This study used inter-regional coherences and their correlations to trace the RRO signal from OB, where it is derived from rhythmic nasal airflow, to higher order brain networks of PFC and HC, where it may potentially contribute to communication between these structures by synchronizing their activities at the respiratory rate. We focused on on-going (i.e. “background”) RRO unaffected by behaviors requiring its short-term fluctuations, e.g. sniffing. We found that this rhythmicity depends on sleep-wake states; it is significantly larger in waking than in sleep. Within arousal states, however, it remains unchanged when the animal is engaged in behaviors or conditions, such as locomotion (AW), and REM sleep associated with theta vs. consummatory behaviors (QW) and SWS associated with non-theta HC activity. In agreement with previous reports 3,4,9,25, RRO was more prominent in PFC with an obligatory transmission of RRO from OB to PFC. This was indicated by parallel variations in dia-OB and OB-PFC coherences in individual experiments, verified by significant correlation between these parameters over the group. In contrast, RRO was relatively low in HC, and the variations of OB-HC coherence did not necessarily follow those between dia and OB. RRO input to HC 5-7, however, lead to strong variations in individual recordings (at odds with the grossly defined states of AW and QW). Importantly, this variability, quantified with OB-HC coherence, was essential for establishing PFC-HC synchrony at the respiratory rate, whereas variations of RRO in OB and PFC had no significant effect.
RRO in waking. Communication and collaboration between HC and PFC and its impairment in psychiatric diseases has been the primary focus of extensive research to date 18,19,24,26. HC-PFC theta synchronization and its role in spatial working memory is relatively well-studied 27-32. On the other hand, PFC is theorized to be the master regulator of working memory and higher-order executive function 33-38, yet the mechanisms by which PFC exerts “top-down” influences remain less clear. Functional coupling of PFC with downstream circuits, including HC, amygdala, ventral tegmental area, by means of delta-range oscillations (2-5 Hz) was recently shown in PFC-specific tasks 14-16. It was proposed that theta and delta oscillations may serve as parallel channels of communication between the HC and the PFC in opposite directions: theta HC-to-PFC and delta PFC-to-HC 21. The results of the present study indicate that the balance and interaction between theta and RRO may also provide a potential mechanism for bidirectional PFC-HC coupling. When RRO is within the delta range, it can contribute to or even drive the PFC 2-5 Hz rhythm 39,40. PFC receives this input whenever OB is driven by RRO and may broadcast it widely. In contrast, the connection of HC to this global 2 rhythm appears more dynamic; when the HC is receptive to RRO input, PFC-to-HC channels would open in the delta range – distinct from the channel established by HC-theta in the opposite direction.
The most striking observation of this study was the marked contrast between the strength and reliability of RRO input, stronger in PFC than in HC (Table 1A), and its effect on PFC-HC coherence, i.e. RRO in HC more influential than in PFC (Table 1B). This raises questions regarding the mechanisms and the functional consequences. Differences in cytoarchitecture, connectivity, and other properties of PFC and HC networks give rise to different intrinsic oscillations in the two structures, which allow them to resonate with rhythmic input at specific frequencies. We propose that, due to these differences, baseline slow RRO recorded in this study may be involved in PFC-HC communication primarily in the of PFC-to-HC direction.
Task-related intrinsic oscillations in PFC appear at frequencies in the delta range 14-16, but these oscillations in waking are markedly different from the wide-band thalamo-cortical delta rhythm of SWS. Specifically, the delta (of waking) is spectrally of narrow-band, is hierarchically nested with gamma oscillations 20,41, and is normally generated in cortico-cortical circuits, associated with various cognitive functions 41-44. Outgoing PFC messages may use the RRO fluctuations in sensitivity of downstream structures when PFC delta and RRO are synchronized. With this mechanism, RRO outside of sniffing 5-7 may make HC networks sensitive to messages arriving assembled in bouts at delta-range frequencies.
On the other hand, background RRO is slower than HC theta rhythm and thus, in order to synchronize with the signature HC oscillation during active states of sniffing, respiratory rate is accelerated and brought within the higher and narrower theta frequency band. The two oscillations, RRO and theta, show distinct characteristics in HC, such as different laminar profiles and theta-modulated gamma bands, and differentially entrain HC neurons 5-7 – even when their frequencies overlap. Olfactory-related activity patterns in OB, such as cell firing and gamma bursts, appear phase locked in these episodes to the synchronized RRO-theta rhythm 45. The exact mechanisms are not completely understood, but rhythmic synchronization of sensory sampling in OB on one hand and excitability of neurons involved in central processing in HC and piriform cortex on the other is considered a “paradigmatic example” of active sensing 11,46. This would serve to optimize odor perception, coordinating it with multiple sensory channels, associated with rhythmic nasal, whisker, and head movements.
Although lacking strong direct projections from the OB 6,47,48, the PFC and HC receive RRO via multisynaptic pathways which includes common connections from the piriform cortex, as well as separate projections, through amygdala (PFC) or entorhinal cortex (HC)3,6. Accordingly, the PFC-HC coherence may emerge from common OB input, or alternatively the RRO may be directed primarily to PFC and then transmitted to HC. Differential nodes in OB output pathways connecting PFC and HC (see e.g. 21,49) may set the balance of RRO between the two structures.
When and how HC couples with slow baseline RRO will require further investigations using specific tasks beyond the sleep-wake states of this study. Data demonstrating the potential role of RRO in non-olfactory processing has accumulated in recent years, not only from rodent studies but also from human studies 10,50,51. A specific challenge for translating the results between species resides with important differences between human and rodents. Brain oscillations, functions, dynamics, and key features including characteristic frequencies are evolutionarily well-preserved 52, but respiratory rate varies widely between species. In humans, respiratory rate (~0.2 Hz) is below the frequencies of the key components of the EEG oscillatory hierarchy which thus cannot establish coherent coupling with respiration. RRO remains however is manifested in humans as respiratory modulation of the amplitude of brain oscillations including slow (delta, theta) as well as fast (beta, gamma) rhythms – involved in cognitive processes 10. This is a different form of coupling which unlike coherence does not require matching the frequencies of rhythms generated by different mechanisms.
RRO in sleep. In contrast to wake states, RRO during sleep appears reduced at the level of OB indicated by relatively low dia-OB coherence (Fig. 2) thus restricting OB-mediated RRO in higher brain structures (PFC, HC) from coupling with oscillations dominant in these networks during sleep. Viczko et al. 53 demonstrated for example that slow oscillations (SO), an archetypical EEG pattern in SWS, emerges separate from respiration even when they overlap in frequency, and argued that it “fit with an SO mechanism as intrinsic emergent property of a deafferented neural network”. Our data are consistent with this concept, suggesting that intrinsic brain oscillations – relevant in sleep-dependent memory consolidation both in SWS (SO 53 and delta 54) and REM sleep (theta 55) – are protected from RRO. It is interesting to note that in humans, where very few studies analyzed RRO during sleep with statistical scrutiny 56,57, subtle changes in EEG linked to respiratory cycles were enhanced in SWS and REM sleep in children with sleep disordered breathing in multiple frequency bands, including delta 56, theta 56,57, alpha, and sigma 57. Adenotonsillectomy, the most common surgical procedure for sleep apnea, which among other benefits improves cognitive function, reduced or normalized these RRO alterations.
It is important to note, however, that our conclusions only concern rhythmic RRO mediated by the OB. It has been reported that, besides RRO, respiration may also pace non-rhythmic events, linking their occurrence to specific phases of respiration (see e.g. 13,50). In sleep, this may include sharp wave/ripples and dentate spikes; that is, intrinsic HC patterns during SWS synchronized with UP-DOWN transitions in cortical networks that are involved in functional PFC-HC interactions serving memory consolidation 58,59. For instance, a recent study 60 found that the post-inspiratory bias of these patterns along with firing of a large number of PFC and HC neurons remained after deafferentation of OB sensory neurons, indicating that mechanisms that bypass the OB play a primary role in their synchronization. The authors hypothesized the contribution of a “so-far undescribed ascending respiratory corollary discharge signal, likely propagating from the brainstem respiratory rhythm generators” could pace limbic networks using a disinhibition-mediated mechanism – consistent with lack of prominent LFPs in the absence of input from the OB 60. The causal model remains to be elucidated. In addition to ascending projections from the pre-Bötzinger complex 61 or the locus coeruleus 62, several other signals from internal organs may be involved –due to respiratory movements and chemosensitive signals from the cardiovascular system 13,63,64.
Potential relevance to COVID-19. Investigation of the modulation of higher brain function by the RRO input generated in the OB may be relevant for understanding COVID-19 pathology. Disturbances in smell have emerged early as the predominant neurological symptom 65,66, and serve as evidence for COVID-19 related neurological abnormalities originating from pathology of the olfactory epithelium. In fact, in the first prospective imaging studies (MRI scans 3-4 months after COVID-19 hospitalization), significant changes in grey matter volume were primarily found in cingulate gyrus, piriform cortex, and hippocampus, correlated with loss of smell and memory loss23. According to current understanding (rev. 22,67), SARS-CoV-2 does not directly infect olfactory sensory neurons; the deficit is mediated instead by the altered microenvironment maintained by cells in the olfactory epithelium expressing ACE2 receptors 68-70 – indicating that a similar mechanism may affect RRO as well. Abnormal PFC-HC coupling would not be immediately noticeable for patients as the more obvious symptom of smell loss, but it may lead to neurological consequences which remain unclear at this time. Further studies are necessary to address the effect of SARS-CoV-2 on this system.