Sensory evoked fMRI paradigms in awake mice

NeuroImage 204 (2020) 116242

Contents lists available at ScienceDirect

NeuroImage journal homepage: www.elsevier.com/locate/neuroimage

Sensory evoked fMRI paradigms in awake mice Xifan Chen a, b, 1, Chuanjun Tong a, c, 1, Zhe Han a, Kaiwei Zhang a, Binshi Bo a, Yanqiu Feng c, Zhifeng Liang a, * a Institute of Neuroscience, CAS Center for Excellence in Brain Sciences and Intelligence Technology, Key Laboratory of Primate Neurobiology, Chinese Academy of Sciences, Shanghai, China b University of Chinese Academy of Sciences, Beijing, China c School of Biomedical Engineering, Guangdong Provincial Key Laboratory of Medical Image Processing, Key Laboratory of Mental Health of the Ministry of Education, Southern Medical University, Guangzhou, China

A B S T R A C T

Mouse fMRI has become increasingly popular in the small animal imaging field. However, compared to the more commonly used rat fMRI, it is challenging for mouse fMRI to obtain robust and specific functional imaging results. In the meantime, in other neuroscience modalities such as optical imaging, functional recording in the awake mice is common and becoming standard. Therefore, in the current study we developed comprehensive setups and analysis pipeline for multi-sensory fMRI paradigms in the awake mice. Customized setups of somatosensory (whisker), auditory and olfactory stimulation were developed for use in the cryogenic coil in the awake mouse fMRI setting. After carefully evaluating head motion and motion artefacts, the nuisance regression approach was optimized for reducing the confounding effect of motion. The high temporal resolution data (TR ¼ 0.35 s) revealed fast temporal dynamics (time-to-peak ~2 s) of evoked BOLD responses in most brain regions. Using the derived awake mouse specific hemodynamic response functions, high spatial resolution data revealed robust, specific and consistent cortical and subcortical activations in response to somatosensory, auditory and olfactory stimulations, respectively. Overall, we present comprehensive methods for acquiring and analyzing sensory evoked awake mouse fMRI data. The establishment of multi-sensory paradigms in awake mouse fMRI provides valuable tools for examining spatiotemporal characteristics and neural mechanisms of BOLD signals in the future.

1. Introduction Mice is perhaps the most widely used model organism in neuroscience and the broader biomedical research. As such, numerous transgenic models and neuroscience tools can be conveniently accessed when using mice for functional magnetic resonance imaging (fMRI) research. However, among the small animal fMRI literature, rats have been the predominantly used species. A large number of rat fMRI studies, both in stimulus evoked or resting states, have been reported starting early days of fMRI (Ogawa et al., 1990). For example, forepaw (or hindpaw) electrical stimulation in rats is known to be a reliable way to produce robust and spatially specific activation under proper anesthesia conditions, and thus has used as a common paradigm to investigate spatiotemporal characteristics and contrast mechanisms of fMRI signals (Herman et al., 2013; Kim et al., 2010; Lu et al., 2016; Masamoto et al., 2007; Schulz et al., 2012). Other sensory modalities, such as visual (Liang et al., 2017), auditory (Zhang et al., 2013) or olfactory stimulation (Li et al., 2014; Poplawsky et al., 2015), were also widely reported in rats to generate robust and localized responses.

In contrast to the wide popularity of rat fMRI, mouse fMRI studies have been sparse in the past and previous studies often generated inconsistent results in the sensory stimulation paradigms. For example, using electrical forepaw stimulation Adamczak et al. reported only ~50% of scans showed positive responses using medetomidine for sedation (Adamczak et al., 2010), and Schroeter et al. reported hindpaw stimulation lead to diffuse activation patterns in all four anesthesia methods including isoflurane, medetomidine, propofol and urethane (Schroeter et al., 2014). The challenge of mouse fMRI might be due to an order of magnitude smaller body size (~25 g) compared to the rat (~300 g) and the resulting difficulty of maintaining stable physiological conditions especially in the anesthetized state. As anesthesia contributes a significant portion of difficulty, efforts have been made to develop and optimize anesthesia methods to achieve robust and specific activation in mouse fMRI (Petrinovic et al., 2016; Schlegel et al., 2015; Sharp et al., 2015; Shim et al., 2018). Combined fentanyl-fluanisone/midazolam and isoflurane was used in CBV weighted intrinsic optical imaging to produce comparable evoked hemodynamic responses to those from the awake state (Sharp et al., 2015). Etomidate was proposed as a novel anesthetic

* Corresponding author. E-mail address: [email protected] (Z. Liang). 1 X.C. and C. T. contributed equally. https://doi.org/10.1016/j.neuroimage.2019.116242 Received 9 July 2019; Received in revised form 8 September 2019; Accepted 2 October 2019 Available online 3 October 2019 1053-8119/© 2019 Elsevier Inc. All rights reserved.

X. Chen et al.

NeuroImage 204 (2020) 116242

skull was adhered to the edge of the skull, and a layer of light-curing (blue light) self-etch adhesive (3M ESPE Adper Easy One) was applied to cover the interparietal bone and the surrounding regions. After the adhesive was light cured, a thick layer of light-curing flowable resin was applied on the same location, then a custom-made head holder was closely placed on the interparietal bone and the resin was cured completely by blue light. Additional thick layer of resin was applied around the head holder. Other exposed regions of the skull were covered with a thick smooth layer of dental cement to prevent the skull from inflammation. After the surgery, mice were given seven days for recovery. A thick layer of Kwik-cast was applied smoothly over the skull immediately before the MRI scanning to reduce distortion caused by air.

in mouse to provide stable brain perfusion and cerebrovascular reserve capacity (Petrinovic et al., 2016). More recently, a combined ketamine and xylazine anesthesia regime was reported to produce reproducible and specific contralateral somatosensory cortex activation in mouse fMRI (Jung et al., 2019; Shim et al., 2018). Despite the above progress in optimizing mouse fMRI, anesthesia still imposes major restrictions in making mouse fMRI a routine and truly translational imaging technique. Many of those anesthesia methods require detailed physiological monitoring and adjustment of anesthetic dosage during the process (Shim et al., 2018), or require mechanical ventilation (Petrinovic et al., 2016; Schlegel et al., 2015, 2018), all of which render the setup difficult or time consuming. Furthermore, the anesthesia or sedation, by definition, alters neural activity and thus could alter functional imaging results. To this end, it has been increasingly common to perform functional imaging without anesthesia or sedation (Gao et al., 2017), particularly in the optical imaging field where the awake imaging is becoming the new default. Awake mouse fMRI has also been reported by several groups in the resting state (Madularu et al., 2017; Yoshida et al., 2016), with optogenetic modulation (Desai et al., 2011; Takata et al., 2018), in disease models (Matsubayashi et al., 2018) or in behaviorally relevant paradigms (Han et al., 2019; Harris et al., 2015). However, the awake mouse fMRI has not yet been fully developed with many unresolved issues. For example, the shapes and potential regional variations of awake mouse specific hemodynamic response functions (HRFs) have not been fully investigated, and also the impact of head motion has not been fully investigated. To this end, the current study aimed to establish a full pipeline of multi-sensory fMRI paradigms in the awake mouse, with optimized experimental setups and data preprocessing and analysis pipeline. Using this pipeline, robust and specific cortical and subcortical BOLD activations were found upon somatosensory (whisker), auditory and olfactory stimulation. With the increasing interest in conducting fMRI in the awake mouse, the current study provides a comprehensive guide to acquire and analyze sensory evoked fMRI in the awake mouse. This will enable neuroimaging investigations of fMRI signal characteristics and neural mechanisms, as well as translational studies of numerous transgenic mouse lines of brain disorders in the future.

2.3. Habituation After recovery from the surgery, mice were then habituated for imaging for seven days. For the first four days of habituation, mice were head fixed on the animal bed with the recorded acoustic MRI scanning noise. Additional sensory stimulation was applied to mice in the last three days, to better mimic the actual imaging condition. The detailed habituation schedules were listed in Table 1. No reward was given during and after the habituation training. 2.4. MRI acquisition All MRI data were acquired with a Bruker BioSpec 9.4T scanner. An 86 mm volume coil was used for transmission and a 4-channel cryogenic phased array mouse head coil (Bruker) was used for receiving. Mouse was head-fixed as described in previous section and was positioned in the scanner. A T2 RARE anatomical image (TR: 3200 ms; TE: 33 ms; matrix Size: 256  256; FOV: 16  16 mm2; slice thickness: 400 μm; resolution: 62.5  62.5 μm2) was acquired for coregistration. After a local shimming was applied, single-shot echo planar imaging (EPI) images were acquired during stimulation. Two sets of EPI images were acquired for each animal: high spatial resolution ones and high temporal resolution ones. Parameters for high spatial resolution EPI were: TR 1500 ms, TE 15 ms, FA 60 , matrix size 100  67 (somatosensory and auditory stimulation) or 105  60 (olfactory stimulation), 150  150 μm2 (somatosensory and auditory) or 100  100 μm2 (olfactory) nominal in-plane resolution, 400 μm (somatosensory and auditory) or 300 μm (olfactory) slice thickness, 15 (somatosensory and auditory) or 12 (olfactory) slices. High temporal resolution EPI images were acquired with the following parameters: TR 350 ms, TE 15 ms, matrix size 75  60 (somatosensory and auditory) or 50  38 (olfactory), 200  200 μm2 nominal in-plane resolution, 400 μm (somatosensory and auditory) or 300 μm (olfactory) slice thickness, 10 slices, GRAPPA acceleration factor 2 for somatosensory and auditory stimulations. 10 EPI scans were acquired for each mouse (5 with high spatial resolution and 5 with high temporal resolution). Acquisition parameters were summarized in Table S1. Respiratory signals were recorded for EPI sessions using a pneumatic pillow (SAII). Sensory stimulation paradigms. The overall stimulation setups are shown in Fig. 1. A total of 40 mice were used (14 for somatosensory (whisker) stimulation; 13 for auditory stimulation; 13 for olfactory stimulation). For all sensory modalities, in each EPI scan, 20 stimuli were delivered with durations of 1 s, 2 s, 4 s or 8 s, and a fixed inter-trial-interval (ITI) time of 15 s. Stimuli of different durations were randomly ordered in each EPI scan. Details of stimulation setups are described below.

2. Methods 2.1. Animals Male adult C57BL/6 mice were used in the current study (weighted between 18 and 30 g). For whisker and olfactory stimulation tasks, eight to nine week old mice were used. For auditory stimulation task, six to seven week old mice were used to prevent age related hearing loss in the C57BL/6 strain (Parham, 1997; Walton et al., 2008). Mice were group housed (5–6/cage) under a 12-h light/dark cycle (light on from 7 a.m. to 7 p.m.) with food and water ad libitum. All animal experiments were approved by the Animal Care and Use Committee of the Institute of Neuroscience, Chinese Academy of Sciences, Shanghai, China. 2.2. Head holder implantation Head holder design and the surgical procedure were similar to our previous study (Han et al., 2019). All surgical procedures were under the standard aseptic condition. Animals were anesthetized with 5% isoflurane for induction and 1–2% for maintenance. Body temperature was maintained with a heating pad (mouseSTAT, Kent scientific cooperation). The scalp over the opened skull was removed, and then the periosteum and other associated soft tissues were removed (3% hydrogen peroxide was used to better clean the soft tissues attached to the skull), as well as the muscle tissue covering the skull posterior to the lambda point. The skull was cleaned by saline with cotton buds, and hemostasis was performed with absorbable gelatin sponge until the bleeding was completely stopped. After the skull was dried, the skin surrounding the

2.5. Whisker stimulation A small brush driven by a step motor (model: 42BYG34-401A, Telesky, China) and a motor driver (model: TB6600, Telesky, China) were used for whisker stimulation. Immediately before the experiment, whiskers were trimmed to 0.7–1 cm to prevent from touching the cryoprobe. The whisker rod was placed close to the mouse’s face (~0.3 cm). 2

X. Chen et al.

NeuroImage 204 (2020) 116242

Table 1 Summary of habituation schedules for awake mouse fMRI.

Duration of Habituation Acoustic Noise Stimulation

day1

day2

day3

day4

day5

day6

day7

30min – –

40min 95 dB –

50min 100 dB –

60min 110 dB –

30min 110 dB þ

45min 110 dB þ

60min 110 dB þ

“/þ” denote absence or presence.

Fig. 1. Illustration of multi-sensory stimulation setups for awake mouse fMRI.

pure air in the delivery port, odors were delivered to the mouse. In the current study, two odors (3-methyl-2-buten-1-ol and hexanoic acid) were used as stimuli to prevent adaptation to a single odor. Two odors were pseudo-randomly delivered, with the same total delivery time. All the odors were supplied as a mixture of air and saturated steam of odor (24  C, 45% humidity). The flow rate of air was 0.60 L/min; the rate of 3-methy-2-buten-1-ol was 1.55 L/min, and the rate of hexanoic acid was 1.13 L/min. The odor delivery was controlled by an Arduino Mega board and Arduino script.

When the whisker rod moved back and forth, the majority of whiskers could be stimulated bi-directionally. Depending on the location of individual whiskers, the angle to which whiskers were stimulated is from ~45 to 90 . The step motor were controlled by an Arduino Uno board (https://www.arduino.cc/) and deliver the whisker stimulation at 10 Hz frequency. The step motor was positioned 4–5 m away from the magnet bore center. 2.6. Auditory stimulation Auditory stimulation was delivered by a MR compatible earphone (Model S14, Sensimetrics) which connected through soft plastic tubes into the mouse ears. The sound pressure of auditory stimulus was set to be ~90 dB. The sound pressures of EPI noise measured at 1 m from magnet bore were ~96 dB for high spatial resolution EPI and ~95 dB for high temporal resolution EPI. Mouse ears, extension tubes and the earphone were covered by a sound-proof silicone (ABR impression silicone, Otoplastik) to reduce the MRI scanning noise. Pure tones (frequency range: 5500–10000 Hz; duration: 50 ms) were used as auditory stimuli. Spectral analysis of delivered sounds and scanner noises can be found in Fig. S1.

2.8. Data analysis The data analysis flow was summarized in Fig. 2. All data were processed using custom scripts in MATLAB (MathWorks, Natick, MA) and SPM12 (http://www.fil.ion.ucl.ac.uk/spm/). After the image format conversion, the mouse brain was extracted manually using ITK-SNAP (http://www.itksnap.org/). The images of each scan were slice-timing corrected and realigned to the first volume for each scan. After field map based EPI distortion correction, EPI images were registered to a mouse brain template (http://www.imaging.org.au/AMBMC/Model) for group analysis and spatially smoothed (0.4 mm isotropic Gaussian kernel for somatosensory and auditory data, 0.1  0.1  0.2 mm for olfactory data). To evaluate the head motion during the imaging, 6 realignment parameters and frame-wise displacement (FD) were obtained. The FD was calculated by the modulus of the differential realignment parameters for each scan assuming a mouse brain radius of 5 mm (Power et al., 2012).

2.7. Olfactory stimulation setup The detailed setup of the olfactory stimulation apparatus can be found in our previous study (Han et al., 2019). Briefly, odors were stored in a 50 ml Eppendorf centrifuge tube. After vaporized and mixed with the 3

X. Chen et al.

NeuroImage 204 (2020) 116242

Fig. 2. Flowchart of data analysis pipeline.

apparent post-stimulus undershoot across three modalities (Fig. S3). The correlation coefficients between observed BOLD responses and the values expected under the model were described as the measurement of the goodness of fitting. Time-to-peak (TTP) and FWHM of the HRFs were also calculated. GLM based statistical analysis was conducted using the above HRFs combined with temporal and dispersion derivatives in SPM12 for high spatial resolution EPI data (TR ¼ 1.5 s). Standard 1st level analysis was done for individual EPI scans and individual animals with FDR corrected p < 0.05 and cluster size >10 voxels. For 2nd level analysis, one sample ttest was conducted to generate the group activation maps with FDR corrected p < 0.005 and cluster size >10 voxels. All activation maps were overlaid on mean EPI images for more accurate visualization.

To assess the impact of motion parameter regression, three regression based motion corrections were compared: (1) 6 regressors: 6 head motion parameters, i.e., realignment parameters from SPM; (2) 12 regressors: 6 head motion parameters and their 1st order derivatives (Satterthwaite et al., 2013); (3) 22 regressors: 12 motion related regressors and 10 non-brain tissue based principal components (PCs). The top 10 PCs of signals outside the brain were estimated, e.g. from the muscle of the jaw, and were used to model non-neural signal variations (Chuang et al., 2019). The impact of nuisance signal regression on stimulus related head motion and resulting fMRI signal variations was shown in Fig. S2. The Pearson correlation coefficients between FD and global signal were calculated to quantitatively reflect the extent to which the motion related signal was reduced by given regressors. GLM based statistical parametric mapping is known sensitive to the hemodynamic response function (HRF) (Schlegel et al., 2015), therefore HRF was empirically estimated from the averaged time series of high temporal resolution EPI images. The high temporal resolution EPI data (TR ¼ 0.35 s) were detrended by a 0.01 Hz high-pass filter. Regions-of-interest (ROIs) were defined according to our previous study (olfactory stimulation: dorsal posterior olfactory bulb (dpOB) and lateral anterior olfactory bulb (laOB)) (Han et al., 2019) and the Allen mouse brain atlas (http://atlas.brain-map.org) (somatosensory stimulation: contralateral and ipsilateral primary somatosensory barrel cortex (S1BF) and contralateral and ipsilateral ventral posterior complex of the thalamus (VP); auditory stimulation: primary auditory cortex (AuD) and medial geniculate complex (MG)). Time-locked averaging was then employed to obtain averaged time series containing baseline (2 s), stimulus (1, 2, 4 or 8 s) and post-stimulus (13 s) epochs. As FD was not significantly correlated with the global signal of high temporal resolution EPI images (Fig. 3), time series were extracted without any regressions. The averaged time series were converted to relative BOLD responses (ΔS(t)/S), and then were modeled to a single gamma variate function x 1 f ðxja; bÞ ¼ ba ΓðaÞ xa1 e b with the corresponding shape parameter a and

3. Results The multi-sensory stimulation paradigms in awake mice was illustrated in Fig. 1. Based on a head holder specifically designed for use with a mouse cryogenic coil (Fig. 1, upper right), whisker, auditory and olfactory stimulation setups were developed and details of those setups can be found in Methods section. After modality specific habituation for 7 days (Table 1), mice were imaged in the awake condition. For each sensory modality, FOV locations were shown in Fig. S4 for both high spatial and high temporal resolution acquisition. Data acquisition parameters were summarized in Table S1, and representative SNR and temporal SNR (tSNR) maps for each acquisition schemes were shown in Fig. S4. Overall SNR and tSNR were highly correlated with each other, and high spatial resolution data showed higher SNR and tSNR than high temporal resolution data (Fig. S4). The data analysis pipeline was summarized in Fig. 2. The head motion was evaluated using realignment parameters and the resulting framewise displacement (Fig. 3 and Fig. S5). Overall the head holder system and habituation yielded low levels of head motion. In the raw data, the median frame-wise displacement was in the range of ~4–8 μm for all three sensory modalities, and the averaged maximum displacement was from 10.8 to 28.5 μm (Fig. 3A and B). Specifically, most head motion occurred in the dorsal-ventral direction (Fig. S5), which is likely due to the design of our head holder system. A-B, frame-wise displacement of three modalities of high spatial resolution (A) and high temporal resolution (B) data. Each dot represented a single EPI scan. C-D, the impact of nuisance regressions on the correlation between frame-wise displacement and global signal of high

scale parameter b. Briefly, the averaged time series were montaged in ascending order according to the stimulation duration. Then, the “montaged” time series were modeled by the convolution between the boxcar waveform and a gamma variate function, as it did not show 4

X. Chen et al.

NeuroImage 204 (2020) 116242

Fig. 3. Characterizations of the head motion and the impact of nuisance signal regressions.

(Fig. S6), and the regression of “12 rp þ 10 PCs” was effective to remove stimulus induced signal changes (Fig. S7). For high temporal resolution data, although there were still global signal fluctuations, its correlation with motion parameters was very low with or without nuisance regression (Figs. S2G and H). Respiration was monitored throughout the imaging session, and respiratory rates showed no significant variation during individual EPI scans in response to stimulation (Figs. S8A and B). The mean respiratory rates remained relatively stable during an imaging session, with a slight decrease over time for somatosensory and auditory stimulation (Fig. S8C). To fully establish the statistical analysis pipeline for awake mouse fMRI, it is essential to characterize the temporal dynamics of evoked BOLD signals. Therefore, we took advantage of the high temporal resolution data (TR ¼ 0.35 s) to describe and model the BOLD temporal profiles (Fig. 4). Whisker (Fig. 4A), auditory (Fig. 4B) and olfactory (Fig. 4C) stimulation all lead to robust increase of BOLD signals from cortical regions and thalamic nuclei involved in each sensory modality, respectively. Parametric modelling of the hemodynamic response function using one gamma variate function indicated fast and similar responses in all four non-olfactory related regions (S1BF, VP, AuD and MG) with time-to-peak ranging from 1.6 to 2.35 s, which are much faster than those of human HRFs (typically 5–6 s). Interestingly, the temporal dynamics in the olfactory bulb in response to two different odors were different from the other two modalities, with slower dynamics (time-topeak around 3.75 s). In addition, with a longer stimulation duration of 8 s, adaption of the evoked responses became apparent in somatosensory and auditory stimulation (Fig. 4A and B), while the olfactory responses of

spatial resolution (C) and high temporal resolution (D) data. Statistical test included one-sample t-test for the correlation between FD and global signal and pair-wise t-test for the impact of nuisance signal regression on the correlation between FD and global signal. maxdev, the maximum of derivatives of frame-wise displacements. C.C., correlation coefficients. n.s., not significant; *, p < 0.05; **, p < 0.01; ***, p < 0.005; ****, p < 0.001. Despite the overall low head motion, head motion is inevitable for functional imaging, especially in the awake animal (or human). Therefore, we examined the effects of nuisance regressor choices for suppressing motion artefacts. The correlation coefficients between the frame-wise displacement and the global signal were found to be nonsignificant for high temporal resolution data (Fig. 3D), indicating known resistance of high temporal resolution data to head motion. However, high spatial resolution data exhibited decreasing correlation between global signals and frame-wise displacement upon additional motion regressors (Fig. 3C). Compared to 6 or 12 realignment parameters, regression of “12 rp þ 10 PCs” significantly reduced the correlation between the global signal and frame-wise displacement. Detailed examination of whole-brain time series showed various motion patterns and motion artefacts, as well as the effects of nuisance regression (Fig. S2). In the high spatial resolution data (Figs. S2A–F), stimulus related head motion and resulting global BOLD signal fluctuations were apparent in the raw data (Figs. S2A–D). Large motion spikes were also seen in some data (Figs. S2E–F). Nuisance regression of “12 rp þ 10 PCs” notably reduced the impact of motion, while the 6 or 12 realignment parameters did not seem to suffice (Figs. S2A–F). The maximum frame-wise displacement was significantly decreased with nuisance regressions 5

X. Chen et al.

NeuroImage 204 (2020) 116242

Fig. 4. Temporal dynamics of evoked time series of high temporal resolution data.

6

X. Chen et al.

NeuroImage 204 (2020) 116242

predicted and actual time series. S1BF(L), left primary somatosensory barrel cortex; VP(L), left ventral posterior complex of the thalamus; AuD, primary auditory cortex; MG, medial geniculate complex; dpOB, dorsal posterior olfactory bulb; laOB, lateral anterior olfactory bulb. TTP, time to peak; FWHM, full width at half maximum. odor 01, 3-methyl-2-buten1-ol. A, illustration of somatosensory related ROIs; B, upper panel, group activation map with FDR corrected p < 0.005 and cluster size >10 voxels; lower panel, group activation map without thresholding. C, time series in response to 1/2/4/8 s(s) stimulation. Vertical black lines indicated the start of each stimulation. S1BF, primary somatosensory barrel cortex; S2, secondary somatosensory cortex; VP, ventral posterior complex of the thalamus; AuD, primary auditory cortex; MG, medial geniculate complex; (L/R), left/right. For bilateral auditory stimulation, brain regions along the auditory pathway showed robust activations, including bilateral auditory cortex (AuD), medial geniculate nucleus (MG) and inferior colliculus (IC) (Fig. 6A and B). Time series of AuD, MG and IC showed robust bilateral activation (Fig. 6C). As expected, olfactory stimulation by two odors (3methyl-2-buten-1-ol and hexanoic acid) all lead significant activation in the olfactory bulb (Fig. 7). The activations were mostly located in the outer portion of the olfactory bulb, clearly suggesting a preferential activation of the glomerular layer. As expected, two different odors elicited responses in different parts of the olfactory bulb, presumably due to the activations of different glomeruli. In addition, activation maps of 1 s, 2 s, 4 s and 8 s stimulation duration were extracted separately and showed overall consistent activations across stimulation durations (Fig. S13) with some differences in response amplitudes. In addition, for all three modalities, similar activation results were obtained from high temporal resolution data (Fig. S14). A, illustration of auditory related ROIs; B, upper panel, group activation map with FDR corrected p < 0.005 and cluster size >10 voxels; lower panel, group activation map without thresholding. C, time series in

8 s stimulation seemed more complicated with delayed peaks (Fig. 4C). As commonly observed in human BOLD responses, nonlinearity also occurred in the awake mice, with longer stimulus leading to less than expected BOLD amplitude (Fig. S9). Nevertheless, overall good fitting of HRF predicted and actual BOLD time series were achieved (Fig. 4A–C, lower right panels and Fig. S10). In addition, we assessed the individual variability of HRF shapes and found overall consistent temporal patterns across mice in each stimulation paradigm (Fig. S11). The above characterization was based on the raw high temporal resolution data, and very similar results were obtained with nuisance signal regression (Fig. S12). Based on the above HRFs derived from the high temporal resolution data, conventional GLM based statistical analysis was conducted in the high spatial resolution data, which revealed robust and specific BOLD activations in all three sensory modalities (Figs. 5–7). For whisker stimulation, the primary regions of the somatosensory path, both the ventral posteromedial and ventral posterolateral nuclei of thalamus (VPM/VPL) and the barrel field of primary somatosensory cortex (S1BF), showed clear and specific unilateral activation as the whiskers were stimulated unilaterally (Fig. 5A and B). Secondary somatosensory cortex (S2), as the downstream cortical target, was also activated. Unexpectedly, bilateral auditory cortex and corresponding medial geniculate nucleus (MG), also showed activation, possibly due to the acoustic noise generated by the mechanical movement of whisker stimulation apparatus. Time series of S1BF and VPM/VPL also indicated strong unilateral evoked BOLD signals, and cortical signals were stronger than thalamic ones (Fig. 5C). A, somatosensory (whisker) stimulation; B, auditory stimulation; C, olfactory stimulation. A-C upper panels, time series in response to 1/2/4/ 8 s(s) stimulation. Time series were extracted from anatomically defined ROIs marked in red in upper right inserts without any regression. Vertical dash lines indicated the start of each stimulation. A-C lower left panels, hemodynamic response functions (HRFs) modeled from a single gamma function. A-C lower right panels, the fitting results between HRF

Fig. 5. Robust and specific cortical and thalamic activations under unilateral whisker stimulation. 7

X. Chen et al.

NeuroImage 204 (2020) 116242

Fig. 6. Robust and specific cortical and subcortical activations under bilateral auditory stimulation.

presented here. B, activation maps without thresholding. Mean t-values were calculated across all scans by anatomically defined ROIs in Fig. 4 (somatosensory: unilateral primary somatosensory barrel cortex; auditory: bilateral primary auditory cortex; olfactory (odor 01): dorsal posterior olfactory bulb). Each dot represented a single high spatial resolution EPI scan. A total of 200 scans were included: 70 somatosensory scans, 65 auditory scans and 65 olfactory scans.

response to 1/2/4/8 s(s) stimulation. Vertical black lines indicated the start of each stimulation. AuD, primary auditory cortex; MG, medial geniculate complex; IC, inferior colliculus; (L/R), left/right. A, illustration of olfactory related ROIs; B, upper two panel, group activation maps in response to two different odors with FDR corrected p < 0.005 and cluster size >10 voxels; lower two panels, group activation maps without thresholding. C, time series in response to 1/2/4/8 s(s) stimulation. Time series were extracted from the significantly activated regions indicated by white arrows in A. Vertical black lines indicated the start of each stimulation. MOB, main olfactory bulb; AOB, accessory olfactory bulb; AON, anterior olfactory nucleus; vpOB, ventral posterior olfactory bulb; dpOB, dorsal posterior olfactory bulb; mpOB, medial posterior olfactory bulb; laOB, lateral anterior olfactory bulb; odor 01, 3methyl-2-buten-1-ol; odor 02, hexanoic acid. Importantly, those activation patterns were highly consistent across EPI scans of an individual animal and across individual animals, indicating high intra-subject and inter-subject consistency (Fig. 8 and Figs. S15–16). No significant difference of BOLD responses among EPI scans were observed (Fig. S17). Finally, we investigated whether and how the nuisance regression approach affected the final statistical results. Overall the head motion (as characterized by mean frame-wise displacement) showed decreasing correlation with t values of individual EPI scans upon more regression, and with regressors of “12 rp þ 10 PCs” the two showed no significant correlation (Fig. 9). For group level statistical maps, three nuisance regression pipelines generated similar activation maps in general, particularly in olfactory stimulation (Fig. S18). However, close inspection suggested at least in somatosensory and auditory stimulation, regression of “12 rp þ 10 PCs” achieved reasonable balance of sensitivity and specificity (Fig. S18). A, activation maps across EPI scans of 14 mice with FDR corrected p < 0.05 and cluster size >10 voxels. The first row were activation maps of each animal and the remaining rows were maps of individual EPI scans of each animal of high spatial resolution data. Only one slice was

4. Discussion Mouse fMRI has become increasingly popular for its rich neuroscience resources and enormous potential for bridging the basic neuroscience research and human fMRI. However, it has been challenging to conduct mouse fMRI, probably due to technical and physiological issues. In addition, awake rodent imaging has gained substantial attention, particularly in mice. Therefore, the current study present robust multisensory fMRI paradigms in awake mice with high temporal (TR ¼ 0.35 s) and high spatial (nominal in plane 0.15 mm, 0.4 mm slice thickness) resolution. With detailed characterization of temporal and spatial characteristics of sensory evoked BOLD responses in awake mice, robust, specific and consistent BOLD activations were found with somatosensory (whisker), auditory and olfactory stimulation. Thus, the current study presented robust multi-sensory fMRI paradigms in awake mice, and paved way for future large-scale and routine applications of awake mouse fMRI. 4.1. Sensory specific stimulation setups for awake mouse fMRI Unlike human fMRI studies where many commercially available stimulation devices exist, few are developed for rodent fMRI, particularly in the context of awake imaging. Therefore, the current study presented comprehensive setups of somatosensory, auditory and olfactory stimulation suitable for awake mouse imaging. Together with the awake 8

X. Chen et al.

NeuroImage 204 (2020) 116242

Fig. 7. Robust and specific activations under olfactory stimulation.

Another important assumption of the conventional fMRI analysis is the linearity of the BOLD response. While it is beyond the scope of the current study to explicitly examine the linearity of BOLD responses in the awake mice, visual inspection indicated apparent adaption with longer stimulation duration of 4 and 8 s. Such adaption is not surprising as it has been reported in both human and animal studies. Therefore, shorter stimulation or event related design will be optimal if quantitative measurement of BOLD amplitudes is of interest for future awake mouse fMRI studies. Notably, HRFs in the olfactory bulb were much slower (time-to-peak ~3.5 s) than other brain regions (time-to-peak ~2 s), suggesting a regional variation of HRFs. Our olfactometer design delivers odors with very short latency less than 100 ms (Han et al., 2018), thus this slower dynamics cannot attribute to the inaccuracy of odor delivery. However, it is still possible that further timing inaccuracy could occur as the awake mice might control their inhalation, which ultimately determines when the odor molecules arrive at the olfactory epithelium. Further detailed examination is required to resolve the discrepancy of HRFs between olfactory bulb and other brain regions.

mouse imaging setup, the design of all custom-made components can be downloaded and 3D printed (HYPERLINK “https://github.com/porridgeJoe” \o “https://github.com/porridgeJoe"https://github.com/po rridgeJoe), along with the scripts for controlling them. Therefore, through such open source approach, sensory based awake mouse fMRI can be established with relative ease and low cost. Additional improvement of those setups could be made, for example additional air suction in for the olfactory stimulation could be beneficial, although we do not expect it would significantly change the olfactory BOLD response. 4.2. Temporal characteristics of evoked BOLD response in awake mice Most fMRI studies employ the GLM based approach for statistical analysis, and usually the stimulation paradigm is convolved with a HRF before entering the design matrix. Thus it is apparent that such analysis heavily relies on the temporal characteristics of evoked BOLD signal, i.e., HRF. With known variability of HRFs across species or even physiological conditions, no specific HRFs were explicitly reported for awake mouse fMRI. Therefore, the current study specifically acquired high temporal resolution (TR ¼ 0.35 s) data with a range of short stimulation durations (1, 2, 4, 8 s) for the purpose of modeling awake mouse specific HRFs. Not surprisingly, the awake mouse HRFs were much faster than ones of human, with a time-to-peak of around 2 s compared to 5–6 s in human. This highlights the need of using awake mouse specific HRF in GLM based analysis.

4.3. Head motion and nuisance removal in awake mouse fMRI It is well known and increasingly recognized that head motion and physiological fluctuations can significantly affect the fMRI results. In the context of awake animal imaging, head motion has been the most notable 9

X. Chen et al.

NeuroImage 204 (2020) 116242

Fig. 8. High intra-subject and inter-subject consistency of activations under unilateral whisker stimulation.

Fig. 9. Impact of differential regression on relationship between median frame-wise displacement and mean t-value.

4.4. Robust and specific activations of somatosensory, auditory and olfactory stimulation

confounding factor. Therefore, in the current study we made considerable efforts to reduce motion in our setup, and characterize the effects of head motion and nuisance regression in the awake mouse. Through the optimized head holder designed for use with the mouse cryogenic coil, low head motion was achieved in the raw data while taking advantage of high SNR offered by the cryogenic coil. However, motion related signal fluctuations were still present in the raw data. In the current study we found that the nuisance regressor of “12 rp þ 10 PCs” provided effective reduction of motion artefacts, based on converging evidence from correlation between the global signal and frame-wise displacement (Fig. 3), whole-brain signal plots (Fig. S2), correlation between frame-wise displacement and statistical values (Fig. 9) and activation maps (Fig. S18). While head motion cannot be absolutely and completely removed, at least the current awake imaging setup and preprocessing steps were sufficient to generate robust and specific sensory evoked activations in awake mice.

For all three sensory modalities, robust and specific activations were obtained along the sensory processing pathways. For somatosensory (whisker) stimulation, it is well known that tactile information is passed through the lemniscal pathway, from medial lemniscus to ventral posterolateral nuclei of thalamus and finally to primary somatosensory cortex. In the current study, such pathway was shown to be activated (Fig. 5) with significant activations in VPM/VPL and S1BF. Further downstream cortical region of S2 was also found to be activated. Compared to the recent anesthetized mouse fMRI study (Jung et al., 2019; Shim et al., 2018), the current study was able to reveal thalamic activation at 9.4 T, which was only reported at 15.2 T but absent at 9.4T in the above two studies. Furthermore, the exact spatial extent of thalamic activation was different. Jung et al. reported a relative diffuse 10

X. Chen et al.

NeuroImage 204 (2020) 116242

aspects of evoked BOLD responses. Ideally, functional imaging with both high spatial and high temporal resolutions could provide the most complete information. Important methodological advances have been made towards this end, such as simultaneous multislice EPI methods. Recently, a fast imaging method (Lee et al., 2019) was also developed specifically for rodent imaging which typically have very limited number of receiver coil elements. However, usually we are still limited by hardware and pulse sequences so that achieving high spatial and high temporal resolution at the same time is still very challenging. Therefore, in the current study we acquired two datasets, one with high spatial resolution and one with high temporal resolution, in an interleaved fashion for each animal. As expected, the high temporal resolution data provided better robustness against head motion (Fig. 3 and Fig. S2), for reasons such as less EPI frames affected by head motion and less intra-volume motion. This feature is particularly advantageous in awake mouse fMRI. However, high temporal resolution data suffered from lower SNR and tSNR (Fig. S4), albeit with lower spatial resolution. Nevertheless, very similar activation maps were achieved with high temporal resolution data (Fig. S14) using the same statistical method. However, the statistical methods for analyzing the sub-second TR data are rapidly developing and likely to be very different from the conventional fMRI data (Chen et al., 2019), therefore would require extra caution when analyzing such data. The detailed comparison or optimization is beyond the scope of the current study, but it is clear that optimal parameters of future studies will be largely dependent on the specific purposes of the studies.

pattern of thalamic activation, including VPL/VPM, posterior nucleus (PO), and non-sensory-related thalamic nuclei such as mediodorsal nucleus (MD). In the current study, thalamic activation was more lateral and more confined, mostly only including VPL/VPM. While this discrepancy might be due to several methodological differences (awake v.s. anesthetized, whisker v.s. forepaw electrical stimulation), the thalamic activation in the current study appears to be more specific and agrees well with the known lemniscal pathway. Somatosensory (whisker or forepaw) evoked fMRI has been extensively utilized in the preclinical imaging field for purposes ranging from method development to neural mechanism investigation, thus the current study provides a viable alternative for those purposes without the negative impact of anesthesia or sedation. Similarly, robust and specific thalamic and cortical activations were found with auditory stimulation in MG and AuD. Given the loud noises, auditory fMRI is generally considered challenging. Several rodent auditory fMRI have been reported in rats (Cheung et al., 2012; Gao et al., 2014, 2015; Zhang et al., 2013) and more recently in mice (Blazquez Freches et al., 2018). The above studies all reported very robust activation in the subcortical structures such as inferior colliculus (IC) and, but the auditory cortex was notably missing in the mouse auditory fMRI study (Blazquez Freches et al., 2018) which was robustly activated in the current study. Several methodological differences might contribute this apparent discrepancy, such as different contrasts (T2* contrast from EPI v.s. T2/T1 contrast from FISP sequence), auditory stimulation difference (5–10 KHz broadband v.s. 5–39 KHz individual frequency sound) and the animal states (awake v.s. medetomidine sedated). In contrast to the excellent cortical activation, the subcortical mapping of the current study was limited by the distortion and signal loss in the ventral part of the midbrain and pons of the EPI images. It should be noted that mice have a markedly different hearing range than humans, with higher sensitivity towards high frequency. In addition, the commonly used mouse strain, C57BL/6J, has a known age-dependent decline of hearing starting at 2–3 months (Parham, 1997; Walton et al., 2008). Therefore, care should be taken to avoid using aging mice for auditory fMRI studies. Additionally, in auditory fMRI sparse sampling is sometimes used as a strategy to reduce the impact of scanner noises (Peelle, 2014). Using preliminary data we also compared the sparse sampling (TR 3.5s, silent period 2s) with the continuous sampling as reported here, and we found similar or even more robust activation using the continuous sampling (Fig. S1). Finally, olfactory fMRI in the awake mice also generated activations that are consistent with our understanding of the olfactory system and previous rodent olfactory fMRI studies. Olfactory fMRI in rodents have long been used for various applications such as layer specific or ultrahigh resolution fMRI and neurovascular coupling (Kida et al., 2002; Li et al., 2014; Muir et al., 2019; Poplawsky et al., 2015, 2019; Xu et al., 2003). As expected, medial and lateral outer layers (mostly olfactory nerve layer (NOL) glomerular layer (GL) were differentially activated by two different odors (3-methyl-2-buten-1-ol and hexanoic acid), as two odors most likely activated different combination of individual glomeruli in the olfactory bulb. Although very high spatial resolution (0.1  0.1  0.3 mm) was achieved in the current study, the physiological nature of the gradient echo BOLD signal limited our ability to achieve layer and glomerulus specific mapping of the odor response, which can be potentially achieved by other more spatially specific functional contrasts such as the CBV weighted contrast. And also, similar to previous studies, FOV was reduced to include mainly the olfactory bulb, due to the difficulty of obtaining good shimming conditions for both the olfactory bulb and the rest of the cerebral cortex. Therefore, the downstream regions of the olfactory pathway were not mapped. This issue will need to be solved in the future to provide a true whole brain mapping of the olfactory pathway.

4.6. Limitations For awake rodent imaging, the stress during the imaging is always a concern due to loud scanner noises and head fixation. Typically, animals are acclimated to noise and immobilization, which was shown to reduce physiological manifestation of stress, such as respiratory rate, heart rate and ultrasonic calls (King et al., 2005; Reed et al., 2013; Yoshida et al., 2016). However, long-term alterations in pain and stress responses after similar habituation procedure have also been reported (Low et al., 2016). In the current study, we adopted a 7-day habituation with progressing addition of stressors (Table 1). During the actual fMRI sessions, respiratory rates were recorded and were shown to be relatively stable with almost no stimulus driven respiratory changes (Fig. S8). As expected, the respiratory rate of around 300 per minute was higher than most of the anesthesia studies, but is still within the range of awake recordings (Lim et al., 2014). There was a slight decrease of respiratory rates over time during imaging sessions for somatosensory and auditory stimulation, but not for olfactory stimulation (Fig. S8). This phenomenon suggest further habituation (possibly with actual imaging sessions) may be beneficial. It would be technically challenging to monitor other stress related parameters during the awake imaging, but measuring serum corticosterone level could help to further characterize the stress response as it was done in the original study that used 7-day habituation (King et al., 2005). Overall, future work is needed to systematically optimize the habituation procedure and characterize the stress response and its impact of imaging results. Ultimately it will be unrealistic to completely rule out any stress during the imaging, which also happens in human fMRI studies. However, in comparison to the massive changes induced by anesthesia, such potentially elevation of stress level probably leads to much less profound changes on a global scale. Nevertheless, it is of less concern in the current study, as we examined the evoked BOLD responses compared to a baseline. Ultimately, the choice of using awake or anesthetized imaging will be dependent on the scientific question. 5. Conclusion

4.5. Comparison between high spatial and high temporal resolution fMRI in awake mice

In the current study we established comprehensive multi-sensory paradigms of awake mouse fMRI, with optimized awake imaging,

Spatial and temporal profiles are two essential and complimentary 11

X. Chen et al.

NeuroImage 204 (2020) 116242

stimulation setups and data analysis pipelines. Spatiotemporal dynamics of evoked BOLD responses were characterized in details, and robust and specific activations were revealed in response to somatosensory, auditory and olfactory stimulation. In conclusion, the awake mouse fMRI can be readily achieved, and is a viable alternative to anesthetized mouse fMRI with great potentials in neuroscience and neuroimaging research.

Lim, R., Zavou, M.J., Milton, P.L., Chan, S.T., Tan, J.L., Dickinson, H., Murphy, S.V., Jenkin, G., Wallace, E.M., 2014. Measuring respiratory function in mice using unrestrained whole-body plethysmography. J. Vis. Exp., e51755 Low, L.A., Bauer, L.C., Pitcher, M.H., Bushnell, M.C., 2016. Restraint training for awake functional brain scanning of rodents can cause long-lasting changes in pain and stress responses. Pain 157, 1761–1772. Lu, H., Wang, L., Rea, W.W., Brynildsen, J.K., Jaime, S., Zuo, Y., Stein, E.A., Yang, Y., 2016 Feb. Low- but not high-frequency LFP correlates with spontaneous BOLD fluctuations in rat whisker barrel cortex. Cerebr. Cortex 26 (2), 683–694. Madularu, D., Mathieu, A.P., Kumaragamage, C., Reynolds, L.M., Near, J., Flores, C., Rajah, M.N., 2017. A non-invasive restraining system for awake mouse imaging. J. Neurosci. Methods 287, 53–57. Masamoto, K., Kim, T., Fukuda, M., Wang, P., Kim, S.G., 2007. Relationship between neural, vascular, and BOLD signals in isoflurane-anesthetized rat somatosensory cortex. Cerebr. Cortex 17, 942–950. Matsubayashi, K., Nagoshi, N., Komaki, Y., Kojima, K., Shinozaki, M., Tsuji, O., Iwanami, A., Ishihara, R., Takata, N., Matsumoto, M., Mimura, M., Okano, H., Nakamura, M., 2018. Assessing cortical plasticity after spinal cord injury by using resting-state functional magnetic resonance imaging in awake adult mice. Sci. Rep. 8, 14406. Muir, E.R., Biju, K.C., Cong, L., Rogers, W.E., Torres Hernandez, E., Duong, T.Q., Clark, R.A., 2019. Functional MRI of the mouse olfactory system. Neurosci. Lett. 704, 57–61. Ogawa, S., Lee, T.M., Kay, A.R., Tank, D.W., 1990. Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc. Natl. Acad. Sci. U. S. A. 87, 9868–9872. Parham, K., 1997. Distortion product otoacoustic emissions in the C57BL/6J mouse model of age-related hearing loss. Hear. Res. 112, 216–234. Peelle, J.E., 2014. Methodological challenges and solutions in auditory functional magnetic resonance imaging. Front. Neurosci. 8, 253. Petrinovic, M.M., Hankov, G., Schroeter, A., Bruns, A., Rudin, M., von Kienlin, M., Kunnecke, B., Mueggler, T., 2016. A novel anesthesia regime enables neurofunctional studies and imaging genetics across mouse strains. Sci. Rep. 6, 24523. Poplawsky, A.J., Fukuda, M., Kang, B.M., Kim, J.H., Suh, M., Kim, S.G., 2019. Dominance of layer-specific microvessel dilation in contrast-enhanced high-resolution fMRI: comparison between hemodynamic spread and vascular architecture with CLARITY. Neuroimage 197, 657–667. Poplawsky, A.J., Fukuda, M., Murphy, M., Kim, S.G., 2015. Layer-specific fMRI responses to excitatory and inhibitory neuronal activities in the olfactory bulb. J. Neurosci. 35, 15263–15275. Power, J.D., Barnes, K.A., Snyder, A.Z., Schlaggar, B.L., Petersen, S.E., 2012. Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. Neuroimage 59, 2142–2154. Reed, M.D., Pira, A.S., Febo, M., 2013. Behavioral effects of acclimatization to restraint protocol used for awake animal imaging. J. Neurosci. Methods 217, 63–66. Satterthwaite, T.D., Elliott, M.A., Gerraty, R.T., Ruparel, K., Loughead, J., Calkins, M.E., Eickhoff, S.B., Hakonarson, H., Gur, R.C., Gur, R.E., Wolf, D.H., 2013. An improved framework for confound regression and filtering for control of motion artifact in the preprocessing of resting-state functional connectivity data. Neuroimage 64, 240–256. Schlegel, F., Schroeter, A., Rudin, M., 2015. The hemodynamic response to somatosensory stimulation in mice depends on the anesthetic used: implications on analysis of mouse fMRI data. Neuroimage 116, 40–49. Schlegel, F., Sych, Y., Schroeter, A., Stobart, J., Weber, B., Helmchen, F., Rudin, M., 2018. Fiber-optic implant for simultaneous fluorescence-based calcium recordings and BOLD fMRI in mice. Nat. Protoc. 13, 840–855. Schroeter, A., Schlegel, F., Seuwen, A., Grandjean, J., Rudin, M., 2014. Specificity of stimulus-evoked fMRI responses in the mouse: the influence of systemic physiological changes associated with innocuous stimulation under four different anesthetics. Neuroimage 94, 372–384. Schulz, K., Sydekum, E., Krueppel, R., Engelbrecht, C.J., Schlegel, F., Schroter, A., Rudin, M., Helmchen, F., 2012. Simultaneous BOLD fMRI and fiber-optic calcium recording in rat neocortex. Nat. Methods 9, 597–602. Sharp, P.S., Shaw, K., Boorman, L., Harris, S., Kennerley, A.J., Azzouz, M., Berwick, J., 2015. Comparison of stimulus-evoked cerebral hemodynamics in the awake mouse and under a novel anesthetic regime. Sci. Rep. 5, 12621. Shim, H.J., Jung, W.B., Schlegel, F., Lee, J., Kim, S., Lee, J., Kim, S.G., 2018. Mouse fMRI under ketamine and xylazine anesthesia: robust contralateral somatosensory cortex activation in response to forepaw stimulation. Neuroimage 177, 30–44. Takata, N., Sugiura, Y., Yoshida, K., Koizumi, M., Hiroshi, N., Honda, K., Yano, R., Komaki, Y., Matsui, K., Suematsu, M., Mimura, M., Okano, H., Tanaka, K.F., 2018 Sep. Optogenetic astrocyte activation evokes BOLD fMRI response with oxygen consumption without neuronal activity modulation. Glia 66 (9), 2013–2023. Walton, J.P., Barsz, K., Wilson, W.W., 2008. Sensorineural hearing loss and neural correlates of temporal acuity in the inferior colliculus of the C57BL/6 mouse. J Assoc Res Otolaryngol 9, 90–101. Xu, F., Liu, N., Kida, I., Rothman, D.L., Hyder, F., Shepherd, G.M., 2003. Odor maps of aldehydes and esters revealed by functional MRI in the glomerular layer of the mouse olfactory bulb. Proc. Natl. Acad. Sci. U. S. A. 100, 11029–11034. Yoshida, K., Mimura, Y., Ishihara, R., Nishida, H., Komaki, Y., Minakuchi, T., Tsurugizawa, T., Mimura, M., Okano, H., Tanaka, K.F., Takata, N., 2016. Physiological effects of a habituation procedure for functional MRI in awake mice using a cryogenic radiofrequency probe. J. Neurosci. Methods 274, 38–48. Zhang, J.W., Lau, C., Cheng, J.S., Xing, K.K., Zhou, I.Y., Cheung, M.M., Wu, E.X., 2013. Functional magnetic resonance imaging of sound pressure level encoding in the rat central auditory system. Neuroimage 65, 119–126.

Acknowledgement This work was supported by the Strategic Priority Research Program of the Chinese Academy of Sciences (Grant No. XDB32030100 to Z. L.), the Shanghai Municipal Science and Technology Major Project (Grant No. 2018SHZDZX05 to Z. L.), the General Program of National Natural Science Foundation of China (Grant No. 81771821 to Z. L, Grant No. 61671228, No.61728107 and No.81871349 to Y. F.), CAS Pioneer Hundred Talents Program (to Z. L.), the Technology R&D Program of Guangdong (Grant No. 2017B090912006 and No. 2018B030333001 to Y. F.). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.neuroimage.2019.116242. References Adamczak, J.M., Farr, T.D., Seehafer, J.U., Kalthoff, D., Hoehn, M., 2010. High field BOLD response to forepaw stimulation in the mouse. Neuroimage 51, 704–712. Blazquez Freches, G., Chavarrias, C., Shemesh, N., 2018. BOLD-fMRI in the mouse auditory pathway. Neuroimage 165, 265–277. Chen, J.E., Polimeni, J.R., Bollmann, S., Glover, G.H., 2019. On the analysis of rapidly sampled fMRI data. Neuroimage 188, 807–820. Cheung, M.M., Lau, C., Zhou, I.Y., Chan, K.C., Cheng, J.S., Zhang, J.W., Ho, L.C., Wu, E.X., 2012. BOLD fMRI investigation of the rat auditory pathway and tonotopic organization. Neuroimage 60, 1205–1211. Chuang, K.H., Lee, H.L., Li, Z., Chang, W.T., Nasrallah, F.A., Yeow, L.Y., Singh, K., 2019. Evaluation of nuisance removal for functional MRI of rodent brain. Neuroimage 188, 694–709. Desai, M., Kahn, I., Knoblich, U., Bernstein, J., Atallah, H., Yang, A., Kopell, N., Buckner, R.L., Graybiel, A.M., Moore, C.I., Boyden, E.S., 2011. Mapping brain networks in awake mice using combined optical neural control and fMRI. J. Neurophysiol. 105, 1393–1405. Gao, P.P., Zhang, J.W., Cheng, J.S., Zhou, I.Y., Wu, E.X., 2014. The inferior colliculus is involved in deviant sound detection as revealed by BOLD fMRI. Neuroimage 91, 220–227. Gao, P.P., Zhang, J.W., Fan, S.J., Sanes, D.H., Wu, E.X., 2015. Auditory midbrain processing is differentially modulated by auditory and visual cortices: an auditory fMRI study. Neuroimage 123, 22–32. Gao, Y.R., Ma, Y., Zhang, Q., Winder, A.T., Liang, Z., Antinori, L., Drew, P.J., Zhang, N., 2017. Time to wake up: studying neurovascular coupling and brain-wide circuit function in the un-anesthetized animal. Neuroimage 153, 382–398. Han, Z., Chen, W., Chen, X., Zhang, K., Tong, C., Zhang, X., Li, C.T., Liang, Z., 2019. Awake and behaving mouse fMRI during Go/No-Go task. Neuroimage 188, 733–742. Han, Z., Zhang, X., Zhu, J., Chen, Y., Li, C.T., 2018. High-throughput automatic training system for odor-based learned behaviors in head-fixed mice. Front. Neural Circuits 12, 15. Harris, A.P., Lennen, R.J., Marshall, I., Jansen, M.A., Pernet, C.R., Brydges, N.M., Duguid, I.C., Holmes, M.C., 2015. Imaging learned fear circuitry in awake mice using fMRI. Eur. J. Neurosci. 42, 2125–2134. Herman, P., Sanganahalli, B.G., Blumenfeld, H., Rothman, D.L., Hyder, F., 2013. Quantitative basis for neuroimaging of cortical laminae with calibrated functional MRI. Proc. Natl. Acad. Sci. U. S. A. 110, 15115–15120. Jung, W.B., Shim, H.J., Kim, S.G., 2019. Mouse BOLD fMRI at ultrahigh field detects somatosensory networks including thalamic nuclei. Neuroimage 195, 203–214. Kida, I., Xu, F., Shulman, R.G., Hyder, F., 2002. Mapping at glomerular resolution: fMRI of rat olfactory bulb. Magn. Reson. Med. 48, 570–576. Kim, T., Masamoto, K., Fukuda, M., Vazquez, A., Kim, S.G., 2010. Frequency-dependent neural activity, CBF, and BOLD fMRI to somatosensory stimuli in isofluraneanesthetized rats. Neuroimage 52, 224–233. King, J.A., Garelick, T.S., Brevard, M.E., Chen, W., Messenger, T.L., Duong, T.Q., Ferris, C.F., 2005. Procedure for minimizing stress for fMRI studies in conscious rats. J. Neurosci. Methods 148, 154–160. Lee, H.L., Li, Z., Coulson, E.J., Chuang, K.H., 2019. Ultrafast fMRI of the rodent brain using simultaneous multi-slice EPI. Neuroimage 195, 48–58. Li, B., Gong, L., Wu, R., Li, A., Xu, F., 2014. Complex relationship between BOLD-fMRI and electrophysiological signals in different olfactory bulb layers. Neuroimage 95, 29–38. Liang, Z., Ma, Y., Watson, G.D.R., Zhang, N., 2017. Simultaneous GCaMP6-based fiber photometry and fMRI in rats. J. Neurosci. Methods 289, 31–38.

12