Sleep deprivation (SD) is a tool to study negative consequences of high homeostatic sleep pressure on brain and behavior. By using functional magnetic resonance imaging (fMRI), several meta-analysis studies (Ma et al., 2015; Yeo et al., 2015; Javaheripour et al., 2019; Saletin et al., 2019; Ning et al., 2022) revealed SD-induced changes in neuronal activity of brain regions, typically in the prefrontal cortex, thalamus, and intraparietal cortex regulating arousal, attention ability or emotional processing. Notably, several resting-state (Rs-)fMRI metrics, including amplitude of low-frequency fluctuations (ALFF) and its normalized version (fractional ALFF, fALFF), regional heterogeneity (ReHo), and degree centrality (DC), were applied to well characterize the differences in the regional properties of spontaneous brain activity (Zang et al., 2004, 2007; Zou et al., 2008; Zuo et al., 2012). In short, ALFF and fALFF are used to reflect the temporal fluctuations of low frequency oscillations (<0.1 Hz) across entire time series. ReHo is a voxel-based parameter for evaluating the similarity or synchronization of time series between a given voxel and its nearest neighbors. DC is an index of local functional connectivity (FC) strength and thus identifies the highest connected nodes by counting the number of direct connections to other nodes. Using these fMRI measures, we are capable to uncover voxel-based spectral and temporal characteristics of brain neuronal activity. For instance, prior Rs-fMRI studies (Gao et al., 2015; Wang et al., 2016; Chen L. et al., 2018) demonstrated reduced ALFF/fALFF values in the frontal and parietal cortex while increased values in the visual cortex and left sensorimotor cortex under the condition of sleep deprivation. Thus, the consistent changes across these fMRI properties should also be further investigated after acute sleep loss.

Generally, a neuronal response requires increased cellular energy metabolism that accelerates local cerebral blood flow or substrate delivery. With application of Positron Emission Tomography (PET), growing evidence (Wu et al., 1991; Thomas et al., 2000; Elmenhorst et al., 2007; Volkow et al., 2008, 2012; Qu et al., 2010; Holst et al., 2017) presented that acute SD remarkably upregulated A₁ adenosine receptor (A₁AR) availability, whereas glucose metabolism and dopamine D2/D3 neurons in human brain were reduced. Adenosine is a neuromodulator and directly linked to the energy metabolism as substrate of the breakdown of adenosine triphosphate. As a crucial mediator for promoting the homeostatic sleep drive, A₁AR can be found in widespread brain regions (Bjorness and Greene, 2009; Huang et al., 2011). Elmenhorst et al. (2017) used [¹⁸F]-CPFPX PET and reported that sleep loss increased the A₁AR availability in some cortical and subcortical brain regions, including the temporal cortex, striatum and insula. Meanwhile, neuronal alterations of these brain regions were also investigated by previous fMRI studies (Bell-McGinty et al., 2004; Venkatraman et al., 2007; Gujar et al., 2011; Krause et al., 2019). For instance, one earlier study (Krause et al., 2019) confirmed that sleep loss extremely amplified pain reactivity within primary somatosensory cortex but reduced pain reactivity in higher-order evaluation and decision-making regions of the striatum and insula. For those above-mentioned overlapped regions, no neuroimaging studies investigated so far whether changes in their neuronal activity are associated with corresponding changes in A₁AR availability. Therefore, we aimed to evaluate the contribution of A₁AR in the specific regional properties of neural activities of some key brain regions by combing these two independent modalities.

Hence, our present study intends to explore the alterations in common Rs-fMRI metrics after 52 h of SD (SD52) and after 14 h of recovery sleep (RS14), as well as whether potential changes in these Rs-fMRI metrics are associated with changes in A₁AR availability and behavioral performance. To this end, we first conducted voxel-based statistical comparisons of ALFF/fALFF, ReHo, and DC after SD52 in comparison with RS14. Then, we correlated the differences (SD52 - RS14) in A₁AR availability with the differences in Rs-fMRI metrics of those significant clusters. Lastly, we tested the correlations between the differences (SD52 - RS14) in the above neuroimaging parameters and corresponding neuropsychological performance.

This study systematically performed whole brain voxel-wise comparisons across local Rs-fMRI metrics of spontaneous brain activity (ALFF/fALFF, ReHo, and DC) comparing SD52 to RS14. After acute sleep loss, we identified both significant higher amplitude of low-frequency fluctuations (ALFF) and ReHo values in cuneus/lingual/calcarine. Meanwhile, ALFF values increased significantly in the superior/middle temporal gyrus, whereas they decreased in right cerebellum. Our findings also showed significantly higher degree centrality in bilateral precentral gyrus and postcentral gyrus, but decrease in thalamus, posterior cingulate cortex, right caudate/putamen, and left cerebellum. Furthermore, our correlation analysis suggested an increase in the differences (SD52 - RS14) of A₁AR availability were correlated with a decrease in the differences of mean ALFF/fALFF in left/middle temporal gyrus and degree centrality in left postcentral gyrus. An increase in the differences (SD52 – RS14) of local Rs-fMRI metrics within several brain regions significantly correlated with an impairment in PVT-speed and high feelings of subjective sleepiness, whereas lower A₁AR availability of most brain regions were associated with larger impairments in cognitive performances (PVT-lapses and 3-back hits).

Of these findings, the most prominent one is that high homeostatic sleep pressure synchronized amplitude of low frequency fluctuations of blood-oxygen-level-dependent (BOLD) activity in primary visual cortex, which was reflected by the increased ALFF/fALFF and ReHo in bilateral cuneus/lingual/calcarine after sleep loss. Similar results of higher ALFF values after 24 or 72 h of sleep loss have already been reported in two Rs-fMRI studies (Gao et al., 2015; Wang et al., 2016). Meanwhile, we also noticed the remarkably decreased degree centrality in this region, which is consistent with earlier findings (Farahani et al., 2019; Xu et al., 2021). After reviewing the behavioral consequences in 21 studies, one earlier study (Waters et al., 2018) concluded that sleep restriction mainly deteriorated the visual domains (90% of these studies) in healthy participants, such as metamorphopsia, illusions, and hallucinations. Additionally, one fluorodeoxyglucose (FDG)-PET study (Thomas et al., 2000) identified a decreased cerebral metabolic rate of glucose uptake within this brain area. In the frequency band of Slow-4, our results extended the sleep-loss related changes to the right fusiform gyrus and left inferior occipital gyrus, which were discovered in other fMRI studies as well (Gao et al., 2015; Chen L. et al., 2018). Though the nature and their pathological functions of low frequency bands were not fully identified yet, some brain regions are more sensitive to one of these two different neighboring frequency bands in the neuroimaging studies of human brain (Hutchison et al., 2004; Han et al., 2011; Li et al., 2017). Combined with their correlations with outcomes of PVT, we propose that the upregulated power spectrum of BOLD signals in these visual areas primarily delayed the response speed in the attentional task. Meanwhile, our results revealed increased ALFF/fALFF values in bilateral superior/middle temporal gyrus but attenuated ALFF values in right cerebellum after sleep loss. Using single-neuron recordings in the human neurosurgical patients, one study (Nir et al., 2017) that reported that sleep deprivation induced prolonged and weakened spiking responses of individual neurons in middle temporal gyrus prior to cognitive lapses during a face/non-face categorization PVT. Lastly, the reduced power spectrum of BOLD activity within the right cerebellum, which was in line with a prior study (Chen L. et al., 2018), may reflect the accelerated oxygen consumption in order to sustain movement, emotional and cognitive functions.

Using the index of degree centrality, we identified hyperconnectivity within the sensorimotor cortex (PreCG and PostCG), but reduced strengths in some hub regions of the default-mode network (PCC) and subcortical network (thalamus and CAU/PUT), as well as left cerebellum. After sleep deprivation, earlier study (Gorgoni et al., 2014) observed enhanced excitability with an amplitude increase of somatosensory evoked potentials. Moreover, they found that voltage changes correlated with post-SD fluctuations of subjects’ sleepiness. Meanwhile, the decrease of thalamocortical connectivity after sleep loss has been shown to be critical for attention and arousal regulation in previous studies (Portas et al., 1998; Chee et al., 2008; Tomasi et al., 2009; Shao et al., 2013; Liu et al., 2018). More specifically, thalamic activity was decreasing with lower arousal level during resting-state but conversely elevated when the subjects were required to perform an attention task. Notably, reductions of FC strengths between thalamus and cortical areas were reported under the conditions of coma, general anesthesia, and non-rapid eye movement sleep (Kaufmann et al., 2006; Akeju et al., 2014; Picchioni et al., 2014; Hannawi et al., 2015). Therefore, these findings suggest that the thalamus acts as “control switch” to regulate human brain consciousness states. In the end, the FC strengths within the posterior cingulate cortex, right PUT/CAU of basal ganglia and left cerebellum were declined because of post-SD sleepiness, which was consistent with previous findings (Lazarus et al., 2013; Tomasi et al., 2016). To sum up, higher connectivity strengths within sensorimotor cortex but a decrease in strength in the default mode and subcortical network after sleep loss were closely associated with attentional deficiencies and sleepiness.

Most importantly, our findings highlight the negative associations between A₁AR distribution volumes and Rs-fMRI metrics in mean ALFF/fALFF values of left superior/middle temporal gyrus and degree centrality of left postcentral gyrus. To our knowledge, this is the first study to investigate their associations by combining PET and Rs-fMRI datasets whereas prior work was limited to separate investigations of the brain’s A₁AR availability and BOLD activity (Gao et al., 2015; Elmenhorst et al., 2017). Noticeably, the A₁AR availability of superior/middle temporal gyrus were elevated in terms of some antidepressant therapies in the patients with major depression disorders, such as electroconvulsive therapy, deep brain stimulation, or transcranial magnetic stimulation (Sakagami et al., 2005; Bekar et al., 2008; Kato, 2009; Hamani et al., 2010). These findings may further indicate that the effectiveness of sleep restriction on emotional regulation of depressive patients might be accomplished by the accumulation of A₁ adenosine receptors density, which was correlated to slower oscillations of BOLD activity. With respect to the left postcentral gyrus, the decreases in A₁AR availability are associated with reduced FC strengths between sensorimotor cortex and other brain regions. Additionally, significant associations of A₁AR availability with PVT-lapses and 3-back hits indicate the high A₁AR bindings within these brain areas might be the molecular mechanism of spatial neglects during sustained attention, which was discussed in a previous publication (Elmenhorst et al., 2017).

Nevertheless, the sample size (N = 14) of our study is relatively small and hence restricted the statistical power for detecting between-conditions differences. To solve this issue, we performed Permutation Analysis Linear Model with 5000 repetitions to produce the statistical inferences. Secondly, we should keep in mind that the neuronal or metabolic activity of the human brain after recovery sleep may not be fully restored to rested baseline level (Wu et al., 2006; Chai et al., 2020). Lastly, several PET studies (Wu et al., 1991; Chuah and Chee, 2008; Volkow et al., 2008, 2012) reported that acute sleep deprivation reduced glucose uptake, D2/D3 neurons, but increased cholinergic neurons in some brain areas. Therefore, the complex relationships among the BOLD activity, different types of neurotransmitters and behavioral outcomes should be investigated further.

In our current study, for the first time, negative correlations between A₁AR availability and BOLD activity in the left superior/middle temporal gyrus and left postcentral gyrus of the human brain provide new insights into the molecular basis of neuronal responses induced by high homeostatic sleep pressure.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the Ethics Committee of the Medical Faculty of the University of Düsseldorf. The patients/participants provided their written informed consent to participate in this study.

CL conceived the presented idea, performed the computations, and wrote the manuscript. DE and E-ME designed and carried out the experiments and revised the manuscript. TK, AM, DA, and AB discussed the results and contributed to the final version of manuscript. All authors contributed to the article and approved the submitted version.