Abstract
The suprachiasmatic nucleus (SCN) is the central clock for circadian rhythms. Animal studies have revealed daily rhythms in the neuronal activity in the SCN. However, the circadian activity of the human SCN has remained elusive. In this study, to reveal the diurnal variation of the SCN activity in humans, we localized the SCN by employing an areal boundary mapping technique to resting-state functional images and investigated the SCN activity using perfusion imaging. In the first experiment (n = 27, including both sexes), we scanned each participant four times a day, every 6 h. Higher activity was observed at noon, while lower activity was recorded in the early morning. In the second experiment (n = 20, including both sexes), the SCN activity was measured every 30 min for 6 h from midnight to dawn. The results showed that the SCN activity gradually decreased and was not associated with the electroencephalography. Furthermore, the SCN activity was compatible with the rodent SCN activity after switching off the lights. These results suggest that the diurnal variation of the human SCN follows the zeitgeber cycles of nocturnal and diurnal mammals and is modulated by physical lights rather than the local time.
- cerebral blood flow
- functional magnetic resonance imaging
- human suprachiasmatic nucleus
- medial preoptic area
- perfusion imaging
Significance Statement
The suprachiasmatic nucleus (SCN) in the hypothalamus is the central clock for circadian rhythms in mammals. However, the circadian activity of human SCN remained elusive due to the difficulty of measuring the activity of such a small nucleus. In this study, we localized the human SCN and investigated the SCN activity using an MRI technique for measuring perfusion. We observed SCN activity patterns of higher activity in daylight time and lower at night and morning. We also observed that the human SCN activity gradually decreased during the night, compatible with the rodent SCN activity after switching off the lights. These results suggest that the diurnal variation of human SCN activity is modulated by physical lights rather than the local time.
Introduction
Circadian rhythms control physiology and behavior with the 24 h cycle of day and night (Hastings et al., 2018) and circadian disruption is related to diseases in various human systems, including autonomic and endocrine systems (Cai et al., 2020; Fishbein et al., 2021). The suprachiasmatic nucleus (SCN) is the central clock for circadian rhythms in mammals (Moore and Eichler, 1972; Stephan and Zucker, 1972; Inouye and Kawamura, 1979, 1982; Sumova et al., 1995; Welsh et al., 1995, 2010; Nagano et al., 2003; Yamaguchi et al., 2003; Saper et al., 2005a,b; Enoki et al., 2012; Musiek and Holtzman, 2016; Hastings et al., 2018). The SCN is a pair of nuclei in the anterior part of the hypothalamus, located above the optic chiasm and lateral to the third ventricle. Intrinsically photosensitive retinal ganglion cells that express the photopigment, melanopsin, and project to the SCN neurons via the retinohypothalamic tract (Berson et al., 2002; Hattar et al., 2002; D. C. Fernandez et al., 2018; Touitou and Point, 2020; Sabbah et al., 2022). Modulated by light exposure, the SCN activity is the master clock for the body clock cycles in animals. The SCN sends the clock information to other parts of the central nervous system and influences the autonomic and endocrine systems. Moreover, the SCN affects motor activity (T. J. Nakamura et al., 2011), emotion (Vandewalle et al., 2010; Legates et al., 2012), and cognition such as memory (Ruby et al., 2008; Legates et al., 2012; F. Fernandez et al., 2014; D. C. Fernandez et al., 2018).
Several previous studies have shown some basic features of the SCN in humans. Postmortem brain studies showed diurnal cycles of neuropeptides in the human SCN (Hofman and Swaab, 1993, 1994; Hofman, 2000). In functional magnetic resonance imaging (fMRI), it has been shown that visual responses of the SCN depend on the light wavelength (Schoonderwoerd et al., 2022). In contrast to animal studies, it is hard to conduct experiments in humans due to the difficulties in strictly controlling the duration of light and darkness. It is also important to note that modern lifestyle influences the internal clock, such as excessive use of bright light until late at night (Touitou and Point, 2020). Moreover, because of the small size of the SCN, measuring the SCN activity in humans remained difficult. However, how the SCN activity changes over one cycle of the circadian rhythm and how the circadian SCN activity differs from those in animals need to be investigated.
This study investigated the diurnal variation of the human SCN activity using perfusion imaging that measures cerebral blood flow (CBF) in the brain with 2 mm spatial resolution. Before the scan, we localized the SCN by using the resting state images collected in our previous study (Ogawa et al., 2022). Brain activity in the SCN over 24 h was then measured. The participants were scanned four times a day, every 6 h, with a within-participant design. Unlike blood oxygenation level dependent (BOLD) imaging, perfusion imaging allows us to compare images taken between long intervals, which overcomes the merit of higher spatial resolution of BOLD images. Brain activity was further scanned every 30 min for 6 h from midnight to dawn to reveal more detailed characteristics of the diurnal variation compared to previous animal data.
Materials and Methods
Experimental designs
In this study, two experiments were conducted, wherein the first investigated the whole cycle of the diurnal activity of the SCN in humans (Experiment 1). Two perfusion images of pseudo-continuous arterial spin labeling (pCASL) in each participant were acquired four times within 24 h (18:00, 24:00, 6:00, 12:00 on local time) and were used to calculate the CBF at a specific time (Fig. 1A). The lights in the MRI room were turned on in all the scans, and participants were instructed to have meals 4.5 h before each scan and to rest at a hotel at night (Extended Data Fig. 1-1). It is a bit unusual to have four meals a day, especially just before sleep at night. Therefore, we instructed the participants to take at least small meals if they do not want to take ordinary meals. To administer scans for multiple (up to four) participants in one 24 h session, the exact scan time was shifted by 30 min for each scan, such as 17:00, 17:30, 18:00, and 18:30 for the evening scans.
Scan schedules. A, Scan schedule of Experiment 1. Scans were conducted in evening (18:00), midnight (24:00), morning (6:00), and daytime (12:00) sessions in that order. Two CBF images were acquired and averaged at each session to increase the signal-to-noise ratio. Between the scan sessions, the participants rested at the accommodation provided. See Extended Data Figure 1-1 for more details. B, Scan schedule of Experiment 2. The room lights were on during the first scan (called pre) and the last scan (called post), whereas the room lights were off during the scans from 0:15 to 5:45. The black and white arrowheads indicate the lights in the MRI room being switched off and on, respectively. The black bar indicates the period in which the light of MRI room was off.
Figure 1-1
Example of schedule in Experiment 1. Participants took a rest at a hotel between MRI scans. Download Figure 1-1, TIF file.
In the second experiment, we investigated the human SCN activity in more detail during the night (Experiment 2). Participants stayed in the scanner throughout the night, except for brief unavoidable interruptions, and were scanned every 30 min from 24:00 to 6:00 (Fig. 1B). The lights in the MRI room were switched off at 0:15 and on at 5:45 am. In each scan, one single perfusion image was acquired. Electroencephalography (EEG) was recorded during scanning using an MRI-compatible 32-channel international 10/20 EEG system. The level of sleepiness during scanning was reported using VAS (from 0 to 10; 10, maximally sleepy or sleeping) after each scan. The participants were and asked in a whisper to report their sleepiness VAS score (0–10) by showing the corresponding number of fingers of the left hand (when the number was 8, 5 and then 3 fingers were shown). When the participants did not respond (58 cases, 27%), we did nothing further. To monitor the stress level, salivary amylase activity was measured before and after the MRI session using a salivary amylase monitor (DM-3.1, NIPRO).
Participants
Twenty-seven right-handed participants without neurological/psychiatric illness or sleep disorders participated in Experiment 1 [13 males and 14 females, age: 22.8 ± 2.7 years (mean ± standard deviation) ranging from 20 to 32 years], while twenty right-handed participants without neurological/psychiatric illness or sleep disorders participated in Experiment 2 (10 males and 10 females, age: 21.5 ± 1.5 years, ranging from 20 to 24 years). The participants were undergraduate and graduate students. Hence, they usually woke up in the morning to attend classes. None of the participants in these experiments engaged in works that may have affected their circadian rhythms, such as night shifts. Written informed consent was obtained from all participants following the Declaration of Helsinki. The Research Ethics Committee, Faculty of Medicine, Juntendo University, approved the experimental procedures.
SCN localization
The location of the SCN was identified using a boundary mapping technique (Margulies et al., 2007; Cohen et al., 2008; Buckner et al., 2011; Eickhoff et al., 2015; Laumann et al., 2015; Poldrack et al., 2015; Glasser et al., 2016; Gordon et al., 2016; Osada et al., 2017, 2021; Gratton et al., 2018; Ogawa et al., 2018, 2020, 2022; Fujimoto et al., 2020, 2022, Suda et al., 2020; Nakajima et al., 2022). The resting-state images collected in our previous study (Ogawa et al., 2022) were used in this analysis. A total of 3,000 volumes were collected from each of 27 participants with 1.25 mm isotropic voxels using an multiband echo planner imaging (MB-EPI) sequence (Feinberg et al., 2010; Xu et al., 2013) (repetition time = 2.3 s, echo time = 20 ms, flip angle = 73°, the in-plane field of view = 180 × 180 mm2, matrix size = 144 × 144, 108 contiguous slices with no gap, phase encoding direction = posterior-to-anterior, parallel acquisition factor = 2, and multiband factor = 3).
The images were preprocessed using the Human Connectome Project pipeline (Glasser et al., 2013) with modifications for a higher resolution. The images were motion-corrected, distortion-corrected, and spatially normalized to the standard space of Montreal Neurological Institute (MNI) coordinates. The time series of the images were projected from the voxel space onto the standard surface (32,492 vertices in each hemisphere) (Glasser et al., 2013), high-pass filtered (cut-off = 2,000 s), and denoised using the ICAFIX method (Salimi-Khorshidi et al., 2014). Surface registration was refined using the MSM-All (Robinson et al., 2018). We included the surrounding voxels of the hypothalamus for the SCN boundary mapping by extending the limits of the ventral diencephalon in a volumetric space. For the processing of pCASL images, we tuned the parameters for skull-stripping to dilate the brain mask and made ROI analyses in a volumetric space. Head motion parameters were not regressed out explicitly. Instead, the ICAFIX that included independent components of head motion excluded the brain activity associated with the movement parameters. The global signal was regressed out from the subcortical data as well as the surface data.
We calculated correlations between each voxel in the hypothalamus, excluding the voxels in the mammillary body, and the vertices in the cerebral surface of each participant. The correlation coefficient in each vertex was transformed to Fisher's z value (i.e., cerebral correlation map), and the spatial similarity of the cerebral correlation maps for hypothalamic voxels was then calculated (i.e., hypothalamic similarity map). Spatial gradients of the hypothalamic similarity maps were computed for each hypothalamic voxel (i.e., hypothalamic gradient map). The hypothalamic gradient maps were averaged across participants to generate the group hypothalamic gradient maps. After a minimal spatial smoothing [full width at half maximum (FWHM) = 1.25 mm] of the group hypothalamic gradient maps, excluding the third ventricle of each side, a three-dimensional watershed algorithm (Vincent and Soille, 1991) was applied to the group hypothalamic gradient maps (i.e., binary watershed maps). The binary watershed maps were averaged across the hypothalamic voxels to generate the probabilistic boundary map (Fig. 2), where the probability value represents how likely the voxel is a boundary of the functional area (parcel), or in other words, how unlikely the voxel is a center of the parcel.
ROI definition of the SCN. A, SCN ROI on the axial section. The probabilistic boundary map of the hypothalamus is shown on the axial section (center panel). High probability (yellow) indicates that the boundaries are likely to exist, while lower probability than the surroundings (red) indicates that the centers of nuclei are likely to exist. The right panel shows the SCN ROI on the axial section. B, ROIs of the SCN on the coronal section. C, ROIs of the SCN on the sagittal section. The center panel shows the probabilistic boundary map, while the right panel shows the ROIs of the SCN. Black lines indicate the border of the hypothalamus. L, left; R, right; D, dorsal; V, ventral; A, anterior; oc, optic chiasm; opt, optic tract; MB, mammillary body; 3V, third ventricle. See Extended Data Figures 2-1, 2-2, and 2-3 for more details.
Figure 2-1
SCN ROI. A. SCN ROI on the coronal section of a high-resolution T1-weighted image of a representative participant. The cyan voxels indicate the SCN ROIs. B. SCN ROIs on the coronal sections of high-resolution T1-weighted images of all participants. Download Figure 2-1, TIF file.
Figure 2-2
F-map and ROIs (SCN, MPO, SO, and PVH) on the coronal (A), axial (B), and sagittal (C) sections in Experiment 1. The black solid lines in the right panel indicate the border of the hypothalamus. The colored voxels indicate ROIs of the SCN, MPO, SO, and PVH. The coordinates of the ROIs are reported in Table 1. Download Figure 2-2, TIF file.
Figure 2-3
Spatial coverage of the SCN using a 2-mm isotropic voxel in relation to another nucleus. In the left panel, when the SCN (approximately 1 mm diameter) is located in the center of the 2 mm voxel, it is unlikely that the adjacent nucleus (approximately 1 mm diameter) located approximately 1 mm apart is included in that voxel. In the middle panel, when the SCN is located 0.5 mm off the center of the voxel, only a small portion of the adjacent nucleus would be included in that voxel. In the right panel, when the SCN is not fully included in the voxel, the adjacent nucleus may be included to an inverse degree. Red and blue disks indicate the SCN and another nucleus of the hypothalamus, such as the SO. The scale bar indicates 1 mm. Download Figure 2-3, TIF file.
The bilateral parcels of the SCN were identified above the optic chiasm, and the center voxel of the SCN parcel was defined as the voxel with the least probability value in the SCN parcel in a 1.25 mm isotropic BOLD image. The region of interest (ROI) of the SCN in a 2 mm isotropic perfusion image was defined as the voxel corresponding to the center voxel in the SCN parcel. Therefore, the volume of the SCN ROI was 8 mm3, consisting of one 2 mm isotropic voxel.
MRI procedures
All MRI data were acquired using a 3-T MRI scanner at Juntendo University Hospital (Siemens Prisma) with a 32-channel head coil. T1-weighted structural images were obtained using 3D magnetization-prepared rapid gradient-echo (resolution = 0.8 × 0.8 × 0.8 mm3) in a separate scan session before the experiments. The imaging parameters for pCASL were determined based on the protocols of the Human Connectome Project (protocols.humanconnectome.org) (Harms et al., 2018). Whole brain perfusion images were acquired using pCASL imaging with MB-EPI (number of measurements = 90, repetition time = 4.0 s, echo time = 25.2 ms, partial Fourier = 6/8, flip angle = 90°, labeling duration = 1.5 s, postlabeling delay = 1.64 s, slice thickness = 1.82 mm, distance factor = 10%, number of slices = 72, slice acquisition order = ascending, in-plane field of view = 212 × 212 mm2, matrix size = 106 × 106, multiband acceleration factor = 6) (X. Li et al., 2015). Two M0 images, which was included in the pCASL sequence, were also acquired after the label/control images series. The mean image of these two M0 images was used for CBF quantification. Two images with one anterior-to-posterior and one posterior-to-anterior encoding direction were acquired using the spin-echo field map sequence before each pCASL imaging, taking less than 1 min. These images were used to perform the top–bottom distortion correction for pCASL images (Andersson et al., 2003). Each scan session took less than 20 min in Experiment 1, and each scan session in Experiment 2 took less than 10 min.
The perfusion images were corrected for motion and distortion. The CBF maps in the standard MNI space were calculated using a ,nd line interface of oxford_asl that is part of the BASIL (Bayesian Inference for Arterial Spin Labeling MRI) toolbox (Chappell et al., 2009) included in FSL (Smith et al., 2004). The structural image was used for a reference in the registration and generating a CSF mask for absolute quantification. Spatially minimal smoothing was applied to the CBF images (full-wise half maximum of Gaussian kernel = 2.0 mm), and the absolute CBF values [ml/100 mg/min] of the SCN ROIs were extracted and sent for statistical analyses. We performed a one-way repeated-measures analysis of variance (ANOVA) for the activity of the SCN over the time (6:00, 12:00, 18:00, and 24:00 on local time) in Experiment 1. Post hoc Tukey–Kramer tests were performed using a statistical threshold (p < 0.05). Based on a reviewer's comment, we performed a post hoc power analysis using G*Power3.1 (Faul et al., 2009). In Experiment 2, we performed a regression analysis using a linear mixed-effects model for the activity of the human SCN and the neuronal activity of the rodent SCN. We also performed a post hoc power analysis for the regression of the human SCN activity. The participants were treated as random effects. To investigate the relation of the SCN activity to sleep, sleepiness during scanning was reported using a visual analog scale (VAS, from 0 to 10; 10, maximally sleepy or sleeping) after each scan. The temporal effect on the sleepiness VAS score was analyzed using a linear mixed-effects model analysis.
EEG procedures
In Experiment 2, EEG was recorded during the scan using an MRI-compatible 32-channel international 10/20 EEG system (Geodesic Sensor Net, Electrical Geodesics, Inc.). Before the experiment, the impedances of all electrodes were adjusted to <50 kΩ. The impedances were checked before each scan. After the experiment, EEG data were band-pass filtered (0.3–30 Hz) and referenced to the electrode Cz. The EEG data around the middle of each scan was used for the classification of the frequency band: β (more than 12 Hz), α (8–12 Hz), θ (4–8 Hz), or δ (1–4 Hz) band. One of the authors classified the frequency band of EEG data of each scan, and then the other author verified the classification. The data of electrodes whose impedance was over 50 kΩ in any scan were not used for the EEG band classification. We performed a one-way repeated-measures ANOVA for the activity of the SCN over the EEG bands.
Results
Two experiments were conducted in this study. In the first experiment, the whole cycle of the diurnal activity of the SCN was investigated by two perfusion images measuring the CBF, which were acquired four times (18:00, 24:00, 6:00, 12:00 on local time) within 24 h (Experiment 1) (Fig. 1A, Extended Data Fig. 1-1). The lights in the MRI room were on in all the scans. On the other hand, the second experiment investigated the temporal trend of human SCN activity at night (Experiment 2). The perfusion images were scanned every 30 min from 24:00 to 6:00 on local time (Fig. 1B), with EEG recorded using an MRI-compatible system. The lights of MRI room were switched off at 0:15 and on at 5:45. To analyze the CBF in the SCN, the SCN was localized by using BOLD images during the resting state collected in our previous study (Ogawa et al., 2022). An areal boundary mapping technique parcellated the anterior part of the hypothalamus into the subregions, including the SCN.
SCN localization
The areal boundary mapping technique generated, from BOLD images (Ogawa et al., 2022), the probabilistic boundary map in 1.25 mm resolution, where the probability value represented how likely the voxel is a boundary of a functional area, a “parcel” that corresponds to a hypothalamic nucleus (Fig. 2). The low probability value in a voxel of the map, therefore, indicates that the voxel is likely the center of a parcel. In the probability map, the SCN was identified bilaterally above the optic chiasm and lateral to the third ventricle (Extended Data Fig. 2-1). We also localized the surrounding nuclei (medial preoptic nucleus: MPO, supraoptic nucleus: SO, paraventricular nucleus: PVH) and listed the coordinates in Table 1. The coordinates of the voxel of the SCN ROI in 2 mm resolution for perfusion imaging were x = −2, y = 2, z = −16 for the left ROI and x = 2, y = 2, z = −16 for the right ROI (Fig. 2).
Centroid coordinates of the SCN and neighboring hypothalamic nuclei and their distances from the SCN
Diurnal variation of SCN activity
Experiment 1 investigated whether the SCN activity had increased or decreased signal changes within 24 h. The activity in the SCN was highest at 12:00 in the daytime and lowest at 6:00 in the morning (Fig. 3A). The SCN activity was significantly modulated among the four scans (F(3,78) = 3.38, p = 0.022). The activity at 12:00 was significantly higher than that at 6:00 (Tukey–Kramer test, p < 0.05). A similar trend was seen when the left and right SCN were analyzed separately (Fig. 3A) and when the first and second scans were analyzed separately (Fig. 3B). We also performed a post hoc power analysis using G*Power 3.1. The power (1 − β) of ANOVA of the SCN activity in Experiment 1 was 0.89. If the power (1 − β) threshold was set to 0.8, the minimum detectable effect size was f = 0.24, and the actual effect size was f = 0.27.
Results of Experiment 1. A, Time course of the SCN CBF. The CBF signal decreases from the daytime (12:00) to the morning (6:00). The rightmost CBF is the same as the leftmost CBF. The dotted line shows the CBF level at 6:00. B, Reproducibility of the SCN CBF. The time course from the first scan is similar to that from the second scan. Error bars indicate standard errors of the mean. The asterisk indicates statistical significance (*p < 0.05). See Extended Data Figures 3-1, 3-2, and 3-3 for more details.
Figure 3-1
Time course of CBF values in the MPO, SO, and PVH in Experiment 1. A. Time course of MPO CBF. There was a significant temporal effect (one-way repeated-measures ANOVA, F(3,78) = 4.35, P = 0.0069). B. Time course of SO CBF. The temporal effect was significant (one-way repeated-measures ANOVA, F(3,78) = 4.78, P = 0.0041). C. Time course of PVH CBF. No significant temporal effect was observed (one-way repeated-measures ANOVA, F(3,78) = 1.73, P = 0.17). The dot lines indicate the CBF levels at 6:00. Error bars indicate standard errors of the mean. Download Figure 3-1, TIF file.
Figure 3-2
Results of one-way repeated-measures ANOVA in the hypothalamic nuclei. Download Figure 3-2, DOCX file.
Figure 3-3
Results of one-way repeated-measures ANOVA in the brain structures defined by the Human Connectome Project (Glasser et al., 2013, 2016). Download Figure 3-3, DOCX file.
The results of ANOVA for the activity in the other nuclei in the hypothalamus are reported in Extended Data Figures 2-2, 3-1, and 3-2. The diurnal variations were observed in the MPO and SO, confined to the anterior part of the hypothalamus. Some different regions, such as the amygdala, showed the diurnal variations (Extended Data Fig. 3-3), consistent with the fact that the SCN innervates other brain regions in and outside the hypothalamus (Pu and Pickard, 1996; Harrington, 1997). The activity pattern was similar to the activity in the adjacent areas of diurnal mammals but not of nocturnal mammals (Sato and Kawamura, 1984; Bano-Otalora et al., 2021a,b).
SCN activity during the night
In Experiment 2, the SCN activity was measured in detail every 30 min from 24:00 to 6:00. A regression analysis indicates that the SCN activity gradually decreased from midnight to dawn (β estimate = −1.40, t(215) = −2.41, p = 0.017) (Fig. 4A). We calculate averages and maximums of framewise displacements (FDs) (Power et al., 2012) in millimeter for each scan in Experiment 2. The results are summarized in Extended Data Figures 4-1 (average FDs) and 4-2 (maximum FDs). The grand average of the FD (averaged across 13 scans and 20 participants) was 0.1 mm. When the outlier participant (S05) was removed, the results were similar (β estimate = −1.46, t(205) = −2.26, p = 0.025). We also performed a post hoc power analysis using G*Power 3.1. The power of the linear regression analysis for the SCN activity was 1.00. If the power (1 − β) threshold was set to 0.8, the minimum detectable effect sizes were f2 = 0.04 (f = 0.19), and the actual effect sizes were f2 = 0.11 (f = 0.33).
Results of Experiment 2. A, Time course of the SCN CBF. The white and black dots indicate that the MRI room was light and dark, respectively. The CBF in the SCN decreased gradually during the night. The gray line is a regression line. See Extended Data Figures 4-1, 4-2, and 4-3 for more details. B, Two hypotheses of correspondence between the SCN activities of humans and mice. One is the time lock to the switch-off of the room lights. The other is local time matched. C, Trends of neuronal activity in the rodent SCN. Graphs of the rodent SCN activity were modified from the data between ZT 12 and ZT 18 and between ZT 18 and ZT 24 extracted from T. J. Nakamura et al. (2011). The gray solid and dotted lines are regression lines. Error bars indicate standard errors of the mean.
Figure 4-1
Averaged framewise displacements (mm) of each scan in Experiment 2. Download Figure 4-1, DOCX file.
Figure 4-2
Maximum framewise displacements (mm) of each scan in Experiment 2. Download Figure 4-2, DOCX file.
Figure 4-3
Time course of CBF values in the MPO, SO, and PVH in Experiment 2. A. Time course of MPO CBF. The MPO activity gradually decreased from midnight to dawn (beta estimate = -1.45, t(215) = -2.86, P = 0.0047). B. Time course of SO CBF. The SO activity showed a gradually decreasing trend from midnight to dawn (beta estimate = -1.37, t(215) = -3.00, P = 0.0031). C. Time course of PVH CBF. A decreasing trend of the PVH activity was observed (beta estimate = -1.64, t(215) = -3.40, P = 0.0008). The white and black dots indicate that the light in the MRI room was on and off, respectively. Error bars indicate standard errors of the mean. Download Figure 4-3, TIF file.
We then compared the activity of the human SCN with the rodent SCN activity published previously (T. J. Nakamura et al., 2011). The previous study is one of very few that collected data, still available to date, from more mature mice that would match in age with the present human study. In the previous study, the multiunit neural activity of the SCN was reported every 30 min. The current study divided the rodent SCN data exposed to darkness into two periods: one from zeitgeber time (ZT) 12 to ZT 18 and the other from ZT 18 to ZT 24, where the time of lights on is defined as ZT 0. Zeitgeber time is a unit of time based on the period of an experimental cue for circadian rhythms, where the start time of the light cycle is ZT 0 and the start time of the dark cycle is ZT 12 in the case of a cycle of 12:12 h light/dark. We examined whether the human SCN activity during the night (24:00–6:00) was more compatible with the activity during ZT 12–18 or with the activity during ZT 18–24 in rodents (Fig. 4B). In other words, we examined whether the human SCN activity during the night is more compatible with the rodent SCN activity when the activity is time-locked to lights off or when the activity was matched to local time. The neuronal activity of the rodent SCN decreased during the first half of the dark period (β estimate = −0.04, t(42) = −4.08, p = 0.0002) (Fig. 4C left), in accordance with the activity of the human SCN during the night revealed in Experiment 1 at 24:00 and 6:00. In contrast, the activity of the rodent SCN showed an increasing trend during the second half of the dark period (β estimate = 0.01, t(42) = 1.59, p = 0.12) (Fig. 4C right). Thus, the activity in the human SCN during the night better matched with the former case in mice.
The temporal effect was significant for the sleepiness VAS scores (β estimate = −0.20, t(216) = −2.37, p = 0.019) (Fig. 5A). Perfusion scans were classified based on the VAS scores of 0–10. No significant difference in the SCN activity was observed among the VAS scores (one-way ANOVA, F(6,92) = 0.17, p = 0.98) (Fig. 5B), suggesting that the level of sleepiness cannot explain the decreasing trend of the SCN activity during the night. Amylase activity that varies diurnally measured before and after the MRI session was not significantly different (two-tailed t-test, t(19) = 1.75, p = 0.096).
Activity of SCN in Experiment 2. A, Sleepiness visual-analogue scale (VAS) score. The score decreased during the night. The white dots indicate that the scores were obtained with the lights on in the MRI room, while the black dots indicate that the scores were obtained with the lights off in the MRI room. B, CBF in the SCN for each sleepiness VAS score. There was no significant difference. See Extended Data Figure 5-1 for more details. C, Percentages of observed EEG bands in each scan. α and θ bands were mainly observed. D, SCN CBF of each EEG band. No significant difference between the EEG bands was observed. See Extended Data Figure 5-2 for more details. Error bars indicate standard errors of the mean.
Figure 5-1
CBF values in the MPO, SO, and PVH for each sleepiness VAS score in Experiment 2. As with SCN, no significant difference was observed in these nuclei (one-way ANOVA, A. MPO, F(6,92) = 0.30, P = 0.94; B. SO, F(6,92) = 0.47, P = 0.83; C. PVH, F(6,92) = 0.57, P = 0.75). Error bars indicate standard errors of the mean. Download Figure 5-1, TIF file.
Figure 5-2
CBF values in the MPO, SO, and PVH in each EEG band in Experiment 2. There is no significant difference between the EEG bands (one-way ANOVA, A. MPO, F(2,43) = 1.01, P = 0.37; B. SO, F(2,43) = 0.92, P = 0.41; C. PVH, F(2,43) = 1.01, P = 0.37). Error bars indicate standard errors of the mean. Download Figure 5-2, TIF file.
EEG was also recorded during the night scans and classified each perfusion scan into β/α/θ/δ bands. The percentage of EEG frequency band during the first half (before 3:00; β, 6.0%; α, 40.0%; θ, 53.0%; δ 1.0%) was not significantly different from that during the second half (after 3:00; β 13.1%; α, 34.3%; θ, 49.5%; δ 3.0%) (χ2(3) = 6.49, p = 0.09) (Fig. 5C). Whether SCN activity was different among the EEG bands was also examined. Due to the small number of events in the δ band, we excluded the data from statistical analyses. A one-way ANOVA showed that the SCN activity was not significantly different among the EEG bands (F(2,43) = 2.50, p = 0.09) (Fig. 5D).
Additional analyses were conducted in the adjacent nuclei. As with the SCN, a regression analysis indicated that the activity in the MPO, SO and PVH showed a gradually decreasing trend from midnight to dawn (Extended Data Fig. 4-3). Sleepiness score was not associated with the activity in the MPO, SO or PVH (Extended Data Fig. 5-1), and the activity in the MPO, SO or PVH was not significantly different among the EEG bands (Extended Data Fig. 5-2).
Discussion
This study investigated the diurnal activity of the SCN in humans using perfusion imaging. In the first experiment, the activity of the human SCN was examined every 6 h within 24 h. The activity of the SCN significantly varied, with the maximum activity recorded at noon and the minimum activity at 6:00. The second experiment examined the human SCN activity in more detail during the night (every 30 min from midnight to 6:00), and it was found that the human SCN activity decreased during this period, and the frequency band of the recorded EEG did not significantly affect the SCN activity. Furthermore, the human SCN activity during the night better matched with the activity of the rodent SCN time-locked to lights off. These results suggest that the diurnal variation of the SCN activity in humans was globally consistent with that in nonhuman mammals and could be influenced locally by lifestyle, such as bright ambient lights late at night.
A boundary mapping technique was employed to allow localization of the SCN in functional images, with a great number of high-resolution (1.25 mm isotropic voxels) BOLD images per participant. The image distortion in the phase-encoding direction around the hypothalamus was corrected using top–bottom correction, and denoising was conducted using an independent component analysis and machine learning. The coordinates of the SCN voxel in this study were (−2, 2, −16) on the left and (2, 2, −16) on the right, which are very close to those in a recent high-resolution fMRI study (Schoonderwoerd et al., 2022): (−5 to −3, 1 to 4, −17 to −15) on the left and (4 to 6, 1 to 4, −17 to −15) on the right. The slight difference in x coordinates appears to have come from the shape of the third ventricle, which is thinner in coronal slices in our present dataset (Extended Data Fig. 2-1).
On the other hand, the size of the SCN seems to be less than one voxel of the perfusion imaging (8 mm3) of the present study. In a recent cellular-resolution atlas, the size of the SCN was estimated to be (1.7 × 1.1 × 1.1) mm3 ∼ 2.1 mm3 (Vimal et al., 2009; Ding et al., 2016; Sharifpour et al., 2022), and in another recent high-resolution in vivo magnetic resonance imaging atlas, the size was estimated to be 4.9 mm3 (Neudorfer et al., 2020). Although the estimated size is still smaller than our 2 mm cubic voxel, the elongated shape of the SCN can be more appropriately covered by a larger ROI, presumably by our 2 mm cubic voxel. Moreover, owing to the 2 mm voxel size larger than the SCN (Meijer et al., 2022), the location of the SCN voxel was largely appropriate in individual structural images (Extended Data Fig. 2-1). It is also to be noted that the areas surrounding the SCN are not classified as nuclei (Mai et al., 2015). The estimated distance was 4.5 mm between the SCN and SO, and 5.5 mm between the SCN and PVH, which is discriminable in the 2 mm voxel/smoothing size. As demonstrated in Extended Data Figure 2-3, it is possible that the adjacent nucleus is included in the 2 mm voxel, but the proportion seems limited.
There are a few limitations regarding the experiments in this study. Exposure to light during overnight time points in Experiment 1 likely increased the SCN activity for the 24:00 and 06:00 time points. The short duration of sleep (about 4 h) may also have affected SCN activity at 06:00 and 12:00 in Experiment 1. For Experiment 2, the interrupted pattern of overnight sleep and slight exposure to light and cognitive demand for the VAS are potential concerns. Finally, no significant correlations were found between the SCN activity and sleepiness/EEG (Fig. 5B,D). As participants did not and likely could not sleep during the experiment, the interpretations of these findings need to be qualified.
An increase in neuronal activity leads to metabolic necessity and increases local blood flow in a large part of the brain. However, previous studies have shown that the BOLD signal in the hypothalamus decreased with events such as light exposure, food cue presentation, drinking water, and glucose ingestion (Smeets et al., 2005; Vidarsdottir et al., 2007; Osada et al., 2017; Ogawa et al., 2022; Schoonderwoerd et al., 2022). Thus, the neurovascular mechanism, which couples the local blood flow and the neural activity, may be inverted in the hypothalamus (Roy et al., 2021; Drew, 2022). The higher blood flow during daytime observed in the present study will not apply to the inverted neurovascular coupling in the hypothalamus; a study of human postmortem tissues showed the diurnal variations of vasopressin production in the SCN, with higher activity in vasopressin neurons during daytime than during the night (Hofman and Swaab, 1994). Another example of such ordinary neurovascular coupling in the hypothalamus can be seen in the resting-state functional connectivity between the hypothalamus and the cerebral cortex, which are anatomically connected and yield positive functional connectivity (Kullmann et al., 2014; Hirose et al., 2016; Zhang et al., 2018; Ogawa et al., 2020). Although precise mechanisms of neurovascular coupling in the hypothalamus remain controversial, it seems reasonable to speculate that neurovascular coupling depends on the nature of the neuronal activity, whether the activity is transiently evoked after an event or spontaneously occurring for a long period.
The CBF in the SCN showed slightly higher activity in the left versus right as revealed by a two-way repeated-measures ANOVA for the SCN activity over the time (06:00/12:00/18:00/24:00) and the laterality (left/right) (F(1,26) = 5.95, p = 0.022). The laterality may suggest that generation of diurnal variations in the brain and body depends on the amplitude of the variation, rather than the absolute level, of the SCN activity. Also, the MPO and SO showed slightly greater variation than the SCN (Extended Data Fig. 3-1). One possible explanation would be that the SCN send diverging output to the neurons in the nuclei (Saper et al., 2005a,b).
Neuronal activity in the SCN, which is high during daytime and low during the night, is generated by an endogenous and autonomous circadian rhythm (Inouye and Kawamura, 1979; Kuhlman and McMahon, 2006; W. Nakamura et al., 2008, T. J. Nakamura et al., 2011; Miyamoto et al., 2012). The daily cycle of activity in the SCN is similar in both nocturnal and diurnal mammals (Bano-Otalora et al., 2021b). The inverse phase difference of neuronal activity was observed within and outside the SCN in nocturnal mammals (Musiek and Holtzman, 2016). In contrast, the neuronal activity outside the SCN in diurnal mammals showed diurnal variations that paralleled the neuronal activity inside the SCN (Sato and Kawamura, 1984; J. Z. Li et al., 2013). Our results showed that the circadian activity in the human SCN was basically similar to that in other diurnal mammals in the SCN and its adjacent areas.
The current study observed a decreasing trend in the SCN activity when the scan room was dark, while increased activity was recorded just after the lights were turned on. Previous animal studies showed that the SCN activity decreased after the lights went out and increased before the lights came on (Schaap et al., 2003; VanderLeest et al., 2007; W. Nakamura et al., 2008, T. J. Nakamura et al., 2011; Miyamoto et al., 2012). The decrease in the SCN activity after turning the lights off is consistent for humans and other mammals. On the other hand, the lights in the scan room were turned on at 5:45 in the “one-time” experiment of this study, presumably earlier than the participants’ usual wake-up time. If the experiment is repeated, the rise of the activity in the SCN would shift earlier (Schaap et al., 2003; VanderLeest et al., 2007, 2009; Brown and Piggins, 2009; Houben et al., 2009). The human SCN activity during the night may adapt to the period of bright hours in our daily life.
Data Availability
The data and code that support the findings of this study are available at Dryad (https://doi.org/10.5061/dryad.280gb5mwz), except for raw image data. The raw image data cannot be deposited in a public repository because sharing raw image data was not included in the informed consent. Any additional information required to reanalyze the data reported in this paper is available from the corresponding author upon reasonable request.
Footnotes
This work was supported by JSPS KAKENHI Grants 22K07334 to A.O., 21K07255 to T.O., and 23H02783 to S.K., a grant from Takeda Science Foundation to S.K, and a Grant-in-Aid for Special Research in Subsidies for ordinary expenses of private schools from The Promotion and Mutual Aid Corporation for Private Schools of Japan.
The authors declare no competing financial interests.
- Correspondence should be addressed to should be addressed to Akitoshi Ogawa at a-ogawa{at}juntendo.ac.jp or Seiki Konishi at skonishi{at}juntendo.ac.jp.
