Animal imaging without anesthesia was met with some resistance, an attitude that persists today based on the many publications proposing “better” anesthetics for functional MRI studies. The contrarians argue that the reduction in stress and motion artifact using anesthesia and sedatives is better than the confound of stress when interpreting brain function. The hesitancy to adopt awake imaging is also founded in the need for specialized technology, additional costs, and time, and for some institutions, ethical issues around the care and treatment of animals. While there are multiple reasons to avoid anesthesia when studying brain function (see reviews Gao et al., 2017; Steiner et al., 2021), many obvious to the layperson, the most compelling is translation to the human condition. We do preclinical studies on brain function in animals to inform the public and guide clinical researchers in their effort to understand the human condition. This communication from bench top to clinic is made seamless if the same imaging modalities, at similar field strengths and under the same physiological conditions is performed by both. Human brain function is never studied under general anesthesia unless you are studying the effects of anesthesia. How antipsychotics, anxiolytics, analgesics, psychostimulants, drugs of abuse, etc., affect brain function cannot be studied under anesthesia. CNS active drugs affect cognition, emotion, and perception of the external and internal milieu all of which are affected by anesthesia.

There are many published variations on acclimation protocols for preparing different animals for awake imaging, e.g., rats (King et al., 2005; Upadhyay et al., 2011; Russo et al., 2021), mice (Ferris et al., 2014; Yoshida et al., 2016; Madularu et al., 2017b; Han et al., 2019), marmoset monkeys (Ferris and Snowdon, 2005; Silva et al., 2011; Ziegler et al., 2020), rabbits (Wyrwicz et al., 2000), dogs (Berns et al., 2012; Cuaya et al., 2016), and rhesus macaques (Goense et al., 2010). Some describe surgical methods and implants for immobilizing the head while many use non-invasive passive restraining systems. In general, animals are lightly anesthetized with a volatile anesthetic, secured in a restraining system, and allowed to recover in a virtual environment meant to mimic the scanner. This virtual simulation with head restraint is usually accompanied by background noise mimicking the actual noise associated with the pulse sequence (Almeida et al., 2021). In our first studies (King et al., 2005) rats were acclimated for 90 min every other day for eight days. Acclimation was assessed by changes in heart rate, respiratory rate, and blood levels of the stress hormone, corticosterone. Physiologic and endocrine measures of autonomic arousal decrease with acclimation. Interestingly, when rats are imaged awake without acclimation and again after several days of acclimation we found no differences in cerebral blood flow (CBF) across multiple brain regions. Essentially baseline CBF is stable in awake rats regardless of stress, a critical point when measuring changes in functional signal dependent upon blood flow. In our more recent studies, we acclimate for five consecutive days, increasing the duration of acclimation with each subsequent day (Sadaka et al., 2021).

In some cases anesthesia can be completely avoided during the acclimation procedure. Rats surgically implanted with a head post that fits into a cradle to prevent any rotational motion can be trained over 8–10 days to enter a “snuggle sack,” slide their head post into the cradle, and lay motionless (Chang P. C. et al., 2016). While time consuming, this procedure eliminates the early use of light anesthesia required for most awaking imaging set-ups and provides excellent functional imaging with a minimum of motion artifact. Silva et al. (2011) described a step-by-step procedure for acclimating marmoset monkeys for each separate phase of the set-up, e.g., magnet environment, restraint, noise, etc., without the use of anesthesia. Ziegler et al. (2020), developed a method for acclimating preadolescent common marmoset monkeys for awake imaging without ever being exposed to anesthetics or sedatives, a condition that better reflects the human imaging experience. The unique design of the head cradle eliminated the need for ear bars or bite bar, helping to minimizing discomfort. The first study on awake dogs trained two animals to remain motionless for enough time to acquire images without the use of anesthetics, sedatives or physical restraint (Berns et al., 2012). Indeed, the absence of any exposure to anesthetics and sedatives with minimum restraint has been the case for all laboratories imaging awake dogs (Andics et al., 2014; Cuaya et al., 2016; Huber and Lamm, 2017).

The purpose of acclimation is to reduce stress and minimize motion artifact. Interestingly this is not true for all animals. We discovered that prairie voles can be imaged without acclimation on their first time in the scanner (Yee et al., 2016). Efforts to acclimate only exacerbated the stress as measured by motion artifact, vocalizations, and recording heart rate/respiratory sinus arrhythmia.

As noted above, MRI is noisy. The rapid changes in electromagnetic forces driving the gradients create the acoustic noise that is pulse sensitive and exacerbated at high field strengths (Mansfield et al., 1995). The problem has been effectively addressed in human imaging with hardware solutions, wave phase reduction, and altered pulse sequences, all of which can reduce the noise by as much as 20 dB (McJury et al., 1997; Hennel, 2001; Katsunuma et al., 2002). Most recently, Paasonen et al. (2020), addressed the problem in awake rat imaging by developing Multi-Band SWeep Imaging with Fourier Transformation (MB-SWIFT) at 9.4 T for EPI sequences. They report a reduction in acoustic noise by 20 to 30-dB as compared to standard EPI sequences accompanied by a decrease in motion artifact providing resting state connectivity data comparable to those reported standard EPI studies (Paasonen et al., 2020).

Our first studies used a volume coil to transmit and a saddle shaped surface coil to receive radiofrequency signals. The head restrainer employed ear bars and a tooth bar, much like a stereotaxis, to immobilize the head. The body was restricted to a fitted tube. Since then, the RF electronics evolved into a single quadrature transmit/receive volume coil that was customized to the size of the rat’s head to optimize brain coverage and space filling for better signal-to-noise (SNR) and homogeneity (Figure 1). The ear bars were eliminated and replaced with a passive head restraining system to minimize the discomfort and stress (see video of rat setup for awake imaging https://www.youtube.com/watch?v=JQX1wgOV3K4). The system is ergonomic enabling setup in minutes, and completely non-invasive with no surgical implants used to immobilize the head. There have been several publications on the construction of RF coils and restraining systems for awake MRI imaging in rats (Martin et al., 2013; Chang J. B. et al., 2016; Madularu et al., 2017a; Stenroos et al., 2018; Derksen et al., 2021), mice (Ferris et al., 2014; Madularu et al., 2017b; Tsurugizawa et al., 2021), and the common marmoset monkey (Papoti et al., 2013, 2017; Schaeffer et al., 2019; Ziegler et al., 2020).

The common read-out for awake imaging in animals and humans is a change in blood oxygen level dependent (BOLD) signal contrast (Kim and Ogawa, 2012) as compared to a baseline obtained prior to the introduction of a drug or a general stimulus. The change in signal can occur within 3–10 s after provocation and can persist for a different duration in time dependent upon the experimental paradigm and the presence or absence of anesthesia (Martin et al., 2006). In some of the earlier studies changes in CBF and contrast enhanced cerebral blood volume (CBV) were measured as markers for functional brain activity. To optimize anatomical fidelity a spin echo pulse sequence can be used. It should also be emphasized that high neuroanatomical fidelity and spatial resolution are critical in identifying distributed neural circuits in any animal imaging study. Many brain areas in a segmented rat atlas have in-plane boundaries of less than 400 μm² and may extend for over 1,000 μm in the rostral/caudal plane. With the development of a segmented, annotated 3D MRI atlas for rats (Ekam Solutions, Boston, MA, United States), it is now possible to localize functional imaging data to 173 precise 3D “volumes of interest” in clearly delineate brain areas. Therefore, it is critical that the functional images are a very accurate reconstruction of the original brain neuroanatomy as shown in Figure 2.

This modality was originally described by Biswal and colleagues as connections between brain areas based on similar low frequency fluctuations in BOLD signal (Biswal et al., 1995). Realistically, resting state in awake animal imaging means no external stimulus. There is undoubtedly some level of stress associated with resting state functional connectivity (rsFC) in animals, particularly rodents. However, following different acclimation protocols, rsFC neural circuits analogous to those reported in human imaging have been identified in animals (for recent review see Pais-Roldan et al., 2021). The first method for rsFC in awake rats was published by Zhang et al. (2010) using a seed-based correlation analysis. Subsequently, studies from different labs used independent component analysis to identify functional clusters or neural circuits, e.g., default mode, salience mode, analogous to those found in humans (Becerra et al., 2011b; Liang et al., 2011; Upadhyay et al., 2011; Belcher et al., 2013; Liu et al., 2019). Tu et al. (2021) combined Designer Receptors Exclusively Activated by Designer Drugs (DREADDs) chemistry and rsFC to suppress the anterior cingulate cortex, a key circuit hub, and showed alterations in whole brain neural circuit organization and integration in the rat. Combining optogenetics with rsFC, Desai et al. (2011), showed they could activate the primary somatosensory cortex and recruit a distributed neural circuit of brain regions that included the motor cortex, striatum, and contralateral somatosensory cortex in mice. rsFC in mice has also been used to identify individual variations in the organization of the cortex that could predict differences in motor behavior (Bergmann et al., 2020). However, there is a dilemma when imaging rsFC in awake animals as noted above. How to choose between the confound of mild stress and arousal associated with the fully awake condition or the use of light anesthesia or sedation which would reflect a true “resting” condition, albeit artificial. The BOLD signal obtained with a combination of isoflurane and medetomidine, an α2 adrenergic receptor agonist, is reported to accurately reflect changes in brain activity and associated neural vascular coupling for use in rsFC (Grandjean et al., 2014; Brynildsen et al., 2017). However, there have been other studies comparing different anesthetics that do not support the use of isoflurane/medetomidine as the best choice for rsFC (Paasonen et al., 2017, 2018), a disparity possibly due to the dosing regimen of medetomidine (Kint et al., 2020).

Over time there have been several publications across numerous mammalian species on methods and applications for awake imaging. These include mice (Ferris et al., 2014), rats (Febo, 2011), voles (Yee et al., 2016), rabbits (Weiss et al., 2018), cats (Ma et al., 2013), dogs (Berns et al., 2012), pigeons (Behroozi et al., 2020) common marmosets monkeys (Ferris et al., 2001; Papoti et al., 2013; Silva, 2017), and rhesus macaques (Goense et al., 2010). Each of these species, most obviously rhesus monkey, have general or unique attributes that provide a better understanding of the human condition. For example the special relationship between dogs and humans and the communication between them [see seminal studies by Andics et al. (2014, 2016) and Huber and Lamm (2017), Karl et al. (2020, 2021)]. However, much of the work today is focused on rodents. From the early CSI-II 2.0T produced by GE to the ultra-high field 11.7T system developed by Bruker, the bore size was kept small to reduce the cost of the technology while pushing the limits of spatial resolution. Hence these small-bore systems (≤30 cm) favor the use of small animals. To that point, this review is primarily focused on rodents and their many applications in awake imaging.

One of the major applications of awake imaging in rodents has been driven by pharmaceutical companies to aid in characterizing or “fingerprinting” drug candidates in early drug discovery. Does a drug candidate get into the brain and interact with its target in a dose-dependent manner? The general protocol is simple and usually involves three different doses of drug plus a vehicle control. These are usually small molecules with specific CNS targets. Scientists from Abbott Laboratories (Abbott Park, IL, United States) pioneered the work on Pharmacological MRI (phMRI) in awake rats. Using a potent agonist for α4β2 nicotinic acetylcholine receptor they mapped the change in brain activity that fit the know localization of the receptors demonstrating target engagement and activation of neural circuitry that fit the behavioral data (Skoubis et al., 2006). In another study they used phMRI to show dose-dependent activation of neural circuitry involved in emesis as a model for predicting emetic liability of early drug candidates (Chin et al., 2006). Addressing the topic of pain regulation through cannabinoid signaling they used a non-selective CB1/CB2 and selective CB2 receptor agonists to affect BOLD signal and concluded CB1 and not CB2 was the primary target of cannabinoid signaling (Chin et al., 2008). They also mapped the BOLD activation pattern of subanesthetic doses of ketamine showing both positive and negative BOLD changes over the brain (Chin et al., 2011). These studies included drugs that would block the effects of ketamine as a potential imaging method for screening new anti-psychotics (Baker et al., 2012). Tang et al. (2018) used phMRI to map brain activation in awake rats in response to ketamine and traxoprodil, a putative antidepressant and selective GluN2B NMDA receptor antagonist. Traxoprodil produced a dose-dependent change in specific brain regions many that overlap with the activation pattern of ketamine (Tang et al., 2018).

In a just published study in mice treated with the phytocannabinoid cannabidiol (CBD) we reported a dose-dependent polarization of BOLD signal change along the rostral-caudal axis of the brain with the forebrain showing positive signal while the brainstem and cerebellum presented with negative signal (Sadaka et al., 2021). The brainstem ascending reticular activating system (ARAS) was decoupled to much of the brain but was hyperconnected to the olfactory system and prefrontal cortex. In one of the more compelling phMRI studies, Becerra et al. (2013) ran a parallel experiment between rats and humans comparing species equivalent doses of buprenorphine, a partial mu-opioid receptor agonist on circuit activation and showed comparable patterns of BOLD signal change. This study demonstrated the significance of phMRI in translating drug function between rats and humans in early drug discovery. In a subsequent study, published only a year later, this translation of buprenorphine activity across species using phMRI was also demonstrated using cynomolgus monkeys (Seah et al., 2014).

Not all small molecules with CNS targets present with a dose-dependent change in brain activity. To this point, is evidence from our lab on a study using the neuropeptide oxytocin, a neurochemical signal involved in social and emotional behavior (Carter et al., 2020). Systemically administered oxytocin failed to induce a dose-dependent change in BOLD signal in a study comparing central vs peripheral routes of oxytocin administration (Ferris et al., 2015). While a single dose of intracerebroventricular (ICV) oxytocin activated many of the brain areas with a high density of oxytocin receptors, peripheral oxytocin failed to do so. However, peripheral oxytocin had its own unique “fingerprint” affecting the olfactory bulb, cerebellum, and several brainstem areas relevant to the ARAS. The patterns of brain activity suggest that peripheral oxytocin may interact at the level of the olfactory bulb and, through sensory afferents from the autonomic nervous system, influence brain activity This raises an interesting consideration when performing any functional brain imaging study in response to a systemically administered drug. While the drug may cross the blood brain barrier and show target engagement, the final profile of BOLD activation may have contributions from the peripheral nervous system.

There have been many imaging studies on awake animals following changes in brain activity to natural stimuli like olfaction, visual cues, sound, touch, and taste. Mice given light pulses of different durations and frequencies showing activation in visual pathways (Dinh et al., 2021). The most common stimulus in rodents is the presentation of an odor. Kulkarni et al. mapped changes in brain activity to four different odors presented to rats during the scanning session (Kulkarni et al., 2012). As expected, all odors activated olfactory pathways; however, one odor, benzaldehyde, the smell of almond, increased activity in brain areas associated with emotion, memory, and reward. Indeed, mammals have genes that code for the benzaldehyde receptor in olfactory epithelium (Bozza et al., 2002). That evolution would favor the smell of almond is not surprising given its caloric density and durable structure. Hence the smell of almond elicits a positive emotional response in rats. What makes this study so interesting is that rats challenged with benzaldehyde were “odor naïve” never having been exposed to almond. The response was innate! The brain is wired to recognize benzaldehyde as a rewarding stimulus. As a validation for awake imaging in prairie voles, animals were presented with different odors and the change in brain activity mapped to a 3D MRI vole atlas (Yee et al., 2016). Zhifeng and coworkers developed an awaking imaging system in mice that allows behavioral responses, e.g., licking, to odor cues (Han et al., 2019). Using an implanted head holder, mice can be trained to lick for water during in a go/no-go conditioning paradigm driven by odor cues. This same awake mouse imaging system can also be used to follow brain activity in response to sensorimotor and auditory stimuli during the scanning session (Chen et al., 2020). Male marmoset monkeys presented with the smell of sexual receptive females show positive BOLD activation in many areas of the brain (Ferris et al., 2001) that is reversed, i.e., negative BOLD, when the stimulus comes from an ovariectomized female (Ferris et al., 2004).

With respect to marmoset monkeys, there are awake imaging studies mapping brain activity in response to touch, sound, and visual stimuli. The application of light touch by a puff of air delivered to the face, arm or leg produces cortical and subcortical maps of topographical body representation (Clery et al., 2020a). Presentation of mixed and pure tones revealed the auditory cortex has tonotopic organization for processing complex and simple stimuli (Toarmino et al., 2017). There are several studies looking at the response of the visual cortex and associated neural circuitry to various types of visual stimulation paradigms. Marmosets can be taught to carry out visual tasks during the scanning session in response to the presentations of different images, e.g., conspecific faces, conspecific body parts and general objects, resulting in the activation of visual cortex and subcortical areas (Hung et al., 2015b). The regional pattern of activity is selective for different categories of visual stimuli. For example, the occipitotemporal pathway shows the strongest preference for faces over other stimuli (Hung et al., 2015a). Objects in motion, coming toward or receding, evoke activity across a large cortical network involving frontal, parietal, temporal, and occipital areas. The activation is greatest as objects come closer suggesting a network for processing stimuli entering a peri personal space (Clery et al., 2020b). Marmosets can be imaged while watching movies to record patterns of brain activity that reflect a more naturalistic, complex stimulus. Patterns of brain activity in movie driven fMRI between marmoset monkeys and humans show similar functional correlates and circuit organization (Hori et al., 2021).

The original “vivarium” studies included the resident/intruder model (Ferris et al., 2008), pup suckling model (Ferris et al., 2005; Febo et al., 2008) (see Figure 3) and another imaging lactating dams responding to the visual presentation of their own pups with and without a male intruder (Nephew et al., 2009). This staged environmental situation using other animals to elicit changes in brain activity was recently used by Gilbert and coworkers to model social interactions between marmosets by simultaneously measuring brain activity in awake marmosets during the scanning session (Gilbert et al., 2021). Each restraining system was equipped with its own RF electronics. Each monkey was positioned facing each other providing the means for visual and olfactory communication. The activation maps generated by this social interaction were associated with visuomotor tasks.

Studies on pain necessitate the use awaking imaging. The data collected from acute and chronic pain models provide insights into the integrated neural circuits processing nociceptive stimuli and the neuroplasticity of these circuits to chronic pain and their responsiveness to therapeutics. Becerra et al. (2011a) applied a thermal stimulus of 48°C to the hindpaw of awake rats and reported activation of cortical, subcortical and brainstem areas consistent with the know literature on pain neural pathways. These researchers also studied a unique rat model for migraine pain that entails the application of a cocktail of inflammatory agents to the dura of awake rats to activate nociceptive fibers in the meninges (Becerra et al., 2017). This inflammatory soup (IS) model was used to look at the altered sensitivity to mechanical and thermal stimulation and resting state FC during the scanning session. With this model of intracranial pain they found hypersensitivity to mechanical stimulation of the face as evidenced by an increase in BOLD signal in hypothalamus, hippocampus, and somatosensory cortex with altered rsFC. In a subsequent study sumatriptan, 5HT-1B/D-receptor agonist and naproxen sodium, a COX-1 and COX-2 inhibitor, drugs used to treat migraines were tested in the IS model and shown to alter whole brain neural circuits that included the default mode, salience, sensorimotor, and cerebellar networks (Bishop et al., 2019). The IS model has also been used by others to show altered rsFC in the periaqueductal gray and anterior cingulate to brain regions involved in the sensory and emotional aspects of migraine headaches (Jia et al., 2017, 2019). Yee et al. (2015) looked at the response to intradermal capsaicin injection into the hindpaw of rats. The specificity of the pain response to capsaicin was tested in a transgenic rat model with a deletion of the TRPV1 gene, the target of capsaicin. The putative neural circuitry of pain is activated in wild-type but not transgenic rats. In addition, wild type but not KO rats show activation of Papez (Kulkarni et al., 2012) and habenular neural circuitry both of which play important roles in emotional experience, learning and memory of aversive information. Awake imaging has been used to test and demonstrate the efficacy of a DALDA analogue, a potent mu-opioid peptide delivered as an encapsulated nanoparticle on capsaicin-induced pain (Shah et al., 2014). Chang et al. (2017) showed the development of neuropathy was associated with changes in BOLD activity in prefrontal ctx and accumbens. With respect to awake imaging and pain it should be noted that the process of acclimating animals to the stress of restraint may affect the pain response, a point of concern that should be considered when interpreting the data (Low et al., 2016).