Brainstem cholinergic modulation of the thalamocortical activity in urethane anesthetized mice

In mammals, sleep consists in the recurrence of two main stages the rapid eye movement (REM) sleep and the slow wave sleep (SWS). The full expression of sleep rhythms requires an intact thalamocortical loop, and its modulation by neuromodulators such as acetylcholine. A high tone of acetylcholine is observed during REM sleep while a low tone of acetylcholine modulates the cortical slow waves during SWS. Brainstem Cholinergic neurons activity correlates with cortical sleep stages but these neurons do not project directly to the cortex. Instead, they could contribute to cortically-recorded sleep stage modulation via a thalamic relay, in particular via the midline thalamic nuclei. Focusing on the brainstem LDTg cholinergic neurons, I investigated how midline thalamic single unit activity and cortical sleep-like stages are modulated during optogenetic-induced activation or silencing of LDTg cholinergic neurons in urethane anesthetized mice. Thalamic neurons were more active during REM-like than SWS-like stages. Bursting activity predominated during SWS-like while tonic firing was prominent during REM-like stage. Optogenetic silencing of the brainstem LDTg cholinergic neurons abolished REM-like stages and reduced tonic spiking of thalamic neurons. Moreover, during SWS-like, silent Down states were prolonged and thalamic tonic spiking during Up states was reduced. Stimulation of the brainstem LDTg cholinergic neurons had a mild effect on thalamic activity even though tonic discharge was increased. Surprisingly, optogenetic stimulation abolished as well REM-like stages maybe through compensatory mechanisms.

correlates with cortical sleep stages but these neurons do not project directly to the cortex.
Instead, they could contribute to cortically-recorded sleep stage modulation via a thalamic relay, in particular via the midline thalamic nuclei. Focusing on the brainstem LDTg cholinergic neurons, I investigated how midline thalamic single unit activity and cortical sleep-like stages are modulated during optogenetic-induced activation or silencing of LDTg cholinergic neurons in urethane anesthetized mice. Thalamic neurons were more active during REM-like than SWS-like stages. Bursting activity predominated during SWS-like while tonic firing was prominent during REM-like stage. Optogenetic silencing of the brainstem LDTg cholinergic neurons abolished REM-like stages and reduced tonic spiking of thalamic neurons. Moreover, during SWS-like, silent Down states were prolonged and thalamic tonic spiking during Up states was reduced. Stimulation of the brainstem LDTg cholinergic neurons had a mild effect on thalamic activity even though tonic discharge was increased. Surprisingly, optogenetic stimulation abolished as well REM-like stages maybe through compensatory mechanisms.

INTRODUCTION
Sleep and wake states are characterized by the amplitude and frequency of the electrical signal recorded at the surface of the brain and by the muscular activity. In mammals, one distinguishes two main stages, slow wave sleep (SWS) and rapid eye movement (REM) sleep. During REM sleep, cortical neurons are depolarized and show little synchronicity, which gives rise to low amplitude oscillations recorded in the cortical local field potential (LFP). During SWS, cortical neuron activity fluctuates synchronously between depolarized, active periods, called Up states, and hyperpolarized, silent periods, called Down states.
Cycles of Up and Down states (UDS) are reflected in the LFP as slow oscillations (0.5-4 Hz) of large amplitude (Steriade et al., 1993a). The full expression of SWS and REM sleep requires an intact thalamocortical loop but is also modulated by neuromodulators from subcortical areas, and in particular by cholinergic inputs.
The activity of cholinergic neurons is also correlated with arousal states, cholinergic neurons are active during REM sleep and wake and virtually silent during SWS (Boucetta and Jones, 2009), and manipulation of their spiking activity can promote or inhibit arousal states (Steriade et al., 1993e;Van Dort et al., 2015). The thalamocortical loop receives cholinergic inputs from the basal forebrain and the brainstem. Brainstem cholinergic neurons are of 4 of 31 particular interest in the regulation of sleep wake cycle as they are part of the historical ascending reticlular activating system (ARAS; Moruzzi and Magoun, 1949). These neurons are located in the laterodorsal tegmental area (LDTg) and pedunculopontine tegmental area (PPTg; Mesulam et al., 1983;Mena-Segovia et al., 2017). LDTg cholinergic neurons only sparsely project to the neocortex (Mesulam et al., 1983;Cornwall et al., 1990) and act on the cortical arousal states through their rostral relays: the non-specific thalamus and the basal forebrain (Steriade et al., 1993e). Here I explore how LDTg neurons modulates the thalamic activity and how this affects cortical LFP.

Anaesthetised recordings
Mice, aged at least 2 months, were deeply anesthetized via intraperitoneal injections of urethane (1.5 to 1.8 g/kg injected in several shots until loss of reflexes is assessed by paw pinching). While restrained in a stereotaxic apparatus (David Kopf instruments, Phymep, France) mice temperature was monitored and maintained at 34-36 °C using a heating blanket (Basi). A small burr hole was drilled at LDTg coordinates AP = -4.92 mm and ML = 1 mm and an optical fiber (300 μm, 0.29 N.A.; Thorlabs) was descended above the LDTg nuclei at a depth of 3 mm. Light delivery was performed by collimating the optic fibre to a blue laser light (488nm, 15 ± 3 mW at fiber tip; laser 2000) or to a green laser light (532 nm, 10 ± 5 mW; laser 2000). Light delivery was controlled by a shutter whose aperture was 6 of 31 driven by custom-made Igor 6.5 procedures. Cortical local field potentials (LFP) were recorded from the prefrontal cortex at coordinates AP = + 1.96 mm, ML = 0.4 mm and DV = 1.2 mm using an extracellular parylene-C insulated tungsten microelectrode (127 μm diameter, 1 MΩ; A-M Systems). Single unit were recorded from the thalamus at coordinates AP = -1.58 mm, ML = 1.3 mm and DV = 3.4 -3.9 mm, using a 20º angle to reach midline thalamic neurons without crossing through the superior sagittal sinus and the ventricles.

Electrophysiology and Analysis:
Cortical LFP signal was low-pass filtered (<120 Hz) and the signal was downsampled from 20 kHz to 1 kHz. Each stimulation procedure was repeated at least 10 times with a minimum of 50 s interval between each trial. LFP was analyzed using custom-made procedures in Igor Pro. LFP traces were subjected to a Morlet wavelet analysis followed by the z-scoring of the power spectrum intensity for each frequency. An instantaneous gamma intensity index was computed by summing up the z-scored intensity in the gamma range (30-80 Hz). REMlike bouts and Up states were identified when the gamma intensity index reached 3-5 SD of the mean gamma intensity index. A aggregation step was performed to remove potential artefacts and aggregate segments of Up and Down states. All Up and Down states were manually assessed before being included in the final analysis and only bouts of well-defined SWS-like were included. REM-like sleep episodes showed comparable gamma intensity index value as Up states and were distinguished from Up states based on their duration (> 3 s), the potential presence of a hippocampus-mediated theta (4-8 Hz) component and a 7 of 31 reduced power intensity for low frequency (<4 Hz) initiated in the seconds preceding the REM-like episode.
Unit activity analysis was performed on high-pass filtered signal (300 -3000 Hz). Spikes were detected when their amplitude reached a manually adjusted threshold corresponding to 10-15 SD of the mean signal amplitude. Inter-spike interval plots were computed to discriminate spikes occurring within a burst from those fired tonically. Classically, action potential emitted within 5-20 ms of another spike were classified as belonging to the same burst and action potential emitted at > 30 ms of any other spike being classified as tonic action potential. The temporal relationship between spikes and LFP was evaluated using custom-made procedures in R version 3.4.4 (R Core Team, 2018).

Statistics
Results are presented as mean ± sem unless otherwise stated. Statistics were performed using R (R Core Team, 2018). Non-parametric Mann-Whitney rank and Kendall rank correlation tests were used throughout the manuscript. Paired tests were performed wherever possible.

MTN neurons activity during SWS-like and REM-like.
I aimed at describing the dynamics of the thalamocortical loop during sleep and the influence of brainstem cholinergic neurons in shaping the thalamocortical dialogue. During urethane anesthesia, brain activity recapitulates sleep stages, which consists in the spontaneous alternation of SWS-like and REM-like stages. I first asked how the spontaneous firing activity of midline thalamic neurons varies in relation to SWS-like and REM-like stages. Local field potentials (LFP) in the prefrontal cortex and single unit activity in the midline thalamus nuclei were recorded simultaneously ( Figure 1A). I compared the activity of 40 thalamic neurons from 21 animals during REM-like and SWS-like sleeps. During SWS-like, the cortical LFP showed low frequency and large amplitude oscillations, while REM-like activity was characterized by low amplitude oscillations, with in some recordings a relative bump in the theta range of frequency (5-10 Hz) probably due to distant field activity from the hippocampus ( Figure 1A and 1B). Bouts of SWS-like and REM-like sleep were discriminated based on the slow oscillation power and the duration of active events (see Methods).
There was no significant correlation between SWS-like and REM-like firing frequency (Kendall rank correlation coefficient Tau = 0.17; p = 0.13). Thalamic neurons exhibit two modes of firing, phasic discharge of bursts of action potentials fired at 50-300 Hz supported by the voltage-dependent window current T and tonic discharge of isolated action potentials fired at a lower frequency (0.1-30 Hz). Bursting activity and tonic spikes were distinguished by plotting the inter-spike interval for each neuron ( Figure 1D). During REM-like sleep, tonic 9 of 31 activity prevailed (87.4 ± 2.7 %) while the proportion of spikes fired in bursts was significantly higher during SWS-like sleep (55.4 ± 4.3 %; paired Mann-Whitney test, p = 1.2·10 -10 ; Figure   1E). Tonic firing was detected during SWS-like in all 40 neurons while burst firing was absent in 11 out of 40 neurons during REM-like sleep. Consequently, burst frequency was higher during SWS-like compared to REM-like (SWS-like: 0.83 ± 0.15 Hz; REM-like: 0.46 ± 0.11 Hz; paired Mann-Whitney test, p = 0.002; Figure 1F), while tonic firing frequency was higher during REM-like than during SWS-like (SWS-like: 0.70 ± 0.16 Hz; REM-like: 3.04 ± 0.49 Hz; paired Mann-Whitney test, p = 6.5·10 -7 ; Figure 1G). Last, there was no significant difference between REM-like and SWS-like in the number of spikes emitted within a burst (SWS-like: 2.49 ± 0.08 spikes / burst; REM-like: 2.47 ± 0.15 spikes / burst; ; paired Mann-Whitney test, p = 0.20). Hence, these results demonstrate that midline thalamic activity is correlated with sleep-like stages.
Next I investigated the thalamic spiking relative to the phase of the UDS. UDS were identified based on cortical multi unit activity, gamma power intensity and slow oscillation waveform (see Methods and Figure 2A). An UDS cycle lasted 1.38 ± 0.08 s and Up and Down states were of similar duration (Up: 0.80 ± 0.03 s; Down: 0.79 ± 0.05 s; n = 21 animals; paired Mann-Whitney test, p = 0.68; Figure 2B); 66.3 ± 2.6 % of the spikes were fired during Up states (incidence Up states : 2.23 ± 0.21 Hz; incidence Down states: 1.54 ± 0.22 Hz; n = 46; paired Mann-Whitney test, p = 7.10 -4 ; Figure 2C), indicating that thalamic neurons fire preferentially during the active phase of the cortical UDS. An additional 6 neurons were analyzed in this sample that corresponds to recordings for which too little REM-like sleep was observed to be quantified. In this sample, 46.9 ± 4.7 % spikes were fired tonically and  Figure   10 of 31 2D). Thus thalamic neurons preferentially fire during Up states, with no specific difference between tonic and bursting activity. but taking all neurons together this decrease was not significant (Ctrl: 1.83 ± 0.58 Hz; ArchT:
Thus, LDTg ACh neurons stimulation increased tonic firing but not burst firing, and tonic firing increased specifically during Up states, but this increase of thalamic activity had no directly measurable impact on the cortical LFP.

MTN activity across cortical states of vigilance
The results presented here demonstrate that midline thalamic neurons activity is correlated with the cortical sleep-like states in the urethane anesthetized mouse (Figure 1 and 2). As reported during natural sleep, midline thalamic neurons were in average more active during REM sleep than SWS (Gent et al., 2018;Hauer et al., 2019). However, the difference of firing rate between SWS-like and REM-like sleep (twofold in this study, Figure 1) was not as marked as during natural sleep, when the mean activity was up to five times higher during REM sleep (Honjoh et al., 2015;Gent et al., 2018). This difference might be due to the fact that, in my sample, one third of the recorded neurons decreased their activity during REMlike stage. Because thalamic neurons are more active during REM sleep (Honjoh et al., 2015;Gent et al., 2018), the presence of silenced neurons in this sample was surprising.
How a neuron would fire during REM-like state was not predictable from its activity during SWS-like sleep (Figure 1), nor from its location in the MTN. Three main reasons then can account for the presently reported silenced neurons: 1) how neurons were sampled, 2) the difference between natural REM sleep and urethane anesthesia REM-like sleep, and 3) the sensitivity of midline thalamic neurons to neuromodulators. 1) In this report, the mean firing rate was 5 times lower for both REM-like and SWS-like state than reported in naturally 18 of 31 sleeping mice (Gent et al., 2018). However, the firing rate here was similar as in other anesthetized studies (Zimmerman and Grace, 2018) and neurons with very low baseline activity during sleep have been reported in sensory thalamic nuclei (Urbain et al., 2019). So despite the fact that SWS-like firing rate was not predictable of the REM-like firing rate, it is possible that neurons with low activity were missed in some natural sleep studies. 2) On the other hand, there is little doubt that urethane anesthesia, despite being the most accurate anesthetics at recapitulating sleep-like stages (Pagliardini et al., 2012(Pagliardini et al., , 2013, presents different features from natural sleep. In particular, the neuromodulatory state might be different and one can hypothesized that the heterogeneity reported here arises from different sensitivity to neuromodulators and/or varying modulatory state across recordings. Variable modulatory state is however unlikely as each neuron was recorded for at least 30 minutes to sample several REM-like episodes and neurons showed a consistent behavior across REM episodes. Thus it unlikely that a fluctuating neuromodulatory state explains the presence of neurons silent during REM in this study. Moreover, neurons active and silent during sleep were recorded from the same mouse ruling out inter-animal variability. Thus one might speculate that these neurons are a feature of MTN, maybe due to their differential sensitivity to neuromodulators, but this would need to be confirmed in natural sleep.

Preventing sleep stage transition with brainstem cholinergic neurons silencing
Cholinergic neurons from the LDTg and PPTg are part of the historical reticular arousal system whose stimulation promotes arousal (Moruzzi and Magoun, 1949;Steriade et al., 1993e). LDTg/PPTg cholinergic neurons firing rate is high during REM(-like) and awake state and low during SWS(-like; Menia-Segovia et al., 2008;Boucetta et al., 2014;Cissé et al., 2018) and these neurons show an increase of activity that precedes SWS to REM/wake transition (Petzold et al., 2015). This has suggested a causal role of brainstem cholinergic  (Marrosu et al., 1995;Carracedo et al., 2013;Wester et al., 2013;Lörincz et al., 2015;Hay et al., 2021). However in these studies, acetylcholine was locally applied to the cortex, while here, the cholinergic nucleus targeted only weakly projects to the cortex and only to the caudal area. This suggests that an intermediate structure was involved in modulating the cortical activity.

Mild influence of optogenetic activation of LDTg neurons on thalamocortical loop.
In this study, the optogenetic activation of cholinergic neurons had a limited effect on both LFP and thalamic firing ( Figure 5 and 6). This was surprising and in stark contrast with a study in natural sleep, that reports SWS to REM or SWS to awake transition upon light

MTN relays partially brainstem cholinergic inputs to the neocortex through tonic spikes
The thalamus is, with the basal forebrain, the main recipient of brainstem cholinergic fibers in the forebrain, and more specifically the LDTg projects to the MTN (Hallanger et al., 1987;Kitt et al., 1994;Mena-Segovia and Bolam, 2017;Huerta-Ocampo et al., 2020). Non-specific thalamic nuclei receive twice as many cholinergic fibers as the relay nuclei and are specifically sensitive to neuromodulation (Parent and Descarries, 2008;Bayer et al., 2002;Sherman, 2007, 2009;Hay et al., 2019). Activation of brainstem pedunculopontine area promotes cortical activated states that is prevented by sectioning the thalamus (Steriade et al., 1993).
We show that MTN neurons fire primarily bursts of action potentials during SWS-like and tonic action potentials during REM-like sleep (Figure 2), consistently with previous reports (Zimmerman and Grace, 2018). However, bursting was not suppressed during REM-like state, which contrasts from the reported firing of relay thalamic neurons and could be a feature of higher-order thalamic neurons (Fuster and Alexander, 1973;Weyand et al., 2001;Ramcharan et al., 2005). The proportion of spikes fired in burst during SWS-like was lower than reported by others in the primary sensory thalami but comparable with reports from higher-order nuclei (Zimmerman and Grace, 2018;Urbain et al., 2019). Similarly, burst firing during REM-like state was strikingly high in our sample. In primary sensory thalamic nuclei, burst has been hypothetized to act as a salience detector, and can powerfully activate the cortical primary sensory areas (Swadlow and Gusev, 2001). The role of MTN burst is not yet fully understood. It could allow for maintaining attention during wake (Ramcharan et al., 2005) and during SWS it could promote state transition. Indeed burst firing has been associated with either the initiation of the termination of Up state (Gent et al., 2018;Varela and Wilson, 2020). We did not observe any effect of stimulation or silencing of LDTg neurons on thalamic bursting (Figure 4 and 6). We could extrapolate from these results that at least 23 of 31 during sleep-like stage, the cholinergic information relayed by the MTN to the cortex is not meant to promote state transition. Conversely, we observed a reduction of MTN tonic firing when LDTg neurons were silenced and an increase when the cholinergic neurons were stimulated. Tonic firing supposedly acts on the cortical dynamics by promoting state maintenance. During wake, it maintains cortical attention (Schmitt et al., 2017), by promoting the sustained depolarization of the prefrontal cortical neurons. In agreement with these results, the reduced MTN tonic firing in the silencing experiment was associated with shorter Up states (Figure 4). Thus MTN tonic firing could act as a relay for LDTg inputs to regulate cortical arousal state.

Conclusion
In conclusion, this study aimed at identifying how cholinergic inputs modulate the thalamocortical loop dynamics. It showed that LDTg inputs primarily act by modulating tonic firing in the MTN. As suggested by Peever and Fuller (2016), it is critical to identify how the different structures involved in state regulation communicate with one another. Even though the results presented here do not exclude the possibility that another forebrain relay (for instance the basal forebrain cholinergic neurons) relay brainstem arousal regulating signals to the neocortex, I show evidence that MTN neurons contribute to the dialogue between brainstem arousal centres and the neocortex. It would be essential to continue this work in natural sleep to get a better understanding of the modulation of the thalamocortical loop by neuromodulators.
24 of 31 ACKNOWLEDGMENTS I am grateful to Prof Paulsen for his unconditional support, the insightful discussions and to have given me the opportunity to use the resources of his lab to carry this work. This work was supported by a BBSRC grant (BB/S015922/1) and by an Herchel Smith Fellowship. The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.