Abstract
Sensory perception emerges from the confluence of sensory inputs that encode the composition of external environment and top-down feedback that conveys information from higher brain centers. In olfaction, sensory input activity is initially processed in the olfactory bulb (OB), serving as the first central relay before being transferred to the olfactory cortex. In addition, the OB receives dense connectivity from feedback projections, so the OB has the capacity to implement a wide array of sensory neuronal computation. However, little is known about the impact and the regulation of this cortical feedback. Here, we describe a novel mechanism to gate glutamatergic feedback selectively from the anterior olfactory cortex (AOC) to the OB. Combining in vitro and in vivo electrophysiological recordings, optogenetics, and fiber-photometry-based calcium imaging applied to wild-type and conditional transgenic mice, we explore the functional consequences of circuit-specific GABA type-B receptor (GABABR) manipulation. We found that activation of presynaptic GABABRs specifically depresses synaptic transmission from the AOC to OB inhibitory interneurons, but spares direct excitation to principal neurons. As a consequence, feedforward inhibition of spontaneous and odor-evoked activity of principal neurons is diminished. We also show that tunable cortico-bulbar feedback is critical for generating beta, but not gamma, OB oscillations. Together, these results show that GABABRs on cortico-bulbar afferents gate excitatory transmission in a target-specific manner and thus shape how the OB integrates sensory inputs and top-down information.
SIGNIFICANCE STATEMENT The olfactory bulb (OB) receives top-down inputs from the olfactory cortex that produce direct excitation and feedforward inhibition onto mitral and tufted cells, the principal neurons. The functional role of this feedback and the mechanisms regulating the balance of feedback excitation and inhibition remain unknown. We found that GABAB receptors are expressed in cortico-bulbar axons that synapse on granule cells and receptor activation reduces the feedforward inhibition of spontaneous and odor-driven mitral and tufted cells' firing activity. In contrast, direct excitatory inputs to these principal neurons remain unchanged. This study demonstrates that activation of GABAB receptors biases the excitation/inhibition balance provided by cortical inputs to the OB, leading to profound effects on early stages of sensory information processing.
Introduction
Sensory systems use prior experience and expectation to interpret the outside world. The integration of external information requires combining the bottom-up flow of sensory information with top-down signals from higher brain areas. In the olfactory system, odorant information from sensory neurons are first integrated in the main olfactory bulb (OB) before broadcasting to the olfactory cortex through OB principal cells, the so-called mitral and tufted (M/T) cells (Poo and Isaacson, 2009; Franks et al., 2011). In turn, the olfactory cortex, and mainly the anterior piriform cortex and anterior olfactory nucleus (respectively APC and AON, collectively called the anterior olfactory cortex, AOC), projects back to the OB (Haberly and Price, 1978a, 1978b; Davis and Macrides, 1981). Cortico-bulbar projections mostly synapse with axonless OB interneuron granule cells (GCs) and, to a lesser extent, with M/T cells. This glutamatergic input onto GC proximal dendrites initiates action potentials and mediates feedforward inhibition onto M/T cells (Balu et al., 2007). In addition, GCs receive glutamatergic input from M/T cells on their apical dendrites, triggering locally reciprocal GABA release back onto M/T cells, thereby producing recurrent or lateral inhibition (Isaacson and Strowbridge, 1998).
In addition to the great number of glutamatergic cortico-bulbar inputs from the AOC, the OB also receives top-down inputs from neuromodulatory centers, including serotoninergic fibers from the raphe, noradrenergic fibers from the locus coeruleus, and cholinergic and GABAergic fibers from the basal forebrain (Matsutani and Yamamoto, 2008; Linster and Fontanini, 2014). Given the abundance and diversity of top-down inputs to the OB and their strong impact on OB functions (Shea et al., 2008; Petzold et al., 2009; Boyd et al., 2012; Ma and Luo, 2012; Markopoulos et al., 2012; Soria-Gómez et al., 2014), deciphering how these inputs are modulated is essential to understanding their physiological role and how they regulate the OB network.
In this study, we demonstrate that GABAB receptors (GABABRs), G-protein-coupled receptors of GABA, regulate specific cortico-bulbar excitatory synaptic transmission. Using conditional genetics to selectively knock out GABABR expression in the AOC, together with a combination of in vitro and in vivo electrophysiology, optogenetics, and fiber-photometry-based calcium imaging, we characterized the functional role of GABABR modulation at cortico-bulbar terminals. We show that presynaptic activation of GABABRs strongly depresses the AOC-to-GC synapse, resulting in diminished feedforward inhibition onto M/T cells' spontaneous and odor-evoked activity. However, the direct AOC-to-M/T cell excitation remains unchanged. In addition, activation of GABABRs also reduces OB spontaneous beta oscillations (15–40 Hz). Collectively, these data uncover a mechanism by which the cortical top-down influence to the OB can be refined precisely.
Materials and Methods
Animals
Adult wild-type (WT) C57BL/6RJ, GABAB(1)fl/fl (Haller et al., 2004), and Tbet-Cre (Haddad et al., 2013) male mice (maintained on a C57BL/6RJ background; 2–5 months old at the time of injection) were used in the study. This work was performed in compliance with the French application of the European Communities Council Directive of September 22, 2010 (2010/63/EEC) and approved by the local ethics committee (CETEA 89, project #01126.02).
Viral injection
Adeno-associated virus [AAV; capside serotype 2/9 for Channelrhodopsin-2 (ChR2), ChRimson, and Cre viruses, and 2/1 for GCaMP6f] were generated by the Penn Vector Core or produced by the Institut National de la Santé et de la Recherche Médicale (INSERM, UMR 1089, IRT1 Vector platform Nantes, www.atlantic-gene-therapies.fr) from ChR2- (K. Deisseroth; catalog #26969 and #20297; Addgene), ChRimson- (E. Boyden; catalog #62723; Addgene), Cre recombinase-, or GCaMP6f (Penn Vector Core)-encoding plasmids. For electrophysiology experiments, high-titer stock of AAV containing the CaMKIIa-hChR2(H134R)-eYFP-WPRE construct (viral titer, 9.4 × 1012 genome copies per milliliter, n = 14 mice for in vitro recordings, n = 15 for in vivo recordings) or a 1:6 mixture of an AAV containing CaMKII-Cre-WPRE (viral titer, 1.1 × 1014) and an AAV containing EF1a-DIO-hChR2(H134R)-mCherry-WPRE (viral titer, 1.4 × 1013) were injected in WT (n = 5) and age-matched GABAB(1)fl/fl mice (n = 4). A separate cohort of animals were also injected with an AAV containing hSyn-hChR2(H134R)-mCherry-WPRE [n = 3 for in vitro recordings, n = 5 for in vivo field EPSP (fEPSP) characterization solely]. No significant difference between using the Syn or CaMKIIa promoter was seen and results were pooled. For photometry experiments, high-titer stock of AAV containing the hSyn-ChRimson-TdTomato-WPRE (viral titer, 2.2 × 1013) or hSyn-DIO-GCaMP6f-WPRE construct (viral titer: 1.1 × 1013) were injected in Tbet-Cre mice (n = 15 odor-recording pairs from 2 mice). For viral injections, mice were deeply anesthetized with a ketamine and xylazine mixture (150 mg/kg Imalgene and 5 mg/kg Rompun, respectively, i.p.) and placed in a stereotaxic apparatus. A small craniotomy was performed and viral solution was injected into the AOC (stereotaxic coordinates: 2.1 mm anterior from bregma, 1.9 mm lateral, and at a depth of 3.3 and 3.7 mm from the brain surface; 150–200 nl/site, 300–400 nl total), allowing virus diffusion in the anterior APC and latero-posterior AON, or into the OB (AP: 5 mm, ML: 1.7 mm, DV: 0.7–1.5 mm, 300 nL total) through a glass micropipette attached to a Nanoinjector system (Nanoject II).
Electrophysiology
Slice recording.
Tissue preparation was performed as described in Valley et al. (2013). Briefly, tissue was dissected in artificial CSF (ACSF) and 300μm-thick slices were vibrosectioned. Recordings were made with borosilicate glass pipettes with a tip resistance between 3 and 6 MΩ. Recordings were discarded if the access resistance exceeded half the input resistance of the cell or if the access resistance varied by ∼30% during the experiment. Data were digitized at 10 kHz (EPC9double; HEKA). ChR2 stimulation used a 470 nm light-emitting diode array (Bridgelux). Light duration was controlled using a digital Sequencer (Master-8; A.M.P.I.) and all stimulations were given with an interstimulus interval of 20 s.
In vivo recording.
Awake recordings were performed as described previously (Lepousez and Lledo, 2013; Soria-Gómez et al., 2014). Mice were anesthetized and an L-shaped metal bar and a silver reference electrode were fixed to the caudal part of the skull. Optic fibers [multimode, 430 μm diameter, numerical aperture (NA) 0.39, Thorlabs] were bilaterally implanted above the anterior commissure (400 μm posterior to the sinus of the olfactory bulb, 0.9 mm lateral, and at a depth of 2.3 mm from the brain surface with an angle of ∼30°). After 1 week of recovery, mice were slowly and progressively trained for head restraint habituation and a 5% sucrose solution was given as a reward. The craniotomy was performed the day before recording and protected with silicone sealant (KwikCast). An array of 4 tungsten electrodes (∼3 MΩ; FHC) glued to one or two miniature cannulas (polymide tubing, 0.0035 inch, Neuralynx; positioned 100–200 μm above the electrode tips, connected to a 10 μl Hamilton syringe) was slowly lowered into the OB and a drop of silicone sealant was applied to the brain surface to increase recording stability. Both LFP and spiking signals were continuously recorded 40 min before and 60 min after local drug microinjection through the miniature cannula (injection speed: 0.05 μl/min; 0.15–0.3 μl total). Signals were preamplified (HS-18; Neuralynx), amplified (1000×; Lynx8, Neuralynx) and digitized at 20 kHz (Power 1401 A/D interface; CED). The identity of M/T cells units were established on the basis of several criteria: (1) stereotaxic coordinates of the mitral cell layer; (2) decrease in both gamma oscillation amplitude and light-evoked fEPSP in the mitral cell layer compared with the LFP recorded in the GC layer (GCL) or external plexiform layer (EPL), where the current sources/sinks are localized (Neville and Haberly, 2003); (3) increase in background spiking activity in a narrow band of 100–150 μm; (4) typical spontaneous activity patterns coarsely time locked to the respiration rhythm; and (5) odor-evoked responses. Light stimulation of AOC axons was performed using either an optic fiber placed on the OB brain surface or with implanted optic fibers coupled to a DPSS laser (473 nm, 150 mW; CNI Lasers; output fiber intensity, 20 mW) via a custom-built fiber launcher and controlled by a PS-H-LED laser driver connected to the CED interface. Light stimulation consisted in single, paired (40 Hz), or train stimulation (10–100 Hz) of 5-ms-long light pulses. The respiration signal was recorded using a thermocouple (0.005 inch Teflon-coated thermocouple, 5TC-TT-JI-40-1M; Omega) placed in front of the animal's nostril, amplified (10,000×), and band-pass filtered (0–10 Hz). The craniotomy was cleaned and covered with Kwik-Cast between sessions. Four recording sessions per mouse (2 per hemisphere) were made at least 1–2 d apart.
Characterization of the light-evoked fEPSP (see Fig. 2B,C) was performed in anesthetized mice. Animals were anesthetized using ketamine/xylazine and positioned in a stereotaxic frame. The animal's body temperature was maintained at 37.5°C by a heating pad and the respiration was monitored to control the anesthesia. LFP recordings were then performed as described above.
Odor presentation.
We used a custom-built flow-dilution olfactometer controlled by the CED interface. Pure monomolecular odorants (Sigma-Aldrich) were diluted in mineral oil (10%) in odorless glass vial. Saturated odor vapor was further diluted with humidified clean air (1:10) by means of computer-controlled solenoid pinch valves. Odor presentation dynamics were monitored and calibrated using a mini-photoionizator detector (mini-PID, Aurora). Cycles of odor, light, and odor + light presentations were repeated at least seven times for each condition. Stimuli were applied for 1 s and a given odorant was presented every 50 s to reduce sensory adaptation. The odorants used in the final dataset were as follows: valeraldehyde (n = 11 responses), acetophenone (n = 5), butyric acid (n = 3), 2-hexanone (n = 3), (S)-limonene (n = 3), ethyl tiglate (n = 2), ethyl butyrate (n = 1), ethyl valerate (n = 1), and 1,4 cineole (n = 1) and a binary (1:1) mixture of 1-pentanol and 1,4 cineole (n = 4), ethyl butyrate and ethyl valerate (n = 4), valeraldehyde and ethyl tiglate (n = 1), (S)-limonene and 2-hexanone (n = 1).
Calcium imaging using fiber photometry
A fiber photometry system adapted from Gunaydin et al. (2014) was used (see Fig. 7A). Immediately after GCaMP6f virus injection in the OB, optic fibers (multimode, 430 μm diameter, NA 0.48, LC zirconia ferrule) were implanted bilaterally in the dorsolateral part of the OB above the virus injection site. Three weeks after injection, GCaMP6f was excited continuously using a 473 nm DPSS laser (output fiber intensity, 0.4–0.5 mW; CNI Lasers) reflected on a dichroic mirror (452–490 nm/505–800 nm) and collimated into a 400 μm multimode optic fiber (NA 0.48) with a convergent lens (f = 30 mm). The emitted fluorescence was collected in the same fiber and transmitted by the dichroic mirror, filtered (525 ± 19 nm), and focused on a NewFocus 2151 femtowatt photoreceptor (Newport; DC mode). Reflected blue light along the light path was also measured with a second amplifying photodetector (PDA36A; Thorlabs) to monitor light excitation and fiber coupling. Signals from both photodetectors were digitized by a digital-to-analog converter (Power 1401; CED) at 5000 Hz and recorded using Spike2 software. For AOC stimulation using ChRimson, an optic fiber (multimode, 430 μm diameter, NA 0.39, with LC zirconia ferrule; Thorlabs; 5–10 mW output fiber intensity) were implanted bilaterally above the AOC and connected to a DPSS laser (589 nm, 200 mW; CNI Lasers) via a custom-built fiber launcher. For drug injection, bilateral acute intrabulbar injections were done through implanted guide cannulas (injection volume, 0.5 μl; speed, 0.1 μl/min via a 33-gauge cannula connected to a 10 μl Hamilton syringe). For odor presentation, mice were placed in a small, ventilated cage (∼0.5 L). Pure monomolecular odorants (Sigma-Aldrich) were diluted in mineral oil (1%) in an odorless glass vial and saturated odor vapor was delivered directly into the ventilated cage at a flow rate of 3 L/min. Odors were presented every 30 s and odor presentation dynamics in the cage were monitored constantly using a mini-PID (Aurora). The odorants used in the final dataset were as follows: valeraldehyde (n = 3), ethyl tiglate (n = 3), pentyl acetate (n = 3), ethyl valerate (n = 2), 2-hexanone (n = 1), ethyl butyrate (n = 1), linalool (n = 1), and pentanol (n = 1).
Pharmacology
Lidocaine (2-diethylamino-N-(2,6-dimethylphenyl)acetamide, 2% in vivo), NBQX (2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide; 1 mm in vivo), baclofen ((RS)-4-Amino-3-(4-chlorophenyl)butanoic acid; 250 μm in slice and 2.5 mm for in vivo experiments), and CGP 52432 ([3-[[(3,4-dichlorophenyl)methyl]amino]propyl] (diethoxymethyl)phosphinic acid; 10 μm for slice and 100 μm for in vivo recording) were obtained from Sigma-Aldrich or Tocris Bioscience and dissolved at a final concentration in either sterile saline for in vivo experiments or ACSF for slice experiments. Analyzing the changes in fEPSPs at different depths in the GCL after baclofen injections allowed an estimation of the drug diffusion to be <500–600 μm (see “Results” section).
Histology
For postrecording histological analysis of electrode positioning and ChR2 expression, animals were intracardially perfused [4% paraformaldehyde (PFA) in 0.1 m phosphate buffer] and the brains were removed and postfixed in the same fixative overnight. Sixty-micrometer-thick brain sections were cut on a vibratome, rinsed in PBS, counterstained with the nuclear dye 4,6-diamidino-2-phenylindole (DAPI), and mounted on slides. Viral expression at the injection site was confirmed and OB sections were inspected to check for proper axonal expression, absence of virus diffusion into the OB, and the absence of significant somatic labeling in the OB. To amplify the eYFP fluorescent signal, immunohistochemistry was performed with a chicken anti-GFP primary antibody (1:4000, 06-896, Millipore Bioscience) and rabbit anti-chicken secondary antibody conjugated to Alexa Fluor-488 (1:1000, 1-11039; Life Technologies). In some experiments, the position of the recording electrode was confirmed using a fluorescent DiI (Life Technologies).
GABABR1 immunohistochemistry was performed as described in Valley et al. (2013) with minor modifications. Live brain tissue sections were cut (300 μm), allowed to recover for 15 min in ACSF, and then quickly transfer to ice-cold 4% PFA for 30 min. Slices were then cryoprotected in 30% sucrose overnight and 12-μm-thick sections were cut using a cryostat the next day. Immunohistochemistry against GABABR1 was performed the same or following day. Slices were rinsed, blocked in normal goat serum for 2 h, and incubated in primary antibody (guinea pig anti-GABABR1, 1:3000, AB2256; Millipore Bioscience) for 48 h at 4°C. The secondary antibody (anti-guinea pig conjugated to A647, 1:1000; Life Technologies) was incubated for 2 h. Slices were then rinsed, counterstained with DAPI, mounted with Mowiol, and imaged with a confocal microscope (LSM 700; Zeiss). Quantification of the GABABR1 immunoreactivity was reported as the fluorescence optical density in an optical plane where GL staining was maximal using ImageJ software.
Data analysis
For light-evoked field potentials, a 10 min time window before and 10 min after drug injection was used to average evoked signals. For the fEPSP characterization, the steepest slope calculated in a 1 ms window was measured to avoid contamination by the fiber volley component. Similar results were found when measuring the slope between 20% and 80% of the descending phase of the peak. When discernable, the amplitude of the fiber volley was also measured.
For measurements of M/T cell-spiking activity, a minimal 10 min time window before and 10 min after drug injection were used for the analysis (up to 40 min before and after drug application). Signals were high-pass filtered (0.3–9 kHz) and spike detection, sorting, clustering, and spike waveform analysis were performed using Spike2 software (CED) followed by manual cluster adjustment. For single-unit validation, all sorted cells displaying >1% of their interspike intervals below a 3.5 ms refractory period were discarded from the analysis. Careful attention was taken to discard any unit that showed some significant change in spike amplitude or waveform caused by the local infusion of drugs.
We determined whether a cell receives significant inhibition or excitation by extracting individual trials and comparing the firing rate during light to the basal firing rate using a Wilcoxon matched-pairs rank-sum test. A 1 s time window was used to detect whether the cell receives significant inhibition after repeated light stimulation. A 15 ms sliding time window after light stimulation and a 5 ms sliding window, respectively, were used to detect inhibition and excitation elicited by a single light pulse.
The change in firing rate to repeated light stimulation (see Fig. 3) was calculated as follows: With the firing rate (stimulus, 1 s) being the M/T cell's firing rate during the stimulation (light, odor, or both) and the basal firing rate the averaged cell's firing rate during the second before the light stimulation.
In Figure 4B, the normalized firing rate in response to a single light pulse was calculated as follows: With the firing rate (light, 2 ms) being the M/T cell's firing rate over a 2 ms period during the light stimulation and the basal firing rate (2 ms) being the cell's firing rate during the 20 ms preceding light stimulation reported to a 2 ms time period.
For the analysis of direct excitation in Figure 4D, the M/T cell's firing activity was normalized as follows: With firing rate (light, 1 ms) being the M/T cell's firing rate over a 1 ms period during the light stimulation and basal firing rate (1 ms) being the cell's firing rate during the 20 ms preceding light stimulation reported to a 1 ms time period. To analyze the coupling between AOC stimuli and M/T cell firing (see Fig. 4E), light pulses from the same stimulation train were pooled and the same calculation as above was performed, with the firing rate (light, 10 ms) being the average number of spikes in a 10 ms time window starting 1 ms after light onset and the basal firing rate being the average number of spikes in the 10 ms time window directly preceding light stimulation.
For phase modulation analysis, the thermocouple signal was down-sampled (0.5 kHz) and filtered (0–10 Hz) to extract the sniffing signal. Oscillation peaks (exhalation end) were identified using an automatic threshold algorithm and a phase histogram (72 bins) of M/T cell spikes relative to the identified peak was computed to measure the phase preference and length of the normalized vector as a measure of modulation strength.
For spontaneous oscillations, signals were down-sampled (5 kHz) and low-pass filtered (0–300 Hz), and 10-min-long epochs excluding 1.5 s after onset of light stimulation (0.5 s after the end of the stimulation) were extracted and subjected to a fast Fourier transformation (Hanning window, 2.44 Hz resolution) to obtain the power spectrum and the spectral power in each frequency band of interest.
For photometry experiments, signals were smoothed (0.02 s window) and down-sampled to 500 Hz. For each trial, the signal was normalized to the averaged fluorescence of the trial using the ΔF/F ratio as follows: With F0 being the average fluorescence over the trial. Sessions with significant averaged changes in the reflected blue light (>1% ΔF/F) were discarded from the analysis.
Statistics
All reported variances are SEM. In all graphs, excluding the box-and-whiskers plots, the mean is represented. In box-and-whiskers plots, the line in the middle of the box represents the median, the box edges represent the 25th to 75th percentiles and the whiskers represent the minimum and the maximum. All two-tailed Wilcoxon signed-rank tests, Mann–Whitney tests, ANOVAs, and curve fit were performed using commercial analysis software (GraphPad Prism) with p < 0.05 considered significant. For circular data, Hotelling paired test (significance p < 0.05) was performed with Oriana (Kovac Computing Services).
Results
Activation of presynaptic GABABRs depresses cortical synaptic transmission onto granule cells
Because AOC projections to the OB predominantly innervate the GCL (Haberly and Price, 1978a, 1978b; Davis and Macrides, 1981) and, because previous immunohistochemical studies reported the presence of GABABR subunits in that layer (Margeta-Mitrovic et al., 1999), we investigated the presence and the functional role of GABABRs at AOC-to-OB synapses. We injected the AOC of adult mice with AAV to express ChR2 in cortico-bulbar axons (Fig. 1A–C). As reported previously, ChR2-eYFP expression was confined mainly to the GCL and found to a minor extent in the mitral cell and glomerular layers (see also Haberly and Price, 1978a, 1978b; Davis and Macrides, 1981; Boyd et al., 2012; Markopoulos et al., 2012). No labeled somas were seen across bulbar layers (Fig. 1B,D) as already reported (Lepousez et al., 2014). Light stimulation of ChR2+ axons in horizontal OB slices (Fig. 1E) evoked monosynaptic EPSCs in voltage-clamped GCs that were abolished upon GABABR agonist R/S-baclofen treatment (hereafter referred as baclofen, 250 μm) and subsequent application of GABABR antagonist CGP-52432 (hereafter referred to as CGP, 10 μm) partially restored the EPSC amplitude (−79.9 ± 3.8% in baclofen and −20.2 ± 15.2% in CGP, one-way ANOVA: p = 0.0009, Dunn's post hoc test: p < 0.05 for baseline vs baclofen and p < 0.01 for baclofen vs CGP; n = 8; Fig. 1F). Previous slice studies showed that baclofen application did not affect GC resting membrane potential, input resistance, or threshold to spike (Isaacson and Vitten, 2003; Valley et al., 2013), suggesting no direct postsynaptic action of baclofen onto GCs. In addition to direct excitatory inputs, cortico-bulbar axons light stimulation produced disynaptic inhibition onto GCs, presumably originating from deep short axon cells (Boyd et al., 2012; Markopoulos et al., 2012). NBQX-sensitive IPSCs were recorded in 8 of the 35 recorded GCs and these IPSCs were blocked by baclofen and restored with CGP application (−80.1 ± 4.3% in baclofen, −20.1 ± 9.7% in CGP and −89.5 ± 2.4%; in NBQX, one-way ANOVA: p = 0.0005, p < 0.01 for baseline vs baclofen and p < 0.05 for baseline vs NBQX and baclofen vs CGP; n = 7; Fig. 1G).
We next evaluated the in vivo functional impact of GABABR modulation on the OB circuit activity. We recorded LFP in vivo and induced light stimulation of ChR2+ axon terminals with an optic fiber positioned at the OB surface (Fig. 2A). A field response composed of an early (N1) and late (N2) component was observed (Fig. 2B,C). Although the voltage depth profile of N1 (peaking at ∼2 ms) was monotonic across OB layers, N2 reversed polarity at the mitral cell layer (Fig. 2B), indicating that N2 encompassed a current sink in the GCL and a current source in the EPL, as reported in studies using electrical stimulation in the APC (Neville and Haberly, 2003; Manabe et al., 2011). These signals were generated mainly by GC depolarization because these cells and their dendrites occupy the vast majority of the space in the GCL and in the EPL and are morphologically organized as a dipole between these two layers (Rall and Shepherd, 1968). Pharmacological characterization in anesthetized animals showed that local microinfusion of AMPA receptor antagonist NBQX (1 mm) into the GCL strongly reduced the N2 slope, with no significant effect on the N1 amplitude (N2: −69.5 ± 10.9%; N1: −8.9 ± 4.9%, n = 3; Fig. 2C). In contrast, local infusion of lidocaine (2%), a voltage-gated Na+ channel blocker, strongly decreased both N1 and N2 (N2: −60.4 ± 1.7%; N1: −79.8 ± 5.9%, n = 4; Fig. 2C; two-way ANOVA: drug effect: F(2,19) = 54.6, p < 0.0001 and drug × negativity: F(2,19) = 6.8, p = 0.006, Holm–Sidak's multiple-comparisons post hoc test: N1: p = 0.48 for NBQX and p < 0.0001 for lidocaine, N2: p < 0.0001 for both drugs compared with vehicle); vehicle injection had no effect on N1 or N2 (p = 0.32, n = 4; p = 0.94, n = 7; respectively). Both N1 and N2 amplitude increased with the power of photostimulation and disappeared in the absence of ChR2. Together with recent studies reporting that light stimulation of cortico-bulbar axon terminals drives direct excitation of GCs (Boyd et al., 2012; Markopoulos et al., 2012), our data strongly suggest that N1 is a ChR2+-induced Na+-spike-dependent fiber volley invading AOC axon terminals, while N2 is a fEPSP generated by AMPAR-dependent depolarization, specific to GCs.
In anesthetized and awake head-restrained mice, local microinfusion of baclofen (2.5 mm) in the GCL did not alter the fiber volley (−3.2 ± 2.7%, n = 4, Wilcoxon matched-pairs signed-rank test: p = 0.63; Fig. 2C), but produced a 2-fold decrease in amplitude and slope of the light-evoked fEPSP. This effect was reversed by subsequent application of CGP (100 μm; fEPSP slope: −65.5 ± 8.6% in baclofen and −32.5 ± 8.5% in CGP; fEPSP amplitude: −56.18 ± 7.3% in baclofen and −20.0 ± 11.5% in CGP; one-way ANOVA: p < 0.0003 for both slope and amplitude, Dunn's multiple-comparisons post hoc test: baseline vs baclofen: p < 0.001 for slope and amplitude, baclofen vs CGP p < 0.05 for amplitude and p = 0.074 for slope; n = 8; Fig. 2D).
To estimate the diffusion of baclofen in the OB, we first validated that, after a first baclofen injection, the second injection at the same site did not depress further the fEPSP amplitude and slope, which is compatible with local saturation of GABABRs. However when baclofen was injected a second time ∼500–600 μm below we observed a strong reduction of the fEPSP, comparable to the reduction observed after the first dorsal injection (−63.3 ± 7.8%, p = 0.23 with Wilcoxon signed-rank test, n = 5). This suggests that the first injection did not reach the second site located ∼500–600 μm deeper in the GCL. To confirm this observation, we injected the same volume and concentration of the nuclear dye DAPI and post hoc histological analysis permitted us to estimate the dye diffusion to be <500 μm. We concluded that, using our injection protocol, baclofen cannot diffuse in an area >600 μm within the OB and therefore cannot significantly diffuse to the AOC. Furthermore, we did not observe any significant difference between anesthetized and awake mice in the structure of the light-evoked fEPSP or in the effect of baclofen (−48.5 ± 4.5% in anesthetized vs −47.0 ± 5.7% in awake mice, n = 8 and 25, respectively, p = 0.98), and data were therefore pooled. However, because cortico-bulbar top-down inputs are sensitive to wakefulness (Boyd et al., 2015; Otazu et al., 2015), the ensuing in vivo experiments were performed solely in awake animals.
To strengthen our in vitro results suggesting a presynaptic localization of GABABRs at the AOC-to-GC synapse (Fig. 1F), we designed a conditional knock-out approach to delete GABABRs in a region-specific manner. A transgenic mouse line that possesses critical exons VII and VIII of the GABAB(1) gene flanked with lox sites (GABABfl/fl mice; Haller et al., 2004) was used. To knock-out the expression of GABABRs and express ChR2 in the same population of cortico-bulbar fibers, two AAVs expressing either Cre recombinase (AAV-CaMKIIa-Cre) and a Cre-dependent ChR2 (AAV-EF1a-DIO-ChR2-mCherry) were co-injected in the AOC of age-matched WT (n = 5) and GABABfl/fl mice (n = 4). Using this strategy, ChR2-eYFP was a reporter for Cre expression and thus identified the population of cells in which GABABRs were conditionally knocked out in GABABfl/fl mice (hereafter named AOCGABAB−/−) (Fig. 2E). As a control, we injected AAV directly expressing ChR2 under the control of the CaMKIIa promoter in WT mice (AAV-CaMKIIa-ChR2-EYFP; n = 7). Three months after viral injection, Cre expression led to ChR2-mCherry labeling of cortico-bulbar axons. In awake animals, baclofen caused a 2-fold decrease in the light-evoked fEPSP slope and amplitude in WT mice, but had no significant effect in AOCGABAB−/− mice (WT: slope: −49.4 ± 6.1%, amplitude: −47.0 ± 5.6%, n = 25 and AOCGABAB−/−: slope: −6.3 ± 5%, amplitude: −7.3 ± 3.2%, n = 21; Two-way ANOVA on slope: baclofen × genotype: F(1,44) = 9.26, p = 0.004; Fisher's post hoc test: p < 0.0001 in WT animals and p = 0.57 in AOCGABAB−/− mice; on amplitude: F(1,44) = 9.37, p = 0.004; Fisher's post hoc test: p < 0.0001 in WT and p = 0.58 in AOCGABAB−/−; saline in WT: −4.8 ± 7.5%, n = 7, Wilcoxon matched-pairs signed-rank test: p = 0.94; Fig. 2F). To confirm that GABABR expression was indeed diminished in cortico-bulbar axons, we performed immunohistochemical labeling for the GABABR1 subunit in OB slices. The immunoreactivity was decreased selectively in the GCL of AOCGABAB−/− mice, but not in the GCL of WT animals (interaction genotype × OB layer: F(2,14) = 4.41, p = 0.033, p = 0.049 for the GCL and p > 0.05 for the other layers, n = 5 WT and n = 4 AOCGABAB−/−; Fig. 2G). Together, these data demonstrate that the expression of presynaptic GABABRs in cortico-bulbar axons allows depression of excitatory feedback onto GCs.
Activation of presynaptic GABABRs depresses cortico-bulbar feedforward inhibition onto M/T cells
We next examined the effect of GABABR activation in cortical fibers on M/T cell spontaneous firing activity. Extracellular recordings of M/T cells (see Materials and Methods section for identification criterions) were performed in awake, head-restrained mice and the same cells were recorded before and after local perfusion of baclofen within the vicinity of the electrode (Fig. 3A). In WT and AOCGABAB−/− mice, baclofen did not alter the M/T cell spontaneous firing rate (−8.3 ± 4.0% in WT, n = 42; −0.3 ± 4.5% in AOCGABAB−/−, n = 29; two-way ANOVA: F(1,69) = 2.75, p = 0.10), consistent with an in vitro study reporting no postsynaptic effect of baclofen on M/T cells (Isaacson and Vitten, 2003). We also investigated whether baclofen changes the temporal relationship between M/T cell firing and the sniff cycle. At the population level, M/T cell phase preference to the sniff cycle did not significantly shift with baclofen application (+40.8 ± 34.7° in WT, n = 17, and −28.9 ± 45.4° in AOCGABAB−/− animals, n = 16, p = 0.33 and p = 0.15 with a paired Hotelling test), whereas it induced a small increase in the strength of the sniffing modulation of M/T cell firing activity in WT, but not AOCGABAB−/− mice (mean vector length: baseline: 0.07 ± 0.01; baclofen: 0.11 ± 0.01; baclofen: F(1,39) = 5.93, p = 0.02; however, baclofen × genotype interaction: F(1,39) = 3.64, p = 0.064; Fisher's LSD post hoc test: WT: p = 0.0013, n = 25, AOCGABAB−/−: p = 0.74, n = 16).
Previous OB slice experiments showed that cortico-bulbar stimulation drives disynaptic inhibition onto M/T cells, mainly mediated by GCs, which is abolished by glutamatergic blockers (Balu et al., 2007; Boyd et al., 2012; Markopoulos et al., 2012). Here, we applied a 1 s light train stimulation on cortico-bulbar axons while recording M/T cell activity using an optic fiber positioned either on top of the OB surface or implanted above the anterior commissure. Because the olfactory cortex send back information to the OB at various regimes (beta, 15–40 Hz, by Gray and Skinner, 1988; Neville and Haberly, 2003; Martin et al., 2006; gamma frequencies, 40–100 Hz, by Boyd et al., 2012; theta, 1–10 Hz, by Youngstrom and Strowbridge, 2015), we decided to span the whole spectrum of cortical axon activity with stimulation frequencies ranging from 10 to 100 Hz. Figure 3B shows the response of an example M/T cell to three frequencies of light stimulation (10, 33, and 67 Hz) and Figure 3C illustrates the inhibition triggered by the different light stimulation patterns on each individual recorded M/T cells. The percentage of change in firing rate is color coded (blue represents inhibition and red excitation). 21/22 of the recorded M/T cells (∼95%) showed reduced firing activity upon cortical stimulation (Wilcoxon matched-pairs rank-sum test for each cell, light stimulation vs prestimulation, p < 0.05; Fig. 3C, left), as reported previously (Markopoulos et al., 2012; Soria-Gómez et al., 2014). The percentage of inhibition was not related to the cell's spontaneous firing rate (y = 0.0041x − 0.73, R2 = 0.012, slope not different from 0: p = 0.63, n = 21), and maximum inhibition was distributed from 33 to 50 Hz (Fig. 3D). Figure 3E represents the effect of 33 Hz light stimulation on individual M/T cells. At the M/T cell population level, firing inhibition as a function of stimulation frequency followed a bell-shaped curve with 33–50 Hz driving maximum inhibition (from −13.0 ± 6.7% at 10 Hz to −55.9 ± 6.4% at 33 Hz; n = 21 cells; Fig. 3F, left). When the same cells were recorded in the presence of baclofen, the magnitude of the light-evoked inhibition of M/T cells decreased significantly. For example, the powerful inhibition induced by 33 Hz light delivery in baseline conditions was abolished in presence of baclofen (−53.2 ± 6.8% in basal condition and −1.3 ± 12% in baclofen; baclofen: F(1,20) = 24.92, p < 0.0001; baclofen × light interaction: F(1,20) = 20.31, p = 0.0002, Fisher's LSD post hoc test: p < 0.0001 in basal condition and p = 0.23 in baclofen; Figure 3E). CGP application partially restored the light-induced inhibition (e.g., M/T cell firing change at 33 Hz: −92.4 ± 4.7% in baseline, −8.0 ± 20.1% in baclofen and −47.0 ± 12.3% in CGP, n = 2). Across all frequencies, baclofen blocked the light-induced inhibition of M/T cells (baclofen and baclofen × light frequency interaction: F(1,21) = 20.31, p = 0.0002 and F(7,147) = 8.41, p < 0.0001, respectively; Fisher's LSD post hoc test, p < 0.05 except for 83 Hz light stimulation, where p = 0.31, n = 21; Fig. 3C,F), whereas vehicle application had no effect (F(1,7) = 0.80, p = 0.40, n = 8).
Because M/T cell inhibition could also be blocked by GABABR action at the GC dendrodendritic synapse (Isaacson and Vitten, 2003; Valley et al., 2013), the same experiments were performed in AOCGABAB−/− mice. In these transgenic animals, light stimulation delivered at various frequencies decreased M/T cell firing with a similar bell-shaped relationship (n = 29 cells; Fig. 3F, right) and individual M/T cells displayed light-evoked inhibition comparable to WT mice (e.g., for 33 Hz stimulation: −48.8 ± 6.1% in M/T cell firing rate, n = 29 cells; Fig. 3C,E). However, baclofen did not alter the light-evoked inhibition of M/T cell in these animals (F(1,28) = 0.46, p = 0.50, n = 29; Fig. 3C,F). Therefore, these results demonstrate that presynaptic GABABR activation at cortico-bulbar axon terminals blocks M/T cell feedforward inhibition.
During these experiments, the influence of cortical inputs on the OB activity was revealed using long-lasting and repeated light stimulation. To examine M/T cell responses to transiently active AOC inputs, we analyzed M/T cell firing activity in response to a single, 5-ms-brief light pulse. In WT mice, 18/22 M/T cells (∼82%) showed a transient suppression of their firing after such a brief light stimulation (Wilcoxon matched-pairs rank-sum test for each cell, p < 0.05). Figure 4A shows a representative cell responding to a single 5 ms light pulse before and after baclofen treatment (top and bottom, respectively). This rapid, transient inhibition of M/T cell firing peaked at 10 ms and lasted ∼30 ms (Fig. 4A,B). Baclofen treatment decreased this inhibition (two-way ANOVA: F(1,17) = 12.39, p = 0.003; baclofen × time interaction, F(34,578) = 9.55, p < 0.0001, Sidak's multiple-comparisons test: 0.0001 < p < 0.05 between 4 and 14 ms after light onset; n = 18; Fig. 4A,B) and dampened the peak amplitude (−20.7 ± 8.1% in baclofen; Wilcoxon match-pairs rank-sum test: p = 0.009, n = 18), resulting in only 13 of 22 recorded cells still showing significant inhibition after baclofen treatment. This feedforward inhibition decayed with a time constant of 17.3 ms in basal conditions and 20.9 ms with baclofen application (r2 = 0.96 and 0.65, respectively; not significantly different, p = 0.29, n = 18) and recovered after CGP infusion (n = 2). In AOCGABAB−/− animals, the light-evoked inhibition was observed in all M/T cells recorded under basal conditions (29/29) and in 28 of 29 recorded cells with baclofen treatment. Baclofen application did not change the time course of inhibition (F(1,28) = 0.612, p = 0.44, n = 29) or the peak amplitude (−2.4 ± 4.3%, p = 0.49, n = 29) (Fig. 4B). These results demonstrate that brief light stimulation of AOC axons is sufficient to elicit feedforward inhibition onto M/T cells and GABABR activation can depress this disynaptic inhibition driven by single cortical inputs.
Cortical feedback excitation to M/T cells is insensitive to presynaptic GABABR modulation
Because AOC glutamatergic afferents to the OB also excite M/T cells directly (Markopoulos et al., 2012; Fig. 1E), we next examined direct excitation from AOC axons in the same M/T cell population. Figure 4C shows an example M/T cell responding with a rapid and precise increase of firing activity in response to a single 5-ms-long light pulse in both basal and baclofen conditions. We found that 7/22 (∼32%) of M/T cell cells received significant direct excitatory input in basal conditions (Wilcoxon match-pairs rank-sum test for each cell). In the 7 M/T cells exhibiting direct excitation before and after baclofen treatment, we observed a slightly prolonged excitation in baclofen conditions, although not significant (two-way ANOVA: interaction time × baclofen: F(29,174) = 1.01, p = 0.46; baclofen: F(1,6) = 5,12, p = 0.064; n = 7; Fig. 4D). No difference was observed in the peak amplitude or in the latency to peak (peak amplitude: +4.6 ± 4.3% in baclofen, p > 0.99; peak latency: 3.0 ± 0.6 ms in baseline vs 2.9 ± 0.5 ms in baclofen, p > 0.99, Wilcoxon match-pairs tests; n = 7; Fig. 4D). We also observed that 2/14 cells showed direct excitation with baclofen treatment, but not in basal conditions. In AOCGABAB−/− mice, we detected a significant direct excitation in 17/29 (∼59%) of the recorded M/T cells. After baclofen administration, 20/29 (∼70%) M/T cells received significant excitation, but baclofen treatment had no effect on this excitation (time × baclofen interaction: F(29,464) = 0.73, p = 0.85; peak amplitude: −11.4 ± 15.9% in baclofen, p = 0.0984; and peak latency: 4.35 ± 0.56 ms in baseline and 4.59 ± 0.54 ms in baclofen, p = 0.47; n = 17; Fig. 4D).
In vitro, light activation of AOC axons failed to reveal fast excitatory synaptic responses on M/T cells, which would have supported the fast evoked firing activity observed in vivo. Instead, we observed small, slow inward currents (average amplitude −9.3 ± 0.9 pA, Vh = −70 mV; n = 9) blocked by NBQX (−89.5 ± 2.4%; p = 0.0003; n = 6), as described previously (Boyd et al., 2012; Markopoulos et al., 2012). Moreover, these evoked currents were frequent in vitro (18/24) and were blocked by baclofen (−78.5 ± 3.1% in baclofen and −23.8 ± 6.7% in CGP; F(1.356,10.85) = 36.31, p < 0.0001, one-way ANOVA, p < 0.05 for all Holm–Sidak's post hoc test; n = 9). Conversely, light-triggered M/T cell spiking in vivo was rare (7/22) and insensitive to baclofen (Fig. 4D). Furthermore, the slow kinetics of the in vitro EPSCs (time to peak from light onset: 5.4 ± 0.5 ms) are incompatible with the in vivo sharp light-evoked spiking (time to peak from light onset: 3.0 ± 0.6 ms). In addition, these slow EPSCs were shown to be unable to trigger M/T cell spiking in slices (Boyd et al., 2012). Collectively, these discrepant observations suggest that these slow currents recorded in vitro do not underlie the fast spiking that we observed in vivo. The recorded EPSCs in vitro could reflect glutamate receptor activation in electrotonically remote regions of M/T cell lateral dendrites or gap junctional coupling with cells receiving direct synaptic inputs.
In vivo, because the efficiency of cortical inhibition of M/T cells is dependent on the stimulation frequency (Fig. 3F), we investigated whether GABABR differentially influenced M/T cell excitatory/inhibitory biphasic response driven at different frequencies. In baseline conditions, M/T cell spiking activity after cortical stimulation decreased at frequencies >10 Hz (Fig. 4E). Baclofen extended the increase in spiking activity after cortical stimulation to higher frequencies (up to 50 Hz) in WT, but not in AOCGABAB−/− animals (WT: frequency × baclofen: F(7,42) = 5.87, p < 0.0001, Fisher's post hoc test: p < 0.01 for 10 to 50 Hz; n = 7; AOCGABAB−/−: F(7,91) = 0.68, p = 0.69; n = 14; Fig. 4E). This result indicates that GABABR activation extends the functional coupling between cortical excitation and M/T cells response in the 10–50 Hz activity band. In summary, activation of cortical feedback triggers both fast direct excitation and feedforward inhibition onto M/T cells, but GABABR activation selectively depresses the inhibitory tone while sparing excitation, thereby reformatting the ratio between excitation and feedforward inhibition received by M/T cells.
Activation of presynaptic GABABRs modulates OB oscillatory activity
Previous studies showed that oscillations and temporal activity might be under the control of extrinsic top-down inputs (Engel et al., 2001). Oscillatory rhythms are prominent in the OB of awake mice (Fig. 5A,B). On the top of breathing-related theta oscillations (1–10 Hz), which are largely driven by olfactory sensory inputs, gamma oscillations (40–100 Hz) are generated by the dendrodendritic synapse (Rall and Shepherd, 1968; Gray and Skinner, 1988; Neville and Haberly, 2003; Kay et al., 2009; Lepousez and Lledo, 2013), whereas beta oscillations (15–40 Hz) are thought to be driven by interactions between the olfactory cortex and the OB (Gray and Skinner, 1988; Neville and Haberly, 2003; Martin et al., 2006). In light of this circuit segregation, we sought to determine whether GABABR-mediated depression of AOC inputs to the OB would alter specific oscillatory frequencies. We found no change in theta power in presence of baclofen in WT or in AOCGABAB−/− animals (WT: −0.8 ± 15.4%, n = 21; AOCGABAB−/−: −17.6 ± 7.0%, n = 29; Two-way ANOVA, F(1,48) = 5.59, p < 0.02, but Fisher's LSD post hoc test: p = 0.11 and 0.09 in WT and AOCGABAB−/− animals, respectively; vehicle injection in WT: p = 0.55, Wilcoxon matched-pairs rank-sum test, n = 9; Fig. 5B,C). Together with the absence of a significant effect of baclofen on M/T cell spontaneous activity, these results suggest that local baclofen application in the GCL did not permit baclofen diffusion superficially to the GL. In contrast, baclofen strongly decreased gamma oscillations in a similar fashion in WT and AOCGABAB−/− animals (WT: −42.8 ± 5.6%, n = 21 and AOCGABAB−/−: −44.0 ± 4.3%, n = 29, F(1,48) = 20.18, p < 0.0001; Fisher's LSD post hoc test: p = 0.017 in WT and p = 0.0002 in AOCGABAB−/−, vehicle in WT: −12.2 ± 10.4%, p = 0.50; n = 9; Fig. 5B,C). Therefore, the reduction of gamma rhythms likely reflects GABABR activation at GC-to-MC synapses (Isaacson and Vitten, 2003; Valley et al., 2013) that are unaltered by our conditional knock-out approach (Fig. 2G). In contrast to gamma oscillations, baclofen strongly decreased spontaneous beta oscillations in WT animals, but not in AOCGABAB−/− mice (respectively, −54.0 ± 6.9%, n = 21 and −17.8 ± 7.0%, n = 29, F(1,48) = 9.60, p < 0.005, Fisher LSD post hoc test: p < 0.05 in WT and p = 0.08 in AOCGABAB−/−; vehicle in WT: −7.2 ± 3.6%, p = 0.50, n = 9; Fig. 5B,C). Therefore, presynaptic GABABR activation on cortico-bulbar inputs regulates spontaneous beta but not gamma oscillations in the OB.
AOC feedforward inhibition of odor-evoked M/T activity is depressed by GABABR activation
We next investigated the impact of GABABR activation at AOC axons on sensory-evoked activity in M/T cells. By stimulating AOC axons during odor presentation, we analyzed the effects of light stimuli on odor-evoked responses (Fig. 6A). We used frequencies of 10, 33, and 67 Hz to deliver light pulses because these frequencies recruit distinct degree of inhibition (Fig. 3F) and correspond to different regimes of cortical activities (respectively theta, beta, and gamma). Figure 6B shows an example M/T cell response to odor and simultaneous odor + light stimulation at baseline or in the presence of baclofen.
Across the population of M/T cells, odor stimulation resulted in either excitation or inhibition in awake mice (n = 16 and n = 24 odor-unit pairs, respectively, six mice; Fig. 6C,D). Baclofen application had no effect on the population response (−1.76 ± 4.19%; two-way ANOVA: baclofen × stimulation: F(3,117) = 12.92, p < 0.0001; Sidak's multiple-comparisons test: p = 0.99; n = 40; Fig. 6E). Simultaneous odor + light stimulation produced a significant decrease in odor-evoked M/T cell firing activity at light frequencies of 33 and 67 Hz, whereas 10 Hz light stimulation did not alter the neuron's evoked activity (one-way ANOVA: F(3.72,144.9) = 22.4, p < 0.0001; Holm–Sidak's multiple-comparisons test: p = 0.061 for 10 Hz, p < 0.0001 for 33 and 67 Hz stimulation; n = 40; Fig. 6Di,Dii,Diii), regardless of whether the odor was excitatory or inhibitory (p > 0.05 at 10 Hz and p < 0.0001 at 33 and 67 Hz for both odor responses, n = 16 and n = 24, respectively; Fig. 6E). The light-induced inhibition of M/T cell odor-evoked activity was diminished with baclofen for both 33 and 67 Hz stimulation, whereas baclofen had no effect on 10 Hz light stimulation (baclofen-induced diminution of light inhibition: 10 Hz: −3.1 ± 4.2%, 33 Hz: −34.4 ± 5.4%, 67 Hz: −12.3 ± 6.2%; baclofen F(1,39) = 4.38, p < 0.043, baclofen × stimulation: F(3,117) = 12.92, p < 0.0001; p = 0.92, p < 0.0001 and p < 0.05 at 10, 33 and 67 Hz; n = 40; Fig. 6E). When separately analyzing odor-inhibited and odor-excited neurons, we observed that, for odor-inhibited responses, baclofen depressed the inhibition of M/T cell activity induced by simultaneous odor and light application at 33 or 67 Hz. For odor-excited responses, baclofen produced a significant decrease in M/T cell inhibition only when odor stimulation was paired with 33 Hz light (odor-inhibited neurons: F(1,23) = 29.0, p < 0.0001, baclofen × stimulation: F(3,69) = 27.9, p < 0.0001; p < 0.0001 for 33 and 67 Hz, n = 24; odor-excited neurons: F(1,15) = 0.67, p = 0.043, baclofen × stimulation: F(3,45) = 36.1, p < 0.0001; p < 0.01 for 33 Hz and p = 0.49 for 67 Hz, n = 16). Therefore, in the context of odor-evoked M/T cell activity, GABABRs activation at AOC axon terminals depresses the cortico-bulbar inhibition on M/T cells.
To confirm the impact of cortical GABABR presynaptic activation on odor-evoked activity of the M/T cell population, we performed calcium imaging in freely behaving animals using fiber photometry (Fig. 7A). To specifically record M/T cell Ca2+ transients, we injected a Cre-dependent AAV expressing GCaMP6f (Chen et al., 2013) in the dorsolateral region of Tbet-Cre mice OBs (Haddad et al., 2013). GCaMP6f was excited continuously at low intensity (0.4–0.5 mW) and the volume fluorescence was collected using an optic fiber implanted above the injection site, spectrally separated using a dichroic mirror, and emission intensity was measured with a femtowatt photodetector (Fig. 7A,B). To gain independent light control of AOC axons and to avoid cross-excitation between GCaMP6f and ChR2, we injected the red-shifted channelrhodospin ChRimson (Klapoetke et al., 2014) into the AOC and targeted red light stimulation with an optic fiber implanted above the AOC, which was connected to a 589 nm laser (Fig. 7A,B). Using this technique, AOC light stimulation at 10, 33, or 67 Hz with red light (589 nm, 5–10 mW) produced a global reduction of spontaneous fluorescence at the recording site (ΔF/F: −3.7 ± 0.9% at 10 Hz, −9.2 ± 2.7% at 33 Hz, and −4.8 ± 0.7% at 67 Hz, n = 6; Fig. 7C), confirming the sensitivity of fiber photometry to detect a population decrease in Ca2+ transients during spontaneous activity in freely behaving mice. These effects were not observed when using blue light (473 nm) at the intensity used for GCaMP6f excitation (0.4–0.5 mW), validating the absence of cross-excitation between ChRimson and GCaMP6f (Fig. 7C). With local OB infusion of baclofen, this inhibition was reduced (two-way ANOVA: baclofen F(1,5) = 13.05, p = 0.015, Holm–Sidak's multiple-comparisons test: p < 0.01 for all light frequencies; Fig. 7C,D). Upon odor presentation, M/T cells responses were characterized by a strong elevation in fluorescence superimposed on a robust breathing modulation of the signal (Fig. 7E). We observed an increase in the GCaMP6f signal in all the odor-recording site pairs in basal conditions (Fig. 7F) and these odor-evoked transients remained unchanged in presence of baclofen (baseline: +4.9 ± 0.8%, baclofen: + 4.7 ± 0.7%; paired t test: p = 0.69; n = 15; Fig. 7G). When the AOC was light stimulated in addition to odor presentation, we observed a decrease in the M/T cell odor-driven responses (one-way ANOVA: F(1.461,20.46) = 13.25, p = 0.0006, Holm–Sidak's multiple-comparisons test: p < 0.01 for 10, 33, and 67 Hz frequencies, n = 15; Fig. 7G). Similar to the electrophysiological recordings, we found that baclofen depressed the light-induced reduction of M/T cell odor-evoked activity and this effect was significant at all frequencies (baclofen: F(1,14) = 4.7, p = 0.048, baclofen × stimulation: F(2,28) = 1.9, p < 0.17, p < 0.001 post hoc for 10, 33, and 67 Hz frequencies, n = 15; Fig. 7G). Together with our in vivo electrophysiological recordings, these data show that activation of presynaptic GABABRs at AOC axon terminals profoundly remodels M/T cell responses to simultaneous sensory and top-down inputs.
Discussion
Cortical projections influence olfactory information processing as early as in the first central relay of the olfactory system, namely the OB. In this region, cortico-bulbar feedback transfers information about, for example, brain states, attention, and prior sensory experience. Because these top-down inputs convey signals relative to dynamic internal states, their regulation must be an essential feature for their precise function. In this study, we revealed a GABABR-dependent mechanism to modulate cortico-bulbar feedback. Using a combination of genetics, pharmacology, electrophysiology, and Ca2+ imaging of neuronal population, we found that: (1) activation of presynaptic GABABRs reduces the direct glutamatergic inputs onto GCs, but not M/T cells; (2) GABABR activation blocks cortical-driven feedforward inhibition of M/T cells' spontaneous and odor-evoked firing activity; (3) GABABR activation biases M/T cell excitatory/inhibitory response ratio to cortical stimulation toward excitation; and (4) depressing glutamate release from AOC axons reduces beta, but not gamma oscillations. Interestingly, GABABR activation does not shunt the overall AOC feedback to the OB, but instead refines the functional connectivity between the AOC and OB.
In this study, we introduced ChR2 into the two olfactory primary cortices, the AON and APC, to gain control over the main source of cortico-bulbar projections. Despite the anatomical distinction of the two olfactory cortices, there is no clear evidence that they affect OB function differentially. Both areas mainly target the GCL and GL of the OB (Boyd et al., 2012; Markopoulos et al., 2012; Lepousez et al., 2014) and, even though distinction in the precise connectivity patterns seems to exist, such subtle differences can also be found within specific area subdivisions (Haberly and Price, 1978a, 1978b; Davis and Macrides, 1981). Moreover, recent studies investigating either AON or APC inputs reported a comparable connectivity pattern (NBQX-sensitive inputs to GCs, MCs, and GL neurons such as periglomerular neurons and superficial short axon cells) and similar functional impact on M/T cell odor-evoked responses (Boyd et al., 2012; Markopoulos et al., 2012). Therefore, we chose here to consider the AON and APC as a single functional entity that we collectively named the anterior (primary) olfactory cortex (or AOC). Further work would be required to investigate potential differences in top-down functions of AON and APC.
GABABRs are widely expressed in the OB. In addition to cortico-bulbar terminals, GABABRs are expressed at olfactory sensory neuron terminals to depress glutamate release. Sensory inputs drive M/T cell activity (Cang and Isaacson, 2003; Margrie and Schaefer, 2003; Phillips et al., 2012) and generate theta OB oscillations (for review, see Kay et al., 2009). In our condition, neither M/T cell spontaneous firing rate nor theta rhythms was sensitive to local baclofen infusion in the GCL. Therefore, it seems apparent that baclofen did not diffuse superficially to sensory axon terminals in the GL. Recordings of fEPSPs at different depths in the GCL further confirmed the drug diffusion area to be <600 μm. GABABRs are also expressed at GC apical dendrites, where they depress GABA release, as reported in vitro (Isaacson and Vitten, 2003; Valley et al., 2013). Surprisingly, we did not observe any effect of baclofen on light-evoked feedforward inhibition onto M/T cells in AOCGABAB−/− mice (Fig. 3F). Given the remoteness of the virus injection site to the OB, the unaltered GABABR1 immunoreactivity in the EPL of AOCGABAB−/− mice and with the near absence of labeled GCs in the OB, it is highly unlikely that the virus diffused to the OB and altered GABABR expression at GC dendrites. Moreover, spontaneous gamma oscillations, which rely on dendrodendritic reciprocal synapses (Rall and Shepherd, 1968; Lepousez and Lledo, 2013), were strongly reduced after GABABR activation in AOCGABAB−/−. Therefore, the lack of effect of baclofen on M/T cell feedforward inhibition in AOCGABAB−/− mice suggests that distal stimulation of GCs by M/T cell dendrites triggers GABABR-sensitive GABA release, whereas AOC terminal proximal stimulation of GCs triggers GABA release in a GABABR-independent manner. An alternate hypothesis could be that AOC axon stimulation preferentially engages adult-born GCs, which have been proven to be GABABR insensitive (Valley et al., 2013). In any case, because GABABR activation had no effect in AOCGABAB−/− mice, we further reasoned that AOC inputs trigger feedforward inhibition but do not drive significant dendrodendritic recurrent inhibition between M/T cells.
To our knowledge, the present study is the first to show the presence of GABABRs in a cortico-bulbar synapse. Using selective GABABR knock-down in cortico-bulbar projections, we discovered an additional site of GABABR expression that adds more insight into the understanding of GABABR-dependent modulation of OB activity. We showed that activation of presynaptic GABABRs depresses the AOC-to-GC excitatory synapse, thereby blocking AOC-driven feedforward inhibition onto M/T cells. In contrast, we did not find any evidence for GABABR-dependent modulation of the AOC-to-M/T cell synapse. The target-dependent expression of presynaptic GABABR could reflect the diversity of AOC-projecting cells or it could be determined by the activity or the nature of the postsynaptic target (i.e., glutamatergic or GABAergic), as reported in cultured hippocampal neurons (Schinder et al., 2000). This target-dependent functional expression of GABABRs modifies the balance between cortical excitation and feedforward inhibition received by M/T cells. Because the temporal window M/T cells use to integrate cortical excitatory events is tightly controlled by GABABR-sensitive feedforward inhibition, GABABR activation enlarges this integration windows as observed in thalamo-cortical feedforward circuits (Chittajallu et al., 2013). Using patterned light stimulation of cortico-bulbar inputs at different frequencies, we observed that this differential GABABR sensitivity extends the coupling between M/T cell responses and AOC axon stimulations to beta frequencies (10–50 Hz; Fig. 3F). Given that cortical activity can operate at such rhythms, we propose that GABABR modulation participates in gating the transfer of the beta regimes on M/T cell firing activity. Consistent with this, we found that GABABR activation at cortico-bulbar inputs selectively depresses spontaneous beta oscillations.
Recent evidence suggests a role for top-down inputs on temporal activity of the targeted structure. Top-down inputs might affect ongoing oscillations, synchronicity of postsynaptic cells, and coherence between brain areas (Engel et al., 2001). In the olfactory system, by acting at the AOC axon terminals, we demonstrated that GABABRs are well positioned to regulate coherence between distant structures, such as the AOC and the OB, a phenomenon likely to emerge in behaviorally relevant tasks (Chabaud et al., 1999; Kay and Beshel, 2010; Cohen et al., 2015). In particular, beta oscillations were depressed by cortical GABABR activation and have been proposed to be supported by a reentry of cortical input in the OB (Gray and Skinner, 1988; Neville and Haberly, 2003; Martin et al., 2014). GABABR could thereby regulate the shift in coherence between cortical structures and the OB under different sensory experiences (adaptation, learning, memory, etc.). In the near future, it will be of great interest to investigate the behavioral impact of this GABABR-dependent presynaptic modulation of top-down activity and to decipher in which context it is engaged.
Footnotes
This work was supported by the life insurance company AG2R-La-Mondiale, the Agence Nationale de la Recherche (ANR-15-CE37-0004), and the Laboratoire d'Excellence Revive (Investissement d'Avenir, ANR-10-LABX-73). Our laboratory is part of the Ecole des Neurosciences de Paris (ENP) Ile-de-France network and is affiliated with the Bio-Psy Laboratory of Excellence. C.M. is a recipient of a fellowship from the French Ministère de l'Education Nationale et de la Recherche. We thank Carine Moigneu and Laurent Cotter for viral injections; all members of the Lledo laboratory for their insights during the course of these experiments; Kurt Sailor for editing the manuscript; Manuel Mameli for helpful discussions; Bernhard Bettler for providing GABABflox/flox mice; Anne Lanjuin and Catherine Dulac for the Tbet-Cre mice; Karl Deisseroth and Edward Boyden for optogenetic tools; Gaël Moneron for help in setting up the fiber photometry path and the Genetically-Encoded Neuronal Indicator and Effector (GENIE) Project; and the Janelia Farm Research Campus of the Howard Hughes Medical Institute for sharing GCaMP6f constructs.
The authors declare no competing financial interests.
- Correspondence should be addressed to Pierre-Marie Lledo, Laboratory for Perception and Memory, Institut Pasteur and CNRS, 25 rue du Dr. Roux, 75 724 Paris Cedex 15, France. pmlledo{at}pasteur.fr