead represent one extreme of a continuous and normally distributed dataset, we plotted histograms of burstiness (Fig. 1Ji), deviation from Poissonian spiking (Fig. 1Jii), and regularity (Fig. 1Jiii) for the entire sample population (n = 466). Our analyses indicate that oscillating AOB neurons indeed form a distinct group. Together, these results demonstrate that rhythmic patterns of slow oscillatory activity characterize a subpopulation of AOB neurons in vivo.
A subpopulation of AOB mitral cells displays slow spontaneous oscillations
AOB neuron oscillatory discharge may result from interactions within an intact network and potential top-down modulation or, alternatively, from intrinsic rhythmogenic properties of individual neurons. To resolve the mechanistic basis of mitral cell rhythmicity and avoid the experimental drawbacks that complicate the biophysical interpretation of in vivo whole-cell patch-clamp data (Maier et al., 2011), we turned to an in vitro model. We recorded spontaneous activity from individual AOB mitral cells in sagittal sections of the mouse olfactory bulb (Fig. 2A). When continuously monitoring mitral cell Vmem for prolonged periods of time under control conditions (0 pA current injection; S2, S5; see Materials and Methods), the vast majority of neurons generated spontaneous discharge [308 of 324 cells; resting Vmem (Vrest) = −73.2 ± 0.6 mV; firing rate, 2.9 ± 0.4 Hz; means ± SEM]. Similar to our in vivo findings, AOB mitral cells displayed one of two distinct activity patterns (Fig. 2B): either irregular firing with no apparent periodicity (irregular discharge; 105 of 308 cells; Vrest = −71.9 ± 0.7 mV; Fig. 2Bi) or “phasic” firing patterns with alternating periods of activity and silence (oscillatory discharge; 203 of 308 cells; Vrest = −74.1 ± 0.8 mV; Fig. 2Bii). These slow oscillations of recurring up and down states (−63.9 ± 2.5 vs −74.5 ± 3.7 mV, mean ± SD) usually remained stable throughout the recording (≤60 min). Typically, bursts of action potentials were superimposed on the slow depolarizing envelope (burst duration, 6.0 ± 5.3 s; IBI, 11.3 ± 7.6 s; within-burst firing rate, 4.2 ± 2.3 Hz, mean ± SD). Essentially identical patterns of spontaneous activity were observed when we recorded mitral cell activity in “loose-seal” cell-attached configuration (n = 39) to prevent dialysis of cytosolic components and maintain unperturbed Vrest (Fig. 2Biii). As a measure of regularity and bursting behavior (Moore et al., 1966), we inspected the ISI distribution and the ACHs (Fig. 2C–F). Figure 2C exemplifies an irregular nonbursting neuron, whereas oscillatory mitral cells are depicted in Figure 2D–F. While discharge periodicity for individual cells was strong, patterns were heterogeneous across the population, revealing no predominant rhythm. We then asked whether these two patterns of spontaneous activity correlate with distinct morphological phenotypes (Fig. 2G). Volume-rendered 3D reconstructions of individual biocytin-filled mitral cells (Fig. 2Gi,Gii) revealed large polymorphic somata and several branched primary dendrites that terminate as multiple tufts within the homonymous AOB half (Wagner et al., 2006). However, morphometric analysis documented no obvious morphological differences between irregularly firing and oscillating neurons (Fig. 2Giii). Together, these data demonstrate that, both under in vivo and in vitro conditions, infraslow oscillatory discharge represents the default activity pattern of a substantial population of AOB mitral cells.
A group of AOB mitral cells are intrinsically rhythmogenic
Neural rhythmogenesis is either a consequence of network activity and thus results from regular recurrence of excitatory and inhibitory synaptic barrages (Crunelli and Hughes, 2010) or, alternatively, oscillations are generated intrinsically by the repetitive, rhythmic discharge of pacemaker-like neurons (Blankenship and Feller, 2010). To distinguish between these mechanisms, we tested whether spontaneous oscillations in AOB mitral cells depend on synaptic drive. Pharmacological inhibition of GABAergic synaptic transmission did not qualitatively affect rhythmic discharge (Fig. 3A). Furthermore, when both GABAergic and glutamatergic fast synaptic transmission were blocked (Fig. 3B,C), stable oscillations persisted in a substantial fraction of AOB mitral cells (43 of 185 cells; 23.2%). While selective block of inhibitory input altered some oscillation parameters (Fig. 3D), complete isolation from fast synaptic transmission did not change oscillatory patterns in those AOB neurons that proved resistant to pharmacological treatment (Fig. 3E). Thus, we next examined whether these neurons shared another hallmark of pacemakers, i.e., a positive causal correlation between oscillation frequency and “baseline” Vmem (Crunelli and Hughes, 2010). Similar to intrinsic oscillators described in other circuits (Hayar et al., 2004; Blethyn et al., 2006; Tazerart et al., 2008), oscillation frequency changed as a function of depolarizing or hyperpolarizing current injection in synaptically isolated oscillating AOB mitral cells (Fig. 3F–H). Hyperpolarization increased, whereas depolarization reduced IBIs (Fig. 3F,Hi). Moreover, each neuron (nine of nine cells) exhibited a characteristic Vmem threshold below which the pattern of periodically recurring up and down states switched to a stable resting state (Fig. 3F,G). Notably, within the voltage range from rest to firing threshold, neither positive nor negative current injections changed a given neuron's general discharge type (i.e., from oscillatory to irregular or vice versa; n = 10; Fig. 3Fii). Together, these results suggest that the mouse AOB contains a group of intrinsically rhythmogenic mitral cells that generate slow Vmem oscillations independent of fast synaptic input.
Pattern variability among intrinsically oscillating mitral cells
Since pacemaker-like neurons can exert profound effects on coding and computation in sensory systems (Izhikevich et al., 2003; Mizuseki et al., 2009), we focused on the subpopulation of intrinsically oscillating AOB mitral cells (iAMCs). Accordingly, we performed subsequent experiments under tonic synaptic inhibition (gabazine, AP5, NBQX). First, we asked whether iAMCs and irregularly discharging neurons differ in their passive membrane properties and/or their spike-generation machinery. Neither Cmem nor τmem differed significantly (Fig. 4Ai,Aii), suggesting that basic biophysical properties do not distinguish iAMCs from the “general” AOB mitral cell population. Rinput, however, was slightly increased in iAMCs (Fig. 4Aiii). Comparison of mean instantaneous spike frequencies as a function of stationary current input (f–I curve; 5–100 pA; Fig. 4Aiv) revealed indistinguishable curves for both neuron types with response saturation at amplitudes >80 pA and maximum average firing frequencies of ∼25 Hz. Moreover, averaged spike waveforms from iAMCs and irregularly firing neurons were similar in spike amplitude, duration, and kinetics (Fig. 4Av). Together with the results shown in Figure 2G, these data demonstrate that the unique spontaneous activity of iAMCs does not correspond to a readily distinguishable morphological or biophysical phenotype.
Next, we analyzed discharge variability among the iAMC population. Similar to in vivo observations (Fig. 1I), measures including IBIs, burst durations, and firing rates within a burst varied considerably across neurons and were not normally distributed (Fig. 4B,C). Silent periods between bursts lasted on average 10.7 ± 8.1 s (±SD) with individual IBI values ranging from 1.5 to 49.3 s (Fig. 4B–F). Moreover, we observed bursts as brief as 0.3 s, but also prolonged firing periods of ≤22.2 s (average burst duration, 5.4 ± 3.9 s; mean ± SD; Fig. 4B–F). Increasing the bath solution's temperature from 25°C (RT) to 37°C decreased both IBIs and burst duration (n = 3; data not shown). Distribution analysis of both IBIs and burst durations, however, revealed no obviously segregated subpopulations among iAMCs (Fig. 4G). The vast majority of iAMCs (99.4%) fire >5 spikes per burst, translating into an average within-burst firing rate of 4.6 ± 3.0 Hz (mean ± SD; range, 0.9–18.6 Hz). Notably, the average voltage change during transitions from down to up states is 11.5 mV (−76.1 ± 3.9 to −64.6 ± 4.0 mV; means ± SD), indicating that any current(s) providing the excitatory drive to iAMCs must operate within this voltage regime. To examine whether oscillation parameters are interdependent, we carried out pairwise correlation analyses of IBIs, burst durations, and up-state voltage (Vu), respectively (Fig. 4H). Our results, however, indicate no obvious correlation between any parameter pair, suggesting that the antagonistic mechanisms underlying the recurring transition and maintenance of depolarization and hyperpolarization are phenotypically independent.
Together, these findings show that iAMCs are a physiologically distinct subgroup of AOB mitral cells that share pacemaker-like discharge properties, spanning a wide range of mechanistically independent characteristics within the oscillation parameter space.
INaP promotes rhythmogenesis
The cyclical interplay of depolarizing and hyperpolarizing pacemaker currents drives autorhythmicity and burst generation in a variety of neurons (Grillner, 2006; Blankenship and Feller, 2010). Conceptually, these currents provide either the transitional excitatory drive from down to up state, the translation into a regular firing pattern during the up-state plateau, or the hyperpolarization that terminates the burst (Colwell, 2011). The hyperpolarization-activated current (Ih), the low-voltage-activated T-type Ca2+ current (IT) and INaP are prototypical depolarizing pacemaker currents and, thus, major determinants of autorhythmicity (Crill, 1996; Perez-Reyes et al., 1998; Chan et al., 2004). Ih, in particular, is a predominant driving force of rhythmic oscillatory activity (Maccaferri and McBain, 1996; Liu and Shipley, 2008). Therefore, we tested whether Ih is involved in oscillatory iAMC discharge. The bradycardic agent ZD7288, a specific but isoform-independent blocker of hyperpolarization-activated cyclic nucleotide-gated channels (BoSmith et al., 1993), did not alter iAMC rhythmic activity (Fig. 5A,B). Ih expression typically manifests as a rebound depolarization (sag) at hyperpolarized potentials (Cichy et al., 2015). In current-clamp recordings, we thus measured the hyperpolarization-evoked sag potential in iAMCs under control conditions (Fig. 5C). Plotting the sag potential amplitude as a function of peak hyperpolarization (Fig. 5D) revealed that pronounced voltage sags were only detected upon membrane hyperpolarization to values substantially more negative than the average down-state potential (Vd). Thus, these results indicate that Ih serves a minor, if any, role in iAMC rhythmogenesis.
Another common subthreshold current that is frequently involved in Vmem bistability and burst generation is IT (Bean, 2007; Crunelli and Hughes, 2010). Especially in neurons characterized by relatively hyperpolarized Vrest, these low-threshold currents frequently accelerate the down-state-to-up-state transition (Llinás and Yarom, 1986) by providing additional subthreshold depolarizing drive. Thus, we asked whether iAMCs express low-threshold T-type Ca2+ channels. In fast voltage ramp recordings, Cd2+-sensitive Ca2+ currents first manifested at suprathreshold Vmem (>−40 mV) and I–V curves did not display a characteristic “dent” at subthreshold voltages (Fig. 5E). Moreover, Ca2+ currents were essentially insensitive to mibefradil (Fig. 5F), a drug that preferentially inhibits T-type Ca2+ channels (Martin et al., 2000). Consistent with this observation, rhythmic discharge patterns of oscillating iAMCs remained essentially unchanged upon exposure to mibefradil (Fig. 5G). Thus, our data strongly suggest that iAMCs do not display substantial IT.
As a regenerative depolarizing current between Vrest and spike threshold, TTX-sensitive INaP affects both firing pattern and frequency (Crill, 1996; Tazerart et al., 2008). Thus, we next examined whether TTX, in addition to disrupting action potential discharge, would also affect subthreshold Vmem oscillations. Indeed, incubation with TTX abolished Vmem fluctuations (Fig. 6Ai). In presence of the toxin, iAMCs exhibited stable Vmem values that were statistically indistinguishable from Vd (Fig. 6Aii). Since selective pharmacological block of INaP has proven difficult (Ramirez et al., 2011; Richter et al., 2014), we evoked and isolated TTX-sensitive steady-state currents by slow depolarizing voltage ramps (Carter et al., 2012). A TTX-sensitive current was first evident at ∼−75 mV and increased steeply with voltage (Fig. 6Aiii). As known for INaP (Bean, 2007), maximum steady-state currents were only a small fraction of the maximum INaT (2.2%; n = 7; data not shown). At subthreshold voltages, however, such currents of only tens of picoamperes may prove functionally significant. If so, average Vu and Vd values must be within the voltage-operating range of INaP, which is indeed the case (Fig. 6Aiv).
A shift in voltage dependence caused by changes in extracellular Ca2+ concentration ([Ca2+]ex) is another hallmark of INaP (Su et al., 2001). Thus, we tested whether activation of TTX-sensitive steady-state currents was altered by changes in [Ca2+]ex (Fig. 6Aiv,Av). Likely as a result of surface charge screening (Hille, 2001), increasing [Ca2+]ex indeed shifted INaP activation to more depolarized values. When converted to conductances and fit by a Boltzmann function (Fig. 6Aiv), INaP activation (calculated as 5% conductance) varied as a function of [Ca2+]ex, ranging from −65.2 ± 2.2 mV (5 mm Ca2+) to −80.2 ± 0.4 mV (0.1 mm Ca2+; Fig. 6Av). Accordingly, if INaP is a major determinant of iAMC autorhythmicity, changes in [Ca2+]ex should alter oscillation patterns. Indeed, IBIs progressively decreased with lower [Ca2+]ex levels (Fig. 6Bi–Biii). By contrast, elevated concentrations (5 mm Ca2+) resulted in cessation of oscillations and stable Vmem values (Fig. 6Bi), an effect that was reversible upon restoration of physiological conditions (1 mm Ca2+; Fig. 6Bii). Together, we conclude that TTX-sensitive INaP is the major excitatory element that drives Vd–Vu transitions in iAMCs.
R-type Ca2+ channels determine iAMC oscillation patterns
During the up state, distinct ionic conductances translate the subthreshold excitatory drive into regular trains of action potentials (Colwell, 2011). Thus, we next investigated the mechanistic basis of burst firing during the depolarizing envelope of iAMCs. While the predominant role of INaT in AOB mitral cell spike generation is well established (Castro et al., 2007; Smith and Araneda, 2010; Shpak et al., 2012), voltage-gated Ca2+ (CaV) channels additionally shape various neuronal discharge parameters (Bean, 2007). Using a pharmacological approach, we first characterized the repertoire of high voltage-activated Ca2+ currents expressed in iAMCs. L-type, P-type/Q-type, and N-type Ca2+ currents were isolated using selective dihydropyridine or peptide toxin CaV-channel antagonists (McCleskey et al., 1987; Bossert and Vater, 1989; Adams et al., 1993). Notably, iAMC autorhythmicity persisted essentially unaltered in the presence of either CaV-channel blocker (data not shown). By contrast, addition of SNX-482, a selective R-type Ca2+ channel inhibitor (Bourinet et al., 2001), dramatically altered iAMC oscillations (Fig. 7A,B). While within-burst firing rates remained unchanged, inhibition of R-type Ca2+ currents (IR) significantly decreased IBIs and prolonged burst duration (Fig. 7Aii). These results indicate that, while iAMCs express at least one member of each high voltage-activated Ca2+ channel subfamily, only R-type/CaV2.3 channels play a significant role in orchestrating autonomous iAMC oscillations.
In iAMCs, Vu is below the activation threshold of SNX-482-sensitive IR (>−50 mV; Fig. 7B). Therefore, action potentials must provide the depolarization required for IR activation during the up state. Accordingly, we next examined whether action potential amplitude or shape changed upon IR inhibition (Fig. 7C). Rise time analysis revealed that SNX-482-sensitive currents contribute little to the action potential rising phase (Fig. 7Cii). By contrast, SNX-482 significantly reduced the amplitude and prolonged the FDHM of average iAMC action potentials (Fig. 7Cii). These data indicate that CaV2.3 channels activate near the action potential peak and that IR is largest during the falling phase when the driving force on Ca2+ increases.
Big conductance K+ channel-dependent negative feedback regulates burst duration
A prolonged action potential FDHM as a result of blocking a Ca2+ inward current (Fig. 7Cii) is somewhat counterintuitive. Apparently, the net effect of blocking Ca2+ entry is to inhibit a net outward current. Ca2+-activated K+ (KCa) channels are likely mediators of such effects (Bean, 2007). Therefore, we next asked whether Ca2+-activated K+ conductances are involved in iAMC oscillations.
Underlying the afterhyperpolarization that follows single spikes or bursts, for example in the hippocampus (Alger and Nicoll, 1980), small conductance KCa (SK) channels are major determinants of firing rate. Thus, we first investigated whether the selective SK channel antagonist apamin (Blatz and Magleby, 1986) affects rhythmic iAMC discharge. Surprisingly, SK channel block did not affect iAMC oscillations (Fig. 8A), consistent with the lack of a substantial apamin-sensitive KCa current in voltage-clamp experiments (Fig. 8B).
Other candidates for mediating a Ca2+-activated K+ efflux during discharge are big conductance KCa (BK) channels. BK channels, which are cooperatively activated by depolarization and increased cytoplasmic Ca2+ (Fakler and Adelman, 2008), are inhibited by both TEA and the selective organic blocker paxilline (Brenner et al., 2005). In iAMCs, relatively low TEA concentrations (1 mm) reversibly switched oscillatory discharge to irregular tonic firing (Fig. 8C), suggesting that coupling of Ca2+ entry to activation and “build up” of IBK might cause burst termination and, consequently, the following transition to Vd (Crunelli et al., 2012). To quantify the relative contributions of IBK and ICaV to pacemaking, we then used an iAMC's own representative burst firing pattern as a voltage command and recorded both TEA-sensitive and Cd2+-sensitive currents (Fig. 8Di). When single spike-dependent currents were plotted as a function of event number (Fig. 8Dii–Div), both putative IBK and ICaV monotonically increased and gradually reached saturation after ∼15–20 events. Thus, while the recorded template showed spike-amplitude adaptation, contributions of BK and CaV channels were largest in later stages of up-state bursting. Moreover, TEA-sensitive and Cd2+-sensitive charge transfer was strongly correlated (Fig. 8Dv). Together, these data suggest that, during prolonged firing, a gradual increase in cytoplasmic Ca2+ is coupled to progressive activation of BK channels, which ultimately causes burst termination. If so, selective inhibition of IBK should increase up-state duration and reduce IBIs. Low micromolar concentrations of paxilline indeed caused corresponding changes in iAMC rhythmogenesis (Fig. 8E), supporting the notion that BK channels play a major role in setting iAMC oscillation frequency.
A “triangle” of interdependent distinct conductances is sufficient to drive autorhythmicity
The above results indicate that mitral cell autorhythmicity emerges from coordinated reciprocal interaction of INaP, IR, and IBK. While the molecular correlate of INaP is debated and subthreshold persistent currents might even originate from the same channels that carry suprathreshold transient currents (Carter et al., 2012), the CaV2.3 channel and the BK channel α-1 subunit Sloα1 have been shown to mediate IR and IBK, respectively. Indeed, antibodies raised against CaV2.3 and Sloα1 label a large population of cells in the AOB mitral cell layer (Fig. 9A,B). Notably, the pharmacological profile of IBK in AOB mitral cells—sensitivity to TEA and paxilline, but resistance to iberiotoxin and charybdotoxin (data not shown)—suggests coexpression of an auxiliary β4 subunit (Sloβ4) that confers type-II BK channel properties (Brenner et al., 2005), such as relatively slow gating kinetics. Immunolabeling against Sloβ4 confirmed its specific expression in AOB mitral cells (Fig. 9C). Moreover, both Sloα1 and Sloβ4 were expressed in neurons identified as iAMCs in prior electrophysiological recordings (Fig. 9D,E).
Are INaP, IR, and IBK unique properties of rhythmogenic AOB mitral cells or are these conductances broadly expressed in both iAMCs and nonrhythmogenic neurons? To address this question, we compared electrophysiological profiles of INaP, IR, and IBK in both mitral cell populations (Fig. 10A–G). All three currents are found in both iAMCs and irregularly firing mitral cells. However, with the exception of SNX-482-sensitive IR (Fig. 10D,E), properties of rhythmogenic currents markedly differed between both neuronal populations. The INaP activation threshold was significantly lower in iAMCs than in nonrhythmogenic neurons (Fig. 10A,Bi). In addition to the resulting left shift in the I–V relationship (Fig. 10C), maximum INaP amplitudes—calculated from sigmoidal fits to the individual I–V curves—were significantly increased in iAMCs (Fig. 10Bii). TEA-sensitive K+ currents also differed between oscillating and irregularly firing mitral cells (Fig. 10F,G). Larger IBK amplitudes are recorded from iAMCs and steady-state activation curves in oscillating neurons are considerably shifted toward lower Vmem values (Fig. 10G). These results show that, while not exclusively expressed in the intrinsically oscillating mitral cell population, INaP, IR, and IBK confer a distinct electrophysiological phenotype on iAMCs.
We next asked whether the interplay of INaP, IR, and IBK is sufficient to drive autorhythmicity in a model AOB mitral cell. Based on the volume-rendered 3D reconstruction of a representative iAMC (Fig. 10Hi), Vmem simulations were run using the Neuron simulation environment (Hines and Carnevale, 1997). In addition to INaP, IR, and IBK, the model neuron also expressed IKdr and IKA (based on our own voltage-clamp experiments; see Materials and Methods), as well as INaT (Migliore et al., 2005). Strikingly, model-based simulations qualitatively reproduced the experimental data, with relatively small changes in Kdr channel density accounting for much of the experimentally observed heterogeneity in both IBIs and burst duration (Fig. 10Hii–Hiii). Together these data demonstrate that iAMCs express a “triangle” of conductances with distinct biophysical properties—INaP, IR, and IBK —that are sufficient to generate autorhythmicity in mouse AOB mitral cells.
Discussion
The accessory olfactory system is central to social information processing in various mammalian species. However, many basic physiological principles underlying sensory processing in the AOB remain poorly understood (Dulac and Wagner, 2006). Here, we investigate spontaneous activity in AOB mitral cells, the sole output/projection neurons of the AOB. We report that a group of AOB mitral cells displays slow stereotypical rhythmic discharge both in vitro and in vivo. Among these neurons, a subpopulation is intrinsically rhythmogenic. In these pacemaker-like cells, reciprocal interactions between depolarizing and hyperpolarizing conductances generate stable patterns of Vmem oscillations within a broad spectral window.
The prevalence of slow oscillatory discharge was larger in AOB slices than observed in vivo. While we recorded rhythmic bursting in 16.7% of all units in vivo, 69.7% of AOB mitral cells displayed periodic discharge in vitro. Several factors could account for this apparent discrepancy. First, AOB slices are isolated from both peripheral sensory input and top-down modulation. When connectivity is intact, either factor could add substantial “noise” to a given mitral cell's output. Second, experimental in vitro conditions might favor oscillatory discharge. While we performed loose-seal extracellular recordings in parallel to whole-cell patch-clamp measurements to prevent dialysis of intracellular components, extracellular in vitro conditions might not exactly mirror endogenous ionic concentrations. For example, our data indicate that small differences in [Ca2+]ex could exert profound effects on rhythmogenesis. Regardless of the exact prevalence, slow/infraslow mitral cell rhythmicity will likely have considerable physiological impact on sensory processing along the accessory olfactory pathway.
At the lower end of the “bandwidth” scale (Buzsáki, 2006), infraslow oscillations with a periodicity of tens of seconds have been described in a variety of neurons and circuits (Hughes et al., 2002; Blethyn et al., 2006; Crunelli and Hughes, 2010; Buzsáki et al., 2013). In iAMCs, rhythmogenesis appears to be driven by cyclical interplay of ≥3 voltage-dependent and/or Ca2+-dependent conductances. Low threshold INaP functions as the major excitatory element that drives iAMC transition from Vd to Vu. R-type CaV channels play a significant role in oscillation maintenance and shape, while the resulting increase in cytoplasmic Ca2+ is coupled to progressive activation of BK channels which, in concert with slow voltage-dependent INaP inactivation (Jasinski et al., 2013), ultimately causes burst termination. Here, we observed that INaP becomes evident in iAMCs at voltages ≥−80 mV. This finding supports recent reports of INaP activation at more negative voltages than previously appreciated (Huang and Trussell, 2008; Carter et al., 2012). The INaP pacemaking “engine” can therefore be engaged or disengaged by small shifts in net current at the down-state Vmem range (Yamada-Hanff and Bean, 2013). In nonrhythmogenic neurons, the INaP activation threshold is shifted to ∼−75 mV, a Vmem value that is more positive than the down state in most iAMCs (median Vd = −76.4 mV). An interplay of R-type CaV and BK channels also supports relatively prolonged bursting. IR, which contributes to bursting in hippocampal (Metz et al., 2005) and thalamic reticular neurons (Zaman et al., 2011), is similar to IT but activates above threshold (Foehring et al., 2000). Importantly, among high-voltage-activated Ca2+ channels, only R-type channels do not associate with BK channels in Ca2+ nanodomains (Fakler and Adelman, 2008), indicating the requirement of a slower, more global increase in cytoplasmic Ca2+ to activate substantial IBK. Together with the pronounced left shift in the Vmem dependence of IBK activation that we observed in iAMCs, the specific IR–IBK combination expressed in oscillating neurons appears ideally suited to shape autorhythmicity.
Our results do not exclude a functional role of other conductances or electrogenic pumps (Krey et al., 2010; Zylbertal et al., 2015). While we could not demonstrate a significant contribution of Ih, IT, or L-type, P-type/Q-type, and N-type Ca2+ currents or SK channel-mediated K+ currents, the activity-dependent cytoplasmic build-up of Na+ and/or Ca2+ during the burst could additionally “recruit” other currents. Potential candidates include Na+-activated K+ currents (Hage and Salkoff, 2012) and Ca2+-activated nonselective cation currents (Peña et al., 2004; Blethyn et al., 2006; Shpak et al., 2012). The Na+-activated K+ current, in particular, could prove important by mitigating INaP-dependent depolarization, effectively decelerating the Vd-to-Vu transition. Indeed, the biophysical foundation of pacemaker-like activity (e.g., in central pattern generators) is often redundant, such that identical firing patterns can be achieved by multiple current combinations (Marder and Goaillard, 2006). Notwithstanding, when added to a model mitral cell that otherwise expresses “standard” Hodgkin–Huxley-type voltage-gated Na+ and K+ channels, the combination of INaP, IR, and IBK is sufficient to drive autorhythmicity.
To what extent intrinsically rhythmogenic mitral cells shape information processing and sensory coding in the AOB remains so far unexplored. Future experiments will thus have to address the impact of iAMC oscillatory discharge on the intact AOB network. The wide spectrum of discharge patterns, as well as the absence of a dominant oscillation frequency among iAMCs, argues against a distinct prevalent AOB rhythm. Instead, parallel pacemaker-like activity of phenotypically different iAMCs might generate synchronous discharge in several AOB microcircuits. By temporally linking AOB neurons into functional assemblies, synchronous firing could facilitate synaptic plasticity and input selection (Buzsáki and Draguhn, 2004). Rhythmic cycles between high and low postsynaptic excitability states would add a temporal dimension to a given circuit's sensory coding space (Schroeder and Lakatos, 2009). Similar to intrinsic theta frequency oscillations that entrain to the sniffing cycle in ET cells of the MOB (Hayar et al., 2004, 2005), operation of the vomeronasal pump (Meredith and O'Connell, 1979) could entrain iAMC oscillations. In the hamster vomeronasal organ, peristaltic vasoconstriction cycles of 0.2–0.5 Hz were recorded in vivo (Meredith, 1994). In addition to input entrainment, orchestration of AOB neurons into functional ensembles could ensure communication reliability and selectivity (Izhikevich et al., 2003) by controlling signal flow among anatomically connected networks. Notably, downstream processing modules for vomeronasal stimuli include several nuclei that mediate pulsatile neuroendocrine release by synchronized slow rhythmic bursting of, for example, gonadotropin-releasing hormone neurons (Chu et al., 2012) or vasopressin magnocellular neurosecretory cells (Brown, 2004).
In summary, we identify a subpopulation of AOB mitral cells that exhibit slow/infraslow spontaneous rhythmicity both in vivo and in vitro. Some of these projection neurons are intrinsically rhythmogenic. The mechanistic basis of mitral cell autorhythmicity is the coordinated interplay of a regenerative subthreshold Na+ current, a high voltage-activated R-type Ca2+ current, and a burst-terminating Ca2+-activated K+ current that is cooperatively activated by depolarization and cytoplasmic Ca2+ accumulation.
Footnotes
This work was supported by grants from the Volkswagen Foundation (I/83533; to M.S.), the Deutsche Forschungsgemeinschaft (SP724/8-1, to M.S.; MA1184/20-2, to R.M.), DFG priority program SPP 1392: Integrative Analysis of Olfaction, Marie Skłodowska-Curie Actions (PCIG10-GA-2011-303785, to Y.B.-S.), the Lady Davis Fellowship Trust (to A.K.), and the German-Israeli Foundation for Scientific Research and Development (1-1193-153.13/2012, to M.S. and Y.B.-S.). M.S. is a Lichtenberg Professor of the Volkswagen Foundation. We thank both Corinna H. Engelhardt and Susanne Lipartowski (RWTH Aachen University) for excellent technical assistance.
The authors declare no competing financial interests.
- Correspondence should be addressed to Marc Spehr, RWTH Aachen University, Institute for Biology II/Dept. Chemosensation, Worringerweg 3, D-52074 Aachen, Germany. m.spehr{at}sensorik.rwth-aachen.de