The (Un)Conscious Mouse as a Model for Human Brain Functions: Key Principles of Anesthesia and Their Impact on Translational Neuroimaging

Abstract

In recent years, technical and procedural advances have brought functional magnetic resonance imaging (fMRI) to the field of murine neuroscience. Due to its unique capacity to measure functional activity non-invasively, across the entire brain, fMRI allows for the direct comparison of large-scale murine and human brain functions. This opens an avenue for bidirectional translational strategies to address fundamental questions ranging from neurological disorders to the nature of consciousness. The key challenges of murine fMRI are: (1) to generate and maintain functional brain states that approximate those of calm and relaxed human volunteers, while (2) preserving neurovascular coupling and physiological baseline conditions. Low-dose anesthetic protocols are commonly applied in murine functional brain studies to prevent stress and facilitate a calm and relaxed condition among animals. Yet, current mono-anesthesia has been shown to impair neural transmission and hemodynamic integrity. By linking the current state of murine electrophysiology, Ca²⁺ imaging and fMRI of anesthetic effects to findings from human studies, this systematic review proposes general principles to design, apply and monitor anesthetic protocols in a more sophisticated way. The further development of balanced multimodal anesthesia, combining two or more drugs with complementary modes of action helps to shape and maintain specific brain states and relevant aspects of murine physiology. Functional connectivity and its dynamic repertoire as assessed by fMRI can be used to make inferences about cortical states and provide additional information about whole-brain functional dynamics. Based on this, a simple and comprehensive functional neurosignature pattern can be determined for use in defining brain states and anesthetic depth in rest and in response to stimuli. Such a signature can be evaluated and shared between labs to indicate the brain state of a mouse during experiments, an important step toward translating findings across species.

Keywords: Ca2+ imaging; EEG; MIND signature; anesthesia; brain functional connectivity; fMRI–functional magnetic resonance imaging; mouse; translational mouse-human.

Figure 1

Principal mechanisms of anesthesia. (A) Brain states are governed by a highly interconnected assembly of subcortical (arousal) nuclei. These nuclei employ distinct transmitter systems, including glutamate (Glu), noradrenaline (NA), serotonin (5HT), dopamine (DA), acetylcholine (ACh), histamine (His), and orexin (Ox). They project to the cortex directly and via higher-order nuclei of the thalamus. Thalamus and cortex are densely interconnected and heavily exchange information. Synchronous rhythmic thalamocortical activity can set the phase relations of distant cortical areas. Similar phase relations facilitate information transfer across the cortex and from frontal to parietal regions. All anesthetics generate distinct thalamocortical rhythms and alter the phase relationship of transmitting and receiving areas, leading to a successive breakdown of cortico-(thalamo-)cortical communication and eventually loss of consciousness. (B) GABA_A agonists, including volatile ethers, affect brain states both top-down and bottom-up by inhibiting excitatory neurons in subcortical nuclei and directly in the cortex. (C) α2AR agonists exert their effects bottom-up mainly by inhibiting the locus coeruleus, which leads to a disinhibition of sleep promoting neurons in the preoptic area of the hypothalamus. (D) N-methyl-d-aspartate (NMDA) antagonists act primarily top-down in a dual mode: at low doses by inhibiting inhibitory interneurons leading to cortical excitation, and at higher dosages by also inhibiting excitatory cortical pyramidal neurons. Further suppression of nociception and arousal is mediated by blocking the parabrachial nucleus. These three routes present the major principles of anesthesia. All anesthetics in current use in preclinical fMRI act in one of these ways. Inspired by Franks (2008), Lee and Dan (2012) and Akeju and Brown (2017), plotted on an MR reference template of the Allen mouse brain atlas (Bakker et al., 2015).

Figure 2

Cortical oscillatory signatures of brain states under anesthesia. (A) Changes in oscillatory signatures for increasing doses of propofol and dexmedetomidine measured from the human anterior cortex. With increasing anesthetic depth, oscillation frequencies decrease and synchronize. (B) Distinct oscillatory signatures induced by different anesthetics (upper panel) and dosages (lower panel). The spectrogram represents the spectrum of frequencies in a time and frequency domain, which facilitates the identification of structured frequency bands. (C) Loss of responsiveness (LOR) and connected consciousness (LOC) induced by increasing doses of propofol. LOR is defined by suppression of responses to click and verbal commands (upper panel). Note the occurrence of alpha bands and the absence of slow-delta bands in the phase between LOR and LOC in the spectrogram. The same is true for return of consciousness (ROC) and responsiveness (ROR). In rodents, alpha waves coincide with the loss of righting reflex (LORR), which corresponds to LOR, and slow-delta waves are indicated by a complete loss of movement (LOM), which corresponds to LOC. (D) Brain states may not be fully stable when maintained via anesthesia. Transitions between two or more intermediate states occurred over longer periods of constant isoflurane concentration in rats. This feature has been defined as metastability. Note the logarithmic scale used to highlight the transitions that occur primarily in the lower frequency band (0–10 Hz). Local field potentials (LFP) recording was conducted in the anterior cingulate cortex (ACC). (E) Stimulation (white bars) of dopaminergic neurons of the ventral tegmental area in the rat shifts cortical states under sedation from slow delta (<4 Hz) towards θ power (isoflurane), or towards θ and β power (propofol). Oscillations: slow (<1 Hz), δ (1–4 Hz), θ (4–8 Hz), α (8–15 Hz), β (15–30 Hz), awake-γ (30–80 Hz), ketamine-γ (25–35 Hz), spindle (9–16 Hz). Adapted with permission from (A,B) Purdon et al. (2015); (B, lower panel) Akeju and Brown (2017); (C) Purdon et al. (2013); (D) Hudson (2017), and (E) Solt et al. (2014).

Figure 3

The thalamocortical circuit as wave generating unit, and cortical information transfer. (A) Scheme of thalamocortical circuit displays major interactions between GABAergic inhibitory interneurons (IN) and excitatory pyramidal cells (PY) in the cortex and thalamic GABAergic inhibitory reticular cells (RE) and excitatory thalamocortical neurons (TC). Arousal nuclei (AN) drive cortical desynchronization via cortical and thalamocortical circuits, including biphasic excitatory–inhibitory modulation (Sun et al., 2013). Anesthetic classes reduce the membrane potential of their respective target neurons (indicated by thin color-coded arrows), leading to inhibition and rebound spiking. Cortical oscillations are produced by synchronized periodic activity between two or more cell types. Transection below cortex (red scissors) leads to slow wave generation in the cortex; transection below thalamus (cerveau isolé, green scissors) leads to generation of sleep-like spindles and slow-delta waves. Anesthesia-induced oscillations: slow (<1 Hz), δ (1–4 Hz), α (8–15 Hz), β (15–30 Hz), γ (25–35 Hz), spindle (9–16 Hz). (B) Fronto-parietal information transfer along the cortical hierarchy. Feedback information (blue) is conveyed from higher-order frontal areas to primary sensory areas in the α/β band, feedforward information flow (red) from primary to higher-order areas in the γ band. In the predictive coding scheme feedback signals are considered predictive models of upcoming sensory states, whereas feedforward information represents the error resulting from a mismatch of a predictive model with an actual state. Feedback transmitted via long-range projections enters cortical columns at the dendritic trees of pyramidal cells in layer 1 (L1). The signal is transmitted to the soma of the same cells in layer 5, which requires driving input of higher-order TC nuclei. Feedback information transfer breaks down with the inhibition of higher-order TC nuclei at increasing depth of anesthesia. (A) Inspired by Ching and Brown (2014) and others. (B) Left panel: adapted from Sikkens et al. (2019). Note that visual higher-order areas in mice differ from those of primates, as reviewed in Glickfeld and Olsen (2017); Pennartz et al. (2019). Right panel: based on Suzuki and Larkum (2020).

Figure 4

Model for cortical signal propagation by nearest neighbor recruitment. In an activated cortical network with desynchronized background activity (as observed in normal waking states), local stimulation elicits waves that weakly entrain neuronal spiking as they travel across the network. In a quiescent cortical network with almost no background activity (as in deeply anesthetized states) local stimulation elicits dense traveling waves, which recruit nearly all cells as they pass. Spheres represent neurons whose membrane potential is indicated by color. Based on Muller et al. (2018).

Figure 5

Subcutaneous electrostimulation of the paw in spontaneous breathing and mechanically ventilated mice under two different anesthetic regimens. (A) Blood oxygenation level-dependent (BOLD) patterns show unilateral responses in the contralateral paw region of S1 for spontaneous breathing mice under ketamine–xylazine. Increasing the field strength to 15 T reveals additional pattern in S2 and thalamic regions not visible at 9.4 T. In contrast large bilateral patterns were observed in ventilated mice under 1% isoflurane. (B) Signal time courses depict substantial surges in mean arterial blood pressure (MABP) and heart rate (HR) in ventilated, but not in spontaneous breathing animals. Data were acquired in-house reproducing findings from Reimann et al. (2018) and Shim et al. (2018). Functional map superimposed on gray shading (15 T) is adapted with permission from Jung et al. (2019).

Figure 6

BOLD patterns evoked by abrupt surges in MABP. (A) Electro-stimulation evokes abrupt surges in MABP and HR. Mimicking these surges by intravenous injection of vasoconstricting drugs leads to BOLD patterns co-locating with large veins. Injection time point is indicated by dashed line. (B) Nociceptive stimuli or stress can elicit abrupt surges in MABP that increase the influx of oxygenated blood into the brain vasculature. MABP and BOLD patterns are highly correlated with each other and with the applied stimulus. This causes the hemodynamic response function (HRF)—that is modeled based on the stimulus paradigm—to reveal MABP-induced effects as significant patterns. (C) Significant patterns co-localize with large veins and can therefore be identified and corrected for, whereas widespread patterns remain below the statistical threshold. (D) At the group level, significant effects persist at more liberal thresholds, and few minor clusters even at more conservative standards. Adapted with permission from Reimann et al. (2018).

Figure 7

Murine nociception induced by heat stimuli applied to the paw. (A) BOLD patterns along the spinothalamic (LST) and spinoreticular tract (SST) depict well-known nociceptive routes in the mouse with high spatial accuracy, including thalamic nuclei (TN), primary (S1) and secondary (S2) somatosensory cortex, the barrel field (S1_BF), and insular (IC) and ACC. (B) Involvement of the habenula nociceptive pathway (HNP) has also been observed, including amygdala (AM), hippocampal areas (HC), entopenduncular nucleus (EP), and the small nuclei of the lateral habenula (L_Hb). These highly precise patterns represented the tip of the iceberg of underlying BOLD effects, likely induced by MABP surges and vasodilative projections from subcortical arousal nuclei. Adapted from Reimann et al. (2016), http://creativecommons.org/licenses/by/4.0.

Figure 8

Functional connectivity (FC) in the mouse brain. (A) FC is defined by the correlation coefficient (cc) of remote brain areas. The left panel shows an assembly of important areas in murine FC, including the default mode network (DMN), cortical sensory-motor networks, and subcortical networks. (B) Murine FC correlation maps for five anesthetic protocols. Seed voxels in the anterior primary somatosensory cortex show anticorrelated time courses in the cingulate cortex. Anticorrelation is an essential criterion for anesthetic states in fMRI in which the FC is not restricted to its structural scaffold. (C) Correlation matrices in the rat for six different anesthetic protocols referenced against awake rats. Asterisks indicate statistically significant differences compared with the awake group (t-test, p < 0.05, false discovery rate corrected). (D) Optogenetic stimulation of the dorsal raphé nuclei that send serotinergic projections into large parts of the brain leading to wide patterns of negative BOLD effects. Adapted from (A) Grandjean et al. (2019a); (B) based on Grandjean et al. (2014), reproducible using original data from the open repository, https://central.xnat.org; (C) adapted with permission from Paasonen et al. (2018), (D) Grandjean et al. (2019b).

Figure 9

The functional characteristics (qualifiers) and MIND signature (identifier) of a specific brain state induced by an anesthetic protocol. The aim of anesthesia in fMRI is to induce stable sedation, while maintaining a brain state that approximates an awake, calm state. The protocol may be either mono-anesthetic (drug a; dose x) or multimodal (drug a, b, c; dose x, y, z). Qualifiers are defined as functional characteristics of an intermediate brain state that has been induced and maintained by the use of anesthesia: The cortical EEG signature permits to monitor anesthetic depth, to evaluate arousability by challenging the stability of a brain state via nociception or stress, and to probe stimulus-response features of sensory evoked potentials (SEP). Hemodynamic translation can be evaluated via fMRI, whereas shape and magnitude of the hemodynamic response function (HRF) provide information about anesthesia-induced vasomodulation. Anesthetic effects on functional connectivity (FC) across the brain can be assessed based on correlation matrices that depict interrelations of cortical and subcortical areas, including anticorrelations, and the dynamic repertoire of recurring sets of FC patterns over time. FC has great value as an identifier of a particular brain state and the integrity of its qualifiers. A fingerprint-like FC signature pattern can be used to unambiguously identify maintained intermediate brain states under anesthesia and communicate them across labs. This signature pattern is termed the maintained, intermediate neurophysiologically-determined (MIND) signature. Flanking tasks in fMRI with resting-state scans will permit a researcher to monitor the maintenance or variations in the brain state under anesthesia throughout the fMRI experiment.