Sleep Loss Impacts the Interconnected Brain-Body-Mood Regulation of Cardiovascular Function in Humans

Poor sleep is associated with hypertension, a major risk factor for cardiovascular disease1,2. However, the mechanism(s) through which sleep loss impacts blood pressure remain largely unknown, including the inter-related brain and peripheral body systems that regulate vascular function3. In a repeated-measures experimental study of 66 healthy adult participants, we demonstrate four core findings addressing this question. First, a night of sleep loss significantly increased blood pressure—both systolic and diastolic, yet this change in vascular tone was independent of any increase in heart rate. Second, sleep loss compromised functional brain connectivity within regions that regulate vascular tone. Third, sleep-loss related changes in brain connectivity and vascular tone were significantly inter-dependent, with changes in brain nodes explaining the shift towards hypertension. Fourth, sleep-loss related changes in mood, specifically reductions in positive and amplification in negative states, each demonstrated an interaction with the impairments in brain connectivity and blood pressure. Together, these findings support an embodied framework in which sleep loss confers increased risk of cardiovascular disease through interactions between brain homeostatic control, mood-state and blood pressure.

6 diastolic blood pressure, which also increased significantly following sleep loss ( Fig. 1A; 114 Diastolic: sleep-rested mean = 71.1 mmHg ± 7.3 SD, sleep-deprived mean = 73.7 115 mmHg ± 7.9 SD, p = 0.03). 116 These changes in blood pressure occurred in the absence of an increase in heart 117 rate. Specifically, there was an overall reduction of resting heart rate under conditions of 118 sleep loss, relative to the sleep rested condition (Fig 1B; sleep-rested mean = 67.5 bpm 119 ± 11.6 SD, sleep-deprived mean = 62.9 bpm ± 10.2 SD, p < 0.001), a finding that is 120 consistent with previous reports regarding sleep loss 6,26-29 , but see 30 . 121 Corresponding to the decrease in heart rate, there was an increase in heart rate 122 variability under conditions of sleep deprivation, relative to the sleep-rested condition 123 (Fig 1C; pNN50: sleep-rested mean = 27.5% ± 20.5% SD, sleep-deprived mean = 124 42.0% ± 19.1%, p < 0.0001). Since the pNN50 reflects vagal modulation of the heart 31 , 125 this observed increase in heart rate variability under conditions of sleep loss is 126 consistent with increased parasympathetic input to the heart linked to the state of 127 extended wakefulness 26,28 . 128 Further advancing a dissociation between increased peripheral vascular blood 129 pressure in the absence of accelerated heart rate, there was no significant association 130 between the differential changes in blood pressure and heart rate between conditions 131 ([sleep-rested] -[sleep deprivation]) for either systolic or diastolic measures (Fig 2A,B; 132 Systolic: r = 0.06, p = 0.61; Diastolic: r = 0.03, p = 0.84). A similar lack of association 133 was observed between the sleep-loss changes in heart rate variability and in blood 134 pressure (Fig 2C,D; Systolic: r = -0.09, p = 0.48; Diastolic: r = -0.05, p = 0.69). In 135 contrast, there was a significant association between the sleep-loss related reduction in 136 heart rate and the corresponding increase in heart rate variability (r = -0.48, p < 0.0001). 137 8 138

Sleep Loss & Viscerosensory Resting Brain Connectivity 139
Next, we assessed functional connectivity within the a priori network known to be 140 associated with cardiovascular control in humans and non-human primates, including 141 Figure 2. Associations between cardiovascular outcomes. Scatterplots represent non-significant correlations between sleep-loss changes (deprivedrested) in cardiovascular outcomes, including (A) heart rate and systolic blood pressure, (B) heart rate and diastolic blood pressure, (C) heart rate variability (pNN50) and systolic blood pressure, and (D) heart rate variability (pNN50) and diastolic blood pressure. the amygdala, anterior cingulate, insula, and ventromedial and medial prefrontal 142 cortices 3,9-13 . Within this network, twelve connections were found to be significantly 143 altered following FDR-correction for multiple tests under conditions of sleep deprivation, 144 relative to the sleep rested state (q = 0.05) 32 . These included connectivity pathways 145 between the insula and anterior cingulate, the insula and MPFC, the anterior cingulate 146 and the MPFC, the anterior cingulate and the amygdala, the ventromedial frontal and 147 MPFC, and MPFC and amygdala (Fig 3A,B, Table S1). 148 Of these twelve, there was a significant reduction in connectivity between the 149 ventromedial frontal and MPFC following sleep deprivation, relative to a night of normal 150  We next tested whether these two sets of identified changes-alterations in brain 164 connectivity and alterations in peripheral blood pressure were significantly inter-related. 165 First, greater sleep-loss-related increases in insula connectivity were significantly 166 associated with a greater rise in systolic blood pressure, especially between the insula 167 and the MPFC (r(60) = 0.25; p = 0.05; Fig. 4A). Second, sleep-loss-related decreases in 168 MPFC connectivity with the amygdala was significantly related to increases in diastolic 169 blood pressure (r(61) = -0.29, p = 0.02; Fig. 4B). Therefore, sleep-loss changes in 170 systolic and diastolic blood pressure were associated with alterations in brain 171 connectivity, with a clustering around changes in insular and MPFC connectivity, 172 respectively. Unlike the association between brain and blood pressure, no associations 12 between sleep-loss related changes in heart rate or heart rate variability and changes in 174 functional connectivity were observed (all p > 0.05). 175 Salience Network (VSN), and Somatomotor Network (SMN) 33 . Neither differences in 180 connectivity within or between these large-scale resting-state networks showed 181 significant associations with the sleep-loss related increases in blood pressure, the 182 decrease in heart rate, or the increase in heart rate variability (all p > 0.05). Therefore, 183 local changes within the specific a priori cardiovascular control network 3,9-13 , rather than 184 global changes in larger-scale brain networks 33 , best accounted for changes in 185 peripheral blood pressure. Building on these inter-connected links, we finally tested the hypothesis that 195 changes in mood and anxiety caused by sleep loss would be significantly associated with, 196 rather than independent of, the sleep-loss related changes in cardiovascular state and 197 cardiovascular control brain networks. 198 As expected 38,39 , sleep deprivation was associated with a worsening of mood: a 199 reduction in positive mood (t(65) = 6.8, p < 0.0001) and an increase in negative mood 200 (t(65) = -5.2, p < 0.0001), relative to the sleep rested condition. Sleep loss also resulted in 201 an increase in state anxiety, relative to the sleep rested state (t(65) = -4.3, p < 0.0001). 202 Demonstrating affective specificity, sleep-loss increases in negative mood state 203 were associated with sleep-loss increases in blood pressure, such that the greater the 204 negative mood increase, the greater the increase in diastolic blood pressure, though this 205 was not significant following FDR correction (r(59) = 0.32, p = 0.18; FDR-corrected). 206 Moreover, sleep-loss changes in heart rate and heart rate variability were not associated 207 with changes in any of the affective measures, (all p > 0.24, FDR-corrected), once again 208 demonstrating specificity to changes in vascular control, rather than heart contractility rate. 209 The association between negative mood and systolic blood pressure was not significant 210  Counter to our experimental hypothesis, however, sleep-loss changes in anxiety 220 were not significantly associated with changes in a priori brain vascular control networks 221 (Table S2). 222 Therefore, a dissociable influence of mood was observed, with sleep-loss 223 amplifications in negative mood state selectively associated with the amplifications in 224 peripheral body blood pressure, and the loss of positive mood state was expressly 225 associated with impairments in viscerosensory brain connectivity. Such findings lend 226 further support to an embodied framework of sleep loss in which changes in brain, body, Taken together, the current study demonstrates that (1) experimental sleep loss 232 significantly and selectively increases peripheral blood pressure, independent of any 233 increase in heart rate, (2) sleep loss simultaneously compromises functional brain 234 connectivity within regions that help regulate vascular tone-the viscerosensory network, 235 (3) the sleep-loss related changes in viscerosensory functional brain connectivity and in 236 peripheral vascular tone were significantly inter-dependent, with the changes in brain

Sleep Loss, Cardiovascular State & Resting Brain Connectivity 291
Adding to the findings of sleep-loss related changes in cardiovascular state and 292 brain-network connectivity, we further demonstrate that these changes are significantly 293 associated. First, the increase in systolic blood pressure was associated with increased 294 connectivity between two regions involved in the regulation of vascular tone: the insula 295 and MPFC. Both the insula and MPFC are involved in the representation and generation 296 of visceral states, including cardiovascular reactions 48 . Thus, these two regions, along 297 with a network of corticolimbic and brainstem systems, play instrumental roles in tuning 298 cardiovascular and autonomic adjustments in response to contextual demands 13,42,48 . 299 Indeed, the insula and MPFC are multi-synaptically connected to pre-autonomic nuclei 300 which innervate both the heart and vasculature 49,50 . There is also evidence that 301 increased resting activity in the insula and MPFC predicts greater evoked blood pressure 302 reactions, interpreted as signaling increased sympathetic-mediated blood pressure 303 reactivity 48 . 304 A different part of the cardiovascular control network was observed to be 305 associated with the sleep-loss increase in diastolic blood pressure. Here, the magnitude 306 of sleep-loss increase in diastolic blood pressure was significantly associated with 307 decreased connectivity between the MPFC and amygdala. The amygdala is instrumental 308 in coordinating behavioral and physiological adjustments through connections to cortical, 309 hypothalamic, and brainstem nuclei involved in stress processing and autonomic-310 cardiovascular control 48 . Additionally, the amygdala can impact blood pressure through 311 influence over the baroreflex, which serves to maintain blood pressure through 312 autonomic regulation of heart rate, cardiac output, and vascular resistance, and is 313 associated with preclinical atherosclerosis 9,51 . Thus, the reduction in amygdala 314 connectivity to other regions involved in cardiovascular control (MPFC) could represent a 315 degree of disorganization of brain homeostatic functions. 316 The vascular system is subject behavioral, autonomic, and endocrine 317 regulation 52 . The complexity of these interacting pathways suggests that altered blood 318 pressure control is due to a failure of a central control system, namely the brain 53 . 319 Notably, in the current data, the decrease in heart rate was not associated with the 320 increase in blood pressure. Furthermore, blood pressure-related changes in functional 321 connectivity in the brain were similarly not associated with heart rate. Taken together, it 322 is likely that multiple, perhaps counteractive, processes underlie the conflicting vascular 323 and cardiac responses to sleep loss, fitting with a loss of central brain regulation.

Experimental Model and Subject Details 362
Sixty-six healthy adults aged 18-24 years (mean ± SD: 20.7 ± 1.7, 52% female) 363 completed a repeated-measures cross-over design (described below). There was no 364 influence of sex on cardiovascular outcomes (p > 0.64). Participants abstained from 365 caffeine and alcohol for 72 hours before and during the entire course of the study and 366 kept a normal sleep-wake rhythm (7-9 hours of sleep per night) with sleep onset before 367 1:00 A.M. and rise no later than 9:00 A.M. for the three nights before the study 368 participation, as verified by sleep logs and actigraphy. 369 Exclusion criteria, assessed using a prescreening semi-structured interview, 370 were as follows: a history of previously-diagnosed sleep disorders, neurological 371 disorders, closed head injury, Axis 1 psychiatric disorders according to the DSM-IV-TR 372 criteria (encompassing mental disorders, including depression, anxiety disorders, bipolar 373 disorder, attention deficit disorder, and schizophrenia), history of drug abuse and current 374 use of antidepressant or hypnotic medication, nicotine use, consumption of more than 375 five alcoholic drinks per week, crossing of time zones in the 3 months before the study, 376 and general contraindications to MRI. Participants also had blood pressures within 377 normotensive ranges, confirmed with a blood pressure reading of less than 120/80 378 mmHg at the time of consenting 55 . Participants who reported sleeping <7 hours per night 379 or consuming three or more daily caffeine-containing drinks were also excluded from 380 entering the study. The Pittsburg Sleep Quality Index 56 was employed to determine the 381 quality of recent sleep history, and participants considered poor sleepers (global score > 382 5) were excluded. The study was approved by the UC Berkeley Committee for the 383 Protection of Human Subjects, with all participants providing written consent. 384

Experimental Design 386
To test the experimental hypotheses, participants entered a repeated-measures 387 study design (Fig 1), including two sessions-one following a normal night of sleep and 388 one after 24 hours of total sleep deprivation. The two sessions were separated by at 389 least 7 days (mean ± SD: 9.8 ± 3.8), with the order of the sleep-rested and deprived 390 conditions counterbalanced across subjects. 391 In the sleep deprived session, participants arrived at the laboratory at 10:00 P.M. In each condition, cardiovascular state was assessed to determine blood 406 pressure, heart rate, and heart rate variability. These assessments were performed 407 approximately 1 hour after awakening, and at circadian-matched times in the sleep 408 recording with rested cuffed arm at heart height, consistent with standard blood pressure 414 recording guidelines 59 . Two consecutive measurements were taken and the mean of 415 these two measurements was used for analyses. 416 Immediately following blood pressure assessment, participants were fitted with 417 electrodes for bipolar electrocardiography (ECG), and cardiac activity was recorded for 418 five minutes. Electrodes were applied in a lead II format with the reference electrode 419 placed below the right clavicle parallel to the right shoulder and a second electrode 420 placed on the torso at the fourth intercostal space on the left side parallel to the left hip.