Beta2 Oscillations in Hippocampal-Cortical Circuits During Novelty Detection

Novelty detection is a core feature of behavioral adaptation and involves cascades of neuronal responses—from initial evaluation of the stimulus to the encoding of new representations—resulting in the behavioral ability to respond to unexpected inputs. In the past decade, a new important novelty detection feature, beta2 (~20–30 Hz) oscillations, has been described in the hippocampus (HC). However, the interactions between beta2 and the hippocampal network are unknown, as well as the role—or even the presence—of beta2 in other areas involved with novelty detection. In this work, we combined multisite local field potential (LFP) recordings with novelty-related behavioral tasks in mice to describe the oscillatory dynamics associated with novelty detection in the CA1 region of the HC, parietal cortex, and mid-prefrontal cortex. We found that transient beta2 power increases were observed only during interaction with novel contexts and objects, but not with familiar contexts and objects. Also, robust theta-gamma phase-amplitude coupling was observed during the exploration of novel environments. Surprisingly, bursts of beta2 power had strong coupling with the phase of delta-range oscillations. Finally, the parietal and mid-frontal cortices had strong coherence with the HC in both theta and beta2. These results highlight the importance of beta2 oscillations in a larger hippocampal-cortical circuit, suggesting that beta2 plays a role in the mechanism for detecting and modulating behavioral adaptation to novelty.


INTRODUCTION
Novelty detection is a crucial feature for behavioral adaptation and ignites cascades of neuronal responses, from the initial evaluation of the stimulus to the encoding of new representations, resulting in the behavioral ability to respond appropriately and adaptively to unexpected stimuli (van Kesteren et al., 2012;Kafkas and Montaldi, 2018). Over recent decades, an important novelty detection feature, beta2 oscillations (∼20-33 Hz), has been described in the hippocampus (HC; Berke et al., 2008;França et al., 2014;Kitanishi et al., 2015). In particular, beta2 power transiently increases during spatial novelty (Berke et al., 2008;França et al., 2014;Kitanishi et al., 2015) and its generation is implicated with AMPA and NMDA receptors plasticity between the connections of CA3 and CA1 hippocampal regions (Berke et al., 2008;Kitanishi et al., 2015). However, the interaction between beta2 with other hippocampal rhythms remains unknown. Furthermore, the HC is not alone in detecting novelty: evidence in both humans and rodents points to a larger hippocampal-cortical circuit for detecting and adapting to novelty, including the mid-prefrontal cortex (mPFC) and posterior parietal cortex (PAR; Spellman et al., 2015;Kafkas and Montaldi, 2018;Pho et al., 2018). It seems plausible that beta2 oscillations are a mechanism of communication across these regions, but there is currently no empirical evidence for or against this possibility.
Here, we tested three novel hypotheses concerning the role of beta2 in novelty detection: First, whether beta2 power increase is associated with different forms of novelty (spatial and object); second if slower hippocampal oscillations can modulate beta2 power, similarly to the phase-amplitude coupling of thetagamma oscillations during memory encoding in the HC; and third, whether the novelty integration hubs in the cortex (PAR and mPFC) synchronize with hippocampal beta2 oscillations during novelty exploration.
Combining behavioral tasks where the animal is exposed to environments with different levels of novelty, and recordings from local field potential (LFP) and multi-units targeting the CA1 region of the HC, PAR, and mPFC, we aimed to describe the interactions among these regions involved with novelty detection processing. Using power spectral analysis, weighted phase lag index (WPLI), mean phase vector length (MPVL), Granger causality, and cross-frequency phase-amplitude coupling (CFC) as indices of local and long-range synchronization (Canolty and Knight, 2010;Vinck et al., 2011;Hyafil et al., 2015), we found that transient beta2 power increases are observed only during interactions with novel contexts-environment or object-and not with familiar contexts. During novelty exploration, robust CFC was observed between theta and multiple gamma subbands. Unexpectedly, beta2 had robust coupling with delta-range oscillations. Finally, the PAR and mPFC cortices exhibited strong coherence with both theta and beta2 during novelty exploration. Within the PAR and mPFC, a similar pattern of coupling between delta-ranged and beta2 was seen as in the HC. The results reported in the present study also suggest that beta2 is an oscillatory feature independent of slow gamma oscillations, showing different dynamics of power and CFC, and related to novelty detection. The synchronization among HC, mPFC, and PAR in beta2 during novelty detection reveals its importance to understanding novelty exploration and its implication in a broad hippocampal-cortical circuit for novelty detection.

Animals
The data shown in this article is from nine male mice with Black57 background. All the animals were recorded in all the experimental sessions described in Figure 1A. The animals had free access to food and water. All experiments were approved by the Centrale Commissie Dierproeven (CCD) and it is according to all indications of the local Radboud University Medical Centre animal welfare body (Approval number 2016-0079).

Electrode Implant Procedures
The self-made electrode arrays used in the present work were custom-designed to target three different regions of the mouse brain: CA1-HC, PAR, and mPFC. A detailed description of the arrays and the manufacturing process can be verified (França et al., 2020b). Briefly, there were 16 channels aiming at mPFC (spread in the coordinates AP: 0.5 and 1.5; ML: 0.25 and 0.75; in three columns of electrodes in different depths −2.0, 1.5 and 1.0), eight channels at PAR (AP: −2 and −2.25; ML: 1.0 and 1.75; DV: 0.5) and eight channels at HC (AP −2.5 and −2.75; ML: 1.0 and 1.75; DV: 1.5).
For surgery, 10-16 week old mice were anesthetized with Isoflurane (induction at 5% Isoflurane in 0.5 L/min O 2 ; maintenance at 1-2% Isoflurane in 0.5 L/min O 2 ; Teva). Mice were fixed in the stereotaxic instrument (Neurostar Stereotaxic). After shaving, the skin was disinfected with ethanol (70%). The local anesthetic Xylocaine [2%, Adrenaline 1:200,000 (AstraZeneca)] was injected subcutaneously at the incision site before exposing the skull. Peroxide [10-20% H 2 O 2 ; (Sigma)] was applied to the skull with a cotton swab for cleaning and visualization of Bregma and lambda. The windows in the skull through which the electrodes would be lowered into the brain were drilled specifically to accommodate the type of arrays to be implanted. To avoid contact between the dental cement and the brain, vaseline was applied to those windows after the implant. Electrodes and screws were fixated onto the skull with dental cement (Super-Bond C&B; Supplementary Figure 1). Approximately 40 min before the end of the surgery, saline and analgesic (Carprofen injected subcutaneous 2.5 mg/Kg) were injected to facilitate the animal recovery.
After the experiments, animals were euthanized for post-mortem histological confirmation of electrode location. The majority of electrodes in mPFC were distributed across the anterior cingulate and secondary motor cortex. The majority of the PAR electrodes were placed among layers 2 to 5. In the HC, all electrodes were located in CA1, within the region enclosed by the stratum pyramidale and the stratum lacunosum-moleculare. Electrode tracing can be verified in Supplementary Figure 1.

Behavioral Task
The experiments were designed to expose the animal to different HC-dependent novelty content (environment and novel object). The experiment consisted of four main different sessions of 10 min recording-two sessions at Open field and two sessions at Open field with Objects-interspersed by 5 min Home Cage recordings (see Figure 1A).
Because our goal was to evoke and investigate noveltyrelated oscillatory features, our task did not require or provide detailed behavioral performance output. However, to investigate how the oscillatory features investigated here were correlated with locomotor activity and behavioral exploration, the average velocity and the object exploration time were extracted. The data was computed by automated tracking of video recordings in the program Ethovision. We labeled time windows as being ''object exploration'' if the animal's nose was within a quadrant draw around the object (∼3 cm of the object).

Data Inspection
Electrophysiology data were acquired using Open Ephys with a sampling rate of 30 kHz. During preprocessing, data were downsampled to 1,000 Hz, and EEGlab (Delorme et al., 2011) was used for visual inspection and cleaning artifacts (open channels were removed from the analysis; high-frequency noises were removed by Independent Component Analysis; Segments containing large deflections in all channels were used as a criterion for recording session exclusion). Six animals had high-quality data in all recording sessions, therefore statistical analysis concerning all sessions was performed in six animals. Analyses in different sessions, therefore, have a different number of animals (varying between seven and nine).

Data Analysis
The data analysis was performed using custom-written and built-in routines in MATLAB (R2015b). Before analyses, the multichannel data from each region were re-referenced to that region's local average.
Spectral and time-frequency analysis was performed via convolution with complex Morlet wavelets (defined as a frequency-domain Gaussian with a 3 Hz full-width at halfmaximum) that ranged in peak frequency from 2 to 80 Hz in 100 linearly spaced steps. We reduced the dimensionality of the multichannel data by implementing a frequencyspecific guided source-separation method based on generalized eigendecomposition. The goal was to create a linear weighted combination of channels (separately per region) that maximized the multivariate energy between the data covariance matrix from the narrowband filtered data, vs. the broadband filtered data. This results in a single time series from each region, which was subjected to further analyses. We and others have shown that this method increases signal-to-noise characteristics while reducing computational costs and multiple-comparisons issues, and is more accurate than other sources separation methods such as principal components analysis and independent components analysis in M/EEG and LFP data (Haufe et al., 2014;de Cheveigné and Arzounian, 2015;Cohen, 2017a;Morrow et al., 2020). An advantage of generalized eigendecomposition over independent components analysis is that it optimizes the spatial filter for narrowband activity, which was a primary goal here. Various spatial filters can produce similar or distinct results, depending on their optimization criteria, and rigorous comparison of the performance of spatial filters is beyond the scope of this article (Cohen, 2017b). Nonetheless, possibly the analysis of the independent components could provide comparable results (Fernández-Ruiz and Herreras, 2013).
The Hilbert transform was then applied to these narrow-band filtered component time series to extract time-varying power and phase estimates. The Hilbert transform was then applied to these narrow-band filtered component time series to extract time-varying power and phase estimates. For analyzing beta2 and theta power on a cycle-by-cycle basis, we first detected the beta2 cycles using the instantaneous phase extracted by the Hilbert transform. We then computed the average amplitude envelope of beta2 and theta in each beta2 cycle. Our analyses were restricted to the 10% beta2 cycles with higher and lower energy.
The power spectrum was computed by averaging over the time-frequency power time series from all time points within each larger time window. CFC was performed in sliding windows of 5 s. The phase of delta-range and theta frequency (2-12 Hz) and the amplitude of beta2 to mid-gamma (20-100 Hz) were extracted. The raw CFC values were transformed into standard deviation (z) values by computing the normalized distance away from a null-hypothesis surrogate distribution, created by 500 permutations in which the phase angle time series were randomly cut and swapped. To decrease the influence of possible volume conduction, we performed coherence computations utilizing WPLI (Vinck et al., 2011). Statistical analyses were performed using the routine RMAOV1-Repeated Measures Single-Factor Analysis of Variance Test (α = 0.05).
Conditional spectral Granger causality was applied using the MultiVariate Granger Causality toolbox (Barnett and Seth, 2014). As this relies on the time-domain signals and not alreadyfiltered data (because the causality spectrum is computed from the autoregression terms), we dimension-reduced each region using principal components analysis, taking the time series of the largest component from each region. Data were downsampled to 250 Hz and a model order ranging from 100 to 200 ms (varied over animals to best fit each dataset) was used for the autoregression model fitting. The advantage of the conditional Granger analysis is that it allowed us to isolate the unique contributions of one region to a second region while accounting for a possible shared variance with the third region.
For detecting spiking activity, the electrophysiological signal was first band-pass filtered between 500 Hz and 6 kHz. Then, waveforms were detected using a threshold of eight times the median absolute deviation as in Quiroga et al. (2004) and aligned by their interpolated peak. We used the wavelet and weighted-PCA approach described in Souza et al. (2019) to automatically sort the waveforms of each channel. Although we could not separate spiking activity into single units, the different MUA clusters found in the same channel presented unique activity patterns, and we, therefore, analyzed their activity separately. To access the phase coupling of spikes to beta2 in the first time-window (30 to 150 s), we first selected beta2 cycles in which the mean amplitude envelope of beta2 and delta-range were both in the highest quartile. Then, for each MUA we computed the MPLV of the spikes occurring on those cycles. MUAs with fewer than 30 spikes were excluded from further analyses. The significance of each MPVL value was assessed using an equivalent surrogate distribution, computed using 500 surrogates with the same number of spikes as the original MUA. For significantly modulated MUAs we also assessed the mean phase of spiking.

Beta2 Power Increases With Both Spatial and Object Novel Content
The experiment consisted of four sessions of successive 10 min recordings. Two sessions at Open field (OF1 Novel; OF2 Familiar) and two sessions at Open field with Objects (OBJ1 Novel; OBJ2 Familiar)-interspersed by 5 min Home Cage recordings (see Figure 1A). We first verified that both spatial and object novelty evoked the same pattern of power dynamics previously described (Berke et al., 2008;França et al., 2014). As expected, beta2 power increased only during novelty (OF1 and OBJ1; Figure 1B), but not during any of the familiar contexts (Home cage, OF2, and OBJ2; Figure 1B, Supplementary  Table 1). We also verified the transient aspect associated with novelty-related beta2, whereby the power returned to initial levels after around 3 min of novelty exposition ( Figure 1C). as previously reported (França et al., 2014). For all further analyses, we utilized four time windows based on the power dynamics of beta2 in the HC verified during the exploration of novelty ( Figure 1C). The comparison of the normalized beta2 power (normalized by the power of the last time window) across the different time windows showed that beta2 power was higher in the first compared to the later time windows (Figure 1D; Supplementary Table 1).
A possible confound for these analyses would be if beta2 power was simply increased during locomotor activity, instead of reflecting novelty. We, therefore, conducted several analyses to rule out this confound. First, we computed the average velocity of the animal in the different time windows defined above, and entered velocity into a two-way ANOVA using ''time window'' and ''session'' as independent variables (Interaction: F (3,93) = 27.44, p < 0.0001; Time Window: F (3,93) = 2.41, p = 0.0501, and Session: F (3,93) = 2.41, p = 0.0398), post hoc Tukey-Kramer test showed that OF1 had significantly higher velocity than OBJ1, OBJ2, OF2, p < 0.05; Figure 1E). Importantly, we saw significant interaction between the time window and session. However, this result does not directly link beta2 to movement. Therefore, we computed the Pearson correlation between the normalized beta2 power and animal velocity per session. We verified a strong correlation between the mean velocity and normalized beta2 power when looking at all experimental sessions together (OF1, OF2, OBJ1, and OBJ2; Figure 1E). To verify if such correlation could explain by itself the previous changes in beta2 power, we computed the correlation separately in novelty (OF1 and OBJ1) and familiar sessions (OF2 and OBJ2). We found that beta2 power correlated with mean velocity in the novelty sessions, but not in the familiar session ( Figure 1E). However, possibly the correlation in the novelty sessions was biased by the higher velocity in the OF1 sessions. Indeed, the correlation between beta2 power and velocity vanished when analyzing OBJ1 session only (r = 0.306, p = 0.216; Figure 1F). Nevertheless, animal speed in OBJ1, OBJ2, and OF2 sessions showed similar distributions ( Figure 1F), despite having different beta2 power values ( Figure 1B). In other words, beta2 was not trivially correlated with movement, showing that velocity cannot explain the novelty-related changes in beta2 power. There was no relationship between beta2 power and theta power on a cycle by cycle basis (t (2,8) = 0.06, p = 0.94; Supplementary Figure 2). Third, we tested whether the amount of object exploration predicted beta2 power. We found no correlation between the total time spent exploring the objects and the beta2 normalized power (r = −0.195, p = 0.25; Figure 1G).
Together, these results show that beta2 power increases only during novelty exposure. Beta2 had transiently higher power in the initial phase of the novelty exposure, which waned towards the end of the session. Beta2 power was correlated with velocity exclusively during novelty exposure, but not in a familiar environment.

Delta Modules Beta2 Amplitude During Spatial and Object Novelty
Given the role of CFC in the HC during spatial navigation, learning, and memory retrieval, we next explored whether there was any CFC coupling between beta2 power and the phase of slower frequencies and if this coupling was modulated during novelty processing. We used the frequency ranges of 2-12 Hz for extracting the phase and 20-100 Hz for computing the amplitude envelope. Because the Open field exploration does not present any well-defined time event to trigger time windows for the CFC, we calculated the modulation index (MI) in sliding time windows, which allowed us to examine both the overall CFC and the temporal dynamics of CFC. We observed two key features of novelty-related CFC: first, we verified theta-gamma CFC for both low-gamma (30-50 Hz lowG) and mid-gamma (60-80 Hz midG); second, CFC was present between theta phase and beta2 power and between delta-range phase and beta2 and lowG power in the same time window of higher beta2 power (Figure 2A). A similar pattern of CFC was also observed in the first exploration of OBJ1 ( Figure 2B).
Then, we verified the temporal dynamics of the most prominent CFC patterns: delta-beta2, delta-lowG, theta-beta2, theta-lowG, and theta-midG. For sessions with novelty (OF1 and OBJ1), the couplings of all those frequency bands were higher in the first time window, suggesting that most of the novelty detection and encoding computation happens during the initial part of the session when beta2 power is higher (Figures 2A,B, Supplementary Table 2-time-window comparison), all pairs of coupling in all time-windows (the exception to OF1 win4-theta/lowG and OBJ1 win3-theta/beta) shown MI higher than chance (Figures 2A,B, Supplementary  Table 2). Conversely, during the re-exposure to the OF2 and OBJ2 we observed small or no changes in the temporal dynamics of beta2 coupling (Supplementary Figure 3, Supplementary  Table 4), but higher MI in all time-windows, exhibiting strong MI for delta-beta2/theta-beta2 and delta-lowG/theta-lowG for the OF2 session (Supplementary Figure 3, Supplementary Table 4). We also found a strong modulation of delta/theta in both novel and familiar sessions, where no effect among time windows was found, but all time-windows in all sessions analyzed exhibited MI higher than chance (Supplementary Figure 3, Supplementary  Table 4). Together, these results show that besides the largely reported theta/gamma coupling (Lasztóczi and Klausberger, 2014;Schomburg et al., 2014;Fernández-Ruiz et al., 2017;Gereke et al., 2017;Lopes-dos-Santos et al., 2018), exposure to novelty is followed by delta-range/beta2 coupling with a similar transient characteristic as seen in the beta2 power dynamics, stronger MI is present in the first time window. Such temporal dynamics are not observed in the familiar sessions. However, all time-windows exhibit a MI higher than chance, indicating that beta2 events  are modulated by delta-range oscillations during the novel and familiar contexts.

Novelty Modulates Oscillatory Coherence in Hippocampal-Cortical Circuitry
To investigate our third key hypothesis of whether beta2 oscillations play an important role in the hippocampalcortical novelty detection system, we computed a measure of pairwise coherence WPLI between the three regions.
We observed consistent theta-band coherence among all pairs of areas (see Figures 3A-C) in all the sessions. Besides, sessions with novelty content also exhibited increased coherence in the beta2 range (Figures 3A,B). In contrast to the increase in beta2 power, this increase in coherence was not restricted to the first time window. In fact, besides the home cage 1 session (when the animal never experienced any novelty), a high beta2 coherence could also be observed even in familiar sessions, suggesting maybe the existence of a prolonged effect on synchronization among areas after the first novelty session. We also note that in the last Home exploration session the coherence between mPFC and PAR, the two cortical areas, were already at similar levels compared to the first Home exploration, while coherence between HC and mPFC, and between HC and PAR was still high (Figure 3C). This might be explained by the mechanisms underlying beta2 synchronization in the circuitry.
We next applied conditional spectral Granger causality to further investigate this and to determine the causal flow of interactions around this circuit. During the OF1, PAR, and HC provided input into the mPFC in theta, beta2, and lowG ranges ( Figure 3D). While in the Object 1 session, PAR exhibited a higher gain in the lowG frequency band (Figure 3E). HC dominated the gain values towards both cortices in both theta and beta2 (Figures 3D,E), exhibiting especially strong gain with theta and beta2 during the OBJ1 session ( Figure 3E). Similar to the coherency, the Granger analysis showed no strong variation across different time windows, and suggest less stable connectivity between mPFC and PAR cortices.
Finally, we verified the presence of beta2 burst in the raw data in all regions analyzed (Figures 4A,B). More specifically, we could also see bursts of beta2 and lowG that happened independently ( Figure 4A). To verify that cortical beta2 was not driven by volume conduction, we also analyzed the multi-unit activity in the three regions of interest, computing the MPVL of the spikes in high cycles of beta2. In Figure 5A, we show examples of multi-units in the mPFC, PAR, and HC that are strongly coupled with the phase of beta2 ( Figure 5A). We found that the multi-unit spikes couple (i.e., showed significant MPVL values) to beta2 events detected in each of the three different regions (Figures 5B,C).
Altogether, these results show that PAR, mPFC, and HC synchronize in beta2, after the first novelty exposure and also in the following familiar sessions, suggesting the existence of a prolonged effect on synchronization. The contribution for such synchronization is dominated by HC towards the cortices. Finally, multi-unit activity coupled with beta2 in the three regions analyzed suggests that the beta2 events are not explained by volume conduction.

Parietal and Mid-prefrontal Cortices Exhibit Strong Delta-Beta2 Coupling During Novelty Exploration
Lastly, to further characterize the participation of PAR and mPFC cortices in processing novelty information, we verified both power and the coupling dynamics in the cortices during novelty detection exploration. We found that the mPFC exhibited similar beta2 power dynamics as in the HC, where the increase of beta2 power was verified in the first time window. The power spectrum density (PSD) revealed an increase in beta2 frequency specifically during the novelty exploration ( Figure 6A, Supplementary Table 3). Such an increase in beta2 could be seen in the raw data and was also independent of bursts in the HC (see Figure 4).
Unexpectedly, the PAR and mPFC cortices not only exhibited strong coherence among each other and the HC, but also presented a similar pattern of coupling as seen in the HC (Figures 2A-C, 6A-F, Supplementary Table 3), in which the novel content induced a strong MI between delta/beta2, theta/beta2, and theta/lowG in time-window 1. In contrast, familiar sessions did not exhibit an increase in the delta/beta2 in any time window (Supplementary  Figures 3B,C, Supplementary Table 4) most of the time windows analyzed did not present any coupling during the familiar session (Supplementary Figure 3C, Supplementary  Table 4). These results suggest that although the three areas are synchronized in beta2 during the novel and familiar sessions, the coupling dynamic involving delta-range modulation is specific to novelty exposure, similar to the beta2 power dynamics. Moreover, this modulation engages both mPFC and PAR pointing to the active participation of the associative cortices in the processing information during the novelty detection sessions.

DISCUSSION
In this work, we used simultaneous extracellular recordings of the HC CA1 region, PAR, and mPFC to characterize, for the first time, beta2 oscillations in the hippocampal-cortical novelty detection circuit of mice. We found that beta2 hippocampal power increases during both spatial and object novelty, but not during the exploration of familiar contexts. We have shown that delta-range oscillations modulate beta2 and lowG during the exploration of new and familiar environments, while theta modulates beta2, lowG, and midG. Also, we found strong coherence in theta and beta2 bands during novelty exploration among the areas recorded, in which the higher Granger gain for beta2 and theta came mostly from HC. Such coherence was translated into the increase of beta2 power in the mPFC but not in the PAR, even though bursts of beta2 could be identified in the raw trace of both mPFC and PAR, as well as beta2-modulated multi-units. Finally, we have observed similar coupling characteristics in the cortex to what is described in the HC, showing that beta2 is also modulated by delta-range activity in the cortex. Taken together, these results highlight the importance of beta2 oscillations in a larger hippocampal-cortical circuit, suggesting that beta2 reflects the mechanism for detecting and modulating behavioral adaptation to novelty.
To coordinate the activity of such diverse brain areas during the process of novelty detection, oscillations are suggested to play a key role in the integration and coordination of the information (Buzsáki and Draguhn, 2004;Fries, 2005). Theta oscillations are  thought to coordinate neural networks during memory encoding within and across different areas (Tort et al., 2009;Benchenane et al., 2011;Colgin, 2015). The close relationship between HC and mid-prefrontal areas as it relates to memory encoding and retrieval has been extensively reported (Benchenane et al., 2010(Benchenane et al., , 2011Samuel, 2019), and theta plays an important role in mediating the function of these two areas (Benchenane et al., 2010(Benchenane et al., , 2011. But until now, no specific oscillatory dynamic responsive to novelty content was reported playing a role in the coordination of different brain areas responsible to process the novelty information. We and others have identified beta2 as an oscillatory feature in the HC related to novelty detection in mice (Berke et al., 2008;França et al., 2014) and in rats (Kitanishi et al., 2015). The previous and current findings begin to elucidate a picture of beta2. Its spectral peak is around 20-30 Hz in mice (Berke et al., 2008;França et al., 2014), and is slightly faster in rats-25 to 48 Hz. It is elicited by spatial/environmental novelty but is not associated with novel olfactory stimuli. Beta2 has been related to the stability of place fields, as well as impairments in memory consolidation of novel recognition. This novelty-linked oscillation is transient, reaching its peak during the first 2 min after the novelty presentation and decreasing in amplitude thereafter. Beta2 likely originates in the projections of CA3 towards CA1 and seems to drive the synaptic delivery of GluR1-containing AMPA receptors and CA3 NMDA receptors. Finally, despite the prominence of beta2 in the HC and inter-connected association cortical areas, it is absent in  ), but for the parietal channels. Note in the first time-window the increase in coupling with beta2 and delta, and theta, but also mid-gamma and slow-gamma with theta. Panel (D) shows the same as in the right panel of (B), but for PAR. Panel (E) shows the same as in (A,B), but with Object 1 session. Note the very high MI for delta and beta2 and mid-gamma and theta. Panel (F) shows the same as in (C,D), but with the Object 1 session. *p < 0.05. primary sensory and motor cortices-regions not associated with novelty detection.
In the present work, we replicate the main features in the power dynamics reported before (Figures 1A-C,  Supplementary Table 1). Similar to the previous reports, we have shown that beta2 can be verified at the raw signal of the HC channels (Figure 4). The results reported in the present work, following what has been previously described, shows a delay between the beginning of the novelty exposition and the peak of beta2 (Figure 1C; Supplementary Table 1). This latency period may reflect the generation of a mismatch from previous expectations (Grossberg, 2009) or the time that animals take to perceive the experience as novel. Another possibility is the delay being related to the stability of the place field that is followed by the dynamic of beta2 (Berke et al., 2008). Although, we would not expect a gender effect on beta2 and novelty, none of the previous studies-and also not the present work-investigated beta2 oscillations in females' brains. As the females outperform males in recognition tasks (Bettis and Jacobs, 2012), possible gender differences in beta2 characteristics remains an open question for future research.
We also replicated the relation between beta2 normalized power and the mean velocity of the animal (Figure 1E; França et al., 2014). We found that this correlation is only present in the novelty exposition sessions, and not in the familiar exploration sessions ( Figure 1E). However, this correlation vanished when analyzing the Object 1 session individually, and might be a spurious effect driven by the higher velocity values in the novel open field (Open field 1) session. In either case, differences in beta2 power cannot be fully explained by the animal speed, since beta2 power is stronger in Object 1 in comparison to Object 2 and Open field 2 even though those sessions have similar velocity distribution (Figures 1B,E,F). Further investigation of the instantaneous power and animal speed might help to establish a more conclusive relationship between beta2 and the animal velocity. Unfortunately, this could not be done in this dataset due to a synchronization problem between those two signals (which does not affect our other analyses). Finally, we found no difference in the mean theta power of high-and low-energy beta2 cycles, suggesting the occurrence of beta2 might be independent of locomotion (Supplementary Figure 2).
As previously reported, no correlation between object exploration time and beta2 normalized power was found ( Figure 1E; França et al., 2014). Because mice have an innate exploratory behavior when they are exposed to novel environments, it is expected to see an increase of the total distance traveled and thus the mean velocity in novel environments. However, except for one pair of time-windows in OF1 (2nd and 5th windows, in which the animal should be more habituated to the novelty), the mean velocity did not statistically change within the exploration session, while beta2 power varied along the session (Figure 1) suggesting that the correlation with velocity might reflect the behavior output expected of novelty sessions, as opposed to velocity directly driving beta2 activity.
One of the novel results reported here was the cross-frequency modulation between a slow frequency range within the delta-range activity and the power of beta2 during novelty detection. This set of results was surprising, and not anticipated for the experimental design. As recently shown, delta oscillations have been related to the respiration rhythm (Lockmann and Tort, 2018;Tort et al., 2018a). However, the only way for checking if the phase of the slow oscillation reported here is indeed related to a delta oscillation was implanting electrodes in the olfactory bulb. Therefore, the results present here were reported as a delta-range oscillation, and future research is needed to further investigate the relationship between respiration and novelty detection.
CFC has been implicated in different brain computations, from modulating different assemblies of neurons, facilitating communication between brain regions, and coordinating local cortical processing required for effective computation and synaptic modification (Canolty and Knight, 2010;Lisman and Jensen, 2013;Hyafil et al., 2015). The relation between theta and gamma, from the involvement of different GABAergic interneurons, and its function has been extensively reported in the hippocampal formation (Fernández-Ruiz and Herreras, 2013;Lasztóczi and Klausberger, 2014;Schomburg et al., 2014;Gereke et al., 2017), It has been suggested that the coupling of different gamma frequencies to different theta phases would serve as a mechanism underlying the communication of CA1 with CA3 or entorhinal cortex (Colgin et al., 2009;Schomburg et al., 2014). Although the report of CFC during tasks without specific time epochs to trigger the analysis are uncommon, previous reports indicate changes in the dynamics of gamma and its relation to theta over the experience exposition (Gereke et al., 2017) and promote encoding of memories for novel object-place associations (Zheng et al., 2016).
Here, the sliding time-window CFC analysis in the HC, especially during the first 150 s (window 1), revealed thetanested spectral components, consistent with previous reports (Lopes-dos-Santos et al., 2018). The CFC revealed the peak of theta/beta2 around 22 Hz (instead of the 25 Hz of beta2 power increase; Figure 2A), theta/lowG at 35 Hz, and theta/midG around 70 Hz (Figure 2A) or 54 Hz (Figure 2A). We observed an increase of theta/midG coupling during the exploration of novel environment and objects (Figures 2A,B), while theta/lowG coupling was more prevalent in the ''retrieval'' at the familiar session (Supplementary Figure 3A, Supplementary Table 4) following previously reported theta/midG coupling increases during learning and retrieval of memory (Tort et al., 2008(Tort et al., , 2009Lisman and Jensen, 2013;Zheng et al., 2016;Gereke et al., 2017;Lopes-dos-Santos et al., 2018).
One might be concerned that the beta2 oscillations are simply a harmonic of theta or a reflection of slow-gamma. However, several considerations suggest that beta2 is a unique spectral signature and not a confound of slower non-sinusoidal rhythms. First, if beta2 were a harmonic oscillation of theta, we would expect the first harmonic (around 16 Hz) to be present and stronger than the second (around Beta2), but this is not observed in our results (Figures 1B-D). Second, it is clear that theta, beta2, and lowG have distinct temporal patterns and characteristics ( Figure 1C). Third, It would also be expected a stronger beta2-theta-harmonic during higher theta in any exploration sessions (novel or familiar), but that is not the case, with strong beta2 appearing only in the novel exposition (OF1 and OBJ1), specifically at the beginning of the session. Fourth, there is no relation between theta and beta2 instantaneous power (Supplementary Figure 2) Firth, the spike-field coherence analyses showed that distinct populations of neurons coupled to beta2 bursts vs. theta (Figure 5).
Interestingly, beta2 and lowG were strongly modulated by the delta-range phase (Figures 2A,B). Although this could initially point to beta2 and lowG as being part of the same oscillatory regimen, beta2 and lowG have different spectral peaks (25 Hz vs. 35 Hz), beta2 has a transient power characteristic and lowG does not ( Figure 1C, Supplementary Table 1) and inspecting the raw signal reveals that these two dynamics can be observed independently of each other (Figures 4A-C). Furthermore, beta2 and lowG have different coherence peaks (Figure 4) and exhibit different temporal coupling dynamics during novelty exploration ( Figure 4D). The usage of the same nomenclature (lowG to describe beta2 and lowG) may create difficulties in the characterization of the function behind these different oscillations, which could also be the reason for beta2 being reported only twice in the past decades (França et al., 2014). Instead of only the band of frequency, in which authors constantly change the frequency range for the same nomenclature, the oscillations ideally should be classified based on different characteristics, from the species been recorded to wave-shape, origin, and physiological function (Cole and Voytek, 2017;Tort et al., 2018b). We also report a strong delta-range phase modulating theta (Supplementary Figure 3A,  Supplementary Table 4). The modulation of delta/theta was not modulated by time-window or novelty, exhibiting high MI values through all time-windows and sessions. Delta-theta coupling was previously reported in both rodents and humans during novelty exposure (Isler et al., 2008;Fujisawa and Buzsáki, 2011;Jirsa and Müller, 2013;Roy et al., 2017). As suggested in previous works, delta-theta coupling in the HC could be involved with multiplexed timing mechanisms inherent to the support processing of information necessary during the acquisition and retrieval of memories (Fujisawa and Buzsáki, 2011).
The distinction between beta2 and lowG is also important in the perspective of a complex network involving different brain regions because beta oscillations are implicated in long-range synchrony between different areas of the brain, a feature not shared with gamma oscillations (Kopell et al., 2000). For the first time, we revealed that during the novelty exploration sessions the HC has a strong coherence in the beta2 frequency band (Figures 3A,B), such coherence is not seen when the animal never faced the novelty content before (Figure 3C). On top of that, Granger causality revealed that the highest Granger gains come from HC and PAR cortex towards mPFC in theta, beta2, and lowG frequency band during OF1 exploration, while in the object novelty session the Granger gain comes mostly from HC (Figures 3D,E). Note that during the subsequent familiar exploration sessions the beta2 coherence in all three areas remains strong, probably carrying novelty content information towards the cortices, which may act as hubs for comparing the familiarity/novelty contents. In contrast to the beta2 power dynamics, which increase only at the beginning of novelty sessions, this suggests a more cumulative effect on coherence. We also notice that in the last HC session, the coherence between the two cortices decreased while their coherence with the HC was still high. This might be explained by the strong hippocampal influence in the generation of beta2, or memory trace retrieval characteristics previously described between mPFC and HC (Jin and Maren, 2015). Further investigation is needed to reveal detailed aspects of these interactions. In summary, the coherence and Granger results presented here point to the close communication among the three areas recorded, showing that all three areas communicate via theta and beta2 during novelty and familiarity exploration.
We also have shown for the first time that beta2 has similar transient power dynamics also in the mPFC, increasing at the beginning of the session and fading towards the end of the session (Figure 6A). Although PAR did not exhibit a statistically significant increase in beta2 power, the beta2 bursts can be verified in the raw signal of mPFC, and PAR LFP was coherent with other areas (Figure 4). Furthermore, all three areas involved showed multi-unit coupling with beta2 bursts events among the three areas analyzed, including PAR multi-unit activity coupled to the beta2 bursts of mPFC, PAR, and HC. The couplings of these three areas had different phase preferences of beta2 events of each region, supporting the interpretation of independent bursts in each area (Figure 5). This corroborates the results of coherence and Granger causality analyses, showing that the cortical beta2 is not a result of volume conduction from the HC. Taken together, our results of: (1) raw traces showing independent bursts in the HC and both cortices; (2) local referencing; (3) phase-lag-based LFP coherence; (4) Granger causality; and (5) phase-diverse long-range spike-field coherence demonstrates for the first time that each region exhibits independent bursts of beta2, with the HC appearing to be the main drive.
We also have shown that similar to HC coupling dynamics, (1) both cortices exhibit strong coupling between theta/midG and theta/lowG during novelty exploration; and (2) that both cortices show the same coupling between delta/beta2 as exhibited in HC in the first time window that beta2 exhibited higher power (Figures 6B,D-F). These couplings are only found during the novelty exploration (Figure 6), and not during familiar exploration (Supplementary Figures 3B,C). Even though there is a trend in the delta-beta2 coupling to be higher in the first time-window, this effect is stronger in the OF1 session (time-window effect in mPFC, PAR, and HC; see Supplementary Tables 2,3). Thus it is not clear whether this modulation follows the temporal dynamics of beta2 power, coherence, or a mix between them. Similar coupling was previously reported in the mid-prefrontal cortex during recording in freely behaving rodents (Andino-Pavlovsky et al., 2017) or during learning and working memory (Canolty and Knight, 2010;Samuel, 2019). However, for the first time, we show that the local delta oscillations modulate the beta2, not only in the HC but also in the PAR and mPFC during novelty exploration. Importantly, we found a dissociation between the time courses of beta2 coherence among the three areas and beta2 power within-area: whereas local beta2 power was transient and primarily observed early in the novelty sessions, interregional coherence was more sustained and remained robust through the sessions. This unveils the existence of multiple processes influenced by beta2 oscillations: one in a shorter timescale, revealed by the transient presence of hippocampal beta2 bursts during novelty exposure; and another, in a longer timescale, is characterized by the beta2 synchrony across the HC, mPFC, and PAR that in our data extends through the entire session of novelty exposure and even further into familiar sessions. Those two mechanisms might be associated with different steps of memory encoding. For example, Grossberg (2009) suggests that initial beta2 bursts could be a mechanism for the fast stabilization of the memory traces (during memory acquisition), explaining the rapid emergence of place cells in the HC (Berke et al., 2008;Grossberg, 2009). It has also been shown that inhibition of protein synthesis in the HC impairs reconsolidation of memory traces only when the memory reactivation involves novelty (Rossato et al., 2007;Radiske et al., 2017)-that is, in the presence of beta2 bursts. Both of those processes, memory acquisition and reconsolidation, involve first setting the memory into an active state, which requires further stabilization towards an inactive memory state (Nader, 2015). Thus, there might be a link between the acquisition/activation of memory traces and the initial beta2 bursts. On the other hand, beta2 coherence between HC and the two cortices stays higher for a longer time after novelty exposure, which could indicate a role in the stabilization of the memory traces and the LTP induction that happens in the HC (Clarke et al., 2010). Despite this being an interesting hypothesis, new experiments are needed to specifically investigate the direct relation of beta2 to the different memory trace processes. Finally, in between these two temporal dynamics of beta2 there is the modulation of beta2 amplitude by delta-range oscillations, which seems to follow a short timescale in the cortex only during novelty, similar to the transient beta2 bursts, and a longer timescale in the HC, even though the modulation tends to be higher in the first time-window.
Together, these results highlight and further support the relation of beta2 oscillations and novelty extending it to a larger hippocampal-cortical circuit and suggesting beta2 as a mechanism for detecting and communicating information among the areas involved in behavioral adaptation to novelty.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors upon request.

ETHICS STATEMENT
The animal study was reviewed and approved by Centrale Commissie Dierproeven (CCD) and it is according to all indications of the local Radboud University Medical Centre animal welfare body (Approval number 2016-0079).

AUTHOR CONTRIBUTIONS
AF conceived and designed the experiments, performed the experiments, analyzed and interpreted the data, and wrote the article. NB conceived and designed the experiments, and wrote the article. BS analyzed and interpreted the data, and wrote the article. MC contributed reagents, materials, analysis tools or data, analyzed and interpreted the data, and wrote the article. All authors contributed to the article and approved the submitted version.