Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 78 of the 78 references it cites were found in a registry or an index; 42 of the 68 sources its claims cite were checked for support.

Energy paradox in REM sleep: balancing supply and consumption in brain metabolism

Open access
Takahashi, Yusuke, Ikoma, Yoko, Matsui, Ko · 2026 · Communications Biology

Abstract

The brain’s capacity for information processing depends on precisely regulated energy dynamics. Yet how metabolic supply adapts to shifting computational demands across brain states remains unclear. Using wide-field fluorescence imaging through the intact skull of live mice, we simultaneously monitored brain blood volume (BBV), astrocytic pyruvate, and neuronal ATP levels during natural sleep. We found that large-scale metabolic dynamics are coupled to neuronal activity but reorganize in a state-dependent manner. During non-rapid eye movement (NREM) sleep, theta-band electrocorticogram (ECoG) activity predicted subsequent blood volume changes, accompanied by rapid anterior-to-posterior vascular waves. In contrast, REM sleep was marked by a pronounced increase in BBV, originating in the posterior cortex and slowly propagating across the brain. This was accompanied by elevated astrocytic pyruvate; paradoxically, however, neuronal ATP levels declined sharply. These findings reveal a dynamic interplay among neurons, astrocytes, and the vasculature, suggesting that distinct energy-allocation strategies underlie the brain’s computational flexibility.

Overview

CoChat Summary

## TL;DR

Using intact-skull wide-field imaging in live mice, the paper shows that brain blood volume, astrocytic pyruvate, and neuronal ATP are tightly coupled to sleep state but reorganize differently in NREM versus REM sleep. REM sleep is marked by a posterior-originating blood-volume surge and increased astrocytic pyruvate, yet neuronal ATP drops sharply.

## Key findings

  • In NREM sleep, theta-band ECoG activity predicted BBV changes about 4–5 s later; the optimized FIR-BRF peaked at 4–5 s and was estimated from 13 NREM episodes across 3 mice.
  • During REM sleep, BBV showed a pronounced increase that began about 50 s before REM onset and propagated from the posterior cortex/RSC to anterior regions about 15 s later.
  • Fast BBV fluctuations (0.05–0.2 Hz) were more spatially synchronized in NREM than REM: using the same clustering threshold, the cortex partitioned into 5 clusters in NREM versus 7 clusters in REM.
  • In REM sleep, astrocytic pyruvate increased as expected, but neuronal ATP decreased sharply; after pharmacological vasodilation with SNP, both astrocytic pyruvate and neuronal ATP increased.

## Methods

  • Design/sample: Within-animal study in 15 male mice older than 15 weeks, using Thy1-ATeam mice for neuronal ATP and Mlc1-tTA::tetO-PYRS mice for astrocytic pyruvate; the same animals were tracked across NREM, REM, and wake and before/after SNP.
  • Measurements: Wide-field macro-zoom fluorescence imaging through the intact skull, with ECoG and neck EMG for sleep staging, plus fluorescent BBV proxy imaging using Alb-mScarlet and dYFP.
  • Analysis: Cross-correlation, FIR-BRF estimation, hierarchical clustering, lagged cross-correlation/temporal delay mapping, seqNMF motif extraction, and paired/ANOVA-based statistics in Python.

## Limitations

  • The paper states that the mechanism linking theta activity to BBV is unclear, and that the contribution of neuromodulators such as noradrenaline and acetylcholine was not directly measured.
  • The authors note that the REM-associated metabolite correction relies on a linear relationship fit during NREM, so applying it to REM involves extrapolation.
  • Evident from the text: the study is correlational for most sleep-state findings, so it cannot by itself establish causality between BBV changes, astrocytic pyruvate, and neuronal ATP.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s42003-026-10646-6

Cites

78 of 78 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Theta-band power is tightly coupled with brain blood volume changes. a Wide-field fluorescence images were acquired through the intact skull using a macro-zoom microscope in head-fixed mice. Shown is an example of a whole-cortex fluorescence image from a Thy1-ATeam mouse. Typically, only the left hemisphere was imaged, as the implanted screw electrode obstructed a part of the contralateral field of view. b Fluorescence images of albumin-mScarlet expressed in blood plasma and YFP expressed in neurons. Blood vessels appear as dark regions in the direct YFP (dYFP) image. c REM sleep was characterized by prominent theta oscillations (6–9 Hz) in the ECoG and low EMG amplitude. During REM sleep, mScarlet fluorescence intensity increased while dYFP intensity decreased. d Scatter plots demonstrating the negative correlation between mScarlet and dYFP fluorescence intensities during NREM, REM, WAKE, and all combined states (CC, Pearson correlation coefficient). e Summary of the correlation coefficients (14 episodes from n = 3 mice; error bars, SEM). The robust negative correlation indicates that the dYFP signal provides a reliable proxy for inverted BBV changes. f Theta-band (6–9 Hz) power of ECoG signal correlated with the inverted dYFP signal. Top to bottom: ECoG power spectrogram, z-scored theta-band power, z-scored idYFP. g The theta-band power of ECoG was convolved with a finite-impulse-response brain-blood-volume response function BRF (FIR-BRF) and compared with idYFP. The FIR-BRF that minimized the mean-squared error between the two waveforms was calculated. h Scatterplot of idYFP versus theta-band power of ECoG from one NREM sleep episode (left). Cross-correlation functions were calculated across episodes and averaged (right, mean ± SEM, 13 episodes from n = 3 mice). The coefficient at zero-time lag (CC) is marked in red; the peak correlation with temporal delay is marked in black (XCTD). i Scatterplot of idYFP versus theta-band power convolved with FIR-BRF (left). The average cross-correlation function (right) shows a higher peak at zero-time lag compared with the raw theta-band power. j Summary of correlation coefficients between idYFP and theta-band signals (13 episodes from n = 3 mice; error bars, SEM). The FIR-BRF-convolved theta-band power showed the strongest correlation with idYFP. k Time delay at which the cross-correlation between the raw theta-band power and idYFP reached its maximum (13 episodes from n = 3 mice; error bars, SEM). Negative values indicate that theta-band power led idYFP. l, Peak latency of the calculated FIR-BRF (13 episodes from n = 3 mice; error bars, SEM).
Fig. 1 Theta-band power is tightly coupled with brain blood volume changes. a Wide-field fluorescence images were acquired through the intact skull using a macro-zoom microscope in head-fixed mice. Shown is an example of a whole-cortex fluorescence image from a Thy1-ATeam mouse. Typically, only the left hemisphere was imaged, as the implanted screw electrode obstructed a part of the contralateral field of view. b Fluorescence images of albumin-mScarlet expressed in blood plasma and YFP expressed in neurons. Blood vessels appear as dark regions in the direct YFP (dYFP) image. c REM sleep was characterized by prominent theta oscillations (6–9 Hz) in the ECoG and low EMG amplitude. During REM sleep, mScarlet fluorescence intensity increased while dYFP intensity decreased. d Scatter plots demonstrating the negative correlation between mScarlet and dYFP fluorescence intensities during NREM, REM, WAKE, and all combined states (CC, Pearson correlation coefficient). e Summary of the correlation coefficients (14 episodes from n = 3 mice; error bars, SEM). The robust negative correlation indicates that the dYFP signal provides a reliable proxy for inverted BBV changes. f Theta-band (6–9 Hz) power of ECoG signal correlated with the inverted dYFP signal. Top to bottom: ECoG power spectrogram, z-scored theta-band power, z-scored idYFP. g The theta-band power of ECoG was convolved with a finite-impulse-response brain-blood-volume response function BRF (FIR-BRF) and compared with idYFP. The FIR-BRF that minimized the mean-squared error between the two waveforms was calculated. h Scatterplot of idYFP versus theta-band power of ECoG from one NREM sleep episode (left). Cross-correlation functions were calculated across episodes and averaged (right, mean ± SEM, 13 episodes from n = 3 mice). The coefficient at zero-time lag (CC) is marked in red; the peak correlation with temporal delay is marked in black (XCTD). i Scatterplot of idYFP versus theta-band power convolved with FIR-BRF (left). The average cross-correlation function (right) shows a higher peak at zero-time lag compared with the raw theta-band power. j Summary of correlation coefficients between idYFP and theta-band signals (13 episodes from n = 3 mice; error bars, SEM). The FIR-BRF-convolved theta-band power showed the strongest correlation with idYFP. k Time delay at which the cross-correlation between the raw theta-band power and idYFP reached its maximum (13 episodes from n = 3 mice; error bars, SEM). Negative values indicate that theta-band power led idYFP. l, Peak latency of the calculated FIR-BRF (13 episodes from n = 3 mice; error bars, SEM).CC BY · PubMed Central
Fig. 2 Brain blood volume (BBV) fluctuations globally synchronize during REM sleep. a dYFP fluorescence waveforms from three tiles in the left cortical hemisphere (locations shown in f). During REM sleep (blue shading), a prominent decrease in dYFP signal, corresponding to increased BBV, was synchronously observed across the cortex. b Scatterplots comparing dYFP signals between tile pairs. The spatially closer pair (tiles 2 and 3) showed higher correlations than the distant pair (1 and 3), with correlation strengthening during REM compared to NREM sleep. c Correlation matrix of dYFP signals across all tiles during a representative NREM sleep episode. Each row shows correlations between one tile and all others. The raw dYFP waveforms of tiles 1–3 are shown in (a), and their spatial location is shown in (f). Rows were reordered by hierarchical clustering, with the dendrogram (right) indicating cluster hierarchy. Applying a clustering threshold divided the cortex into 19 regions (f, left). d Representative correlation matrix during REM sleep. Using the same threshold, fewer (5 regions) but larger clusters emerged, reflecting widespread synchrony of dYFP signals across tiles (f, right). e Relationship between clustering distance (x-axis) and the number of clusters (y-axis). For any distance threshold, fewer clusters were detected during REM sleep (blue) than NREM sleep (black), indicating stronger synchrony of BBV fluctuations during REM sleep. In this example, clustering into 5 clusters corresponded to distances of 14 (NREM) and 3 (REM). f, Spatial clustering of cortical tiles during NREM and REM sleep. g Across episodes, the cluster distance yielding 5 clusters was smaller during REM sleep than NREM sleep (p = 0.0017, Cohen’s d = 1.28,18 episodes from n = 3 mice, paired t-test; **p < 0.01; error bars, SEM). h Forcing division of the cortex into two clusters consistently revealed an anterior cluster (blue; motor and somatosensory areas) and a posterior cluster (red; occipital and medial cortex) across animals (n = 3) in both NREM and REM sleep.
Fig. 2 Brain blood volume (BBV) fluctuations globally synchronize during REM sleep. a dYFP fluorescence waveforms from three tiles in the left cortical hemisphere (locations shown in f). During REM sleep (blue shading), a prominent decrease in dYFP signal, corresponding to increased BBV, was synchronously observed across the cortex. b Scatterplots comparing dYFP signals between tile pairs. The spatially closer pair (tiles 2 and 3) showed higher correlations than the distant pair (1 and 3), with correlation strengthening during REM compared to NREM sleep. c Correlation matrix of dYFP signals across all tiles during a representative NREM sleep episode. Each row shows correlations between one tile and all others. The raw dYFP waveforms of tiles 1–3 are shown in (a), and their spatial location is shown in (f). Rows were reordered by hierarchical clustering, with the dendrogram (right) indicating cluster hierarchy. Applying a clustering threshold divided the cortex into 19 regions (f, left). d Representative correlation matrix during REM sleep. Using the same threshold, fewer (5 regions) but larger clusters emerged, reflecting widespread synchrony of dYFP signals across tiles (f, right). e Relationship between clustering distance (x-axis) and the number of clusters (y-axis). For any distance threshold, fewer clusters were detected during REM sleep (blue) than NREM sleep (black), indicating stronger synchrony of BBV fluctuations during REM sleep. In this example, clustering into 5 clusters corresponded to distances of 14 (NREM) and 3 (REM). f, Spatial clustering of cortical tiles during NREM and REM sleep. g Across episodes, the cluster distance yielding 5 clusters was smaller during REM sleep than NREM sleep (p = 0.0017, Cohen’s d = 1.28,18 episodes from n = 3 mice, paired t-test; **p < 0.01; error bars, SEM). h Forcing division of the cortex into two clusters consistently revealed an anterior cluster (blue; motor and somatosensory areas) and a posterior cluster (red; occipital and medial cortex) across animals (n = 3) in both NREM and REM sleep.CC BY · PubMed Central
Fig. 3 Slow and fast components of cortical blood volume dynamics across sleep stages. a Representative snapshots of dYFP fluorescence during the transition from NREM to REM sleep. Anterior and posterior regions of interest (ROIs) are defined in Fig. 5i. Bottom traces: raw dYFP signals from anterior (orange) and posterior (blue) ROIs. At the NREM-to-pre-REM transition, dYFP signals decreased nearly simultaneously in both ROIs, whereas at the pre-REM to REM transition, the posterior signal decreased ∼15 s earlier than the anterior. b Schematic of time delay (Δt) quantification. After low-pass filtering and normalization of dYFP signals, Δt was defined as the difference in half-decay times between anterior and posterior signals. c The posterior dYFP signal reached its half-decay point ∼15 s earlier than the anterior (15 episodes from n = 3 mice; error bars, SEM). d Representative Δt map showing propagation of the dYFP signal decrease (reflecting BBV increase) across the cortex. The dYFP signal decrease originated in posterior regions and propagated anteriorly over ∼20 s. e Top trace: dYFP fluorescence signal recorded from the primary motor cortex (MOp). Bottom traces: fast components (0.05–0.2 Hz band-pass filtered dYFP signals) from MOp, sensorimotor barrel field (SSb), and primary visual cortex (V1) (locations indicated by red tiles in f). The amplitude of the fast component was markedly reduced during REM sleep in MOp and V1. f Quantitative analysis of fast-component amplitude (ΔF/F, %). Top: spatial maps of amplitude during NREM, REM, and the REM/NREM amplitude ratio. Amplitude was globally reduced during REM sleep but relatively preserved in SSb. Bottom: comparison of amplitudes during NREM (gray) and REM (blue) for the whole cortex (Whole), retrosplenial cortex (RSC), and SSb. ROI boundaries are outlined in red in (b). Amplitude was reduced in all regions during REM sleep (Whole, p = 1.9 × 10–5, Cohen’s d = 1.95; RSC, p = 1.7 × 10–6, Cohen’s d = 2.34; SSb, p = 0.0017, Cohen’s d = 1.29; 18 episodes from n = 3 mice, paired t-test; **p < 0.01, ***p < 0.001; error bars, SEM).
Fig. 3 Slow and fast components of cortical blood volume dynamics across sleep stages. a Representative snapshots of dYFP fluorescence during the transition from NREM to REM sleep. Anterior and posterior regions of interest (ROIs) are defined in Fig. 5i. Bottom traces: raw dYFP signals from anterior (orange) and posterior (blue) ROIs. At the NREM-to-pre-REM transition, dYFP signals decreased nearly simultaneously in both ROIs, whereas at the pre-REM to REM transition, the posterior signal decreased ∼15 s earlier than the anterior. b Schematic of time delay (Δt) quantification. After low-pass filtering and normalization of dYFP signals, Δt was defined as the difference in half-decay times between anterior and posterior signals. c The posterior dYFP signal reached its half-decay point ∼15 s earlier than the anterior (15 episodes from n = 3 mice; error bars, SEM). d Representative Δt map showing propagation of the dYFP signal decrease (reflecting BBV increase) across the cortex. The dYFP signal decrease originated in posterior regions and propagated anteriorly over ∼20 s. e Top trace: dYFP fluorescence signal recorded from the primary motor cortex (MOp). Bottom traces: fast components (0.05–0.2 Hz band-pass filtered dYFP signals) from MOp, sensorimotor barrel field (SSb), and primary visual cortex (V1) (locations indicated by red tiles in f). The amplitude of the fast component was markedly reduced during REM sleep in MOp and V1. f Quantitative analysis of fast-component amplitude (ΔF/F, %). Top: spatial maps of amplitude during NREM, REM, and the REM/NREM amplitude ratio. Amplitude was globally reduced during REM sleep but relatively preserved in SSb. Bottom: comparison of amplitudes during NREM (gray) and REM (blue) for the whole cortex (Whole), retrosplenial cortex (RSC), and SSb. ROI boundaries are outlined in red in (b). Amplitude was reduced in all regions during REM sleep (Whole, p = 1.9 × 10–5, Cohen’s d = 1.95; RSC, p = 1.7 × 10–6, Cohen’s d = 2.34; SSb, p = 0.0017, Cohen’s d = 1.29; 18 episodes from n = 3 mice, paired t-test; **p < 0.01, ***p < 0.001; error bars, SEM).CC BY · PubMed Central
Fig. 4 Distinct spatiotemporal dynamics of the fast component of the dYFP signal during NREM and REM sleep. a Example of lagged cross-correlation analysis. Band-pass filtered (0.05–0.2 Hz) dYFP signals from MOp and V1 were compared. The peak of the cross-correlation function defines the temporally delayed cross-correlation (XCTD), and the corresponding time lag defines the temporal delay (TD). Examples are shown for NREM (left) and REM (right). b XCTD matrix of fast-component dYFP signals across cortical regions during a representative NREM sleep episode. Rows in the matrix were ordered by hierarchical clustering, and a distance threshold (gray line) yielded 5 clusters (right map). Seed regions are marked with triangles and remapped in (f). c XCTD matrix and clustering during REM sleep from the same recording. Using the same threshold resulted in 7 smaller clusters, indicating reduced long-range synchronization of BBV fluctuations compared to NREM sleep. d Relationship between cluster distance (x-axis) and number of clusters (y-axis). For any threshold, fewer clusters were detected during NREM sleep (black) than REM sleep (blue), indicating greater synchrony of the fast BBV fluctuations in NREM sleep. e Across episodes, the distance yielding 5 clusters was smaller in NREM sleep than REM sleep (p = 0.0048, Cohen’s d = 0.95,18 episodes from n = 3 mice, paired t-test; **p < 0.01; error bars, SEM). f XCTD maps during NREM sleep seeded from MOp, SSb, RSC, and V1 (top). Bottom: cortical areas with XCTD > 0.8 and the percentage of total cortical area indicated below. g Corresponding XCTD maps during REM sleep. h Comparison of cortical areas with XCTD > 0.8. Highly correlated areas were significantly larger during NREM sleep than REM sleep for MOp, SSb, and RSC, but not for V1 (MOp, p = 0.015, Cohen’s d = 1.06; SSb, p = 0.00036, Cohen’s d = 1.89; RSC, p = 0.042, Cohen’s d = 0.74; V1, p = 0.33, Cohen’s d = 0.36; 15 episodes from n = 3 mice, paired t-test; *p < 0.05, **p < 0.01, ***p < 0.001; error bars, SEM). i, Representative fast-component traces from anterior (A) and posterior (P) cortex (top). Sliding-window cross-correlation (50 s window) shows dynamic temporal relationships. The heatmap represents the cross-correlation function at each time point, and the black line tracks the time lag of the peak cross-correlation. j Additional examples of anterior–posterior temporal delay sequences. k Average time lag between anterior and posterior activity. During NREM sleep, anterior signals consistently preceded posterior signals, whereas REM sleep showed variable patterns (NREM, p = 4.5 × 10–7, Cohen’s d = 1.85; REM, p = 0.32, Cohen’s d = 0.24; 18 episodes from n = 3 mice, one-sample t-test, ***p < 0.001; error bars, SEM).
Fig. 4 Distinct spatiotemporal dynamics of the fast component of the dYFP signal during NREM and REM sleep. a Example of lagged cross-correlation analysis. Band-pass filtered (0.05–0.2 Hz) dYFP signals from MOp and V1 were compared. The peak of the cross-correlation function defines the temporally delayed cross-correlation (XCTD), and the corresponding time lag defines the temporal delay (TD). Examples are shown for NREM (left) and REM (right). b XCTD matrix of fast-component dYFP signals across cortical regions during a representative NREM sleep episode. Rows in the matrix were ordered by hierarchical clustering, and a distance threshold (gray line) yielded 5 clusters (right map). Seed regions are marked with triangles and remapped in (f). c XCTD matrix and clustering during REM sleep from the same recording. Using the same threshold resulted in 7 smaller clusters, indicating reduced long-range synchronization of BBV fluctuations compared to NREM sleep. d Relationship between cluster distance (x-axis) and number of clusters (y-axis). For any threshold, fewer clusters were detected during NREM sleep (black) than REM sleep (blue), indicating greater synchrony of the fast BBV fluctuations in NREM sleep. e Across episodes, the distance yielding 5 clusters was smaller in NREM sleep than REM sleep (p = 0.0048, Cohen’s d = 0.95,18 episodes from n = 3 mice, paired t-test; **p < 0.01; error bars, SEM). f XCTD maps during NREM sleep seeded from MOp, SSb, RSC, and V1 (top). Bottom: cortical areas with XCTD > 0.8 and the percentage of total cortical area indicated below. g Corresponding XCTD maps during REM sleep. h Comparison of cortical areas with XCTD > 0.8. Highly correlated areas were significantly larger during NREM sleep than REM sleep for MOp, SSb, and RSC, but not for V1 (MOp, p = 0.015, Cohen’s d = 1.06; SSb, p = 0.00036, Cohen’s d = 1.89; RSC, p = 0.042, Cohen’s d = 0.74; V1, p = 0.33, Cohen’s d = 0.36; 15 episodes from n = 3 mice, paired t-test; *p < 0.05, **p < 0.01, ***p < 0.001; error bars, SEM). i, Representative fast-component traces from anterior (A) and posterior (P) cortex (top). Sliding-window cross-correlation (50 s window) shows dynamic temporal relationships. The heatmap represents the cross-correlation function at each time point, and the black line tracks the time lag of the peak cross-correlation. j Additional examples of anterior–posterior temporal delay sequences. k Average time lag between anterior and posterior activity. During NREM sleep, anterior signals consistently preceded posterior signals, whereas REM sleep showed variable patterns (NREM, p = 4.5 × 10–7, Cohen’s d = 1.85; REM, p = 0.32, Cohen’s d = 0.24; 18 episodes from n = 3 mice, one-sample t-test, ***p < 0.001; error bars, SEM).CC BY · PubMed Central
Fig. 5 Extraction of spatiotemporal motifs of brain blood volume (BBV) fluctuations. a Preprocessing pipeline for unsupervised motif extraction using seqNMF. Time-series image sequences including NREM, REM, and WAKE episodes were collected. dYFP signals from cortical tiles were band-pass filtered (0.05–0.2 Hz) along the time axis (top), and the upper envelope was obtained with a Hilbert transform (pink, middle). Each frame was then min-max normalized, and tiles were vectorized into a time-series matrix (tiles × frames). b Using seqNMF, the time-series matrix was decomposed into spatiotemporal motifs (left, 6 s components) and corresponding temporal weights (right). Motifs highlighted in red (3, 6, 8, 10) are shown in detail in (c). c Four representative motifs (left) with corresponding temporal weights (right). Active areas in motifs 6, 10, 8, and 3 corresponded mainly to RSC, SSb, AC, and MOp, respectively. Their temporal occurrence patterns were distinct: motif 6 (NREM-high), motif 10 (REM-high), motif 8 (REM-low), and motif 3 (non-specific). Notably, motif 6 activity in the RSC gradually increased during NREM and peaked just before the transition into REM (see Supplementary Fig. 4).
Fig. 5 Extraction of spatiotemporal motifs of brain blood volume (BBV) fluctuations. a Preprocessing pipeline for unsupervised motif extraction using seqNMF. Time-series image sequences including NREM, REM, and WAKE episodes were collected. dYFP signals from cortical tiles were band-pass filtered (0.05–0.2 Hz) along the time axis (top), and the upper envelope was obtained with a Hilbert transform (pink, middle). Each frame was then min-max normalized, and tiles were vectorized into a time-series matrix (tiles × frames). b Using seqNMF, the time-series matrix was decomposed into spatiotemporal motifs (left, 6 s components) and corresponding temporal weights (right). Motifs highlighted in red (3, 6, 8, 10) are shown in detail in (c). c Four representative motifs (left) with corresponding temporal weights (right). Active areas in motifs 6, 10, 8, and 3 corresponded mainly to RSC, SSb, AC, and MOp, respectively. Their temporal occurrence patterns were distinct: motif 6 (NREM-high), motif 10 (REM-high), motif 8 (REM-low), and motif 3 (non-specific). Notably, motif 6 activity in the RSC gradually increased during NREM and peaked just before the transition into REM (see Supplementary Fig. 4).CC BY · PubMed Central
Fig. 6 SNP-induced vasodilation increases astrocytic pyruvate and neuronal ATP. a A FRET-based cytosolic pyruvate sensor (PYRS) was expressed in astrocytes (Mlc1-tTA::tetO-PYRS mice). The direct YFP signal (dYFP, red; ∼510 nm excitation) served as an indicator of brain blood volume (BBV). The FRET signal (fYFP, green; ∼430 nm excitation) decreases with rising pyruvate. Both dYFP and fYFP decreased after SNP injection (gray bar, 60 s), reflecting increased BBV, but the larger decrease in fYFP indicated elevated astrocytic pyruvate signal estimated by—(fYFP – dYFP). b A FRET-based cytosolic ATP sensor (ATeam) was expressed in neurons (Thy1-ATeam mice). Here, FRET efficiency increases with ATP binding, and neuronal ATP signal was estimated by fYFP – dYFP. After SNP injection, the smaller decrease in fYFP relative to dYFP indicated increased neuronal ATP signal. c Trajectory of astrocytic pyruvate versus inverted dYFP (idYFP). Pyruvate increased in BBV but reached saturation ∼3 min after SNP injection. d Trajectory of neuronal ATP versus idYFP. ATP increased with BBV, but unlike pyruvate, showed no saturation. Averaged pyruvate (e) and ATP (f) signals across trials (n = 9 trials in 3 animals, respectively). Analysis windows: Post1 (1–2 min) and Post2 (4–5 min) after SNP injection. g Pyruvate signal increased significantly in both Post1 and Post2 (one-sample t-test with Holm correction; Post1, p = 0.00071, Cohen’s d = 1.74; Post2, p = 0.0029, Cohen’s d = 1.41; error bars, SEM). h ATP signal increased significantly in both periods (Post1, p = 0.028, Cohen’s d = 0.89; Post2, p = 0.00020, Cohen’s d = 2.14), with a stronger rise in Post2. The Post1/Post2 ratio differed significantly between pyruvate and ATP signals (one-sample t-test, p = 0.0047; error bars, SEM). Two-way mixed-design ANOVA revealed a significant interaction between signal type and time point, with post-hoc tests showing significance only for ATP between Post1 and Post2 (ATP Post1 vs Post2, p = 0.00019; Pyr Post1 vs Post2, p = 0.099; Post1 ATP vs Pyr, p = 0.13; Post2 ATP vs Pyr, p = 0.13).
Fig. 6 SNP-induced vasodilation increases astrocytic pyruvate and neuronal ATP. a A FRET-based cytosolic pyruvate sensor (PYRS) was expressed in astrocytes (Mlc1-tTA::tetO-PYRS mice). The direct YFP signal (dYFP, red; ∼510 nm excitation) served as an indicator of brain blood volume (BBV). The FRET signal (fYFP, green; ∼430 nm excitation) decreases with rising pyruvate. Both dYFP and fYFP decreased after SNP injection (gray bar, 60 s), reflecting increased BBV, but the larger decrease in fYFP indicated elevated astrocytic pyruvate signal estimated by—(fYFP – dYFP). b A FRET-based cytosolic ATP sensor (ATeam) was expressed in neurons (Thy1-ATeam mice). Here, FRET efficiency increases with ATP binding, and neuronal ATP signal was estimated by fYFP – dYFP. After SNP injection, the smaller decrease in fYFP relative to dYFP indicated increased neuronal ATP signal. c Trajectory of astrocytic pyruvate versus inverted dYFP (idYFP). Pyruvate increased in BBV but reached saturation ∼3 min after SNP injection. d Trajectory of neuronal ATP versus idYFP. ATP increased with BBV, but unlike pyruvate, showed no saturation. Averaged pyruvate (e) and ATP (f) signals across trials (n = 9 trials in 3 animals, respectively). Analysis windows: Post1 (1–2 min) and Post2 (4–5 min) after SNP injection. g Pyruvate signal increased significantly in both Post1 and Post2 (one-sample t-test with Holm correction; Post1, p = 0.00071, Cohen’s d = 1.74; Post2, p = 0.0029, Cohen’s d = 1.41; error bars, SEM). h ATP signal increased significantly in both periods (Post1, p = 0.028, Cohen’s d = 0.89; Post2, p = 0.00020, Cohen’s d = 2.14), with a stronger rise in Post2. The Post1/Post2 ratio differed significantly between pyruvate and ATP signals (one-sample t-test, p = 0.0047; error bars, SEM). Two-way mixed-design ANOVA revealed a significant interaction between signal type and time point, with post-hoc tests showing significance only for ATP between Post1 and Post2 (ATP Post1 vs Post2, p = 0.00019; Pyr Post1 vs Post2, p = 0.099; Post1 ATP vs Pyr, p = 0.13; Post2 ATP vs Pyr, p = 0.13).CC BY · PubMed Central
Fig. 7 Theta-band power of the ECoG and dYFP changes precede the onset of REM sleep. Representative recordings from Mlc1-tTA::tetO-PYRS (a) and Thy1-ATeam (b) mice. From top to bottom: ECoG spectrogram, theta/delta power ratio, delta-band power (0.5–4 Hz), theta-band power (6–9 Hz), and optical signals (dYFP, fYFP, and pyruvate or ATP) in RSC. Arrowheads indicate automatically detected change points (purple, theta-band; red, dYFP; black, pyruvate/ATP). c Timing of changes in theta-band power of ECoG, dYFP, and astrocytic pyruvate signals relative to REM sleep onset (13 episodes from n = 3 mice). Increases in theta-band power and decreases in dYFP occurred significantly earlier than REM sleep onset (one-sample t-test with Holm correction, theta-band power, p = 3.0 × 10–8, Cohen’s d = 2.99; dYFP, p = 7.9 × 10–7, Cohen’s d = 2.12), whereas pyruvate did not (p = 0.101, Cohen’s d = 0.47). Latencies were similar for theta-band power and dYFP changes, but both preceded pyruvate signal change (one-way ANOVA with Tukey’s post-hoc; theta vs dYFP, p = 0.896; dYFP vs pyruvate, p = 4.1 × 10–5; theta vs pyruvate, p = 9.0 × 10–6; error bars, SEM). d Corresponding analysis for theta-band power, dYFP, and ATP signals (14 episodes from n = 3 mice). Theta-band power and dYFP again preceded REM sleep onset, whereas ATP signal did not (one-sample t-test with Holm correction, theta-band power, p = 4.5 × 10–4, Cohen’s d = 1.35; dYFP, p = 5.0 × 10–5, Cohen’s d = 1.76; ATP, p = 0.303, Cohen’s d = 0.39). No latency difference was found between theta-band power and dYFP, but both preceded ATP signal change (one-way ANOVA with Tukey’s post-hoc; theta vs dYFP, p = 0.257; dYFP vs ATP, p = 0.00318; theta vs ATP, p = 0.0035; error bars, SEM).
Fig. 7 Theta-band power of the ECoG and dYFP changes precede the onset of REM sleep. Representative recordings from Mlc1-tTA::tetO-PYRS (a) and Thy1-ATeam (b) mice. From top to bottom: ECoG spectrogram, theta/delta power ratio, delta-band power (0.5–4 Hz), theta-band power (6–9 Hz), and optical signals (dYFP, fYFP, and pyruvate or ATP) in RSC. Arrowheads indicate automatically detected change points (purple, theta-band; red, dYFP; black, pyruvate/ATP). c Timing of changes in theta-band power of ECoG, dYFP, and astrocytic pyruvate signals relative to REM sleep onset (13 episodes from n = 3 mice). Increases in theta-band power and decreases in dYFP occurred significantly earlier than REM sleep onset (one-sample t-test with Holm correction, theta-band power, p = 3.0 × 10–8, Cohen’s d = 2.99; dYFP, p = 7.9 × 10–7, Cohen’s d = 2.12), whereas pyruvate did not (p = 0.101, Cohen’s d = 0.47). Latencies were similar for theta-band power and dYFP changes, but both preceded pyruvate signal change (one-way ANOVA with Tukey’s post-hoc; theta vs dYFP, p = 0.896; dYFP vs pyruvate, p = 4.1 × 10–5; theta vs pyruvate, p = 9.0 × 10–6; error bars, SEM). d Corresponding analysis for theta-band power, dYFP, and ATP signals (14 episodes from n = 3 mice). Theta-band power and dYFP again preceded REM sleep onset, whereas ATP signal did not (one-sample t-test with Holm correction, theta-band power, p = 4.5 × 10–4, Cohen’s d = 1.35; dYFP, p = 5.0 × 10–5, Cohen’s d = 1.76; ATP, p = 0.303, Cohen’s d = 0.39). No latency difference was found between theta-band power and dYFP, but both preceded ATP signal change (one-way ANOVA with Tukey’s post-hoc; theta vs dYFP, p = 0.257; dYFP vs ATP, p = 0.00318; theta vs ATP, p = 0.0035; error bars, SEM).CC BY · PubMed Central
Fig. 8 Astrocytic pyruvate increases while neuronal ATP decreases during REM sleep. a, b Representative recordings from an Mlc1-tTA::tetO-PYRS mouse expressing a pyruvate sensor in astrocytes (panels a, c, d, g, i, k) and a Thy1-ATeam mouse expressing an ATP sensor in neurons (panels b, e, f, h, j, l). ECoG spectrograms, raw ECoG, and EMG traces indicate sleep-wake states. c–f Fluorescence signals from four cortical regions: secondary motor cortex (MOs), somatosensory upper limb cortex (SSu), primary visual cortex (V1), and retrosplenial cortex (RSC). Astrocytic pyruvate (d) and neuronal ATP (f) signals were calculated by correcting FRET signal (fYFP) with dYFP to remove blood volume and pH change effects (pyruvate: – (fYFP – dYFP; ATP: fYFP – dYFP). Astrocytic pyruvate (d) increased during REM sleep and declined gradually after awakening, whereas neuronal ATP (f) decreased during REM sleep and recovered rapidly upon awakening. g, h Trajectories astrocytic pyruvate (g) or neuronal ATP (h) signals versus inverted dYFP (idYFP, a proxy for BBV) during pre-REM (100 s before REM onset; black), REM (last 50 s before awakening; light blue), and WAKE (50 s after awakening; orange) periods. Red circles indicate the mean values for each period. Arrows indicate the temporal progression direction of trajectories. Cortical maps of idYFP and astrocytic pyruvate (i) and neuronal ATP (j), averaged over each sleep period. All cortical maps represent absolute ΔF/F values. k, l Scatter plots summarizing mean astrocytic pyruvate (k) and neuronal ATP (l) signal changes against idYFP in eight cortical regions (26 episodes from n = 3 mice) for pre-REM, REM, and WAKE periods. Data are presented as mean ± SEM. During REM sleep, posterior regions (e.g., RSC, V1) showed larger increases in idYFP and astrocytic pyruvate and stronger decreases in neuronal ATP. Individual data points for each episode underlying the regional means shown in (k and l) are provided per cortical region in Supplementary Fig. 7.
Fig. 8 Astrocytic pyruvate increases while neuronal ATP decreases during REM sleep. a, b Representative recordings from an Mlc1-tTA::tetO-PYRS mouse expressing a pyruvate sensor in astrocytes (panels a, c, d, g, i, k) and a Thy1-ATeam mouse expressing an ATP sensor in neurons (panels b, e, f, h, j, l). ECoG spectrograms, raw ECoG, and EMG traces indicate sleep-wake states. c–f Fluorescence signals from four cortical regions: secondary motor cortex (MOs), somatosensory upper limb cortex (SSu), primary visual cortex (V1), and retrosplenial cortex (RSC). Astrocytic pyruvate (d) and neuronal ATP (f) signals were calculated by correcting FRET signal (fYFP) with dYFP to remove blood volume and pH change effects (pyruvate: – (fYFP – dYFP; ATP: fYFP – dYFP). Astrocytic pyruvate (d) increased during REM sleep and declined gradually after awakening, whereas neuronal ATP (f) decreased during REM sleep and recovered rapidly upon awakening. g, h Trajectories astrocytic pyruvate (g) or neuronal ATP (h) signals versus inverted dYFP (idYFP, a proxy for BBV) during pre-REM (100 s before REM onset; black), REM (last 50 s before awakening; light blue), and WAKE (50 s after awakening; orange) periods. Red circles indicate the mean values for each period. Arrows indicate the temporal progression direction of trajectories. Cortical maps of idYFP and astrocytic pyruvate (i) and neuronal ATP (j), averaged over each sleep period. All cortical maps represent absolute ΔF/F values. k, l Scatter plots summarizing mean astrocytic pyruvate (k) and neuronal ATP (l) signal changes against idYFP in eight cortical regions (26 episodes from n = 3 mice) for pre-REM, REM, and WAKE periods. Data are presented as mean ± SEM. During REM sleep, posterior regions (e.g., RSC, V1) showed larger increases in idYFP and astrocytic pyruvate and stronger decreases in neuronal ATP. Individual data points for each episode underlying the regional means shown in (k and l) are provided per cortical region in Supplementary Fig. 7.CC BY · PubMed Central

Verification

Claims the paper makes

28 Valid · 6 Not valid · 14 Not checked · 186 Own results and interpretations

121 more claims · 11 with their cited sources checked

Create a free account to see all 234 claims and to verify the 1 reference still open.

See all claims — free account
  • It may state two separate claims as one. 90
  • Probably not a substantial claim. 34
  • Borderline: one of the tests for a substantial claim is unclear. 2

Create a free account to see these claims.

Citations

Cites

Show all 78 references

Metadata

Authors
Takahashi, Y., Ikoma, Y., & Matsui, K.
Affiliations
Tohoku University (JP)
Corresponding author
Matsui, Ko
Venue
Communications Biology
Published
2026-07-27
Volume
9
Issue
1
ISSN
2399-3642 (ISSN-L)
OpenAlex
W7171433647
Cited by
1
References
78
Citations by year
2026: 1
FWCI
4.11
Type
journal article
Language
en
Open access
Licence
Locations
  • Communications Biology (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
Funders
  • MEXT | Japan Society for the Promotion of Science (23H04659, 25H01713, 26H01665, 18H05110, 20H05046, 19H03338, 22H02713, 25K02373, 22K15218, 24K18234, 26K09537)
  • Takeda Science Foundation (n/a)
  • Uehara Memorial Foundation (n/a)
  • Daiichi Sankyo Foundation of Life Science (n/a)
  • Research Foundation for Opto-Science and Technology
  • Data Science Program in Tohoku University
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Sleep and Wakefulness Research
  • Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Neuroscience and Neuropharmacology Research
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural dynamics and brain function
Keywords
REM sleep, NREM sleep, brain energy metabolism, cerebral blood volume, wide-field fluorescence imaging
MeSH terms
Adenosine Triphosphate, Animals, Astrocytes, Brain, Energy Metabolism, Male, Mice, Inbred C57BL, Neurons, Sleep, REM, Pyruvic Acid, Mice

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-09, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-09 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Energy paradox in REM sleep: balancing supply and consumption in brain metabolism. https://app.cochat.ai/papers/10.1038/s42003-026-10646-6
The paper
Takahashi, Y., Ikoma, Y., & Matsui, K. (2026). Energy paradox in REM sleep: balancing supply and consumption in brain metabolism. Communications Biology. https://doi.org/10.1038/s42003-026-10646-6

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.