Brainwide blood volume reflects opposing neural populations

Abstract

The supply of blood to brain tissue is thought to depend on the overall neural activity in that tissue, and this dependence is thought to differ across brain regions and across brain states. Studies supporting these views, however, measured neural activity as a bulk quantity, and related it to blood supply following disparate events in different regions. Here we measure fluctuations in neuronal activity and blood volume across the mouse brain, and find that their relationship is consistent across brain states and brain regions but differs in two opposing brainwide neural populations. Functional Ultrasound Imaging (fUSI) revealed that whisking, a marker of arousal, is associated with brainwide fluctuations in blood volume. Simultaneous fUSI and Neuropixels recordings showed that neurons that increase vs. decrease activity with whisking have distinct hemodynamic response functions. Their summed contributions predicted blood volume across states. Brainwide Neuropixels recordings revealed that these two opposing populations coexist in the entire brain. Their differing contributions to blood volume largely explain the apparent differences in blood volume fluctuations across regions. The mouse brain thus contains two neural populations with opposite relation to brain state and distinct relationships to blood supply, which together account for brainwide fluctuations in blood volume.

Figure 1.. Brainwide blood volume is strongly modulated by arousal events.

a. We measured blood volume with functional ultrasound imaging (fUSI), varying the coronal imaging planes (red outline) across sessions to image a volume of brain (black outline). Imaged volumes were aligned to the Allen Brain Atlas and voxels were averaged within brain regions (colours). b. Behavioural measurements — whisking and locomotion — from an example session. c. Corresponding fluctuations in blood volume in four example regions in isocortex (RSPagl), hippocampus (CA1), thalamus (ATN) and zona incerta (ZI, see Table S1 for abbreviations), showing predictions from whisking (black) with the filters in j. d. Average blood volume 0.9 s before the onset of whisking bouts in the example session, after subtracting the baseline (0.5–2 s before whisking onset). e. Same, for brief whisking bouts (1.3–3.5 s, without locomotion), 0.9 (left) and 3.0 s (right) after whisking onset. f. Same, for longer whisking bouts (> 3.5 s, without locomotion), 0.9 s after whisking onset (left), or offset (right). g. Average z-scored changes in blood volume across all sessions and mice, in each imaged region during brief whisking bouts (1.3–3.5 s) without locomotion. Arrows indicate the times of the images in e. Black dots below whisking indicate bout onset and average offset ± 1 s.d. Results are averaged first within mice and then across mice (number of sessions and events indicated in Figure S1). Asterisks denote significant increases (red) and decreases (blue) at 0.9 s and 3.0 s (permutation test). h. Same, for longer whisking bouts without locomotion (> 3.5 s, average = 12.1 s), aligned to bout onset (left) and offset (right). See Figure S4 for whisking bouts associated with locomotion. i. Fraction of regions with significant change in blood volume at each timepoint, determined using a session permutation test. Dotted lines indicate the timepoints used to measure significance in g–h. j. Best-fitting filters to predict blood volume from whisking, for all brain regions (Figure S2, Figure S3). These filters model changes in blood volume for a notional delta function (impulse) of whisking (top). Example predictions are shown in a (black).

Figure 2.. Bulk firing rate fails to predict changes in blood volume with whisking events.

a. While imaging blood volume with fUSI, we used Neuropixels probes to record from hundreds of neurons in visual cortex and hippocampus. b. Snippet from a typical recording session in visual cortex, with whisking (gold), blood volume (cyan) along with the normalized firing rate of all neurons recorded bilaterally across two probes in visual cortex. The firing rate for each neuron is divided by the 95^th percentile of firing rate across time. Neurons are sorted by depth on each probe. Bulk firing rate (grey) is obtained by taking the average over all neurons. c. The filtered bulk firing rate (grey) to predict blood volume (cyan) throughout the session. d. The best-fitting filter to predict blood volume from bulk firing rate. e. We identified whisking bouts and looked at brief (left) and longer whisking bouts, aligned to onset (middle) or offset (right). f. Average changes in bulk firing rate during whisking events. g. Average changes in blood volume during whisking events (cyan), and prediction from bulk firing rate in visual cortex (grey). d–g show results averaged across all recording sessions in visual cortex (see Figure S8 for similar results in hippocampus). Shaded area shows the s.e. across sessions.

Figure 3.. The firing rate of two arousal−defined neural populations best predicts blood volume.

a. We analysed the same simultaneous Neuropixels-fUSI recordings as in Figure 2. b. Top: Same snippet of recording as in Figure 2, with neurons sorted by correlation with whisking. The firing rate for each neuron is divided by the 95^th percentile of firing rate across time. Bars indicate Arousal+ neurons (red, correlation with whisking > 0.05), and Arousal− neurons (blue, correlation with whisking < −0.05). Bottom: Average firing rate for Arousal+ neurons (red), and Arousal− neurons (blue). The firing rate of each neuron is z-scored before averaging. c. Firing rate of all neurons recorded in visual cortex across all sessions, aligned to and averaged across whisking events. The traces (top) show the corresponding average whisking (gold), and bulk firing rate (black). Each row is z-scored. Neurons are sorted by their correlation with whisking. d. Same experiment as b, showing the simultaneously measured blood volume (cyan), and the combined prediction from the firing of Arousal+ and Arousal− neurons (purple). e. The best-fitting filters to predict blood volume from Arousal+ (red) and Arousal− (blue) neurons. f. Traces of whisking for brief (left) and longer whisking events, aligned to onset (middle) or offset (right). See Figure S7 for the briefest whisking bouts. g. Average changes in firing rate for Arousal+ neurons during whisking events. h. Same, for Arousal− neurons. i. Average change in blood volume during whisking events (cyan), and prediction from Arousal+ and Arousal− neurons in visual cortex (purple). Data in e–i is averaged across all recording sessions in visual cortex (see Figure S8 for results in hippocampus). j. Coherence between blood volume and the predictions from bulk firing rate (grey, as in Figure 2) or from Arousal+ and Arousal− neurons (purple), in visual cortex. k. Same, for recordings in hippocampus. In e-k, shaded area shows the s.e. across sessions.

Figure 4.. The firing rate of the two populations, but not the bulk firing rate, accounts for blood volume across states.

a. Snippet from a simultaneous Neuropixels-fUSI recording session in visual cortex, as in Figure 3b. a. Normalized firing rate of all neurons recorded in visual cortex sorted by correlation with whisking. The firing rate for each neuron is divided by the 95^th percentile of its firing rate across time. Bars indicate Arousal+ neurons (red, correlation with whisking > 0.05), and Arousal− neurons (blue, correlation with whisking < −0.05). b. We use a normalized version of pupil size for each recording session to define an arousal index, from which we derive four states: putative rapid eye-movement sleep (REM), putative non-rapid-eye-movement sleep (NREM), quiet wake, and active wake. This snippet shows an episode of putative REM, with very constricted pupil, presence of eye movements, downcast and nasal eye position, increased activity of both Arousal+ and Arousal− neurons, as well as elevated blood volume. c. Behavioural metrics as a function of arousal index, defined as a scaled version of pupil size in 3 s intervals. 0 corresponds to the transition to putative REM, with increased eye movements, and characteristic eye positions. 1 corresponds to the transition to active wake, with increased whisking and eye movements. NREM and quiet wake are defined as transition states between these extremes. d. Firing rate of Arousal+ (red) and Arousal− (blue) neurons as a function of arousal index. e. Actual blood volume (cyan) and predictions from bulk (grey) and combined (purple) models, as a function of arousal index. f. Cross-correlation between bulk firing rate and actual blood volume, blood volume predicted from the bulk or combined model, for each state. Dots indicate the centre of mass of the cross-correlations for each state. g. Prediction accuracy, computed as the Pearson correlation between actual and predicted blood volume, depending on state. State is computed based on arousal index in 10 s intervals. Asterisks indicate significant differences between combined and bulk prediction (paired t-test, p < 0.05, n = 10 recording sessions).

Figure 5.. The two arousal-defined neural populations are present throughout the brain.

a. Example Neuropixels recording with neurons in visual cortex (VIS, top) and lateral group of the dorsal thalamus (LAT, bottom) from Ref. , showing the firing rate of all recorded neurons for each region, ordered based on their correlation with whisking (gold). The firing rate for each neuron is divided by the 95^th percentile of firing rate across time. Bars indicate Arousal+ neurons (red, correlation with whisking > 0.05), and Arousal− neurons (blue, correlation with whisking < −0.05). Bottom: Average firing rate for all (grey), Arousal+ (red) and Arousal− neurons (blue). The firing rate of each neuron is z-scored before averaging. Arousal+ and Arousal− neurons are identified on half of the data, and activity is shown for a snippet of the other half. b. A map with all the Neuropixels probe insertions that we included for analysis (18,791 neurons across the brain). c. Comparison of the fraction of Arousal+ and Arousal− neurons across regions. Arrows indicate the example regions shown in a. d. Left: Proportion of Arousal+ neurons across regions. Right: Average firing rate during whisking events for Arousal+ neurons within each region. Activity is aligned to the onset of brief bouts (left), the onset (middle) or the offset (right) of long bouts. The average activity of Arousal+ neurons across regions peaked at the time of whisking onset (0 s offset, dashed line). e. Same, for Arousal− neurons. The average activity of Arousal− neurons across regions hit its lowest mark after whisking onset (dashed line).

Figure 6.. The two arousal-based populations better predict brainwide blood volume around arousal events.

a. Bulk prediction of whisking-evoked changes in blood volume, obtained by convolving whisking-related bulk firing rate (as shown in Figure S12) with the filter from Figure 2e, averaged between visual cortex and hippocampus (Figure S8). b. Combined prediction of whisking-related blood volume from Arousal+ neurons and Arousal− neurons, obtained by convolving the firing rate of the two populations (as shown in Figure 5d, e) with their respective filters from Figure 3e, averaged between visual cortex and hippocampus (Figure S8) and summing their contributions scaled by the proportion of Arousal+ and Arousal− neurons in each region. c. Mean squared error (MSE) between actual and predicted whisking-related blood volume for the bulk prediction, plotted against the MSE for the combined prediction. The MSE is computed across time for the brief and longer whisking bouts (0–4.5 s window). Each dot is a brain region, accompanied by its Allen acronym. d. Average blood volume (from the brainwide fUSI recordings) 1.2–1.8 s after whisking onset plotted against the relative bias to Arousal+ vs. Arousal− neurons (computed as the difference in the number of Arousal+ and Arousal− neurons over their sum (from the brainwide Neuropixels recordings). Each dot is a brain region, accompanied by its Allen acronym. The linear fit (black line) highlights the strong Pearson correlation (r = 0.63).