Breathing modulates cortico-hippocampal dynamics during o ﬄ ine states

Network dynamics have been proposed as a mechanistic substrate for the information transfer across cortical and hippocampal circuits. During sleep and o ﬄ ine states, synchronous reactivation across these regions underlies the consolidation of memories. However, little is known about the mechanisms that synchronize and coordinate these processes across widespread brain regions. Here we address the hypothesis that breathing acts as an oscillatory pacemaker, persistently coupling distributed brain circuit dynamics. Using large-scale recordings from seven cortical and subcortical brain regions in quiescent and sleeping mice, we identiﬁed a novel global mechanism, termed respiratory corollary discharge, that co-modulates neural activity across these circuits. Analysis of inter-regional population activity and optogenetic perturbations revealed that breathing rhythm couples hippocampal sharp-wave ripples and cortical DOWN/UP state transitions by jointly mod-ulating excitability in these circuits. These results highlight breathing, a perennial brain rhythm, as an oscillatory sca ﬀ old for the functional coordination of the limbic circuit, supporting the segregation and integration of information ﬂow across neuronal networks during o ﬄ ine states. detailed investigation of the mechanism and tested for a possible modulatory role of the breathing rhythm on brain circuits. to system-level resonance at breathing frequency. This sets the stage for future investigations of the interaction between the RCD and ROR in limbic networks and, in turn, the top-down modulation of breathing and emotional responses. Breathing rhythm contributes to synchronization of the cortico-hippocampal dynamics - Implications for models of memory consolidation mechanisms of RCD and ﬁne-timescale, closed-loop optogenetic perturbations. Our causal optogenetic manipulation of 442 one side of this complex dynamical system identiﬁes an excitability proﬁle of hippocampo-cortical interaction within a 443 breathing cycle and paves the way for future respiration phase-resolved gain of function studies. 444 While we show here that the respiratory dynamics bias the prefrontal DOWN/UP states via RCD, slow oscillations 445 can emerge in isolated cortical slubs 121 or slices 105 and thus slow oscillations observed at a similar frequency in the tha- 446 lamocortical system across all mammalian species 18 are not necessarily forced by the breathing rhythm 68,126 . Rather, 447 we suggest considering their interaction from the mechanistic dynamical systems perspective of coupled bistable non- 448 linear dynamical systems. Indeed, leading models of the generation of neocortical UP states from DOWN states 25 or 449 inspiratory bursts from expiratory silence in preBötzinger circuits 32 suggest that both phenomena rely on regenerative 450 avalanches, due to recurrent connectivity, that are followed by activity-dependent disfacilitation. Given that neocortical 451 slow oscillations can be locally generated 112,128 , are globally synchronized by the thalamic input 65 and propagate across 452 the neocortex 51,76 , RCD biasing of the cortical DOWN/UP state complex could be considered as an extension of a global 453 system of mutually-coupled nonlinear oscillators. The persistent synchronous output of the respiratory oscillator and its 454 marginal independence of the descending input might provide a widespread asymmetric bias to both cortical DOWN/UP 455 states and hippocampal SWR across o ﬄ ine states of di ﬀ erent depth. It is likely, however, that via descending cortical 456 projections, cortical SO provides feedback to the pontine respiratory rhythm-generating centers and thus the interaction 457 between respiratory dynamics and slow oscillations could be bidirectional.


Introduction
Prefrontal oscillations are related to breathing throughout behavioral states. (a) Example traces of simultaneously recorded respiratory EOG and medial prefrontal local field potentials (LFP) (See also Supplementary Fig. 1). (b) Example time-frequency decomposition of respiratory and mPFC LFP signals, revealing the reliable relationship between the two signals. (c) Distribution of peak frequency bins of the spectrally decomposed respiration (left; darker colors) and mPFC LFP (right; lighter colors) during slow-wave sleep, quiescence, exploratory behavior and self-initiated wheel running (N = 9 freely-behaving mice) (See also Supplementary Fig. 2). (d,e) Averaged normalized power spectral density of respiration (d) and mPFC LFP (e) across states as in (c). (f ) Frequency-resolved comodulation of respiration and mPFC LFP oscillation power, across mice and behaviors (N = 9 mice). (g) Example coherence spectrum between respiration and mPFC LFP during offline states. Inset, average coherence value in the 2-5 Hz band (N = 9 mice). (h) Phase shift of 2-5 Hz filtered respiration and mPFC LFP signals during offline states for an example animal (blue histogram) and overlaid magnitude of phase modulation (logZ) and average phase shift for all animals (red dots; N = 9 mice). Black arrow depicts the average phase and logZ of the phase shift for the example and the red arrow for the population. (i) Example spectral Granger causality between respiration and mPFC LFP for both causal directions. Inset, group statistics of the average Granger causality for the 4 Hz band (2)(3)(4)(5) between respiration and mPFC LFP for both causality directions (N = 9 mice, Wilcoxon signed-rank test, resp ! mPFC versus mPFC ! resp, ** P<0.01). a.u., arbitrary units; s.d., standard deviations. Shaded areas, mean ± SEM.  hippocampal CSD, enabling the identification of synaptic inputs into dendritic sub-compartments. Although the LFP 147 profile only highlights the prominence of the respiratory band in the DG hilus region ( Supplementary Fig. 5a), depthresolved coherence analysis identifies coherence with breathing in the dendritic CA1 and DG layers(Supplementary 149 Fig. 5b,c), while high-resolution CSD analysis revealed more accurately the presence of two distinct and time-shifted 150 respiratory-related inputs to DG dendritic sub-compartments (Fig. 3d,e; Supplementary Fig. 5e). Inspiration was 151 associated with an early sink in the outer molecular layer of DG, indicative of input from the layer II (LII) of the lateral 152 entorhinal cortex (LEC), followed by a sink in the middle molecular layer of DG, indicative of input from the layer II of 153 the medial entorhinal cortex (MEC) (Fig. 3d,e; Supplementary Fig. 5e). 176 To explore the extent of cortical and subcortical entrainment by breathing, we further recorded LFP and single-unit 177 activity in the BLA, NAc, V1, as well as somatic and midline thalamus (Fig. 3f-s; Supplementary Fig. 4e). Similar to 178 mPFC, LFP in both BLA and NAc was comodulated with breathing across a range of frequencies, with most prominent 179 modulation at ⇠4 Hz, the main mode of breathing frequency during quiescence ( Fig. 3f-i), and exhibited reliable cycle-180 to-cycle phase relationship with the respiratory oscillation ( Supplementary Fig. 6a,b). Given the nuclear nature and 181 lack of lamination of these structures, which obfuscates the interpretation of slow LFP oscillations, we examined the 182 modulation of single-unit activity by the phase of breathing, revealing that a large proportion of BLA, NAc, and thalamic 183 neurons were significantly modulated by respiration, notably firing in distinct phases of the breathing cycle ( Fig. 3j-n).

184
In the visual cortex, both LFPs (Fig. 3p) and a large fraction of neurons (Fig. 3q) were phase-modulated by breathing. 185 Interestingly, the magnitude of phase modulation was maximal in layers 4-6, which is consistent with both the phase 186 modulation of sensory thalamic neurons (Fig. 3r), the presence of strong coherence of the LFP with respiration in the 187 thalamus ( Supplementary Fig. 5b) and respiration-related CSD sinks in the L4/5 where thalamic inputs arrive ( Fig.   188 3s).

Reafferent origin of limbic gamma oscillations
A prominent feature of prefrontal LFP is the presence of fast gamma oscillations (⇠80 Hz), believed to emerge due to local 190 synchronization 113,118 (Fig. 4a). To investigate the relationship of prefrontal gamma oscillations to the breathing rhythm 191 and well-known OB gamma oscillations 3,40,42,66 , we recorded simultaneously from the two structures and calculated the 192 phase-amplitude coupling between breathing and gamma oscillations (Fig. 4a,b; Supplementary Fig. 6c,d). Both 193 OB and mPFC fast gamma oscillations were modulated by the phase of breathing, such that gamma bursts occurred 194 predominantly simultaneously and in the descending phase of the local LFP (Fig. 4c). The simultaneously occurring 195 OB and mPFC gamma oscillations matched in frequency and exhibited a reliable phase relationship with a phase lag 196 that suggests directionality from the OB to the mPFC ( Fig. 4d; Supplementary Fig. 6d). To examine the underlying 197 synaptic inputs mediating the occurrence of these oscillations in the mPFC, we calculated CSD across mPFC layers, 198 triggered on the phase of the OB gamma bursts. This analysis revealed a discrete set of sinks across prefrontal layers 199 associated with OB bursts (Fig. 4i,j). Similar to the slow time scale LFP signals, these results suggest that fast gamma  and NAc cells (red, n = 76 cells). (q) Distribution of the mean preferred gamma phase for each significantly modulated BLA and oscillatory signals in mPFC are generated by OB-gamma rhythmic polysynaptic inputs to mPFC and are not exclusively 201 attributable to a locally generated rhythm.

202
Examination of the OB ⇠80 Hz gamma-triggered dorsal hippocampal CSD revealed a DG outer molecular layer sink 226 (Fig. 4k,l), a profile distinct from the similar frequency CA1 lm gamma ( Supplementary Fig. 6e). This would be 227 consistent LEC LII inputs, a region known to exhibit olfactory-related activity 64 . In parallel, slow BLA gamma (⇠40 228 Hz) and fast NAc gamma (⇠80 Hz) oscillations were modulated by the phase of breathing, occurring predominantly in 229 the trough and ascending phase of breathing respectively (Fig. 4m-o).

230
To examine whether these breathing-modulated OB-mediated gamma oscillations have a functional role in driving 231 local neuronal activity, we quantified coupling of local single units to mPFC gamma signals, revealing that ⇠40% of 232 principal cells and ⇠55% of interneurons increased their firing rate in response to local gamma oscillations (Fig. 4e,f ). of these LFP patterns after OB lesion or tracheotomy 9,53,83,123 . However, the distributed, phase-specific and massive 240 modulatory effect that breathing had on unit activity across wide brain regions is at odds with the anatomically-specific 241 synaptic pathways that we identified as responsible for slow and fast currents.

242
To causally test whether OB reafferent input is the sole origin of the LFP patterns and unit entrainment, we employed 243 a pharmacological approach, that enables selective removal of the reafferent input. A well-characterized effect of systemic 244 methimazole injection is the ablation of the olfactory epithelium ( Supplementary Fig. 7a) that hosts the olfactory 245 sensory neurons 8 , that respond to both odors and mechanical stimuli 46 . Effectively, this deprives the OB of olfactory and 246 respiratory input, while leaving the breathing rhythm generators ( Supplementary Fig. 7b-e), as well as the bulbar 247 circuits intact, enabling us to study the contribution of re-afferent input on the brain activity in freely-behaving mice.

248
This manipulation eliminated the respiration-coherent and spectrally-narrow prefrontal slow oscillatory LFP component Surprisingly, the olfactory deafferentation (OD) left most prefrontal and thalamic neurons modulated by breathing, 255 although the strength of modulation was reduced (Fig. 5h,i), suggesting that a so-far undescribed and yet-to-be-256 determined anatomical-physiological mechanism of centrifugal efference copy provided by brainstem circuits (termed 257 ascending respiratory corollary discharge (RCD)) is responsible for the massive entrainment of limbic neurons. Interest-258 ingly, following OD, mice exhibited intact memory and fear expression, suggesting that the RCD might be underlying 259 the behavioral expression ( Supplementary Fig. 7i-k). Dorsal hippocampal neurons were somewhat stronger affected 260 by the ablation, yet more than 40% of cells were still significantly phase modulated by breathing (Fig. 5i,j), indicating a 261 differential degree of contribution of RCD and ROR to unit firing across mPFC, hippocampus, and thalamus. In contrast 262 to the prefrontal CSD, the olfactory deafferentation led to a strong reduction of the outer molecular layer current-sink 263 originating in LEC LII in the respiration-locked CSD (Fig. 5k,l), while leaving MEC LII sink and other non-respiration 264 related CSD patterns intact ( Supplementary Fig. 7h), suggesting that the LEC input is driven by ROR, while MEC 265 input is driven by RCD.   (Fig. 6i). Both UP and DOWN state onsets were strongly modulated by the breathing phase and time from inspiration 289 ( Fig. 6j-m), while the magnitude of modulated UP states followed the profile of UP state onset probability (Fig. 6k).

290
In line with the results on ripples and prefrontal units, UP and DOWN state modulation was not affected by olfactory 291 deafferentation, suggesting that RCD is the source of this modulation (Fig. 6m,n). Example (black) and distribution of preferred breathing phase of SWR occurrence, for SWRs that are terminating an UP state (red dots, N = 8 mice). (q) Probability of SWR occurrence as a function of time from UP or DOWN state onset after OD (N = 7 mice). Note that the observed pattern is identical to pre-OD shown in (o). s.d., standard deviations; a.u., arbitrary units; n.s., not significant; OD, olfactory deafferentation. Shaded areas, mean ± SEM. Stars indicate significance levels (* P<0.05; ** P<0.01; *** P<0.001).
Previous observations during sleep in rats identified a temporal correlation between ripple occurrence and cortical 294 DOWN/UP state complexes 98,110,112 , with large fraction of ripples preceding DOWN state transition, however the 295 mechanism underlying this correlation remains unknown. We found that ripples preceded the termination of prefrontal 296 UP states and onset of the DOWN states both before (Fig. 6o) and after deafferentation (Fig. 6q), with ripples 297 associated with UP state termination and immediately preceding a DOWN state onset tended to occur in the early 298 post-inspiratory phase (Fig. 6p). This is in line with the RCD-driven synaptic inputs to the DG middle molecular 299 layer preceding ripple events, which are suggesting an RCD-mediated coordination of both cortical and hippocampal 300 excitability favoring SWR co-occurrence with the cortical UP states ( Supplementary Fig. 8c,d).

301
Ripple output is known to recruit prefrontal neural activity 98,110 . In agreement with this, hippocampal ripples evoked 302 a response in prefrontal LFP and gave rise to an efferent copy detected as a local increase in fast oscillatory power in 303 the PFC LFP (Fig. 7a) 59 . In response to ripple events, ⇠14% of prefrontal PNs and ⇠42% of INs exhibited increased 304 firing (Fig. 7b,d,e Supplementary Fig. 8i,l), while a small fraction of these cells was also rhythmically modulated 305 by the ripple phase (Fig. 7c). In parallel, ⇠69% of NAc cells were significantly driven by ripple events (Fig. 7h,i; 306 Supplementary Fig. 8l), while in both mPFC and NAc there was a great overlap between cells that were phase 307 modulated by breathing and those that were responsive to ripples ( Fig. 7j; Supplementary Fig. 8l). Importantly, the 308 phase of breathing modulated the excitability of both mPFC and NAc, as revealed by the modulation of ripple-evoked 309 activity magnitude by the phase of breathing (Fig. 7e,i), as well as the fine-timescale (10ms) co-firing between CA1 and 310 mPFC (Fig. 7f ) and within prefrontal regions ( Supplementary Fig. 8e).

311
Given the observation that prefrontal population activity is limited by the respiratory modulation on a low-dimensional 312 manifold (Fig. 2d), we investigated the effect of ripples occurring during inspiration on the trajectory of the neural 313 population activity. Ripples transiently perturbed cortical dynamics, which quickly returned to the respiration-driven 314 limit cycle (Fig. 7g). 315 Given the mutual connectivity between the cortical networks and the hippocampus, delineating the causal role of the 316 joint respiratory modulation for the coordination of SWR and UP/DOWN dynamics only by passive observation of the 317 tripartite correlation of cortical, hippocampal and respiratory dynamics is difficult. We thus opted for optogenetically 318 generating ripple oscillations, an experimental manipulation that enabled us to decorrelate the timing of hippocampal 319 ripples from the breathing cycle and thus allowed us to investigate the effect of respiratory modulation of excitability 320 of prefrontal circuits and its sensitivity to SWR-driven inputs. To achieve that, we expressed excitatory opsins in the 321 dCA1 (Fig. 7l) and delivered low-intensity half-sine wave light stimulation (Fig. 7k,m). Light stimulation resulted in 322 the rhythmic depolarization of dCA1 neurons (Fig. 7n) and the generation of short-duration, high-frequency oscillations 323 (termed opto-ripples) ( Fig. 7m; Supplementary Fig. 8f ) 37,115 with peak frequency 75Hz (Fig. 7o). These oscillations induced evoked activity in the mPFC qualitatively similar to the one described during intrinsic ripples, both at the LFP 325 level, that was consistent with opto-ripple evoked K-complex ( Fig. 7p; Supplementary Fig. 8g), unit level (Fig.   326 7r; Supplementary Fig. 8h), and population dynamics (Fig. 7q). Interestingly, the magnitude of the prefrontal 327 depolarization in response to opto-ripples was modulated by the ongoing phase of the breathing cycle when opto-328 ripple was generated, similar to intrinsic ripples (Fig. 7d, Supplementary Fig. 8i). A qualitatively similar, though

Discussion
The propagation of information across distinct neuronal networks is facilitated by the coordination of these dynamics 366 between brain regions 7,112 . In this study, we demonstrate that the respiratory rhythm, via a centrifugal corollary dis-367 charge, acts as a functional oscillatory scaffold and provides a unifying global temporal coordination of neuronal firing 368 and network dynamics across cortical and subcortical networks during offline states. The comprehensive phase-resolved 369 picture of this synchronization provides the basis for mechanistic theories of information-flow across the limbic system 370 (Fig. 8b).

371
Respiratory corollary discharge couples global brain circuits Using pharmacological manipulations paired with large-scale recordings, we identified a joint mechanism of respiratory 372 entrainment, consisting of an efference copy of the brainstem respiratory rhythm or vagal re-afferents (respiratory corollary 373 discharge; RCD) that underlies the neuronal modulation of brain regions and a respiratory olfactory reafference (ROR) that contributes to the modulation and accounts for respiration-locked LFP signals (Fig. 8a). We have demonstrated the global entrainment of neuronal activity by the respiratory rhythm that is mediated by the intracerebral RCD, likely 376 originating in the brainstem rhythm generator circuits and being unaffected by olfactory deafferentation. tracing and activity-dependent labeling studies to identify its anatomical substrate. 395 We suggest that centrifugal modulation by breathing is analogous to the predictive signaling employed in a wide range 396 of neural circuits 28 , such as those underlying sensory-motor coordination 117 and likely extends to other brain structures 397 and brain states. The global outreach of RCD to higher-order areas suggests that it might play an important role in the 398 coordination of multi-sensory processing, in sync with orofacial motor output during both passive and active orofacial  Based on CSD analysis and pharmacological manipulation, we suggest that RCD and ROR inputs, via MEC and 431 LEC respectively, give rise to hippocampal unit modulation, RCD likely responsible for the emergence of dentate spikes 432 and SWR respiratory-entrainment. Depending on its strength, the RCD input, either via feed-forward inhibition of 433 the CA3 1 or forward excitation can, respectively, delay or advance SWRs within the respiratory cycle. In parallel, the  Anatomically-resolved analysis of prefrontal OB-generated current sources and unit activity suggests that deep layers 476 and mostly ventral regions are the main targets of OB reafference and give rise to observed LFP signals. Although the 477 interpretation of these gradients is challenging given the existence of volume conduction from the OB, these findings 478 suggest a potential functional role of the differential modulation of orbital, prefrontal, and cingulate regions and is worthy 479 of future investigation. Importantly, the propagation of respiration-driven excitation to distant structures is jointly driven 480 by both ROR and RCD, the former giving rise to observed LFP signals that disappear after OD, as shown here, or OB 481 ablation 9,53,83 . Thus, due to the fast breathing frequency in mice, slow and/or delta power cortical LFP signals are in 482 part, and depending on the synaptic distance from the OB, contaminated by ROR, making the direct analysis of slow 483 oscillation or delta waves based on LFP signal alone unreliable. Further investigations are required to dissociate the 484 differential functional role of ROR and RCD in recruiting brain circuits to the breathing rhythm.

485
This joint modulation of downstream circuits suggests a model in which synchronous ROR inputs reach the target 486 regions in sync with RCD-coordinated local activity. These oscillations potentially provide a temporally-optimized privi-487 leged route for olfactory reafferent input to affect the ongoing cortical activity, in line with recent reports in humans 135 .

488
This could explain the efficacy of experimental manipulations that bias learning 4 , consolidation 102 , or sleep depth 96 489 using odor presentation during sleep which suggest that olfaction is a royal path to the sleeping brain.

497
In summary, the data provided here suggest that breathing provides a constant stream of rhythmic input to the brain.

498
In addition to its role as the condicio sine qua non for life, we provide evidence that the breathing rhythm acts as a global 499 pacemaker of the brain, providing a persistent corollary discharge signal that enables the integration and segregation 500 of information flow and processing across the distributed circuits by synchronizing local, internally-generated dynamics 501 during offline states. In this emergent model of respiratory entrainment of limbic circuits, we speculate that this perennial  93 . For head-fixed recordings, a 531 craniotomy above the targeted structure and a midline bilateral craniotomy above the mPFC was performed to enable 532 the recording from all cortical layers. Dura was left intact and craniotomies were sealed with Kwik-Cast (WPI, Germany) 533 after surgery and after each recording session. For electromyographic (EMG) and electrocardiographic (ECG) recordings, 534 two 125µm Teflon-coated silver electrodes (AG-5T, Science Products GmBH) were sutured into the right and left nuchal 535 or dorsal intercostal muscles, using bio-absorbable sutures (Surgicryl Monofilament USP 5/0). Wires were connected to 536 a multi-wire electrode array connector (Omnetics) attached to the skull. For the recording of the neural activity of the 537 olfactory epithelium, which was used as a proxy for respiration 91 , a small hole was drilled above the anterior portion 538 of the nasal bone (AP: +3 mm from the nasal fissure, ML: +0.5 mm from midline) until the olfactory epithelium was 539 revealed. A 75µm Teflon-coated silver electrode (AG-3T, Science Products GmBH) was inserted inside the soft epithelial 540 tissue. Approximately 500µm of insulation was removed from the tip of this wire and the other end was connected 541 to the same Omnetics connector as the rest of the electrodes. Two miniature stainless steel screws (#000-120, Antrin 542 Miniature Specialties, Inc.), pre-soldered to copper wire were implanted bilaterally above the cerebellum and served as 543 the ground for electrophysiological recordings and as an anchoring point for the implants. All implants were secured 544 using self-etching, light-curing dental adhesive (OptiBond All-In-One, Kerr), light-curing dental cement (Tetric Evoflow, 545 Ivoclar Vivadent) and autopolymerizing prosthetic resin (Paladur, Heraeus Kulzer). animal was tracked using a 3-axis accelerometer (ADXL335, Analog Devices) incorporated in the headstage, which was 573 used as the ground truth for the head-motion. Accelerometer data were sampled at 30 kHz and the sensitivity of the 574 accelerometer is 340 mV/g (g is the standard acceleration due to gravity; ⇠ 9.8m/s 2 ). The gravity vector is differentially 575 acceleration due to head movement and the static acceleration due to gravity, the first time derivative of the acceleration 577 was calculated (jerk; units: g/s). This measure eliminates the effect of gravity and the dynamic acceleration dominates.

578
The effect of gravity on the different axes is amplified during head rotations. The jerk of each axis was analyzed separately 579 for the quantification of head-motion, however for the behavioral state detection the magnitude of the jerk was quantified 580 as: where ! a k is the acceleration for each axis and smoothed in time using a narrow Gaussian window 581 (2 s, 100 ms s.d.) (Supplementary Fig. 1e). This head micro-motion is not used to extract the respiration signal used 582 throughout the manuscript, which is instead recorded from the nasal cavity. The only exception is Supplementary Fig.   583 1a and Supplementary Fig. 2c,d, where the head-mounted accelerometer signal is also used for purely demonstration 584 purposes, to highlight the fact that respiration is also reflected in the head micromotions.

585
Additionally, the activity of mice was tracked using an overhead camera (Logitech C920 HD Pro). The camera data 586 were transferred to a computer dedicated to the behavior tracking and were acquired and processed in real-time using a 587 custom-designed pipeline based on the Bonsai software 70 . Video data were synchronized with the electrophysiological data 588 using network events. Video was preprocessed to extract the frame-to-frame difference and calculate a compound measure 589 that we found provided an excellent proxy for the behavioral state. Video frames were thresholded and binarized. A logical 590 exclusive OR operation was applied on consecutive frames, a calculation that provides the effective frame difference. The 591 sum of these differences provides a measure of overall change between consecutive frames. We found that the changes 592 in the amplitude of variance of this measure over time are informative for the current state of the mouse. Complete 593 immobility is easily distinguishable using this measure, due to the low amplitude and small variance of the signal. A 594 threshold was set manually such that even small muscle twitches during sleep were captured, but breathing-related 595 head-motion was below threshold. Using the density of head micro-motions and muscle twitches, we were able to classify 596 behavioral segments as active awakening, quiescence, or sleep ( Supplementary Fig. 1f ). For head-fixed recordings, we 597 relied solely on high-resolution video of the mouse snout and body, from which we derived a micro-motion signal that 598 was used in the same way as the jerk-based signal for freely-behaving mice.

692
Conversely, neurons recorded ventrally of this reference channel were characterized as superficial CA1 pyramidal neurons.

693
Given that neuronal spikes can be identified in more than one channel of the polytrode, neurons were assigned to the 694 channel with the highest spike amplitude 30 . Well-described CSD profiles of hippocampal oscillatory patterns were used to 695 assign somatodendritic CA1 and DG layers to channels ( Supplementary Fig. 5e). The middle of stratum radiatum was 696 assigned to the channel with the deepest sharp wave current-source density sink associated with ripple oscillations 20,134 .

697
Stratum oriens was defined as the channels above the pyramidal layer SWR CSD source and below the internal capsule, stimulation. An early sink following visual stimulation between layers 3 and 4 80 , a later sink between layers 5 and 6, and 709 a peak of the high-frequency power in mid-layer 5 103,107 .

Network event detection
Ripples were detected from a CA1 pyramidal layer channel using the instantaneous amplitude of the analytic signal

Statistical analysis
For statistical analyses, the normality assumption of the underlying distributions was assessed using the Kolmogorov-

Data availability
All relevant data that support the findings of this study will be made available from the corresponding authors upon 777 reasonable request.