Network Dynamics in the Developing Piriform Cortex of Unanesthetized Rats

Abstract

The time course of changes in functional cortical activity during early development has been extensively studied in the rodent visual system. A key period in this process is the time of eye opening, which marks the onset of patterned visual input and active vision. However, vision differs from other systems in that it receives limited patterned sensory input before eye opening, and it remains unclear how findings from vision relate to other systems. Here, we focus on the development of cortical network activity in the olfactory system—which is crucial for survival at birth—by recording field potential and spiking activity from piriform cortex of unanesthetized rat pups from birth (P0) to P21. Our results demonstrate that odors evoke stable 10–15 Hz oscillations in piriform cortex from birth to P15, after which cortical responses undergo rapid changes. This transition is coincident with the emergence of gamma oscillations and fast sniffing behavior and preceded by an increase in spontaneous activity. Neonatal network oscillations and their developmental dynamics exhibit striking similarities with those previously observed in the visual, auditory, and somatosensory systems, providing insight into the network-level mechanisms underlying the development of sensory cortex in general and olfactory processing in particular.

Introduction

In most species, cortex undergoes protracted changes in cellular and synaptic structures during early postnatal life. As a result, the circuits that process sensory information—as well as the operations they perform—also undergo changes, but this process remains poorly understood (Avitan and Goodhill 2018). Using in vivo recordings, previous work in rodents demonstrates that cortical activity during the first 1–2 weeks of life consists of coordinated network oscillations that drive plastic changes in thalamocortical connectivity and help establish cortical sensory maps (Cang et al. 2005; Tiriac et al. 2012; Clause et al. 2014; Yang et al. 2016). Work on the visual system has further demonstrated that neonatal cortical network activity patterns undergo rapid changes around the time of eye opening (Rochefort et al. 2009; Colonnese et al. 2010; Colonnese 2014; Hoy and Niell 2015; Murata and Colonnese 2018). However, each sensory system has its own input patterns and developmental milestones, and findings obtained from one sensory system do not necessarily translate to other systems.

This study investigated changes in network activity in the olfactory (piriform) cortex of rats during early development. In contrast to vision, olfaction is crucial for the survival of neonatal rats (Miller and Spear 2009), a function it shares with somatosensation and taste. Previous work has demonstrated that olfactory transduction is functional at prenatal stages (Pedersen et al. 1983), relaying odor-evoked inputs to central structures by the time of birth (Astic and Saucier 1981; Mair and Gesteland 1982; Schwob et al. 1984; Fletcher et al. 2005; Illig 2007; Gretenkord et al. 2019). Moreover, primary olfactory cortex has been identified as an integral node in the network that mediates olfactory behaviors in neonates (Raineki et al. 2009; Morrison et al. 2013). These findings underscore a role for piriform cortex in neonatal odor processing, but the underlying mechanisms at the network level, and how they change over the course of development, remain unknown.

In order to characterize the development of activity patterns that reflect functional network-level operations, we recorded day-by-day changes in spontaneous and odor-evoked local field potential (LFP) and spiking activity from the piriform cortex of unanesthetized rat pups during the first 3 weeks of life. Given the importance of olfaction during early life, we predict that natural odor input evokes stable network activity in the piriform cortex at the time of birth. Neonatal piriform network activity may also have a strong spontaneous component, analogous to the visual (Mooney et al. 1996; Hanganu et al. 2006), auditory (Tritsch et al. 2007, 2010), and somatosensory (Khazipov et al. 2004; Dooley et al. 2020) systems and in line with a recent demonstration of spontaneous network activity in the neonatal olfactory bulb (Gretenkord et al. 2019). Based on the findings from the visual system described above, we further predict that piriform cortical network activity undergoes rapid developmental changes when input patterns change with the emergence of fast sniffing behavior. Alternatively, changes in network activity sensory processing may proceed gradually, in parallel with structural changes (Sarma et al. 2011). Finally, given the unique circuit- and systems-level organization of piriform cortex compared with primary sensory neocortex (Haberly 2001), we predict that the network activity patterns in piriform cortex during early development differ from those previously observed in the developing sensory neocortex. Alternatively, identical network patterns could be realized using different substrates.

Materials and Methods

Animals

Pregnant Long–Evans rats were purchased from Charles River Laboratories (arriving at embryonic days 12–17) and housed under controlled conditions (12-h light cycle, ad lib access to food and water). Until birth, home cages were inspected daily for litters. The first day a litter was present at 9:00 AM was marked postnatal day 1 (P1). Animal care and experimental procedures were in compliance with the National Institutes of Health’s Guide for the Care and Use of Laboratory Animals and were approved by the Institutional Animal Care and Use Committee at Wake Forest School of Medicine.

Electrophysiological Recordings

On the day of recording, pups were removed from the litter and injected with Meloxicam to reduce pain (intramuscular, 10 mg/kg). Anesthesia was induced with 5% isoflurane vaporized in air, checked by tail pinch, and maintained by dynamically adjusting concentration in the range of 0.1–2% based on visual inspection of breathing. Under anesthesia, the scalp was removed and the skull cleaned with hydrogen peroxide. A metal ring was then attached to the skull using superglue (P0–10) or dental cement anchored with skull screws (P11–21), and the animal was placed on a platform with the metal ring rigidly attached to a stereotaxic apparatus (Blumberg et al. 2015). A craniotomy was made overlying the piriform cortex, and multielectrode silicon probes (Cambridge Neurotech acute probes with 2 shanks spaced 250 μm apart; 16 electrode contacts/shanks spaced 25 μm apart) coated with Vybrant DiI cell-labeling solution were inserted into the brain. In a subset of animals, bipolar electrodes (stainless steel, 0.14 mm diameter, spaced 250–300 μm apart) were inserted into ipsilateral olfactory bulb through a second craniotomy. A ground wire (0.14-mm-diameter silver wire) was inserted into or close to the cerebellum and served as the reference for all recordings. The animal was then removed from isoflurane and allowed to acclimatize to head restraint for 1 h before the recording session. Throughout the procedure, body temperature was monitored and manually maintained at a target temperature of 37 °C (range: 35–37 °C) using an infrared heat lamp (measured using a temperature probe placed underneath the animal). Ambient air was humidified by placing damp paper towels underneath the animal. In a subset of recordings, we recorded nuchal electromyography to determine the state of the animal (Blumberg et al. 2005). Animals cycled between sleep and wakefulness, with the majority of the time spent asleep, as evidenced by periods with low muscle tone and muscle twitches, alternating with periods of high muscle tone and intermittent limb movements (see Supplementary Fig. S1), as well as occasional audible vocalizations. Directly after recording sessions, animals were deeply anesthetized with isoflurane, decapitated, and their brains were removed.

Stimuli and Stimulus Delivery

We used n-amyl acetate of >98% purity (Sigma-Aldrich) as an odor stimulus. During recording sessions, odor was presented orthonasally for 5 s (5–10 trials, 30- to 45-s interstimulus interval) and diluted in medical grade air at a flow rate of 300 mL/min, using a constant-flow air-dilution olfactometer (Aurora Scientific, 4-channel olfactometer model 206A). We used a relatively high concentration of odorant (5% saturated vapor) that reliably drives afferent olfactory inputs and sniffing behavior (Uchida and Mainen 2003; Verhagen et al. 2007; Wesson et al. 2008), but it is below threshold for trigeminal stimulation (Silver et al. 1988).

Respiration Monitoring

Respiration was monitored through a metal cannula (25 gauge) inserted into the nasal cavity contralateral to the recording site at the level of the nasal fissure and connected to a pressure sensor using microbore tubing (0.29-mm inner diameter PE tubing). Superglue was used to create an airtight seal between nasal bone and cannula.

Histology

Brains were stored in 10% formalin for 10–15 days, followed by 30% sucrose for 10–15 days, and cut into 30–60 μm sections using a sliding microtome. Sections were mounted on glass slides in DAPI mounting medium and imaged using epifluorescence microscopy to reconstruct electrode locations. Coordinates for electrode placement were initially based on a developmental rat brain atlas (Khazipov et al. 2015) and further refined based on histological reconstruction of recording sites as experiments progressed (see Supplementary Table S1 and Supplementary Fig. S2).

Data Analysis

The broadband extracellular signal from each electrode contact was digitized (25 kHz sampling rate) and stored for off-line analysis, using INTAN headstages (RHD 2132) and amplifiers (RHD 2000). Action potentials were extracted, clustered, and sorted using the Spyking Circus toolbox (Yger et al. 2018). Clusters were considered single units if less than 2% of the waveforms occurred at an interspike interval of 2 ms or less. Only single units that had a baseline firing rate of 1 Hz or more were included in the analyses reported here. All subsequent analyses were performed in Matlab. Broadband signal from one of the electrode contacts was processed to obtain LFP activity by applying a bandpass filter (0.1–300 Hz, fourth-order Butterworth filter), followed by downsampling to 1 kHz. Spectral analyses were performed using the Chronux toolbox for Matlab (http://chronux.org/) (Mitra and Bokil 2008). In order to detect peaks in the spectrum indicative of oscillatory activity, we first calculated a baseline-normalized spectrum by subtracting the spectrum obtained from the baseline period (4 s immediately preceding stimulus onset) from the spectrum obtained from the stimulus period. Note that during the first second after odor onset, the spectrum was sometimes characterized by a broadband increase in power, likely caused by a large transient deflection of the LFP evoked by the first inhalation following odor onset that typically occurs in this time window (see examples in Figs 1 and 2). In order to exclude this nonspecific response from our analysis of oscillatory activity, the stimulus period was defined as the window 1–5 s following odor onset (onset of the analysis window is indicated by dashed lines in Figs 1 and 2). Primary peak was defined as the maximum of the baseline-normalized spectrum. This point was significantly different from baseline in each recording (t-test comparing power during stimulus and baseline period for each frequency bin: P < 0.01) and fell in the 8–30 Hz frequency range. A secondary peak was identified in a subset of sessions as the largest significant difference from baseline in the frequency above 1.7 × primary peak frequency. Primary and secondary peak frequencies were used to calculate band-filtered signals (fourth-order Butterworth filter with a bandwidth of 4 Hz, centered on peak frequency). Amplitude envelope and instantaneous phase of band-filtered signals were then obtained using the Hilbert transform. Coherence (and phase difference) between LFP activity in piriform cortex (signal X) and various other signals (signal Y: LFP in olfactory bulb, respiration, amplitude envelope) was computed using the Chronux toolbox; significance of coherence was determined for each session by comparing trial-averaged coherence values for each frequency bin with the 95% confidence intervals of a shuffled control distribution obtained from random parings of signals X and Y (n = 100). Spectrograms were computed for display purposes only, using Morlet wavelets. Olfactory bulb recordings were obtained and analyzed in an identical manner, except that the 2 signals from the bipolar electrode were first subtracted. For analysis of respiration behavior, the signal from the pressure sensor was differentially amplified (A-M Systems model 1700) and stored as an analog input along with neural signals. Offline, the respiration signal was filtered between 0.1 and 20 Hz (fourth-order Butterworth filter) and downsampled to 1 kHz. For individual sniff analysis, inhalations were detected using an amplitude threshold; frequency was calculated as one over the duration of the period until the next inhalation. Significance of phase locking between spiking activity and the phase of the band-filtered LFP signal was determined using Raleigh’s test.

Figure 1.

LFP activity in piriform cortex in response to odor stimuli recorded during an example session at P0. (A) Single-trial LFP response (in gray) along with respiration activity (in red) obtained at P0, aligned on stimulus onset (t = 0). (B) Magnified representation of the signal in (A), along with the band-filtered LFP signal and its envelope (blue). (C) Average spectrogram (n = 10 trials). (D) Average spectrum during baseline (light) and stimulus (dark) periods. (E) Coherence between primary peak envelope and respiration during the stimulus period relative to shuffled control. Dashed vertical lines in (A) and (B) indicate onset of the stimulus period used for analysis. Green horizontal bars in (D) and (E) indicate significant increases relative to baseline and shuffled, respectively (t-test (D), P < 0.01 or z-test (E), P < 0.05).

Figure 2.

Simultaneous recording from olfactory bulb and piriform cortex in an example session at P5. (A) Top: Broadband LFP signal recorded simultaneously from piriform cortex (red) and olfactory bulb (blue) during a single trial; bottom: band-filtered LFP from both areas. Black horizontal bar indicates odor presentation. (B) Power spectrum of the LFP in the olfactory bulb (top) and piriform cortex (bottom) (n = 5 trials). Black horizontal bars indicate significant increase in power from baseline (t-test: P < 0.05). Peak frequency is marked by the dashed line. (C) Top: Coherence between olfactory bulb and piriform cortex. Black horizontal bars indicate significant increase in coherence compared with shuffled data (z-test: P < 0.05). Bottom: Phase difference between olfactory bulb and piriform cortex in the 5–20 Hz range (shaded area in C). Black line indicates linear regression, slope = 0.07 rad/Hz.

Results

We recorded extracellular neuronal activity from piriform cortex and/or respiration behavior in response to orthonasally presented amyl acetate in a total of 60 unanesthetized rat pups, ranging from birth (P0) to P21 (see Supplementary Table S2). Note that only a single recording was made from each animal and that data from different ages were therefore obtained from different animals. Our main analyses focus on the LFP as a measure of network-level operations (Mitzdorf 1985; Kay and Freeman 1998; Khazipov and Luhmann 2006). Analysis of spiking activity is provided in supplemental figures referred to throughout the text.

Odor Stimuli Evoke Complex Cortical Network Responses at Birth

Given the importance of olfaction and olfactory cortical processing for the survival of neonates, we predict that external odor stimuli evoke complex activity patterns in the newborn piriform cortex. Figure 1 shows multiple representations of LFP activity recorded from piriform cortex during an example session that took place within 12 h of birth. Responses are aligned on odorant presentation (t = 0). Figure 1A shows a single-trial LFP response to n-amyl acetate, with simultaneously recorded respiration activity. After odor onset, the LFP exhibits a slow deflection upon each inhalation. Closer inspection of the activity pattern (Fig. 1B) reveals that slow potentials contain faster oscillations (visualized by band-pass filtering the LFP signal centered on peak frequency, see Methods). To quantify oscillatory structure, we used spectral analysis. Figure 1C shows a trial-averaged spectrogram, giving insight into the frequency composition of the LFP signal as a function of time. Starting at odor onset, a sustained increase in power is visible during the stimulus period around 10–15 Hz, matching the frequency of the oscillation observed in Figure 1B. Power spectra obtained from the stimulus and baseline periods are shown in Figure 1D, revealing spectral peaks in 2 frequency ranges. A prominent peak occurs at low frequencies (<5 Hz) during both baseline and stimulus periods and reflects slow potentials evoked upon inhalation (see Fig. 1A,B). Coherence analysis confirms that 0–5 Hz activity in the LFP is tightly linked to respiration, occurs with approximately equal strength during baseline and stimulus, and is stable across development (see examples in Fig. 3 and Supplementary Fig. S3 for population result), in line with previous reports on LFP activity in the olfactory bulb in neonatal (Gretenkord et al. 2019) and adult animals (Fontanini et al. 2003; Moberly et al. 2018). At higher frequencies, a peak between 10 and 15 Hz is visible as an increase during the stimulus period relative to baseline and reflects the faster oscillation observed in Figure 1B,C. To quantify the relation between this 10–15 Hz oscillation and respiration, we calculated the coherence between its envelope (blue outline in Fig. 1B) and the respiration signal, revealing that the amplitude of the 10–15 Hz oscillation is significantly modulated over the course of the respiration signal (Fig. 1E). Thus, inhalation of odorants triggers slow potentials containing bursts of 10–15 Hz oscillatory activity in the piriform cortex of newborn rats (see Supplementary Figs S4 and S5 for population results). Spiking activity showed a similar relation to the respiration cycle and was phase locked to 10–15 Hz oscillations in the LFP signal (see Supplementary Figs S6 and S7).

Figure 3.

Examples of piriform cortex responses to odor stimuli observed during the first 3 weeks of life. Same conventions as in Figure 1. Note extended voltage (A) and frequency (C, D) ranges compared to Figure 1; red horizontal bar in (D) indicates significant decreases relative to baseline.

In addition to oscillatory activity in response to odor stimuli described above, spectra obtained from pre-stimulus baseline activity also exhibited a peak that overlapped with the peak in stimulus-evoked oscillatory activity (see examples in Figs 1 and 3). We first characterized the potential “bursty” nature of spontaneous activity (i.e., whether it consists of periods of silence interleaved with short intervals of activity) using a burst detection method that has previously been used to analyze spontaneous activity in the neonatal olfactory bulb (Cichon et al. 2014). We found that spontaneous piriform cortical activity exhibits large, infraslow (<1 Hz) amplitude modulations, consistent with a previous report on spontaneous activity in the neonatal olfactory bulb of anesthetized mice (Gretenkord et al. 2019). However, even periods in between high activity levels exhibited substantial spectral structure identical to the structure of stimulus-evoked activity: fluctuations in the amplitude of primary oscillations that are coherent with respiration (Supplementary Fig. S8). Together, these findings suggest that ongoing activity in the neonatal piriform cortex is continuous in nature and primarily modulated by respiration.

Primary Oscillations Originate in the Olfactory bulb

The results presented above demonstrate that neonatal piriform cortex exhibits spontaneous and odor-evoked oscillatory activity in the 10–15 Hz range. Oscillatory in the same frequency range has previously been observed in neonatal visual (Colonnese et al. 2010) and somatosensory cortices (Khazipov et al. 2004; Dooley et al. 2020). In the visual system, neonatal network oscillations are not generated locally in cortex but require thalamocortical interactions. Although the olfactory system lacks thalamic relay, oscillations in piriform cortex may rely on interactions with its own relay structure: the olfactory bulb. This would be consistent with a recent report demonstrating spontaneous oscillatory activity in the olfactory bulb of mice in an overlapping frequency range (Gretenkord et al. 2019). To test whether oscillations in piriform cortex are generated locally or rely on the olfactory bulb, we recorded LFP activity simultaneously from both areas (Fig. 2A). Bulbar and cortical field potential have similar temporal structure (Fig. 2B) and are highly coherent in the 10–15 Hz frequency range (Fig. 2C), suggesting functional interactions between the 2 areas. To determine directionality of this interaction, we determined the phase difference between the 2 areas as a function of frequency (Schoffelen et al. 2005). This analysis revealed that the phase difference OB–PC was consistently increasing in the 5–20 Hz frequency range, indicating a lag of the signal in piriform cortex relative to the olfactory bulb. The observed slope for the recording shown in Figure 2 corresponds to a time lag of 12 ms, consistent with previously reported conduction delays of lateral olfactory tract axons at this age (Schwob et al. 1984; Gretenkord et al. 2019). Identical coherence and phase patterns were observed during baseline (see Supplementary Fig. S9 for population results). Together these data suggest that primary oscillations in the piriform cortex—both spontaneous and stimulus-evoked—are not generated locally in the cortex but originate in the olfactory bulb.

Peak Frequency Exhibits Nonlinear Developmental Dynamics

Previous work on the visual system has shown that cortical network activity patterns undergo nonlinear developmental changes. In order to characterize developmental changes in odor-evoked activity patterns, we compared responses across age on a day-by-day basis. Figure 3 shows 3 more examples of LFP activity in response to odor stimuli recorded from piriform cortex of animals at different ages. The odor-evoked activity pattern at different ages is in essence the same: During odorant presentation, inhalation triggers a slow potential containing a faster oscillation that is visible as the primary peak in the spectrum relative to baseline. Despite the preservation of this basic motif across age, frequency of the primary oscillation increases (Fig. 4; one-way analysis of variance [ANOVA] on peak frequency with factor Age Group [4-day age bins]: F = 28.17; P < 0.01), but not in a gradual manner: Peak frequency remains stable until P15 (pairwise comparisons [Scheffe post hoc test]: P > 0.05), followed by a significant increase between age groups P12–15 and P16–21 (pairwise comparisons: P < 0.01). Curve fitting further revealed that change in peak frequency with age is best described by a piecewise linear fit with a break point at P15, indicating a significantly accelerated developmental trajectory starting at P15 (P < 0.05, see Supplementary Fig. S10 for details on statistical evaluation).

Figure 4.

Primary peak frequency as a function of age. Each data point represents primary peak frequency obtained from a single session. Equal-sized age groups (4 days each) are indicated by dashed lines. Mean ± standard error of the mean (SEM) is plotted for each age group. Lines indicate best fit to the data.

Developmental Changes in Spontaneous Network Activity

The results presented above demonstrate that the frequency of the primary oscillation observed in response to odor stimuli undergoes rapid developmental changes starting at P15. Rapid development of cortical network responses in other systems has been shown to coincide with the establishment of high levels of spontaneous activity and a corresponding decrease in the relative magnitude of sensory-evoked responses (Golshani et al. 2009; Colonnese et al. 2010; Murata and Colonnese 2018). To investigate whether this dynamic is paralleled in the piriform cortex, we determined the relative contribution of spontaneous activity to stimulus-evoked responses using different measures of neural activity. First, we qualitatively compared the structure of ongoing activity between P0–P15 and P16+ age groups and found no differences between age groups: Spontaneous activity exhibits similar fluctuations, and primary oscillations are primarily modulated by respiration across all ages (see Supplementary Fig. S11). We then quantified changes in the overall amplitude of the primary oscillation as a function of age (average spectral power in a window centered on peak frequency; Fig. 5A). Two-way ANOVA with factors Age Group and Epoch (baseline, stimulus) revealed overall increases in amplitude with age (main effect of Age Group: F = 18.37, P < 0.01) and an overall larger amplitude in the stimulus period compared with the baseline period (main effect of Epoch: F = 616.53, P < 0.01). Figure 5B shows stimulus-evoked amplitude (a measure of responsiveness that is independent of absolute amplitude) as a function of age. Importantly, the difference in amplitude between baseline and stimulus periods decreases with age (Age Group × Epoch interaction: F = 17.93, P < 0.01), suggesting that the increase in spontaneous network activity is not simply the result of the network increasing in size. The same dynamic was also observed in single neuron spiking activity (see Supplementary Fig. S12). Together, these analyses demonstrate an overall increase in the relative magnitude of spontaneous network activity with age.

Figure 5.

Stimulus-evoked response magnitude as a function of age. (A) Primary peak spectral amplitude during stimulus (filled squares) and baseline (open squares) periods as a function of age (mean ± SEM, n = 49). (B) Same data as in (A), expressed as percent change from baseline period.

Tracking spontaneous piriform network activity across age further revealed that spontaneous activity not only increases in overall magnitude but also in complexity. The example recording in Figure 3 (right panel) illustrates the emergence of a secondary oscillation at 40–90 Hz, similar in frequency to gamma oscillations previously observed in the adult olfactory cortex (Fig. 6A; see Supplementary Fig. S4 for population analysis) (Kay and Freeman 1998). Secondary oscillations were first observed after P15 and most pronounced in spontaneous activity (i.e., suppressed by odor input), as opposed to primary oscillations (Fig. 6B). Note that before P16, secondary peaks appear in the spectrum (examples in Fig. 3, left and middle panels around 30 Hz). However, further analysis revealed that these peaks, unlike secondary peaks observed after P15, are merely harmonics of the primary peak and likely reflect asymmetries in the primary oscillation rather than a separate oscillation (Fig. 6B,C): t-test comparing absolute distance from harmonics between P0–P15 and P16+ confirmed that harmonic relation was significantly decreased in P16+ compared with P0–P15 (t = 4.40, P < 0.05). Thus, changes in spontaneous piriform cortical network activity during early development are characterized by an increase in magnitude and complexity leading up to P16.

Figure 6.

Dynamics of secondary oscillation. (A) Example trace of spontaneous LFP activity (black), along with band-filtered LFP signal and its envelope (blue), as well as respiration activity (red), recorded from an animal at P17. (B) Normalized envelope of primary and secondary oscillations, aligned on stimulus onset (t = 0), mean ± SEM over all recordings in the P16+ age group that exhibited 2 peaks in the spectrum (n = 8). Normalization was performed using the following formula: x − min/(max − min). (C) Primary peak frequency as a function of secondary peak frequency for each session that exhibited 2 peaks (n = 29). Color represents the age group the recording was obtained from. Dashed line indicates the double harmonic of primary peak frequency.

Rapid Developmental Changes in Cortical Activity Coincide with the Onset of Fast Sniffing Behavior

The results presented above reveal several changes in piriform cortical network activity that occur around P15: Odor-evoked 10–15 Hz oscillations increase abruptly in frequency starting at P15, preceded by a gradual increase in spontaneous activity levels, and coincident with the appearance of high frequency (40–90 Hz) oscillations. Several developmental changes in cortical network activity in the visual system—including rapid changes in input-evoked activity patterns preceded by an increase in a spontaneous activity levels and—occur around the time of eye opening. This constitutes a major developmental milestone that marks a profound change in input patterns to cortex. In the olfactory system, a major change in the pattern of inputs occurs with the onset of active olfaction, characterized by alternating respiration frequency between 2 modes: breathing (~1 to 3 Hz) and fast sniffing (~4 to 9 Hz) (Verhagen et al. 2007; Wachowiak 2011; Moore et al. 2013). In order to test for a possible temporal relation between developmental changes in cortical network activity and the emergence of active olfaction, we analyzed respiration behavior in a subset of animals that yielded reliable respiration data. Figure 7A shows an example of respiration behavior (obtained from an animal at P20) that alternates between breathing (<4 Hz) and fast sniffing (>4 Hz). Whereas animals at all ages displayed some modulation of respiration, including polypnea (bouts of sniffing with increased frequency; Alberts and May 1980; Seelke and Blumberg 2004), respiration rarely exceeded 4 Hz before P15 (Fig. 7B). Quantification of the distribution of inter-sniff intervals revealed a significant change starting at P15 (Fig. 7C; two-way ANOVA with factors Age Group [P0–P14, P15+] and Percentile [50th, 99th]; interaction: F = 13.95, P < 0.01). Moreover, whereas inter-sniff intervals corresponding to frequencies >4 Hz before P15 constituted mostly isolated events, fast sniffs observed after P15 tended to occur in bouts. We quantified this by counting how many fast sniffs followed the occurrence of the first fast sniff in a 1-s window following that sniff (t-test comparing average number of sniffs in a bout between 2 age groups: t = 5.85, P < 0.001; Fig. 7D). Thus, developmental changes in piriform cortical network activity coincide with changes in olfactory sampling behavior.

Figure 7.

Respiration behavior as a function of age. (A) Example of respiration activity obtained from an animal at P20 that switches between slow breathing (<4 Hz) and fast sniffing (>4 Hz) behavior. (B) Percentage of inter-sniff intervals that constitute fast sniffing as a function of age (n = 37). (C) Mean ± SEM of median and 99th percentile inter-sniff intervals for 2 age groups (n = 23 and 14, respectively). (D) Average number of fast sniffs in a bout, as a function of age (n = 37).

Discussion

The present study characterized, for the first time, network activity patterns in the piriform cortex of unanesthetized rats over the course of early development. Olfaction is critical for survival during early postnatal life. For example, neonatal animals learn preferences for maternal odors that in turn elicit orientation and nursing behaviors (Miller and Spear 2009). Piriform cortex is thought to play a crucial role in neonatal odor processing (Morrison et al. 2013). Our results demonstrate that piriform cortex exhibits both spontaneous and odor-evoked activity patterns at the time of birth, consisting of slow respiration-driven oscillations (<5 Hz) that contain faster (10–15 Hz) oscillations. This activity pattern remains stable until P15, after which several aspects of piriform cortical network activity undergo abrupt changes to an adult-like state. The characteristics of the oscillations themselves, as well as their developmental time course, exhibit striking similarities with those previously observed in neocortical systems, providing insight into sensory cortical development in general and the network-level mechanisms governing neonatal olfactory processing in particular.

Our findings regarding the nature of oscillatory network activity in the neonatal olfactory system are in line with previous work on the olfactory bulb during the second week of life. Fletcher et al. (2005) reported that the olfactory bulb of anesthetized rats at P7 exhibits approximately 10 Hz oscillations in response to odor stimuli, and Gretenkord et al. (2019) showed that spontaneous 4–12 Hz oscillations are generated in the mouse olfactory bulb at P8–P10 and relayed to cortical structures, including piriform cortex. Our findings confirm that an olfactory bulb–piriform cortex circuit underlies the generation of network oscillations in the neonatal olfactory system and further demonstrate its functionality at the time of birth.

Neonatal piriform cortical network activity shares several temporal and spectral characteristics with network activity previously observed in neonatal neocortical sensory systems: Salient activity patterns include 10–15 Hz oscillations that are nested in low frequency (<5 Hz) waves, are increased in response to peripheral input, and phase locked to the occurrence of action potentials (Khazipov et al. 2004; Minlebaev et al. 2007; Colonnese et al. 2010; Ackman et al. 2012; Chipaux et al. 2013). In the neonatal neocortex, these network oscillations originate in thalamus and are sustained by feedback excitation interactions between cortex and thalamus (Weliky and Katz 1999; Molnar et al. 2003; Tolner et al. 2012; Murata and Colonnese 2016). Our results demonstrate that the neonatal 3-layer piriform allocortex exhibits highly similar oscillations in the absence of thalamic relay. Future work aimed at identifying the cell type- and layer-specific organization of olfactory bulb–piriform cortex connectivity will determine to what extent the olfactory bulb plays an analogous function to thalamus in early cortical development. It is possible that bidirectional communication between bulb and cortex is necessary for sustaining 10–15 Hz oscillations. Centrifugal projections (Schwob and Price 1984), as well as their adult targets (inhibitory granule cells) (Hinds 1968), exist at the time of birth. Moreover, inhibition in the olfactory bulb is functional at early developmental stages (at least as early as P5) (Wilson and Leon 1986; Seelke and Blumberg 2004). Thus, centrifugal projections may sustain 10–15 Hz oscillations via feedback inhibition, similar to the adult olfactory circuit (Martin et al. 2006, 2007; Lowry and Kay 2007). Alternatively, centrifugal projections may sustain 10–15 Hz oscillations via a feedback excitation loop similar to thalamocortical systems. Functionally, the neonatal olfactory circuit may underlie a developmentally transient type of plasticity that has previously been observed in slice recordings. Isaacson et al. demonstrated that synapses of olfactory bulb projections onto piriform cortex pyramidal cells are highly plastic during the first 2 weeks of life, after which this plasticity declines rapidly (Franks and Isaacson 2005; Poo and Isaacson 2007).

Despite similarities in the circuits that generate network oscillations in neonatal olfactory and neocortex, they differ in the type of inputs that evoke them. In contrast to the odor-driven piriform network responses observed in the present study, the main source of input to sensory neocortex during early development is mostly internally generated—in the form of retinal/cochlear waves and spontaneous muscle twitches in the visual/auditory and somatosensory systems, respectively. The network oscillations triggered by these spontaneous inputs are transiently expressed and disappear with the maturation of the sensory periphery (Mooney et al. 1996; Khazipov et al. 2004; Hanganu et al. 2006; Tritsch et al. 2007; Colonnese et al. 2010; Ackman et al. 2012; Tiriac et al. 2012). The stereotyped peripheral inputs observed in neonatal neocortex are not corrupted by behavioral or environmental variation and thought to support the formation of abstract sensory maps characteristic of primary sensory neocortex, whereas stimulus-driven activity patterns in piriform cortex likely support the formation of individual-specific, experience-driven associations. Differences between sensory cortices in terms of input sources may be responsible for differences in the oscillatory responses they evoke, including differences in duration, frequency of occurrence, and temporal distribution.

In addition to odor-evoked 10–15 Hz oscillations, we found that spontaneous 10–15 Hz oscillations were also present from the time of birth. The results from our simultaneous recordings from olfactory bulb and piriform cortex are in line with previous work in neonatal mice, showing that spontaneous oscillations are generated by mitral/tufted cells in the olfactory bulb (Gretenkord et al. 2019). However, ongoing activity observed here was continuously present and primarily modulated by respiration and thus did not exhibit bursting characteristics of awake neonatal neocortex or anesthetized olfactory bulb. Moreover, we demonstrate that the same ongoing activity patterns increase in amplitude with age—consistent with the existence of ongoing activity in the olfactory bulb of adult rats (Stakic et al. 2011), and thus did not constitute a transient phenomenon as in neonatal neocortex. Although our findings demonstrate that spontaneous activity originates at least in part in the olfactory bulb, additional sources cannot be ruled out. Previous work has shown that ongoing activity in adult piriform cortex is shaped by various top-down inputs (Maier et al. 2015; Sadrian and Wilson 2015). Future work using peripheral deprivation experiments and connectivity tracing will shed light on the contributions of bottom-up and top-down projections to neonatal piriform cortical network activity. Thus, our present findings suggest that olfactory cortex has adult-like ongoing activity at birth that is progressively strengthened over the course of development, likely reflecting increased horizontal (Smith et al. 2018) or network-level connectivity.

The present study is the first to track network activity in olfactory cortex over an extended age range. Over the course of early development, neonatal piriform cortical network responses to odor input undergo changes following a nongradual pattern: primary oscillation frequency remains stable from birth to P15, followed by a rapid increase. Rapid changes in the frequency of odor-evoked oscillations appear to occur in concert with several other changes in network activity. Leading up to P15, we observed an increase in the relative magnitude of spontaneous activity (see above), and immediately following P15, fast gamma oscillations indicative of feedback inhibition (Chattopadhyaya et al. 2004; Goldberg et al. 2011; Pangratz-Fuehrer and Hestrin 2011; Yang et al. 2014) emerged. Nonlinear developmental changes in input-driven cortical network responses have previously been observed in other systems, each following a unique developmental time course. In the visual system, cortical network responses change rapidly around the time of eye opening, likely driven by changes in circuit composition and connectivity (Colonnese et al. 2010) but also by changes in input patterns caused by maturation of the sensory periphery and the onset of patterned vision (Rochefort et al. 2009; Colonnese et al. 2010; Hoy and Niell 2015). In the olfactory system, rapid changes in odor-evoked network responses around P15 cannot be linked to the onset of sensory input per se, which is present by the time of birth. Instead, we speculate that rapid developmental changes in cortical odor responses are linked to changes in input patterns caused by the emergence of fast sniffing—a hallmark of active odor perception (Wachowiak 2011). This is in line with findings from the somatosensory, where rapid changes in cortical sensory responses occur around P9 (Colonnese et al. 2010), when several aspects of body and whisker control start to emerge (Landers and Philip Zeigler 2006; Grant et al. 2012); It remains unknown how the onset of active sensation may drive changes in network activity. Given that inputs to the piriform cortex are structured by respiration, several factors may contribute to the observed developmental dynamics of network activity, including altered input patterns caused by fast sniffing (Verhagen et al. 2007), changes in the interaction between cortex and sensory-specific effector systems that generate sniffing behaviors (Bianchi et al. 1995), and/or the top-down networks that modulate them (Alexandrov et al. 2007). Future experiments manipulating these factors will determine their role in shaping the developmental time course of piriform cortical network activity. The observed coordinated, nonlinear dynamics in the development of cortical network activity is striking given the developmental trajectory of the underlying circuit elements. Pyramidal cell differentiation, excitatory and inhibitory connectivity within the piriform cortex, as well as connectivity of the piriform cortex with the olfactory bulb, extra-olfactory regions, and modulatory systems undergo seemingly uncorrelated, gradual developmental changes (Illig 2007; Sarma et al. 2011). Our findings suggest that these gradual changes do not necessarily affect network-level activity until they collectively reach a critical point, thus offering periods of stable functionality while incorporating new circuit elements.

Finally, the network activity patterns observed in the present study exhibit similarities with those previously observed in adult animals. Neonatal 10–15 Hz oscillations share several characteristics with adult beta oscillations: They are increased in amplitude in response to odor input, modulated by respiration, and phase locked to the occurrence of action potentials. As discussed above, neonatal and mature beta oscillations may be generated by a (partly) overlapping circuit that involves bidirectional connectivity between the olfactory bulb and piriform cortex. Secondary oscillations observed here share characteristics with adult gamma oscillations: They are most prominent in spontaneous activity and not harmonically related to primary oscillations, suggesting independent underlying circuits (Kay and Freeman 1998; Neville and Haberly 2003; Litaudon et al. 2008; Kay and Beshel 2010; Frederick et al. 2016; Osinski et al. 2018). Previous work suggests that gamma oscillations are indicative of local, intracortical feedback inhibition (Bartos et al. 2007). The finding that primary and secondary oscillations increase in frequency after P15—rapidly reaching values of adult beta and gamma oscillations, respectively—further supports the idea that neonatal network oscillations constitute direct developmental precursors of adult network oscillations. Despite the similarities between adult network activity and the activity patterns observed in the present study at P21, it remains unknown how piriform cortical network activity at this point in development compares with network activity previously observed in adult animals. Recordings of piriform cortical network activity in adult animals are typically obtained from freely moving animals in the context of a particular behavioral task (Kay and Freeman 1998; Kay and Beshel 2010; Frederick et al. 2016; Osinski et al. 2018). The absence of task context in the present experiments precludes comparison of the exact network dynamics between neonatal and adult stages.

Notes

Many thanks to Rikkert Hindriks, Don Katz, and Arianna Maffei for valuable discussions and to Emilio Salinas for statistics advice. We also thank three anonymous reviewers for their highly constructive feedback. Conflict of Interest: None declared.

Funding

National Institute of Deafness and Other Communications Disorders of the National Institutes of Health (R01 DC016063 to J.X.M.).