Volume 601, Issue 16 pp. 3605-3630
Research Article
Open Access

Olfactory-driven beta band entrainment of limbic circuitry during neonatal development

Johanna K. Kostka

Corresponding Author

Johanna K. Kostka

Institute of Developmental Neurophysiology, Center for Molecular Neurobiology, Hamburg Center of Neuroscience (HCNS), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Corresponding authors J. K. Kostka and I. L. Hanganu-Opatz: Institute of Developmental Neurophysiology, Center for Molecular Neurobiology, Hamburg Center of Neuroscience (HCNS), University Medical Center Hamburg-Eppendorf, Falkenried 94, 20251 Hamburg, Germany.  Email: [email protected] and [email protected]

Search for more papers by this author
Ileana L. Hanganu-Opatz

Corresponding Author

Ileana L. Hanganu-Opatz

Institute of Developmental Neurophysiology, Center for Molecular Neurobiology, Hamburg Center of Neuroscience (HCNS), University Medical Center Hamburg-Eppendorf, Hamburg, Germany

Corresponding authors J. K. Kostka and I. L. Hanganu-Opatz: Institute of Developmental Neurophysiology, Center for Molecular Neurobiology, Hamburg Center of Neuroscience (HCNS), University Medical Center Hamburg-Eppendorf, Falkenried 94, 20251 Hamburg, Germany.  Email: [email protected] and [email protected]

Search for more papers by this author
First published: 11 July 2023

Handling Editors: Richard Carson & Nathan Schoppa

The peer review history is available in the Supporting Information section of this article (https://doi.org/10.1113/JP284401#support-information-section).

This article was first published as a preprint. Kostka JK, Hanganu-Opatz IL. 2021. Olfactory-driven beta band entrainment of limbic circuitry during neonatal development. bioRxiv. https://doi.org/10.1101/2021.10.04.463041

Abstract

Cognitive processing relies on the functional refinement of the limbic circuitry during the first two weeks of life. During this developmental period, when the auditory, somatosensory and visual systems are still largely immature, the sense of olfaction acts as ‘door to the world’, providing an important source of environmental inputs. However, it is unknown whether early olfactory processing shapes the activity in the limbic circuitry during neonatal development. Here, we address this question by combining simultaneous in vivo recordings from the olfactory bulb (OB), lateral entorhinal cortex (LEC), hippocampus (HP) and prefrontal cortex (PFC) with olfactory stimulation as well as opto- and chemogenetic manipulations of mitral/tufted cells in the OB of non-anaesthetized neonatal mice of both sexes. We show that the neonatal OB synchronizes the limbic circuity in the beta frequency range. Moreover, it drives neuronal and network activity in LEC, as well as subsequently, HP and PFC via long-range projections from mitral cells to HP-projecting LEC neurons. Thus, OB activity shapes the communication within limbic circuits during neonatal development.

Key points

  • During early postnatal development, oscillatory activity in the olfactory bulb synchronizes the limbic circuit.
  • Olfactory stimulation boosts firing and beta synchronization along the olfactory bulb–lateral entorhinal cortex–hippocampal–prefrontal pathway.
  • Mitral cells drive neuronal and network activity in the lateral entorhinal cortex (LEC), as well as subsequently, the hippocampus (HP) and the prefrontal cortex (PFC) via long-range projections from mitral cells to HP-projecting LEC neurons.
  • Inhibition of vesicle release on LEC targeting mitral cell axons reveals direct involvement of LEC in the olfactory bulb-driven oscillatory entrainment of the limbic circuitry.

Introduction

Coordinated neuronal activity during early development refines the neural circuits that account for cognitive processing in the adult brain. During the first two postnatal weeks, discontinuous oscillations in the lateral entorhinal cortex (LEC) already drive similar activity patterns in the hippocampus (HP), which in turn entrains the prefrontal cortex (PFC) (Ahlbeck et al., 2018; Bitzenhofer, Ahlbeck, Wolff et al., 2017; Brockmann et al., 2011; Hartung, Brockmann et al., 2016). Disturbance of these early activity patterns in mouse models of psychiatric risk (Chini et al., 2020; Domnick et al., 2015; Hartung, Cichon et al., 2016; Richter et al., 2019; Xu et al., 2021), as well as through pharmacological (Krüger et al., 2012) or optogenetic manipulations (Bitzenhofer et al., 2021), led to disruption of adult circuits and behavioural abilities. However, so far it is unknown whether these early activity patterns in limbic networks emerge endogenously or whether they are shaped by sensory experience.

Due to the limited functionality of the visual and auditory systems during the first two postnatal weeks – newborn mice are still blind and deaf – their contribution to the development of limbic networks has been considered minor. This hypothesis has been supported by data showing that the synchrony between V1 and the HP–PFC network before eye-opening is rather weak (Brockmann et al., 2011). On the other hand, sensory feedback signals evoked by myoclonic twitches synchronize the primary sensory cortex and the HP in the beta frequency range during sleep (Del Rio-Bermudez et al., 2020; Mohns & Blumberg, 2010) and neonatal hippocampal sharp waves can be triggered by twitches (Karlsson et al., 2006).

Besides these reafferent signals evoked by twitches, little is known about the impact of sensory inputs on the limbic system during development. A prominent sensory system for newborn rodents is the olfactory system, as neonatal mice process olfactory inputs from birth and use this information to instruct the learning and cue-directed behaviours essential for their survival (Logan et al., 2012). Correspondingly, the anatomical pathways from the olfactory bulb (OB) to cortical areas are unique among sensory systems as they lack the relay via the thalamus. Mitral cells (MCs) send afferents directly to the piriform cortex (PiR) and limbic brain areas such as the LEC and amygdala (Igarashi et al., 2012; Luskin & Price, 1983). At adult age, in line with the anatomical connectivity, strong functional coupling during odour processing has been found between OB and these brain areas. Synchronized beta oscillations between the OB and limbic brain areas such as the PiR, LEC and HP are critically involved in olfactory memory processing (Gourévitch et al., 2010; Igarashi et al., 2014; Martin et al., 2006; Ravel et al., 2003). Moreover, beta oscillations in prefrontal–hippocampal networks have been identified to support the utilization of odour cues for memory-guided decision making (Symanski et al., 2022).

The tight and behaviourally relevant coupling between the OB and limbic circuits at adult age leads to the question of which role olfactory activation early in life plays for these circuits. Previously, we showed that discontinuous oscillatory activity in the theta–beta range, emerging as a result of bursting MCs in the neonatal OB, entrains similar oscillatory patterns in the LEC of anaesthetized mice (Gretenkord et al., 2019; Kostka et al., 2020). Moreover, odour stimulation evokes beta oscillations in the PiR of non-anaesthetized neonatal rats (Zhang et al., 2021). However, the role of neuronal and network activity in the OB for the functional entrainment of neuronal activity in downstream areas within the neonatal hippocampal–prefrontal network in non-anaesthetized mice is still largely unknown.

To address these knowledge gaps, we simultaneously monitored single-unit activity (SUA) and local field potentials (LFPs) in the OB, LEC, HP and PFC of non-anaesthetized neonatal mice (postnatal day (P) 8−10) during odour stimulation and manipulation of mitral/tufted cell (M/TC) activity using excitatory opsins and inhibitory designer receptors exclusively activated by designer drugs (DREADDs). We show that odour stimulation and optogenetic activation of M/TCs triggers action potential firing in the LEC and HP as well as prominent beta oscillations that synchronize the OB with the downstream cortical areas. Conversely, blocking MC output specifically diminishes the OB–limbic network coupling in the beta frequency range. These data document the ability of coordinated activity at the sensory periphery of newborns to shape the network activity in circuits accounting for adult cognitive processing.

Methods

Ethical approval

All experiments were performed in compliance with the German laws and the guidelines of the European Union for the use of animals in research (European Union Directive 2010/63/EU) and were approved by the local ethical committee (Behörde für Gesundheit und Verbraucherschutz Hamburg, ID 15/17). All experiments conform to the principles and regulations described in the Editorial by Grundy (2015).

Animals

Time-pregnant C57Bl/6/J (ncre = 68, male: n = 31, female: n = 37) and Tbet-cre mice (The Jackson Laboratory, Strain #:024507) (ncre+ = 52, male: n = 28, female: n = 24) from the animal facility of the University Medical Center Hamburg-Eppendorf were housed individually in breeding cages on a 12 h light/12 h dark cycle and fed ad libitum. Offspring (both sexes) were injected with either AAV9-Ef1a-DIO-hChR2(E123T_T159C)-EYFP (Addgene, Plasmid #35509), AAV9-EF1a-DIO-hM4D(Gi)-mCherry (Addgene, Plasmid #50461) or AAV9-EF1a-DIO-MOS-OPN3-mScarlet virus at P0 or P1. Genotypes were determined using genomic DNA and following primer sequences (Metabion, Planegg/Steinkirchen, Germany) as described previously (Gretenkord et al., 2019): for Cre, PCR forward primer 5′-ATCCGAAAAGAAAACGTTGA-3′ and reverse primer 5′-ATCCAGGTTACGGATATAGT-3′. The PCR reactions were as follows: 10 min at 95°C, 30 cycles of 45 s at 95°C, 90 s at 54°C, and 90 s at 72°C, followed by a final extension step of 10 min at 72°C. In addition to genotyping, enhanced green fluorescent protein expression in the OB was detected using a dual fluorescent protein flashlight (Electron microscopy sciences, Hatfield, PA, USA) prior to surgery. Cre+ and cre littermates were used in the experimental or control group, respectively. At P8–10 cre and cre+ mice underwent odour stimulation, light stimulation or Compound 21 injections and in vivo multi-side electrophysiological recordings.

Surgical procedures and recordings

Virus injection for transfection of M/TCs with ChR2 and hm4D(Gi)

For transfection of M/TCs with the ChR2 derivate E123T/T159C, eOPN3 or inhibitory DREADDs (hM4D(Gi)), P0–1 pups were anaesthetized with isoflurane (induction: 5%, maintenance: 2.5%, Forane, Abbott) and fixed into a stereotaxic apparatus. A dose of 0.5% bupivacaine/1% lidocaine was locally applied to the skin above the OB before mice received unilateral injections of one of two viral constructs (AAV9-Ef1a-DIO hChR2(E123T/T159C)-EYFP, 200 μl at titre ≥1 × 1013 vg/ml, Plasmid #35509, Addgene, MA, USA; AAV9-EF1a-DIO-hM4D(Gi)-mCherry, 200 μl at titre ≥1 × 1014 vg/ml, Plasmid #50461, Addgene, MA, USA; AAV9-EF1a-DIO-MOS-OPN3-mScarlet, 200 μl at titre 8.7 × 1011 vg/ml, Plasmid #131002, Addgene, MA, USA). The virus was produced by Addgene or the Virus Facility of the University Medical Center Eppendorf. A total volume of 200 nl was slowly (200 nl/min) delivered at a depth of around 0.5 mm into the right OB using a micropump (Micro4, WPI, Sarasota, FL). Following injection, the syringe was left in place for at least 30 s to avoid reflux of fluid. Pups were maintained on a heating blanket until full recovery and returned to the dam.

Virus injection for tracing

For the transfection of M/TC axons with enhanced yellow fluorescent protein (EYFP) and the retrograde labelling of HP-projecting neurons with mCherry, P0–1 pups received the viral construct AAV9-hSyn-hChR2(H134R)-EYFP (200 μl at titre ≥1 × 1013 vg/ml, #26973-AAV9, Addgene, MA, USA) into the OB and the retrograde virus AAVrg-CamKIIα-mCherry (80 μl at titre ≥7 × 1012 vg/ml, #114469-AAVrg, Addgene, MA, USA) into the HP. Virus injection was performed similarly to the transfection of M/TCs with ChR2 or hM4D(Gi). After 10 days, the mice underwent terminal anaesthesia using 10% ketamine (Ketamidor, Richter Pharma AG, Germany)/2% xylazine (Rompun, Bayer, Germany) in 0.9% NaCl solution (10 μg/g body weight, i.p.) and the brains of the investigated mice were perfused with 4% paraformaldehyde (PFA), sliced and MC axons and HP-projecting neurons in LEC and PiR were imaged using a confocal microscope.

Surgical procedure for electrophysiology

For acute in vivo recordings, P8–10 mice underwent surgery according to previously described protocols (Brockmann et al., 2011; Gretenkord et al., 2019; Kostka et al., 2020; Xu et al., 2021). Under isoflurane anaesthesia (induction: 5%, maintenance: 2.5%, Forane, Abbott), the skin above the skull was removed and 0.5% bupivacaine/1% lidocaine was locally applied to the neck muscles. The anaesthesia depth was closely monitored throughout the surgery by assessing the breathing rate and testing the toe-pinch reflex. Two plastic bars for head fixation in the recording setup were mounted on the nasal and occipital bones with dental cement. The bone above the right OB (0.5–0.8 mm anterior to frontonasal suture, 0.5 mm lateral to inter-nasal suture), LEC (0 mm posterior to lambda, 6−7.5 mm lateral from the midline), HP (2.5 mm anterior to lambda, 3.5 mm lateral from the midline) and PFC (0.5 mm anterior to bregma, 0.1–0.5 mm lateral from the midline) was carefully removed by drilling a hole of <0.5 mm in diameter. After a recovery period of 20 min, pups were transferred to the setup for electrophysiological recording. Throughout the surgery and recording session, mice were maintained on a heating blanket at 37°C.

Multi-site electrophysiological recordings in vivo

Acute three-side or four-side recordings were performed in head-fixed non-anaesthetized P8–10 mice similar to the method described previously (Chini et al., 2022; Xu et al., 2021). For this, mice were head-fixed into a stereotaxic apparatus and maintained on a heating blanket at 37°C throughout the recording. One-shank electrodes (NeuroNexus, MI, USA) with 16 recording sites (0.4–0.8 MΩ impedance, 50 μm inter-site spacing for recordings in OB and HP, 100 μm inter-site spacing for recordings in LEC and PFC) were stereotactically inserted into OB (0.5–1.8 mm, angle 0°), LEC (for four-side recordings, depth: 2 mm, angle: 180°; for three-side recordings, depth: 2–2.5 mm, angle: 10°), HP (1.3–1.9 mm, angle 20°) and PFC (1.8–2.1 mm, angle 0°) using micromanipulators. For four-side recordings, electrodes were inserted perpendicular to the lateral side of the brain into LEC using a free-standing micromanipulator, as previously described (Xu et al., 2021). For light stimulation, one-shank optrodes (NeuroNexus, MI, USA) with the same configuration as the electrodes were inserted in the OB. Before insertion, the electrodes were covered with DiI (1,1’-Dioctadecyl-3,3,3’,3’-tetramethylindocarbocyanine perchlorate, Molecular Probes, Eugene, OR). A silver wire was inserted into the cerebellum and served as a ground and reference electrode. A piezo-electric sensor placed under the mouse's chest was used to monitor respiration. Before data acquisition, a recovery period of 20 min following the insertion of electrodes was provided. Extracellular signals were band pass-filtered (0.1–9 kHz) and digitized (32 kHz) by a multichannel amplifier (Digital Lynx SX; Neuralynx, Bozeman, MO, USA) and Cheetah acquisition software (Neuralynx). Spontaneous activity was recorded for at least 20 min before light stimulation or Compound 21 (C21, Hellobio, Ireland) injection. After the recording session, all animals were killed. For this, mice underwent terminal anaesthesia using 10% ketamine (Ketamidor, Richter Pharma AG, Germany)/2% xylazine (Rompun, Bayer, Germany) in 0.9% NaCl solution (10 μg/g body weight, i.p.). Subsequently, the brains were perfused with 4% PFA and sliced for post hoc histological assessment of the electrode position. The position of recording electrodes in the OB, LEC, HP and PFC was confirmed after histological assessment post-mortem. For the analysis of LFPs in the OB, the recording site centred in the external plexiform layer was used, whereas for HP the recording site located in the CA1 was considered. For the analysis of LFPs in the LEC, only recording sites that were histologically confirmed to be located in superficial entorhinal layers were used. Similarly, only recording sites confined to the prelimbic sub-division of the PFC were considered. For the analysis of spiking activity, all recording sites confirmed to be located in the areas of interest (OB, LEC, HP and PFC) were considered. When necessary, spikes recorded in the OB were assigned to specific layers according to the location of recording sites.

Morphology

All mice were anaesthetized with 10% ketamine (Ketamidor, Richter Pharma AG, Germany)/2% xylazine (Rompun, Bayer, Germany) in 0.9% NaCl solution (10 μg/g body weight, i.p.) and transcardially perfused with Histofix (Carl Roth, Germany) containing 4% PFA. Brains were postfixed in 4% PFA for 24 h and sliced. Slices (100 μm thick) were mounted with Fluoromount containing DAPI (Sigma-Aldrich, MI, USA). The positions of the DiI-labelled extracellular electrodes in the OB, LEC, HP and PFC were reconstructed to confirm their location. Virus expression was verified by EYFP (for ChR2) or mCherry (for hM4D(Gi)) fluorescence in the right OB. For confocal imaging of EYFP or mCherry fluorescence in M/TCs, HP and LEC, 50 μm thick slices mounted with Vectashield (CA, USA) were used.

Light stimulation

Activation of M/TCs was achieved by either ramp or pulse light stimulation applied using a diode laser (473 nm; Omicron, Austria) which was controlled by an arduino uno (Arduino, Italy). For ramp stimulation, a light stimulus with linear increasing power (3 s rise time) was presented 30−60 times. For pulse stimulation 3 ms light pulses at 2 Hz were delivered. Laser power was adjusted for every recording (1.37–5.15 mW) to reliably induce neuronal firing. For the inactivation of neurotransmitter release from axon terminals in LEC light stimulation (10 s) was applied using a diode laser (473 nm; Omicron, Austria).

Olfactory stimulation

For olfactory stimulation, a custom-made olfactometer was used and odours were applied for 2 s triggered by respiration and controlled by an arduino (Arduino, Italy). Clean air flow at a constant speed (0.9 l/min) was applied. Odour stimulation was triggered by exhalation and reached the mouse with an estimated delay of 50 ms. The odour concentration during the next inhalation was constant. Two to four different odours ((R)-(+) Limonene (Alfa Aesar), 1-Octanol (Acros Organics), Vanillin, and Isoamyl acetate (Sigma-Aldrich), 1% in mineral oil (Sigma-Aldrich)) were delivered in a randomized order using mass flow controllers.

Compound 21 injection

Compound 21 (3 mg/kg dissolved in 0.9% NaCl), a selective DREADD agonist, was injected subcutaneously after >20 min recording of baseline activity, while the mouse was fixed in the stereotaxic apparatus. The activity was recorded for 40−120 min post-injection.

Data analysis

LFP analysis

Data were analysed offline using custom-written scripts in the MATLAB environment (MathWorks, Natick, MA). Data were first low pass-filtered (<100 Hz) using a third-order Butterworth filter before down-sampling by factor 20 to 1.6 kHz to analyse LFPs. All filtering procedures were performed in a manner that preserved the phase information.

Power spectral density

Power spectral density was analysed for either the entire baseline period, 2 s periods before (Pre), and during light ramp stimulation (Stim) for recordings combined with optogenetic manipulation. For recordings involving odour stimulation, 1.5 s periods before, and 1.5 s after the first inhalation during odour stimulation were considered. For recordings paired with DREADD manipulation, the power was either calculated for every minute or averaged for the entire baseline period (19 min) and post-C21 injection period (30 min). Power was calculated using Welch's method with non-overlapping windows of 2 s (ramp periods) or 3 s length and spectra were smoothened with a moving average of 20 samples. Time–frequency plots of power were calculated with a continuous wavelet transform (Morlet wavelet) between scales 1/100 Hz and 1 Hz using the MATLAB function wt provided by the cross wavelet and wavelet coherence toolbox (http://grinsted.github.io/wavelet-coherence/).

Coherence

The imaginary part of coherence, which is insensitive to volume conduction-based effects (Nolte et al., 2004), was calculated for the same time periods as the power by taking the absolute value of the imaginary component of the normalized cross-spectrum:
C X Y f = I m P X Y f P X X f P Y Y f $$\begin{equation}{C_{XY}}\left( f \right) = \left| {{\mathop{\mathrm Im}\nolimits} \left( {\frac{{{P_{XY}}\left( f \right)}}{{\sqrt {{P_{XX}}\left( f \right){P_{YY}}\left( f \right)} }}} \right)} \right|\end{equation}$$ (1)

Spectral dependency ratio

The spectral dependency ratio (SDR) was calculated according to Shajarisales et al. (2015) from the power spectral densities ( P X X ${P_{XX}}$ (f) and P Y Y ${P_{YY}}$ (f)) of the signals X and Y:
S D R X Y = m e a n P Y Y f m e a n P X X f × m e a n P Y Y f P X X f $$\begin{equation}\;SD{R_{X \to Y}} = \frac{{mean\left( {{P_{YY}}\left( f \right)} \right)}}{{mean\left( {{P_{XX}}\left( f \right)} \right) \times mean\left( {\frac{{{P_{YY}}\left( f \right)}}{{{P_{XX}}\left( f \right)}}} \right)}}\end{equation}$$ (2)
S D R Y X = m e a n P X X f m e a n P Y Y f × m e a n P X X f P Y Y f $$\begin{equation}\;SD{R_{Y \to X}} = \frac{{mean\left( {{P_{XX}}\left( f \right)} \right)}}{{mean\left( {{P_{YY}}\left( f \right)} \right) \times mean\left( {\frac{{{P_{XX}}\left( f \right)}}{{{P_{YY}}\left( f \right)}}} \right)}}\end{equation}$$ (3)

The most likely direction of causation is the one having significantly larger SDR values. (https://github.com/OpatzLab/HanganuOpatzToolbox/tree/master/LFPanalysis/getSDR.m)

Spiking analysis

Single units were automatically detected and clustered using the python-based software klusta (Rossant et al., 2016) and manually curated using phy (https://github.com/cortex-lab/phy). The firing rate was computed by dividing the total number of spikes by the duration of the analysed time window. To assess the spike probability, histograms of spike count using 1 ms bins were calculated for periods around the light pulse (50 ms before to 150 ms after) and normalized to the number of delivered light pulses. For peristimulus time histograms of odour-evoked firing rates, even trials were sorted relative to the peaks obtained from odd trials and averaged. Cross-covariance of spike trains was calculated as described previously (Gretenkord et al., 2019; Siapas et al., 2005). To classify the response of spiking activity as activated or inhibited, Wilcoxon's signed-rank tests (P < 0.05) for periods during odour stimulation versus before odour stimulation were performed across odour trails for each unit. Briefly, the cross-covariance for two spike trains N i ${N_i}$ and N j ${N_j}$ , was estimated from the cross-correlation histogram ( J i j T , b ( u ) ) $J_{ij}^{T,b}( u ))$ as follows:
q ̂ i j u = J i j T , b u b T P ̂ i P ̂ j , $$\begin{equation}{\hat q_{ij}}\left( u \right) = \frac{{J_{ij}^{T,b}\left( u \right)}}{{bT}} - {\hat P_i}{\hat P_j},\end{equation}$$ (4)
( b = $b{\mathrm{\;}} = {\mathrm{\;}}$ binsize, u = l a g $u{\mathrm{\;}} = {\mathrm{\;}}lag$ , T observation period, P ̂ i = N i ( T ) T ${\hat P_i} = \frac{{{N_i}( T )}}{T}$ , P ̂ j = N j ( T ) T ${\hat P_j} = \frac{{{N_j}( T )}}{T}{\mathrm{\;}}$ ). The standardized cross-covariance was calculated as
Q i j u = b T P i P j q ̂ i j u , $$\begin{equation}{Q_{ij}}\left( u \right) = \sqrt {\frac{{bT}}{{{P_i}{P_j}}}} \,{\hat q_{ij}}\left( u \right),\end{equation}$$ (5)
with P i , P j ${P_i},{P_j}$ being the mean firing rates. Only pairs of units with firing rates >0.05 Hz and significant standardized cross-variance were considered. The null hypothesis was rejected when | Q i j ( u ) | > Z α $|{Q_{ij}}( u )| &gt; \;{Z_\alpha }$ . ( Z α = 2 e r f 1 ( 1 α N l a g s ) ${Z_{{\alpha}}} = \sqrt 2 \;er{f^{ - 1}}( {\frac{{1 - \alpha }}{{{N_{lags}}}}} )$ ; two-tailed critical z value at level α = 0.01). The standardized mean cross-variance for one unit was calculated as
Q i u = 1 K j = 1 K Q i j u $$\begin{equation}{Q_i}\left( u \right) = \sqrt {\frac{1}{K}} \;\mathop \sum \limits_{j\; = \;1}^K {Q_{ij}}\left( u \right)\end{equation}$$ (6)
(K = number of units in 2. Region) and the mean for all unit pairs as: Q i ( u ) = 1 L 2 i = 1 L Q i ( u ) $\;{Q_i}( u ) = \;\frac{1}{{{L^2}}}\mathop \sum \nolimits_{i\; = \;1}^L {Q_i}( u )$ (L = number of units in 1. Region).

Modulation index

The modulation index (MI) of power and firing rate for light stimulation, odour stimulation, or DREADD manipulation was calculated as
M I = V a l u e S t i m V a l u e P r e V a l u e S t i m + V a l u e P r e . $$\begin{equation}MI = \frac{{Valu{e_{Stim}} - Valu{e_{Pre}}}}{{Valu{e_{Stim}} + Valu{e_{Pre}}}}.\end{equation}$$ (7)

Spike–LFP coupling

Phase locking of spiking units to network oscillations was assessed using a previously described algorithm (Siapas et al., 2005). For this, the LFP signal was bandpass filtered (2–4 Hz (respiration rhythm, RR), 4−12 Hz (theta), 12−30 Hz (beta), 30−80 Hz (gamma)) using a third-order Butterworth filter. The instantaneous phase was extracted using the Hilbert transform on the filtered signal. The coupling between spikes and network oscillations was tested for significance using the Rayleigh test for non-uniformity. For analysis of baseline properties (Fig. 1) only neurons that showed significant phase locking were considered for the analysis of the mean phase angle and the locking strength, which was calculated as mean resulting vector length (RVL). For paired comparison of RVLs (Figs 3FG and 7) all units with a firing rate higher than 0.01 Hz during baseline (18 min) and after C21 injection (18 min, DREADD manipulation) or more than 10 spikes before (Pre) and during (Stim) light ramp pulses were considered (https://github.com/OpatzLab/HanganuOpatzToolbox/blob/master/Spikes-LFPanalysis/getPPC_PLV.m).

Phase–amplitude coupling

Phase–amplitude coupling (PAC) between RR phase in the OB and beta band amplitude in the LEC and HP was calculated as previously described (Tort et al., 2010). Briefly, the LFP signals were bandpass filtered and the Hilbert transform was used to extract the phase and amplitude, respectively. Subsequently, the amplitude of the beta-filtered signal in LEC or HP was determined at each phase of the filtered OB signal. The phase was divided into 16 bins and the mean amplitude for each bin was calculated and normalized to the total number of bins. Normalized PAC was calculated as the deviation between an empirical and uniform amplitude distribution. PAC matrices were z-scored and the average was calculated for RR (2–3 Hz) – beta (12–30 Hz) coupling.

Experimental design and statistical analysis

Statistical analysis was performed in the MATLAB environment or R Statistical Software. Since every pup was recorded only once and manipulations were performed on the mouse level, the statistical analysis considered each mouse as an experimental unit. As none of the data sets were normally distributed, data were tested for significance using a Wilcoxon's rank-sum test (two unrelated samples) or Wilcoxon's signed-rank test (two related samples). Data (except phase values) are presented as median (med) and interquartile range (iqr). Individual numbers of neurons (nUnits) and animals (nanimals) are reported alongside the test statistics in the text. Tables with the number of modulated units per animal, brain area and manipulation are included in the supplementary Excel sheet, SupplementaryDataSheet.xlsx. Outlier removal was applied to paired data points if the distance of their difference from the 25th or 75th percentile exceeds 2.5 times the interquartile interval of their difference. Cre littermates were used as controls. Phase locking was tested for significance using the Rayleigh test for non-uniformity. Phase angles were compared using a circular non-parametric multi-sample test for equal medians. Differences in proportions were tested using Fisher's exact test. Nested data were analysed with (generalized) linear mixed-effect models (GLMER and LMEM) using animals as a random effect. Proportions (e.g. the proportion of locked units) were fitted with generalized linear (mixed-effect) models (family = Binomial, link = logit). Significance levels *P < 0.05, **P < 0.01 and ***P < 0.001 were considered. If not included in the text, values and corresponding statistics for all presented data are included in the Excel table: SupplementaryDataSheet.xlsx.

Results

Oscillatory activity in OB entrains the network activity in limbic circuits of neonatal mice

To get first insights into the impact of OB activity on developing cortical circuits including the LEC, HP and PFC, we performed acute recordings of LFP and SUA simultaneously from all four brain areas in non-anaesthetized head-fixed neonatal (P8–10) mice (nanimals = 56, Fig. 1A, B) and assessed the temporal relationships between network oscillations and neuronal firing. All investigated areas showed discontinuous oscillatory activity in the theta–beta range (Brockmann et al., 2011; Gretenkord et al., 2019; Hartung, Brockmann et al., 2016), accompanied by continuous low amplitude slow frequency oscillations peaking at 2−4 Hz (respiration rhythm, RR). As previously described, the RR is coherent with the breathing rhythm of the mouse and dependent on nasal input (Gretenkord et al., 2019; Kostka et al., 2020). It is most prominent in the OB but can also be found in several cortical and subcortical brain areas including the LEC, HP and PFC (Tort et al., 2018, 2021; Gretenkord et al., 2019). To quantify the coupling of the OB to the cortical areas, we calculated the imaginary coherence (Fig. 1C). While a high level of synchrony linked OB with all investigated cortical areas, the strength of coupling was frequency-dependent, having the highest magnitude in the beta frequency range for OB–LEC and OB–HP and in the RR frequency band for OB–LEC and OB–PFC (Fig. 1C). The beta band imaginary coherence between OB and LEC was higher when compared with the previously reported values in anaesthetized mice (Gretenkord et al., 2019) as anaesthesia shifts the OB–LEC synchronization towards lower frequencies.

Details are in the caption following the image
Figure 1. Functional beta band coupling between neonatal OB, LEC, HP and PFC
A, top, schematic of recording configuration for simultaneous extracellular recordings in the OB, LEC, HP and PFC. The positions of recording sites were displayed superimposed on the corresponding brain areas (Brainrender, Claudi et al., 2021). Bottom, digital photomontages displaying the DiI-labelled (red) electrode tracks in DAPI (blue)-stained slices of the OB, LEC, HP and PFC of a P10 mouse. B, representative beta band-filtered LFP traces recorded in the OB, LEC, HP and PFC displayed together with the wavelet spectra of beta oscillations recorded simultaneously. C, mean spectra of imaginary coherence were calculated for OB–LEC (yellow, nanimals = 42), OB–HP (green, nanimals = 38) and OB–PFC (blue, nanimals = 40). The grey lines correspond to the significance threshold as assessed by Monte Carlo simulation. Bounded lines correspond to the SEM. D, top, polar plots displaying the phase locking of significantly locked units in the OB (red, nunits = 640 from 50 mice), LEC (yellow, nunits = 42 from 19 mice), HP (green, nunits = 51 from 28 mice) and PFC (blue, nunits = 26 from 16 mice) to beta oscillations in the OB. Grey numbers indicate the radius of the inner and outer circles of the polar plot. Bottom, histograms of mean phase angle for significantly phase-locked OB (red), LEC (yellow), HP (green) and PFC (blue) units. Histograms are replicated over two OB beta cycles (grey curve) (***P < 0.001, Rayleigh test for non-uniformity). E, plots of standardized mean spike–spike cross-covariance for OB–LEC (yellow, nunit pairs = 100), OB–HP (green, nunit pairs = 164) and OB–PFC (blue, nunit pairs = 182). Negative lags indicate that spiking in the first brain area precedes spiking in the second brain area. F, spectral dependency ratio calculated for the beta band of OB–LEC (yellow, nanimals = 42), OB–HP (green, nanimals = 40) and OB–PFC (blue, nanimals = 38). Grey dots and lines correspond to individual animals (*P < 0.05, **P < 0.01, ***P < 0.001, Wilcoxon's signed-rank test). HP: hippocampus; LEC: lateral entorhinal cortex; LFP: local field potential; OB: olfactory bulb; PFC: prefrontal cortex. [Colour figure can be viewed at wileyonlinelibrary.com]

Next, we calculated the phase locking of SUA recorded in the OB, LEC, HP and PFC to beta band oscillations (12–30 Hz) in OB (Fig. 1D). Significantly locked OB units fired shortly before the trough of the beta cycle (phase: 0.763 ± 0.006 π, 51.6 ± 3.27% (640/1124) of units from 50 mice, P < 0.0001, Rayleigh test for non-uniformity), while a small proportion of LEC and HP units were also locked to significantly shifted phase angles (LEC: phase: 1.556 ± 0.038 π, 11.2 ± 3.1% (42/384) of units from 19 mice, P = 4.19 × 10−11; HP: phase: 0.315 ± 0.047 π, 17.7 ± 3.3% (51/384) of units from 28 mice, P = 3.68 × 10−5, Rayleigh test for non-uniformity; OB vs. LEC P = 4.58 × 10−4, OB vs. HP: P = 2.33 × 10−5, non-parametric multi-sample test for equal medians) (Fig. 1D). In contrast, units in LEC and HP fire preferentially at the trough of local entorhinal and hippocampal beta oscillations, respectively. These data suggest a phase shift from the OB to the LEC and HP. Solely the prefrontal firing showed no phase preference of locking to the oscillatory phase in the OB (phase: 1.715 ± 0.082 π, 6.9 ± 1.7% (26/318) units from 16 mice, P = 0.605, Rayleigh test for non-uniformity).

To test whether the communication between the OB and cortical areas is directed and whether the OB acts as a driving force within the circuit, we assessed the temporal relationship between the firing of beta-locked OB units and firing in cortical regions by calculating the standardized cross-covariance of unit pairs (Siapas et al., 2005). For unit pairs between the OB and LEC as well as the OB and HP, the peak of cross-covariance was at negative time lags, indicating that spiking in the OB preceded cortical firing (Fig. 1E). The cross-covariance of unit pairs OB–LEC and OB–HP predominantly peaked at negative lags (OB–LEC: negative: 52.3%, positive: 47.7%; OB–HP: negative: 52.9%, positive: 47.1%). Solely more unit pairs between the OB and PFC showed more peak positive lags than peak negative lags (negative: 38.5%, positive: 61.5%). As spike-dependent methods are strongly biased by the firing rate of investigated neurons, which is rather low in neonatal mice, we next used the SDR, a method that infers causal direction from time-series data (Shajarisales et al., 2015; Ramirez-Villegas et al., 2021), to confirm the directed communication between the OB and cortical areas in the beta band. SDR values for the beta band for OB → LEC were significantly higher than for LEC → OB (z = 5.370, nanimals = 42, P = 7.86 × 10−8, Wilcoxon's signed-rank test), supporting the conclusion that the OB drives the LEC. Further, the SDR analysis revealed a spectral dependency of HP as well as PFC on OB (OB → HP: z = 5.511, nanimals = 40, P = 3.57 × 10−8, OB → PFC: z = 5.112, nanimals = 38, P = 3.19 × 10−7, Wilcoxon's signed-rank test), suggesting the contribution of OB beta activity to the oscillatory entrainment of prefrontal and hippocampal circuits (Fig. 1F).

Thus, tight directed interactions between OB and cortical areas led to timed firing and oscillatory entrainment within downstream LEC–HP circuits.

Olfactory stimulation drives firing and network synchronization along the OB–LEC–HP–PFC pathway

In adult mice, olfactory sampling has been reported to induce firing and augment the synchrony within LEC–HP–PFC networks (Gourévitch et al., 2010; Igarashi et al., 2014; Lockmann et al., 2018; Symanski et al., 2022; Vanderwolf & Zibrowsk, 2001; Xu & Wilson, 2012). At neonatal ages, odour stimulation evokes beta oscillations in the PiR of non-anaesthetized neonatal rats (Zhang et al., 2021) and odour exposure led to increased beta band activity in the LEC of anaesthetized neonatal mice (Gretenkord et al., 2019). However, it is unknown whether odour sampling also engages the downstream areas, the HP and PFC, in neonatal mice. To address these questions, LFP and SUA were acutely recorded in the OB, LEC, HP and PFC of non-anaesthetized neonatal mice (n = 19) before, during, and after stimulation (2 s) with 1% limonene, a neutral odour presented 40 times (Fig. 2A). Odour stimulation significantly activated 57.9% of OB units (nunits = 280 from 15 mice, baseline firing rate 1.642 ± 0.109 Hz) (Fig. 2B, C) (P < 0.05, Wilcoxon's signed-rank test Pre vs. Stim for each unit) across the 40 trials. While most OB units (55.9 ± 7.3%, 162/280 units from 15 mice) in each mouse responded with an increase in their firing rate, 9.7 ± 2.7% (30/280 units from 15 mice) of them were inhibited in response to the odour stimulation (Fig. 2C). The augmented firing has also been detected in downstream areas, yet at a different latency from the first inhalation during odour exposure. Units in the LEC (latency max firing med: 140.0 ms, iqr: 93.9–287.1 ms, n = 35) responded faster than units in the HP (latency max firing med: 362.0 ms, iqr: 357.6–408.3 ms, n = 6, pLECvs.HP = 0.0149, Wilcoxon's rank-sum test) and the PFC (latency max firing med: 306.8 ms, iqr: 221.7–404.4 ms, n = 32, pLECvs.PFC = 0.0053, Wilcoxon's rank-sum test), the activation of which was long-lasting and spanned several inhalation cycles (Fig. 2B, D). The percentage of significantly activated units was higher for the LEC (nunits = 92 from eight mice, baseline firing rate 0.927 ± 0.122 Hz) (42.9 ± 11.8%, 35/92 units) and PFC (nunits = 96 from 15 mice, baseline firing rate 0.847 ± 0.074 Hz) (20.9 ± 5.9%, 23/96 units) for each mouse, when compared with HP (nunits = 78 from 14 mice, baseline firing rate 0.532 ± 0.046 Hz) (9.8 ± 7.2%, 6/78 units), where a few significantly inhibited units (14.3 ± 8.4%, 9/78 units) in response to odour stimulation, have also been detected (Fig. 2C).

Details are in the caption following the image
Figure 2. Effects of olfactory stimulation on firing and oscillatory activity in the OB, LEC, HP and PFC
A, representative LFP trace extracellularly recorded in the (a) OB, (b) LEC, (c) HP and (d) PFC during stimulation with 1% limonene accompanied by the corresponding wavelet spectrum. B, plots of mean odour-evoked firing rates for units in response to 1% limonene in (a) OB (nunits = 280 from 15 mice), (b) LEC (nunits = 92 from eight mice), (c) HP (nunits = 78 from 14 mice) and (d) PFC (nunits = 96 from 15 mice). Mean peristimulus histograms from even trials were aligned to the first inhalation during odour stimulation and ordered by peak firing time obtained from odd trials. C, left, MI of SUA of units in response to odour stimulation of the (a) OB (nunitrs = 280 from 15 mice), (b) LEC (nunits = 92 from eight mice), (c) HP (nunits = 78 from 14 mice) and (d) PFC (nunits = 96 from 15 mice). (Significantly activated units are shown in red, whereas significantly inhibited units in grey, P < 0.05, Wilcoxon's signed-rank test, not modulated units are shown in blue, the circle size corresponds to the firing rate of the units). Right, bar graphs displaying the percentage of activated and inhibited units per animal. D, z-scored firing rate of activated unit–odour pairs in the (a) OB (red), (b) LEC (yellow), (c) HP (green) and (d) PFC (blue) in response to the first inhalation during odour stimulation. E, (a) plots of MI for power during stimulation with 1% limonene for the OB (red, nanimals = 18), LEC (yellow, nanimals = 11), HP (green, nanimals = 18) and PFC (blue, nanimals = 14). (b) Mean MI of LFP power in different frequency bands for the OB (red, nanimals = 18), LEC (yellow, nanimals = 11), HP (green, nanimals = 18) and PFC (blue, nanimals = 14). Asterisks represent significance values (*P < 0.05, **P < 0.01, ***P < 0.001, Wilcoxon's signed-rank test to zero, dots correspond to individual animals) calculated for the averaged MI of power for the different frequency bands (shaded areas in (a)). Dotted grey line corresponds to MI = 0. F, plot of coherence before (grey) and during stimulation with 1% limonene for OB–LEC (yellow, nanimals = 12), OB–HP (green, nanimals = 19), and OB–PFC (blue, nanimals = 14). Asterisks represent significance values (*P < 0.05, **P < 0.01, ***P < 0.001, Wilcoxon's signed-rank test) calculated for the averaged coherence for the different frequency bands (shaded areas). HP: hippocampus; MI: modulation index; LEC: lateral entorhinal cortex; LFP: local field potential; OB: olfactory bulb; PFC: prefrontal cortex SUA: single-unit activity. [Colour figure can be viewed at wileyonlinelibrary.com]

To assess the impact of odour stimulation on the network activity in the OB and its downstream areas, we calculated the mean MI of power for each frequency band, a measure of the LFP power change during a manipulation when related to the values before it. This analysis revealed that odour stimulation broadly (2–40 Hz) increased the oscillatory power in all four brain areas (OB: nanimals = 18, RR: med: 0.411, iqr: 0.222–0.521, P = 3.27 × 10−4, theta: med: 0.475, iqr: 0.348–0.604, P = 1.96 × 10−4, beta: med: 0.478, iqr: 0.359–0.533, P = 1.96 × 10−4, gamma: med: 0.134, iqr: 0.084–0.190, P = 1.96 × 10−4; LEC: nanimals = 11, RR: med: 0.123, iqr: 0.002–0.198, P = 0.054, theta: med: 0.286, iqr: 0.241–0.362, P = 0.002, beta: med: 0.252, iqr: 0.188–0.360, P = 0.002, gamma: med: 0.089, iqr: 0.060–0.180, P = 0.002; HP: nanimals = 18, RR: med: 0.046, iqr: 0.004–0.155, P = 0.035, theta: med: 0.238, iqr: 0.102–0.306, P = 0.003, beta: med: 0.246, iqr: 0.124–0.363, P = 0.001, gamma: med: 0.082, iqr: 0.007–0.125, P = 0.039; PFC: nanimals = 14, RR: med: 0.209, iqr: 0.105–0.315, P = 1.22 × 10−4, theta: med: 0.269, iqr: 0.208–0.326, P = 3.66 × 10−4, beta: med: 0.251, iqr: 0.195–0.294, P = 2.44 × 10−4, gamma: med: 0.061, iqr: 0.042–0.091, P = 1.22 × 10−4; Wilcoxon's signed-rank test) (Fig. 2E). Additional synchrony between the OB and LEC (nanimals = 12; RR: P = 0.002; theta: P = 0.007; beta: P = 0.009; gamma: P = 0.016; Wilcoxon's signed-rank test), OB and HP (nanimals = 19, RR: z = −3.541, P = 3.98 × 10−4; theta: z = −2.093 P = 0.036; beta: z = −3.179, P = 0.002; gamma: z = 0.362, P = 0.717; Wilcoxon's signed-rank test) as well as the OB and PFC (nanimals = 14, RR: P = 3.66×10−4; theta: P = 0.009; beta: P = 0.02; gamma: P = 0.952; Wilcoxon's signed-rank test) increased most prominently in the RR and beta band in response to odour stimulation (Fig. 2F). The observed increase in power and synchrony was independent of the specific odour that was presented. These data demonstrate that odour exposure drives the oscillatory coupling of the OB and downstream limbic regions already at neonatal age.

Activation of M/TCs induces beta band oscillations in neonatal OB

To elucidate the mechanisms of directed communication between the OB and downstream cortical areas, we activated channelrhodopsin (ChR2)-transfected M/TCs by light and simultaneously monitored the network and neuronal activity in the LEC, HP and PFC of non-anaesthetized neonatal mice during acute extracellular recordings. ChR2 is a light-gated ion channel that upon blue light activation increases the excitability of the expressing neuron. Transfection of M/TCs was achieved using a cre-dependent virus vector (AAV9-Ef1a-DIO-hChR2(E123T/T159C)-EYFP) that was injected into the right OB of P1 Tbet-cre mice (Fig. 3A). ChR2-EYFP expression was reliably detected in M/TCs and their projections 7 days after injection (Fig. 3B).

Details are in the caption following the image
Figure 3. Effects of M/TC manipulation by light on single-unit entrainment and oscillatory activity in the OB
A, schematic of the experimental protocol. B, top, photograph of the dorsal (left) and ventral side (middle) of a brain from a Tbet-cre+ mouse showing EYFP expression in the OB and M/TC axonal projections (lateral olfactory tract) to the LEC, piriform transition area (APir), and cortical amygdala (CoA) (right). Bottom, digital photomontages displaying the DiI-labelled electrode track in the OB (left) and confocal images displaying the mitral cell layer (MCL) of the right OB at different magnifications (middle and right). C, (a) representative extracellularly recorded LFP in the OB displayed band pass filtered in different frequency bands and accompanied by the corresponding wavelet spectrum during ramp stimulation, as well as by the simultaneously recorded multi-unit activity (MUA) in the MCL. (b) Magnification of a beta oscillation during ramp stimulation and corresponding MUA trace. Grey lines indicate times of action potential occurrence. D, (a) power spectrum for the OB LFP before (orange) and during (red) ramp stimulation. The grey shaded area corresponds to the beta band (12–30 Hz). (b) Mean MI of LFP power in different frequency bands for cre+ (red, nanimals = 32) and cre (black, nanimals = 14) mice (cre+ vs. cre: **P < 0.01, ***P < 0.001, Wilcoxon's rank-sum test, dots correspond to individual animals). E, (a) raster plot of SUA in the OB before, during and after ramp stimulation. (b) Z-scored firing rate in response to ramp stimulation of units recorded in the OB of cre+ (red, nunits = 281 from 18 mice) and cre (black, nunits = 228 from 11 mice) mice. (c) Bar plots depicting the percentage of activated (red) and inhibited (grey) units during (Stim) and after (Post) ramp stimulation for cre+ (left) and cre (right) mice. F, (a) phase locking of OB units to respiration rhythm oscillations in the OB (nunits = 176 from 26 mice). Left, plot showing the difference in the direction of the resulting vector lengths for OB units before (Pre, orange) and during ramp stimulation (Stim, red) (***P < 0.001, circular non-parametric multi-sample test for equal medians). Right, polar plots displaying phase locking of OB units before (Pre, orange) and during ramp stimulation (Stim, red). The mean resulting vectors are shown as blue lines. (***P < 0.001, Rayleigh test for non-uniformity). Grey numbers indicate the radius of the inner and outer circles of the polar plot. (b) Violin plots displaying the resulting vector length of OB units before (Pre, orange) and during ramp stimulation (Stim, red). Grey dots and lines correspond to individual units. (***P < 0.001, linear mixed-effect model). (c) Percentage of significantly locked units before (Pre, yellow) and during (Stim, red) stimulation. (***P < 0.001, generalized linear mixed-effects models). G, same as (F) for beta oscillations (nunits = 176 from 26 mice). LEC: lateral entorhinal cortex; LFP: local field potential; OB: olfactory bulb; PFC: prefrontal cortex; MI: modulation index; M/TC: mitral/tufted cell. [Colour figure can be viewed at wileyonlinelibrary.com]

Ramp light stimuli of increasing intensity (473 nm, total duration 3 s, 30−60 trials) were used to activate M/TCs in the OB of P8–10 mice (Fig. 3A). The stimulation parameters have been set in line with previous data (Bitzenhofer, Ahlbeck, Hanganu-Opatz et al., 2017) to prevent not only firing as a result of tissue heating but also artificially synchronous firing patterns and large stimulation artefacts. Ramp stimulation led to a sustained increase of spike discharge and broadband (4–100 Hz) LFP power augmentation in OB that peaked in the beta frequency range (12–30 Hz) (Fig. 3C, D). In cre+ mice, the MIs for theta, beta and gamma power were significantly increased and different from those calculated for cre animals (cre+ vs. cre: RR: z = 1.349, P = 0.177, theta: z = 3.235, P = 0.001, beta: z = 5.336, P = 9.52 × 10−8, gamma: z = 5.336, P = 9.52 × 10−8, ncre+ = 32, ncre = 14, Wilcoxon's rank-sum test) (Fig. 3Db). Additionally, SUA is strongly augmented during ramp stimulation (Fig. 3E, F). SUA abruptly increased with ramp onset (76.2% of units activated significantly, 2.8% units inhibited significantly, P < 0.05, Wilcoxon's signed-rank test Pre vs. Stim for each unit) in all OB layers and decreased post-stimulus (8.2% of units activated significantly, 29.9% of units inhibited significantly, P < 0.05, Wilcoxon's signed-rank test Pre vs. Post for each unit, activated units: Stim: 72.3 ± 6.9% vs. Post: 10.4 ± 5.9% from 18 mice, main time effect: 4.235, Cl: [3.627 4.916], P < 0.00001, GLMER). Ramp stimulation significantly modulated the neuronal firing for cre+ but not cre mice (Fig. 3Eb, c) (cre+: proportion of modulated units: 0.790, CI: [0.739 0.834], cre: 0.009, CI: [0.002 0.03]; statistical threshold: 0.05, CI: 95% confidence intervals for a binomial model).

To assess the temporal relationship between neuronal firing and beta oscillations in OB, we calculated the locking of SUA firing to the oscillatory phase before (Pre) and during (Stim) light stimulation (Fig. 3Cb). Ramp stimulation caused a significantly stronger locking of OB units to beta oscillations (Pre: med: 0.147, iqr: 0.094–0.227, P = 1.303 × 10−31, Rayleigh test for non-uniformity; Stim: med: 0.180, iqr: 0.103–0.294, P = 3.094 × 10−10, Rayleigh test for non-uniformity; Stim vs. Pre: nunits = 176 from 26 mice, P = 9.39 × 10−5, LMEM) (Fig. 3Ga, b) and an augmentation of the proportion of significantly phase-locked units to the beta rhythm during ramp stimulation (Pre: 19.8 ± 4.8%, 29/176 units from 26 mice, Stim: 51.5 ± 5.8%, 97/176 units from 26 mice, main time effect: 2.054, Cl: [1.521 2.627], P = 2.84×10−13, GLMEM) (Fig. 3Gc). However, the preferred phase did not change (Fig. 3Ga) (Pre: 0.636 ± 0.027 π; Stim: 0.60 ± 0.021 π; P = 0.201, circular non-parametric multi-sample test for equal medians). Of note, the coupling of OB units to the RR phase changed during light stimulation (Pre: 0.785 ± 0.029 π; Stim: 1.428 ± 0.03 π; P = 1.610 × 10−6, circular non-parametric multi-sample test for equal medians) and was significantly weaker than during baseline periods (Pre: med: 0.129, iqr: 0.080–0.220, P = 1.606 × 10−5, Rayleigh test for non-uniformity; Stim: med: 0.097, iqr: 0.057–0.157, P = 2.499 × 10−4, Rayleigh test for non-uniformity; Pre vs. Stim: nunits = 176 from 26 mice, P = 7.782 × 10−6, LMEM) (Fig. 3Fa, b). In contrast, the proportion of locked units (Pre: 14.6 ± 3.2%, 26/176 units from 26 mice, Stim: 13.5 ± 2.5%, 25/176 units from 26 mice, main time effect: −0.046, Cl: [−0.646 0.551], P = 0.879, GLMEM) (Fig. 3Fc) and the power of RR oscillations were not altered upon light stimulation. Therefore, the larger beta power observed during ramp stimulation is accompanied by a change in the entrainment of M/TCs and interneurons from the RR to the beta rhythm.

These data indicate that the activation of M/TCs recruits the local circuitry in the OB and thereby organizes the OB network activity in the beta rhythm.

M/TC activation drives neuronal firing in the LEC and HP

To characterize the downstream effects of beta band entrainment of OB, we first analysed the organization of OB projections in neonatal mice. In line with morphological investigations in adult mice (Igarashi et al., 2012), we previously showed that MC axons are already present in the superficial layers of the LEC at neonatal age (Gretenkord et al., 2019). Entorhinal neurons in layer II/III strongly project to the HP and weakly to the PFC (Hartung, Brockmann et al., 2016; Xu et al., 2021). Here, we performed axonal tracing of M/TCs using the anterograde virus (AAV9-hSyn-hChR2(H134R)-EYFP) injected into the OB at P8. Simultaneously, we monitored the entorhinal neurons that project to the HP by using the retrograde virus (AAVrg-CamKIIa-mCherry) injected into the HP at P8 (Fig. 4A, B). At P18, MC axons expressing EYFP were present in layer I/II of the LEC and PiR (Fig. 4B). Additionally, mCherry-expressing HP-projecting neurons were identified in entorhinal layer II/III. These neurons send their apical dendrites to layer I of the LEC, where they collocate with MC axonal projections (Fig. 4B).

Details are in the caption following the image
Figure 4. Effects of optogenetic manipulation of M/TCs on single-unit firing
A, schematic of the experimental protocol used to trace MC axons and neurons projecting to HP. (Brainrender, Claudi et al., 2021). B, top, digital photomontages displaying EYFP (green) and mCherry (red) fluorescence in coronal slices including the OB (left, injection side of AAV9-hSyn-hChR2-EYFP), HP (right, injection site of AAVrg-CamKIIα-mCherry) and LEC (right). Note the co-expression of EYFP and mCherry in the LEC. Bottom left, EYFP and mCherry co-expression in the LEC are shown at larger magnification (dashed box). Bottom right, EYFP (left), mCherry (middle) and their co-expression are shown at larger magnification for a HP-projecting entorhinal neuron with dendrites targeting layer I. C, (a) spike probability of units in the OB (red), LEC (yellow), HP (green) and PFC (blue) after a 3 ms light pulse (473 nm) delivered to the OB. Numbers indicate the delay of the peak spike probability for each brain area. All OB units (including directly responding granule cell layer and mitral cell layer units and those subsequently activated) were considered. (b) Spike–spike cross-covariance for OB–LEC (yellow), OB–HP (green) and OB–PFC (blue). Negative lags correspond to OB activity driving spiking in other brain areas. HP: hippocampus; LEC: lateral entorhinal cortex; M/TC: mitral/tufted cell; OB: olfactory bulb; PFC: prefrontal cortex. [Colour figure can be viewed at wileyonlinelibrary.com]

Since these morphological data suggest that the OB interacts with downstream cortical areas, in a second step we monitored the functional impact of direct OB projections on limbic circuits. For this, we used pulse (3 ms) and ramp (3 s) blue light stimulations (473 nm) of transfected OB neurons and simultaneously recorded the neuronal activity in the LEC, HP and PFC. Pulse stimulation of M/TCs induced neuronal firing in all investigated brain areas, except the PFC (Fig. 4Ca). While the light-evoked OB firing rate had already sharply peaked 7−8 ms post-stimulus, the responses in the other brain areas were substantially broader and delayed (37 ms in the LEC, 45−60 ms in the HP). A second firing increase was detected in the OB after ∼28 ms and might reflect OB internal processing or feedback activation from downstream areas. To expand on these results, we used normalized cross-covariance analysis to uncover the temporal correlations between light-evoked spike trains in the investigated brain regions. The most prominent interaction was detected for OB–LEC, with OB firing preceding the entorhinal discharges (Fig. 4Cb). While having a similar directionality, the OB–HP cross-covariance peaked later and less precisely. The data give the first insights into the communication pathways relaying the information from M/TCs to LEC and, subsequently, to HP.

Thus, M/TC firing might drive the activation of entorhinal, and subsequently, hippocampal and prefrontal circuits.

M/TC activation boosts beta band coupling within downstream limbic circuits

The long-lasting effects of M/TC stimulation on the neuronal firing of downstream areas, LEC, HP and PFC, suggest that OB activation might act as a driving force for the generation of network oscillations in neonatal limbic circuits. To test this hypothesis, we paired ramp light stimulation of ChR2-transfected M/TCs with LFP recordings in the LEC, HP and PFC of P8–10 mice (Fig. 5A). Ramp stimulation of M/TCs significantly increased the MI of power for the beta band in the LEC, HP and PFC of cre+ compared with cre mice (LEC: z = 2.134, ncre+ = 22, ncre = 10, P = 0.033; HP: z = 2.311, ncre+ = 22, ncre = 11, P = 0.021; PFC: z = 3.635, ncre+ = 22, ncre = 8, P = 2.78 × 10−4; Wilcoxon's rank-sum test) (Fig. 5A and 5).

Details are in the caption following the image
Figure 5. Effects of optogenetic manipulation of M/TCs on network activity in the LEC, HP and PFC
A, representative LFP traces recorded in the OB, LEC, HP and PFC during ramp stimulation of ChR2-transfected M/TCs accompanied by the corresponding wavelet spectra. B, (a) plot of MI for power during ramp stimulation of oscillations in LEC for cre+ (yellow, nanimals = 22) and cre (black, nanimals = 10) mice. Asterisks represent significance values (*P < 0.05, Wilcoxon's rank-sum test) calculated for different frequency bands (shaded areas). (b) Same as (a) for HP (cre+: green, nanimals =22; cre: black, nanimals = 11). (c) Same as (a) for PFC (cre+: blue, nanimals = 22; cre: black, nanimals = 8). C, (a) left, imaginary coherence between the OB and LEC before (grey) and during (yellow, nanimals = 21) light stimulation. Asterisks represent significance values (*P < 0.05, **P < 0.01, ***P < 0.001, Wilcoxon's signed-rank test) calculated for different frequency bands (shaded areas). (b) Same as (a) for the OB and HP (green, nanimals = 21). (c) Same as (a) for the OB and PFC (blue, nanimals = 21). HP: hippocampus; LEC: lateral entorhinal cortex; LFP: local field potential; M/TC: mitral/tufted cell; OB: olfactory bulb; PFC: prefrontal cortex. [Colour figure can be viewed at wileyonlinelibrary.com]

Moreover, we assessed the degree of synchrony between the OB and cortical areas during light stimulation by calculating the imaginary part of coherence, a measure that is insensitive to false connectivity arising from volume conduction (Nolte et al., 2004). The imaginary coherence between the OB and LEC, the OB and HP as well as the OB and PFC increased during light activation of M/TCs, the most prominent effects being detected in the beta band range (Fig. 5C). Coherence was significantly elevated during ramp stimulation for the beta band for OB–LEC and OB–HP (OB–LEC: z = −3.076, nanimals = 21, P = 0.002; OB–HP: z = −2.589, nanimals = 21, P = 0.0096; Wilcoxon's signed-rank test). In contrast, RR, theta and gamma coherence did not change for OB–LEC and OB–HP (OB–LEC: nanimals = 21, RR: z = −0.191, P = 0.848, theta: z = 0.817, P = 0.414, gamma: z = 0.261, P = 0.794; OB–HP: nanimals = 21, RR: z = 0.052, P = 0.958, theta: z = −1.060, P = 0.289, gamma: z = 1.686, P = 0.092; Wilcoxon's signed-rank test). Moreover, OB–PFC coherence in beta and gamma bands increased in response to light stimulation (nanimals = 21; RR: z = −0.365, P = 0.715; theta: z = 0.330, P = 0.741; beta: z = −3736, P = 1.87 × 10−4, gamma: z = −3.076, P = 0.002; Wilcoxon's signed-rank test).

These results indicate that activation of M/TCs not only induces beta oscillations in the OB but also increases the 12−30 Hz oscillatory coupling between the OB and downstream cortical areas.

Inhibition of M/TC output reduces oscillatory power as well as neuronal firing in the LEC and HP

To elucidate whether M/TC activity is necessary for the generation of oscillatory activity in downstream areas, in a first set of experiments, we used inhibitory DREADDs (hM4D(Gi)) that block presynaptic vesicle release when activated by a selective artificial activator, like Compound 21 (Roth, 2016). We expressed hM4D(Gi) in M/TCs by cre-dependent virus vector injection (AAV9-EF1a-DIO-hM4D(Gi)-mCherry) at P1 (Fig. 6A). At P8, M/TC soma as well as their axons forming the lateral olfactory tract, which targets the posterior part of the cerebrum, expressed hM4D(Gi)-mCherry (Fig. 6B). We performed acute extracellular recordings of LFPs and SUA from the OB, LEC and HP of non-anaesthetized P8–10 mice (nanimals = 35) before (baseline, 20 min) and after (40 min) subcutaneous injection of Compound 21 (C21, 3 mg/kg), a synthetic and selective activator of DREADDs (Thompson et al., 2018) (Fig. 6A). Since the impact of OB activation on the PFC was rather weak, we did not monitor its activity during OB silencing.

Details are in the caption following the image
Figure 6. Effects of silencing M/TC output by inhibitory DREADDs on the oscillatory activity in OB, LEC and HP
A, top, schematic of the experimental protocol. Bottom, schematic of recording configuration for simultaneous extracellular recordings in OB, LEC and HP (Brainrender, (Claudi et al., 2021). B, (a) photograph of the dorsal (left) and ventral side (right) of a brain from a P8 Tbet-cre+ mouse showing mCherry (red) expression in the OB and M/TC axonal projections (lateral olfactory tract) to the piriform cortex and LEC. (b) Digital photomontages displaying the DiI-labelled electrode track (red) in DAPI (blue)-stained slices including the OB (left), LEC (middle) and HP (right) from a P10 mouse. (c) Confocal images displaying the MCL of the right OB at different magnifications. MC bodies, as well as dendrites, express mCherry. C, representative LFP traces recorded in the OB, LEC and HP during C21 injection accompanied by the corresponding wavelet spectra for a cre+ mouse (top) and a cre mouse (bottom). D, colour-coded MI of LFP power in OB (left), LEC (middle) and HP (right) of cre+ (top) and cre mice (bottom) before and after C21 injection. Vertical red lines correspond to the C21 injection. E, MI of LFP power averaged for different frequency bands for cre+ (coloured) and cre (black) mice for OB (left, ncre+ = 17, ncre+ = 18), LEC (middle, ncre+ = 13, ncre+ = 12) and HP (right, ncre+ = 10, ncre+ = 9) (cre+ vs. cre: *P < 0.05, **P < 0.01, *** P < 0.001, Wilcoxon's rank-sum test). F, colour-coded minute-by-minute z-scored firing rates before and after C21 injection for cre+ and cre mice in OB (left, cre+: nunits = 409 from 15 mice, cre: nunits = 399 from 18 mice), LEC (middle, cre+: nunits = 128 from nine mice, cre: nunits = 43 from six mice) and HP (right, cre+: nunits = 65 from 10 mice, cre: nunits = 51 from seven mice). Vertical red lines correspond to the C21 injection. G, top, MI of SUA in response to odour stimulation mice in OB (left, cre+: nunits = 409 from 15 mice, cre: nunits = 399 from 18 mice), LEC (middle, cre+: nunits = 128 from nine mice, cre: nunits = 43 from six mice) and HP (right, cre+: nunits = 65 from 10 mice, cre: nunits = 51 from seven mice) of cre+ and cre mice. Significantly activated units are shown in red, whereas significantly inhibited units in grey, P < 0.05, Wilcoxon's signed-rank test, not modulated units are shown in blue, the circle size corresponds to the firing rate of the units. Bottom, bar graphs show the percentage of activated and inhibited units from cre+ and cre mice. HP: hippocampus; LEC: lateral entorhinal cortex; LFP: local field potential; MI: modulation index; M/TC: mitral/tufted cell; OB: olfactory bulb; PFC: prefrontal cortex; SUA: single-unit activity. [Colour figure can be viewed at wileyonlinelibrary.com]

C21 caused broadband power reduction in the OB that reached a maximum magnitude within 5 min after the injection (cre+ vs. cre, RR: z = −1.304, P = 0.192, theta: z = −2.228, P = 0.026, beta: z = −3.812, P = 0.0001, gamma: z = −2.459, P = 0.014, ncre+ = 17, ncre = 18, Wilcoxon's rank-sum test) (Fig. 6CE) and persisted for at least 2 h (data not shown). Solely the continuous RR in OB was not affected by the activation of DREADDs (ncre+ = 17, ncre = 18, z = −1.304, P = 0.192) (Fig. 6E). In line with the main action of inhibitory DREADDs to reduce the vesicle release in the expressing neurons, while having little, if any, impact on their ability to generate action potentials (Roth, 2016; Stachniak et al., 2014), C21 injection decreased the neuronal firing only in a subpopulation of OB units (48.9 ± 7.0%, 193/409 units from 15 mice) (Fig. 6F). This fraction was comparable to the number of units from cre mice which reduced their firing rates in response to C21 injection (43.8 ± 6.8%, 186/399 units from 18 mice, main line effect: 0.402, Cl: [−0.513 1.365], P = 0.383, GLMEM) (Fig. 6G).

To assess the temporal relationship between OB spikes and oscillatory events in the OB, the phase locking of SUA to RR and beta rhythm was calculated. The phase coupling to RR (baseline: med: 0.094, iqr: 0.051–0.153; C21: med: 0.159, iqr: 0.078–0.310; nunits = 524 from 16 mice, P = 2.20 × 10−16, LMEM) (Fig. 7Aa, c) and the fraction of significantly locked units to the RR (baseline: 42.5 ± 5.8%, 241/524 units, C21: 64.0 ± 7.1%, 377/524 units from 16, main time effect: 1.316, Cl. [1.032 1.607], P < 0.00001, GLMEM) were increased after C21 injection (Fig. 7Ab) in cre+ but not cre mice (baseline: med: 0.097, iqr: 0.052–0.187; C21: med: 0.107, iqr: 0.062–0.204; nunits = 382 from 13 mice, P = 0.153, LMEM; baseline: 50.1 ± 5.9%, 206/382 units, C21: 54.1 ± 6.3%, 225/382 from 13 mice, main time effect: 0.236, Cl: [−0.073 0.546], P = 0.135, GLMEM). In contrast, phase locking to beta oscillations was only slightly reduced after C21 injection in cre+ (baseline: med: 0.105, iqr: 0.061–0.152; C21: med: 0.093, iqr: 0.053–0.139; nunits = 524 from 16 mice, P = 0.003, LMEM) but not affected in cre mice (baseline: med: 0.091, iqr: 0.047–0.134; C21: med: 0.090, iqr: 0.053–0.146; nunits = 382 from 13 mice, P = 0.498, LMEM) (Fig. 7Ba, c). Moreover, C21 did not affect the number of significantly locked units (cre+: baseline: 48.6 ± 4.6%, 273/524 units. C21: 48.3 ± 4.7%, 272/524 units from 16 mice, main time effect: −0.008, Cl: [−0.255 0.239], P = 0.95, GLMEM; cre: baseline: 54.0 ± 2.2%, 204/382 units, C21: 48.8 ± 6.3%, 199/382 units from 13 mice, main time effect: −0.054, Cl: [−0.341 0.234], P = 0.714, GLMEM) (Fig. 7Bb).

Details are in the caption following the image
Figure 7. Phase locking of OB firing to LFP oscillations in the OB after silencing the M/TC output by inhibitory DREADDs
A, (a) phase locking of OB units to respiration rhythm oscillations in the OB. Polar plots displaying phase locking of OB units from cre+ (top, nunits = 524 from 16 mice) and cre mice (bottom, nunits = 382 from 13 mice) before (baseline) and after C21 injection (C21). The mean resulting vectors are represented by a blue line. (***P < 0.001, Rayleigh test for non-uniformity). Grey numbers indicate the radius of the inner and outer circles of the polar plot. (b) Bar plots displaying the percentage of significantly locked units for cre+ (coloured, nunits = 524 from nanimals = 16) and cre mice (grey, nunits = 382 from nanimals = 13) before (baseline) and after (C21) C21 injection (*P < 0.05, Fisher's exact test). (c) Violin plots displaying the resulting vector length of OB units for cre+ (top, coloured, nunits = 524 from 16 mice) and cre mice (top, coloured, nunits = 382 from 13 mice) before (baseline) and after C21 injection (C21). Grey dots and lines correspond to individual units (**P < 0.01, ***P < 0.001, linear mixed-effect model). B, same as (A) for beta oscillations in the OB of cre+ (nunits = 524 from 16 mice) and cre (nunits = 382 from 13 mice) mice. LFP: local field potential; M/TC: mitral/tufted cell; OB: olfactory bulb. [Colour figure can be viewed at wileyonlinelibrary.com]

Next, we monitored the effects of chemogenetic silencing of M/TCs on the downstream brain areas, the LEC and HP. Silencing the M/TC output led to a broadband reduction of oscillatory power in the LEC and HP (Fig. 6D and 6) (LEC: ncre+ = 13, ncre = 12, RR: z = −3.455, P = 0.0006, theta: z = −3.40, P = 0.0007, beta: z = −3.455, P = 0.0006, gamma: z = −3.455, P = 0.0006; HP: ncre+ = 10, ncre = 9, RR: P = 0.211, theta: P = 0.013, beta: P = 0.017, gamma: P = 0.028; Wilcoxon's rank-sum test) (Fig. 6E). Moreover, silencing the M/TC output significantly inhibited more LEC units in cre+ than in cre mice (cre+: 68.7 ± 9.9%, 80/129 units from nine mice, cre: 32.3 ± 15.1%, 13/43 from six mice main line effect: 1.831, Cl: [−0.009 4.054], P = 0.048, GLMEM) and strongly reduced the firing of LEC units (cre+: −0.319, iqr: −0.564–0.005, P < 0.0001, LMEM, nunits = 125 from nine mice, cre: med: 0.091, iqr: −0.111–0.474; nunits = 43 from six mice; P = 0.011, LMEM, cre+ vs. cre: P = 0.04, LMEM), the effects lasting >1 h after C21 injection (Fig. 6F). In contrast, silencing of M/TC output had a shorter (∼20 min) and weaker impact on hippocampal firing (Fig. 6F, G) (cre+: med:−0.103, iqr: −0.598–0.111, nunits = 65 from 10 mice, P = 0.0016, LMEM; cre: med: 0.039, iqr: −0.229–0.260, nunits = 51 from seven mice, P = 0.194, LMEM; cre+ vs. cre: P = 0.131, LMEM) with around half (51.5 ± 10.8%) of HP units being significantly inhibited in response to C21 injection (Fig. 6G).

These results indicate that silencing the M/TC output decreases the oscillatory power of beta oscillations and neuronal firing in LEC, as a first downstream station of OB projections. In its turn, the weaker drive from the LEC might lead to poorer oscillatory entrainment of HP activity, yet with a less pronounced impact on its neuronal firing.

Selective inhibition of MC axonal terminals in the LEC by light reduces the oscillatory activity in the HP and PFC

Since the impact of DREADD-induced silencing of OB synaptic outputs on the LEC and HP activity results from both mono- and polysynaptic connectivity, in a second set of experiments we dissected the role of the direct monosynaptic OB-to-LEC pathway using recently developed inhibitory opsins. We expressed the targeting-enhanced mosquito homologue of the vertebrate encephalopsin (eOPN3) that selectively suppresses neurotransmitter release at presynaptic terminals through the activation of the Gi/o signalling pathway (Mahn et al., 2021) into M/TCs (Fig. 8A). Light stimulation (473 nm) of eOPN3-expressing MC terminals in LEC (Fig. 8B) reduced the power of entorhinal beta band activity (med: −0.214, iqr: −0.277 to −0.095, P = 0.008, nanimals = 9, Wilcoxon's signed-rank test), yet not of RR, as well as theta-band and gamma-band oscillations in the LEC (RR: P = 0.496, theta: P = 0.098, gamma: P = 0.074, nanimals = 9, Wilcoxon's signed-rank test) (Fig. 8C and 8). Consequently, the optogenetic silencing led to a reduction of oscillatory power in the HP and PFC in a frequency range spanning from RR to beta (HP: RR: P = 0.016, theta: P = 0.031, beta: P = 0.016, gamma: 0.078, nanimals = 7; PFC: RR: P = 0.016, theta: 0.047, beta: P = 0.047, gamma: P = 0.156, nanimals = 7; Wilcoxon's signed-rank test) (Fig. 8C and 8 and 8). Of note, beta and gamma power in the OB was reduced following the silencing of MC terminals in the LEC (beta: med: −0.139, iqr: −0.256 to −0.074, P = 0.020; gamma: med: −0.075, iqr: −0.141 to −0.041, P = 0.027; nanimals = 9; Wilcoxon's signed-rank test) (Fig. 8C and 8) most likely as a result of reduced centrifugal feedback from the LEC and HP (Kostka & Bitzenhofer, 2022).

Details are in the caption following the image
Figure 8. Light-induced inhibition of MC axonal terminals in the LEC
A, schematic of recording configuration for simultaneous extracellular recordings in the OB, LEC, HP and PFC (Brainrender, (Claudi et al., 2021). B, photograph of a DAPI (blue)-stained slice from a P8 Tbet-cre+ mouse showing mScarlet (red) expression in MC axonal projections (lateral olfactory tract) to the LEC in two different magnifications. C, representative LFP traces recorded in the OB, LEC, HP and PFC during stimulation of OPN3-transfected M/TCs accompanied by the corresponding wavelet spectra. Vertical black lines correspond to 100 μV. D, (a) plot of MI for OB power during optogenetic activation of OPN3 in MC axon terminals (red, nanimals = 9). (b) Mean MI of LFP power in different frequency bands for OB. E, same as (D) for LEC (yellow, nanimals = 9). F, same as (D) for HP (green, nanimals = 7). G, same as (D) for PFC (blue, nanimals = 7). Asterisks represent significance values (*P < 0.05, **P < 0.01, ***P < 0.001, Wilcoxon's signed-rank test to zero, dots correspond to individual animals) calculated for different frequency bands (shaded areas in (a)). Dotted grey line corresponds to MI = 0. HP: hippocampus; LEC: lateral entorhinal cortex; LFP: local field potential; MI: modulation index; M/TC: mitral/tufted cell; OB: olfactory bulb; PFC: prefrontal cortex. [Colour figure can be viewed at wileyonlinelibrary.com]

Thus, the direct axonal projections from MCs to the LEC strongly contribute to the oscillatory entrainment of downstream limbic brain areas, such as the HP.

Inhibition of M/TC output reduces the communication between the OB and downstream cortical areas

To back up the hypothesis that the M/TC activity entrains limbic circuits in beta oscillations, we monitored the communication between the OB and downstream areas during the silencing of M/TC output with DREADDs by using three distinct measures. First, we assessed the synchrony between the OB, LEC and HP by calculating the imaginary coherence in different frequency bands before (baseline) and after C21 injection (C21) (Fig. 9A, B). Beta and gamma coherence between the OB and LEC as well as beta coherence between the OB and HP were significantly reduced after C21 injection (OB–LEC: nanimal = 12, beta baseline: med: 0.048, iqr: 0.037–0.108, C21: med: 0.033, iqr: 0.028–0.050, P = 0.012; gamma: baseline: med: 0.037, iqr: 0.032–0.051, C21: med: 0.044, iqr: 0.032–0.072, P = 0.021; OB–HP: n = 10, beta: baseline: med: 0.063, iqr: 0.042–0.076; C21: med: 0.040, iqr: 0.035–0.046, P = 0.006, Wilcoxon's signed-rank test). In contrast, the coherence in other frequency bands was not affected by C21 injection (OB–LEC: nanimals = 12, RR: baseline: med: 0.065, iqr: 0.020–0.111, C21: med: 0.053, iqr: 0.040–0.074, P = 0.569; theta: baseline: med: 0.060, iqr: 0.033–0.089, C21: med:0.047, iqr: 0.034–0.067, P = 0.424; OB–HP: nanimals = 10, RR: baseline: med: 0.044, iqr: 0.036–0.070, C21: med: 0.077, iqr: 0.040–0.094, P = 0.160; theta: baseline: med: 0.045, iqr: 0.029–0.073, C21: med: 0.046, iqr: 0.039–0.069, P = 1; gamma: baseline: med: 0.028, iqr: 0.024–0.035, C21: med: 0.031, iqr: 0.029–0.043, P = 0.065; Wilcoxon's signed-rank test) (Fig. 9A). Moreover, C21 injection did not change the coherence between the OB and LEC (nanimals = 8, RR: baseline: med: 0.076, iqr: 0.055–0.080, C21: med: 0.054, iqr: 0.034–0.073, P = 0.461, theta: baseline: med: 0.052, iqr: 0.035–0.065, C21: med: 0.039, iqr: 0.029–0.055, P = 0.055, beta: baseline: med: 0.042, iqr: 0.036–0.057, C21: med: 0.052, iqr: 0.035–0.072, P = 0.109, gamma: baseline: med: 0.054, iqr: 0.035–0.072, C21: med: 0.070, iqr: 0.035–0.102, P = 0.109; Wilcoxon's signed-rank test) and OB and HP of cre mice (nanimals = 8, RR: baseline: med: 0.046, iqr: 0.030–0.068, C21: med: 0.054, iqr: 0.038–0.098, P = 0.148, theta: baseline: med: 0.073, iqr: 0.028–0.083, C21: med: 0.064, iqr: 0.054–0.074, P = 0.945, beta: baseline: med: 0.056, iqr: 0.035–0.076, C21: med: 0.036, iqr: 0.028–0.055, P = 0.148, gamma: baseline: med: 0.029, iqr: 0.027–0.038, C21: med: 0.026, iqr: 0.021–0.035, P = 0.055; Wilcoxon's signed-rank test).

Details are in the caption following the image
Figure 9. Modulation of functional communication within olfactory–cortical networks through silencing the M/TC output by inhibitory DREADDs
A, (a) imaginary coherence calculated for OB–LEC in cre+ mice (left, yellow, nanimals = 12), and in cre mice (right, grey, nanimals = 8), before (baseline, light grey) and after C21 injection (C21, dark coloured). (b) Same as (a) for OB–HP (cre+: green, nanimals = 10, cre: grey, nanimals = 8) (horizontal black line: P < 0.05, Wilcoxon's signed-rank test). B, (a) z-scored PAC between OB phase and LEC (top, nanimals = 13) and HP (bottom, nanimals = 10) amplitude, before (baseline) and after C21 injection (C21). (b) PAC averaged for respiration rhythm–beta coupling (black box in (a)) for OB–LEC (top, cre+: nanimals = 12, cre: nanimals = 11) and OB–HP (bottom, cre+: nanimals = 10, cre: nanimals = 10), before (baseline, grey) and after C21 injection (coloured) for cre+ and cre mice. Dotted grey line corresponds to a z-score of 1.96. (*P < 0.05, Wilcoxon's signed-rank test). C, (a) SDR calculated for the band pass-filtered LFP (1–100 Hz) in OB and LEC (nanimals = 13) in cre+ mice. SDR values for OB → LEC and LEC → OB before (baseline, grey) and after C21 injection (C21, yellow). (b) Difference in SDR values for both directions for cre+ (left, yellow, nanimals = 13) and cre (right, grey, nanimals = 11) mice. D, same as (C) for OB and HP (cre+: green, nanimals = 10, cre: grey: nanimals = 9). Black dots and lines correspond to individual animals. (**P < 0.01, ***P < 0.001, Wilcoxon's signed-rank test). HP: hippocampus; LEC: lateral entorhinal cortex; LFP: local field potential; M/TC: mitral/tufted cell; OB: olfactory bulb; PAC: phase-amplitude coupling; SDR: spectral dependency ratio. [Colour figure can be viewed at wileyonlinelibrary.com]

Second, we calculated the PAC to elucidate the role of M/TCs in the modulation of cortical beta oscillations by the RR phase in the OB. C21 injection significantly reduced the z-scored PAC values between the OB RR phase and the amplitude of beta oscillations in the LEC (baseline: med: 2.499, iqr: 1.624–2.883; C21: med: 1.608, iqr: 0.674–2.361, nanimals = 13, P = 0.017, Wilcoxon's signed-rank test) and HP (baseline: med: 2.363, iqr: 2.135–2.764; C21: med: 1.907, iqr: 1.319–2.248, nanimals = 10, P = 0.037, Wilcoxon's signed-rank test) (Fig. 9B). Additionally, fewer mice showed significant RR-beta PAC values after C21 injection (z-score >1.96) in the LEC (baseline: 53.9% vs. C21: 39.8%) and HP (90% vs. 50%). C21 injection did not alter PAC values in cre mice (OB–LEC: baseline: med: 1.932, iqr: 1.544–2.20, C21: med: 1.750, iqr: 1.578–2.418, nanimals = 11, P = 0.738; OB–HP: baseline: 2.025, iqr: 1.645–2.206, C21: med: 1.839, iqr: 1.408–2.129, nanimals = 10, P = 0.625; Wilcoxon's signed-rank test).

Third, we tested the effect of C21 on the directionality of interactions between the OB, LEC and HP (Fig. 9C). We calculated the SDR and found that the prominent drive from the OB to the LEC during baseline periods (OB → LEC: med: 2.164, iqr: 1.769–2.778; LEC → OB: med: 0.366, iqr: 0.249–0.433; P = 0.0002, Wilcoxon's signed-rank test) was absent after silencing of M/TC output (OB → LEC: med: 1.211, iqr: 0.871–1.666; LEC → OB: med: 0.600, iqr: 0.378–0.854; P = 0.080, Wilcoxon's signed-rank test) the values for OB → LEC and LEC → OB being comparable (Fig. 9Ca) showing a significantly smaller difference than during baseline periods (baseline: med: 1.792, iqr: 1.300–2.512; C21: med: 0.619, iqr: 0.087–1.160; nanimals = 13; P = 0.001, Wilcoxon's signed-rank test) (Fig. 9Cb).

Similarly, the drive from OB to HP (OB → HP: med: 1.422, iqr: 1.257–2.076; HP → OB: med: 0.492, iqr: 0.367–0.618; P = 0.002, Wilcoxon's signed-rank test) was disrupted by C21 injection (OB → HP: med: 1.107, iqr: 0.709–1.605; HP → OB: med: 0.531, iqr: 0.430–1.101; P = 0.625, Wilcoxon's signed-rank test) (Fig. 9Da). This leads to a significantly smaller difference between SDR values for both directions after C21 injection (baseline: med: 0.891, iqr: 0.601–1.709; C21: med: 0.627, iqr: −0.303–1.175; nanimals = 13; P = 0.004, Wilcoxon's signed-rank test) (Fig. 9Db). In contrast, C21 injection did not affect the SDR values for OB–LEC (baseline: med: 1.192, iqr:−0.49 to −2.13, C21: med: 1.045, iqr: 0.479–2.531; nanimals = 11, P = 0.175, Wilcoxon's signed-rank test) and OB–HP (baseline: med: 1.348, iqr: 0.557–2.413, C21: med: 1.167, iqr: 0.553–1.708; nanimals = 9, P = 0.496, Wilcoxon's signed-rank test) of cre mice.

Thus, these results show that the M/TC activity is critical for the communication between the OB and its downstream cortical areas.

Discussion

Long-range interactions within limbic circuits emerge early in life (Chini & Hanganu-Opatz, 2021), yet it is still unknown whether the coordinated activity patterns underlying this coupling are endogenously generated or emerge as a result of the driving force of sensory systems. Several distinct activity patterns have been identified in developing cortical–hippocampal networks. Apart from discontinuous theta and beta oscillations synchronizing LEC–HP–PFC networks at neonatal age (Brockmann et al., 2011; Hartung, Brockmann et al., 2016; Xu et al., 2021), spontaneous twitches during active sleep drive early sharp waves in the developing HP (Mohns & Blumberg, 2010; Valeeva et al., 2019). Besides these muscle twitches (Del Rio-Bermudez et al., 2020) and passive tactile sensation, olfactory inputs are likely candidates for the instruction of limbic circuitry development during the first two postnatal weeks. Newborn rodents are not only able to smell from birth but, importantly, also use olfactory information for learning and cue-directed behaviours such as localization of the nipples of the dam (Logan et al., 2012; Welker, 1964). A first piece of evidence for the critical role of olfaction in limbic development is the fact that the neonatal OB shows functional coupling with the LEC, the gatekeeper of the limbic circuitry, during discontinuous network oscillations in the theta–beta frequency range as well as in the continuous respiration-related rhythm (RR) in anaesthetized mice (Gretenkord et al., 2019). Here, we extended these findings and uncovered that MC firing most likely plays a role in the beta band entrainment of downstream areas, such as the HP and PFC in awake neonatal mice (P8–10). The temporal dynamics of oscillatory and firing activity revealed that in periods with and without active odour sampling, the OB drives the activation of limbic circuits.

Generation of beta oscillations in the OB of neonatal mice

SUA analysis revealed that M/TC activation leads to an augmented firing rate in almost all OB units, indicating that OB interneurons, such as GCs, also increase their firing in response to M/TC activation. Experimental and modelling studies have shown that both beta and gamma oscillations in the OB rely on dendro-dendritic interactions between M/TCs and GCs (Fourcaud-Trocmé et al., 2014; Osinski & Kay, 2016; Osinski et al., 2018). In adults, the emergence of gamma and beta oscillations is controlled by different excitability states of GCs as well as their dependency on centrifugal input, with beta oscillations relying on a higher GC excitability and centrifugal feedback projections (David et al., 2015; Martin et al., 2007; Neville & Haberly, 2003; Osinski & Kay, 2016). However, gamma oscillations are absent in the neonatal OB, most likely as a result of the late functional integration of interneurons into local circuits and the different biophysical properties of MCs and GCs during development (Dietz et al., 2011; Fletcher et al., 2005; Yu et al., 2015). Instead, discontinuous beta band oscillations are present, not only in the developing OB, but also in other sensory and limbic areas (Bitzenhofer, Ahlbeck, Wolff et al., 2017). In the neonatal V1 and PFC, they have been shown to accelerate along development until reaching the gamma-band range at juvenile age (Bitzenhofer et al., 2020; Hoy and Niell, 2015; Li et al., 2012). Whether the beta band activity in the OB undergoes a similar transition to faster rhythms and how this process is controlled by interactions within the OB and by feedback projections from the PiR and LEC remain to be elucidated.

Influence of anaesthesia on network synchronization between the OB and LEC

Our previous investigations showed that in anaesthetized mice, the strongest OB-driven activation of neonatal LEC occurs in the theta frequency range (Gretenkord et al., 2019). Here, we show that in awake mice, the MC-driven synchronization of the OB and LEC shifts to the beta frequency range. Urethane anaesthesia has been shown to decrease beta and gamma oscillations in the OB of adult rats and to dampen spontaneous and odour-evoked GC responses, while enhancing MC firing in response to odours (Kato et al., 2012; Li et al., 2012). Thus, the anaesthesia-induced reduction in fast frequency oscillations might be due to decreased dendro-dendritic inhibition on MCs. As GCs are strongly targeted by centrifugal projections and beta oscillations in the OB are dependent on centrifugal inputs (Neville & Haberly, 2003), the absent beta synchronization between the OB and downstream cortical brain areas, during anaesthesia (Gretenkord et al., 2019), might result from reduced centrifugal input to the OB (Nunez-Parra et al., 2014).

Anatomical and functional connectivity between the OB and limbic brain areas

The present data show that the OB network activation entrains neuronal activity in downstream cortical areas of neonatal awake mice (P8–10). In adult rodents, axonal terminals of MCs have been found to target fan and pyramidal neurons in LII/III of the LEC that, in their turn, relay this information to the HP (Schwerdtfeger et al., 1990; Wouterlood & Nederlof, 1983). The axonal projections from layer II/III LEC pyramidal neurons to CA1 are involved in associative odour learning in adults (Li et al., 2017). Already at neonatal age, MC axons reach layer I of the LEC (Gretenkord et al., 2019; Walz et al., 2006). Here, projections of layer II/III LEC neurons that target the HP and might establish synaptic contacts with MC axons were detected. Optogenetic and olfactory stimulation revealed that the activation of M/TCs induced firing of LEC, HP and PFC neurons, indicating that the pathway OB-to-HP is already functional from birth. CA1 receives entorhinal input not only via the direct performant path, but also through the tri-synaptic path, spanning DG and CA3 (Basu et al., 2016). The long latency (∼60 ms) in light-induced CA1 firing might, therefore, be partly mediated by the tri-synaptic path. Further, the LEC receives also indirect inputs via the PiR (Kerr et al., 2007). While the rather long and broad activity increase in the LEC after MC stimulation might argue for PiR involvement, a prominent monosynaptic drive from the OB to LEC is present too, and controls the downstream HP, as demonstrated by selective OPN3-induced silencing.

Oscillatory entrainment of limbic networks upon OB activation during neonatal development

Coordinated activity patterns in the OB organized by MCs promote not only neuronal firing but also entrain downstream cortical areas in beta oscillations. Olfactory stimulation as well as ramp light stimulation of M/TCs leads to an increase in beta power in the OB, LEC, HP and PFC. Moreover, increased beta band coupling between OB and cortical areas emerged during olfactory and ramp stimulation, whereas inhibition of M/TCs vesicle release reduced the drive OB → LEC and OB → HP as well as RR-beta cross-frequency coupling and beta coherence between OB–LEC and OB–HP. Considering that the activation of MCs leads to a similar entrainment of the oscillatory activity in downstream areas as observed during olfactory sampling, these results identify the MC-driven beta rhythm as a potential mechanism of long-range communication between the OB and downstream cortical networks.

Influence of centrifugal inputs on interactions between the OB and limbic brain areas

Olfactory processing in adult mice is dynamically modulated by feedback projections from higher brain regions, which provide information about contextual factors, such as attention or behavioural state. Aside from neuromodulatory areas, limbic brain areas such as the cortical amygdala, CA1 of the ventral HP, and the LEC send direct projections to the OB (Brunert & Rothermel, 2021; Padmanabhan et al., 2019), where they mainly target inhibitory neurons (Boyd et al., 2012; Markopoulos et al., 2012). Even though the specific role of feedback projections from different limbic brain areas is far from being understood, it has been hypothesized that inputs from the LEC might provide information about recent experiences and multisensory events, while HP inputs might provide contextual information (Brunert & Rothermel, 2021). Even less well elucidated is the role of centrifugal inputs during development. However, recent experimental evidence supports their contribution to the here-described beta band coupling. Morphological assessment showed that centrifugal projections from the LEC and HP start to target the OB during the second postnatal week (Kostka & Bitzenhofer, 2022). They might account for the proportion of unit pairs with positive cross-covariance lags during spontaneous activity that we described above. Future studies should reveal the role of limbic feedback projections on odour sampling and subsequent odour processing during neonatal development.

Relevance of OB-controlled beta band activation of cortical circuits during early postnatal development

Beta oscillations have been reported to play a key role in working memory and decision making in adult humans (Spitzer & Haegens, 2017). Further, prominent beta band synchrony between cortical areas has been identified during olfactory-guided memory and decision-making tasks in rodents (Igarashi et al., 2014; Martin et al., 2007; Symanski et al., 2022). The firing of beta-entrained CA1 interneurons during these tasks is related to accurate performance, indicating that beta oscillations enable temporal coordination and recruitment of neurons within functional behaviourally relevant cell assemblies (Rangel et al., 2016; Symanski et al., 2022). In line with these experimental data, modelling confirmed that beta oscillations optimally contribute to the coupling of cell assemblies over long axonal conductance delays (Bibbig et al., 2002; Kopell et al., 2000, 2011). At neonatal age, mouse models of psychiatric risk or epilepsy show reduced beta band activity, which has been found to correlate with later cognitive deficits (Chini et al., 2020; Marguet et al., 2015; Xu et al., 2021). Interfering with beta band oscillations in the PFC during a defined developmental period causes network miswiring and poor behavioural performance in adult mice (Bitzenhofer et al., 2021). Thus, during development, discontinuous beta band events that have been identified in the PFC, HP and LEC might facilitate the formation of initial cell assemblies with relevance for cognitive performance later in life. Here, we identified the olfactory activity as a prominent driver of these early beta oscillations. Thus, olfactory sensory inputs, possibly in conjunction with other sensory signals, such as reafferent signalling following spontaneous twitches, could be important for the synchronization of limbic network activity during the second postnatal week. The results let us hypothesize that transient disturbance of neonatal olfactory processing precludes the functional refinement of entorhinal–hippocampal–prefrontal circuits, ultimately leading to cognitive deficits in adulthood. Further research is warranted to directly test this hypothesis and elucidate the role of early activity patterns in the OB for cognitive development.

Biography

  • image

    Johanna K. Kostka is currently postdoc at the Institute of Developmental Neurophysiology at the Center of Molecular Neurobiology in Hamburg. She received her bachelor's degree in Biophysics at the Humboldt University Berlin, her master's degree in Brain and Cognitive Science at the University of Amsterdam, and her PhD at the University of Hamburg. She is using in vivo electrophysiology, as well as opto- and chemogenetics to investigate network interactions between the olfactory bulb and limbic brain areas, such as the entorhinal cortex, the hippocampus and the prefrontal cortex, during neonatal development.

Data availability statement

The data that support the findings of this study are available in the supplementary material of this article.

Competing interests

The authors declare no competing interests.

Author contributions

I. L. H.-O. and J. K. K. conceived the study and designed the experiments. J. K. K. carried out the experiments and analysed the data. J. K. K. and I. L. H.-O. interpreted the data. J. K. K. and I. L. H.-O. wrote the article. Both authors discussed and commented on the manuscript.

Funding

This work was funded by grants from the German Research Foundation (Ha4466/11-1, Ha4466/20-1 and SFB 936 B5 to I.L.H.-O), European Research Council (ERC-2015-CoG 681577 to I.L.H.-O.), MSCA-ITN H2020-860563, Horizon 2020 DEEPER (101016787 to I.L.H.-O.), and Landesforschungsförderung Hamburg (LFF73 and LFF76 to I.L.H.-O.).

Acknowledgements

We thank A. Marquardt, A. Dahlmann, P. Putthoff and K. Titze for excellent technical assistance, Dr I. Braren from the Vector Facility of the UKE for the virus production as well as Drs M. Chini, S.H. Bitzenhofer and R.L. van den Brink for helpful discussions. Moreover, we thank Dr S.H. Bitzenhofer for building the olfactometer.

Open access funding enabled and organized by Projekt DEAL.

    Additional information