Altered sleep intensity upon DBS to hypothalamic sleep–wake centers in rats

Abstract Deep brain stimulation (DBS) has been scarcely investigated in the field of sleep research. We hypothesize that DBS onto hypothalamic sleep- and wake-promoting centers will produce significant neuromodulatory effects and potentially become a therapeutic strategy for patients suffering severe, drug-refractory sleep–wake disturbances. We aimed to investigate whether continuous electrical high-frequency DBS, such as that often implemented in clinical practice, in the ventrolateral preoptic nucleus (VLPO) or the perifornical area of the posterior lateral hypothalamus (PeFLH), significantly modulates sleep–wake characteristics and behavior. We implanted healthy rats with electroencephalographic/electromyographic electrodes and recorded vigilance states in parallel to bilateral bipolar stimulation of VLPO and PeFLH at 125 Hz and 90 µA over 24 h to test the modulating effects of DBS on sleep–wake proportions, stability and spectral power in relation to the baseline. We unexpectedly found that VLPO DBS at 125 Hz deepens slow-wave sleep (SWS) as measured by increased delta power, while sleep proportions and fragmentation remain unaffected. Thus, the intensity, but not the amount of sleep or its stability, is modulated. Similarly, the proportion and stability of vigilance states remained altogether unaltered upon PeFLH DBS but, in contrast to VLPO, 125 Hz stimulation unexpectedly weakened SWS, as evidenced by reduced delta power. This study provides novel insights into non-acute functional outputs of major sleep–wake centers in the rat brain in response to electrical high-frequency stimulation, a paradigm frequently used in human DBS. In the conditions assayed, while exerting no major effects on the sleep–wake architecture, hypothalamic high-frequency stimulation arises as a provocative sleep intensity-modulating approach.


Introduction
The regulation of physiological sleep and wakefulness relies on the equilibrium of a well-explored network of wake-and sleep-promoting centers in the brain. Mutual inhibition between these centers has been proposed as the main regulatory mechanism in the flip-flop switch model [1,2], where effective sleep requires the suppression of arousal systems by mainly the inhibitory GABAergic ventrolateral preoptic nucleus (VLPO) in the anterior hypothalamus [3,4], whose activation via agonism of adenosine receptors increases nonrapid eye movement (NREM) sleep [5,6]. Additional inhibition of wake-promoting neurons by adenosine builds upon this neurotransmitter's sleep-promoting action [7]. As a counterpart, the perifornical area of the posterior lateral hypothalamus (PeFLH) is densely populated, among other cell populations, by hypocretin (orexin) neurons, implicated in the facilitation of arousal [8]. The PeFLH region may also be an important organizer of the normal succession of stages during the sleep-wake cycle through excitatory reciprocal feedback to monoaminergic centers including the locus coeruleus (LC), a major arousal hub [8].
The complex regulation of sleep and wakefulness by counteracting and heterogeneous hypothalamic and brainstem nuclei, including VLPO and PeFLH, has been demonstrated in the last decades using stereotaxic injections of toxins/receptor agonists, as well as by opto-and pharmacogenetics [9][10][11][12]. Such approaches have been very useful and successful in specifically targeting distinct neuronal populations within heterogeneous brain areas and modulating sleep phenotypes in animals, thus identifying the basic nature of the wake-and sleep-controlling nuclei. However, while precise information is obtained from these approaches when they are acutely applied, their neuromodulatory effects during long-term (hours, days) applications have not been characterized. Additionally, this molecular approach presents some practical limitations toward cell types and species [13], with restricted genetic tools and inaccessibility to larger brains [14], which joins other factors such as the irreversibility of the genetic manipulation and the unknown long-term consequences of viral expression and potential protein accumulation, as caveats for eventual clinical applications.
Overall, clinically accessible approaches, such as electrical deep brain stimulation (DBS), have to be explored to determine whether targeting sleep-wake-controlling brain areas is a valuable therapeutic strategy offering significant neuromodulatory effects.
Here, we aimed to study the effects of high-frequency electrical neuromodulation of the VLPO and PeFLH nuclei, as representative sleep-wake-controlling hypothalamic targets, to determine whether these regions can serve as valuable putative therapeutic targets in further animal and human studies on the treatment of severe, drug-refractory sleep-wake disturbances. We thus tested the effect of continuous electrical high-frequency stimulation (HFS, often implemented in clinical practice for reversible functional ablation of the targeted nuclei; [15]) on the sleep-wake behavior, stability and intensity as assessed by electroencephalographic/electromyographic (EEG/EMG) recordings in light, dark, and 24 h periods. We hypothesize that HFS of the VLPO area will decrease NREM sleep proportion, stability and/or intensity by functionally inhibiting its sleeppromoting action; whereas HFS of the PeFLH region will decrease wakefulness proportion, stability and/or intensity by functionally disabling the arousal/wake-promoting nucleus' function.

Experimental design
A total of 14 animals were chronically instrumented with EEG/EMG headsets and DBS leads into the VLPO, from which 5 animals were excluded from the analysis due to histological determination of mistargeting of the DBS leads, leaving 9 available animals. In those 9 animals, we performed sleep-wake proportions, fragmentation and delta power analyses upon HFS in light, dark and 24 h periods. For each analysis, statistical outliers, defined as cases with behavioral scores >2 standard deviations from each overall group mean and/or animals presenting technical issues with the EEG/EMG, were additionally excluded from the analysis. Thus, for the VLPO, the remaining number of animals analyzed was n = 7 during the light period (1 statistical outlier and 1 technical failure in entire EEG) and n = 6 during the dark period (1 statistical outlier, 1 technical failure in the entire EEG and 1 technical issue during the dark period) for all parameters.
A total of 8 animals were chronically instrumented with EEG/EMG headsets and DBS leads into the PeFLH, from which 1 animal was excluded from the analysis due to histological determination of DBS leads mistargeting. In the remaining 7 animals, we analyzed sleep-wake proportions, fragmentation and delta power analyses also in light and dark periods and per 24 h. No statistical outliers were found in the analyses, whereas again we excluded one animal presenting technical issues in the entire EEG/EMG. Thus, in the PeFLH, the remaining number of animals analyzed was n = 6 for all parameters.

Animals
We included adult male Sprague-Dawley rats (Charles River Laboratories International Inc, Germany) weighing between 290 and 380 g at the time of surgery. The rats were housed individually on a 12:12 light/dark cycle with food and water available ad libitum throughout the experiments. In compliance with ethical regulations and to explore potential side effects that would cause eventual protocol discontinuation, we monitored animals' weight daily while in EEG/EMG recordings/DBS sessions as well as a home-cage activity after interventions. We additionally checked body temperature before and after the interventions. The animal room temperature was constantly maintained at 21-24°C.
Ethical approval: All experiments were approved by the veterinary office of the canton of Zurich and conducted according to the local guidelines for care and use of laboratory animals under license ZH205/12.

Surgical procedures for EEG/EMG and DBS electrode implantation
We anesthetized rats with isoflurane (4.5% for induction, 2.5% for maintenance) and injected them with buprenorphine (0.05 mg/kg, s.c.) for analgesia. We monitored wound healing, body weight and home-cage activity of the animals on a daily basis over the first week after surgeries and weekly thereafter. We performed the EEG/EMG and DBS implantation procedures using adapted versions from previously established protocols [16,17]. Briefly, we positioned the animals in a standard stereotactic apparatus (model 1900, Kopf Instruments, Tujunga, CA, USA) over a temperature-controlled pad, and made a midline incision exposing the skull. We made burr holes over the position of DBS coordinates followed by durotomy. Before insertion of the DBS electrodes, we placed two anchoring screws in the front (M1, one per hemisphere) and inserted four screws for peridural EEG recording, one pair per hemisphere over the parietal cortex. Additionally, we inserted a pair of gold wires into the rats' neck muscles, which served as EMG electrodes for monitoring muscle tone. All EEG/ EMG electrodes were connected to a plug by soldering to stainless steel wires.
We implanted DBS electrodes on targets after adjustment of the initial coordinates: (i) VLPO (AP −0.04 mm, ML 0.8 mm, DV −10 mm) using a factor obtained by dividing the measured individual Bregma-Lambda distance by the reference measure from the Atlas [18] and (ii) PeFLH (AP −2.9 mm, ML 1 mm; DV −9 mm) with AP −2.7 mm applied for measured Bregma-Lambda distance <6.9 mm and AP −3.1 mm applied for measured Bregma-Lambda distance ≥ 6.9 mm. Finally, we cemented all headpieces to the skull as illustrated in Figure 1.

DBS: hardware and characteristics
In our experimental setup, we used bilateral concentric bipolar DBS electrodes as previously described [16], stereotactically implanted into the VLPO and PeFLH. The goldplated electrodes, composed of an inner platinum-iridium wire, functioning as the negative pole and an outer stainless steel layer as the positive one, produce a concentrated current field around the tip of the electrode. The maximum outer diameter of the electrode is about 250 µm with a tip diameter of approximately 50 µm (Technomed, Beek, The Netherlands).
We used a stimulus of rectangular shape in a currentcontrolled paradigm and applied a bipolar biphasic (60 µs negative -60 µs positive) stimulation, which bears the advantage of potential translation into clinical studies.

Selected DBS parameters
Based on the most commonly used parameter in the literature, we selected 125 Hz as the frequency of stimulation, fixing the negative pulse width at 60 μs as regularly used in clinical studies. We based the choice of the pulse amplitude on a priori simulations aiming at investigating the distribution of the electrical field in brain tissues [19]. We modeled the electrode position in the VLPO, as a proxy structure due to its relatively small and well-defined anatomy. For this, we used the final outcome of the finite element method of modeling and simulation, a widely applied numerical technique for calculating approximate solutions of general partial differential equations [19]. The model integrated the electrode configuration (top vs center) and aspects of the local structure of the surrounding region. We split the region of interest into 4 quadrants and neglected overlapping effects of the resulting borders. We applied a physicscontrolled mesh and set the density with an application of a spatial resolution of 1 µm for the region close to the electrode (0-0.15 mm). For the distal region (0.15-10 mm), we used a spatial resolution of 10 µm. We set the tissue conductivity to 0.07 S/m (within common frequency ranges), at body temperature and of bovine origin due to lack of data for the rat [20] and the pulse width to 0.06 ms. Moreover, we accommodated for dielectric tissue properties represented by optic chiasm -och-(medial to the left VLPO) and median forebrain bundle -mfbb-(lateral to the left VLPO). We assessed the distance of mfbb and och by taking the average of two brain slices at the AP level 0.00 and −0.12 mm from bregma [18], where we preferentially targeted the VLPO. We performed simulations at 40, 90 and 150 µA and observed no summation effects (i.e., absence of rest-energy when the second pulse is applied). Accordingly, we neglected the effects of frequency. To increase the comparability between the different simulations using either 40, 90 or 150 µA, the absolute field strength (3.75 V/mm) was kept constant. The simulations were based on relative (and not absolute) electrical field strengths.

Experimental protocols and setup
We single-housed the rats following surgery and granted recovery of at least 2 weeks to all animals before further interventions. We connected the DBS electrodes to a stimulation device (model STG4008-1.6 mA, Multi Channel Systems MCS GmbH, Reutlingen, Germany) in parallel to EEG/EMG electrodes through externalized cables that hang from a rotating swivel at the top of the cage, allowing free motion of the animals inside the experimental cage ( Figure 1). We monitored stimulus delivery using an oscilloscope. Following a setup adaptation period of 1-2 days, we conducted EEG/EMG recordings for 24 h during two consecutive days, one before (DBS OFF) and one during (DBS ON) continuous bilateral electrical stimulation (Figure 2a).

EEG/EMG recording, scoring and analysis
We sampled EEG and EMG at 200 Hz and amplified and processed the signals by an analog-to-digital converter. We used EMBLA hardware and Somnologica-3 software (Medcare Flaga). We discarded activity in the 50 Hz band from the analysis because of power line artifacts. We obtained power spectra of the EEG by discrete Fourier transformation (range: 0.5-100 Hz; frequency resolution: 0.25 Hz; time resolution: consecutive 4 s epochs; window function: Hanning).
We performed blinded visual scoring of acquired EEG manually as described previously [17]. We excluded artifacts by visual review of the raw data and identified and scored three vigilance states based on EEG/EMG patterns: wakefulness (WAKE), NREM sleep and rapid eye movement (REM) sleep. We divided the 24 h scoring sessions into light and dark periods or into 1 h intervals and assessed the proportion of the vigilance states separately for each period. To provide a quantitative measure of sleep fragmentation, we calculated the sleep fragmentation index as follows: a behavioral state bout was defined as a consecutive series of epochs in the same behavioral state without state transitions. The resulting amount of behavioral state bouts was then divided by the total number of 4 s epochs in the same sleep stage, resulting in a comparable measure for fragmentation between 0 and 1 [17,21]. We also determined the total delta power of NREM sleep, calculated as the summarized power in the slow-wave activity (SWA) band (0.5-4 Hz), before and during stimulation. We performed all signal processing and analyses as described using MATLAB (MathWorks).
Furthermore, we analyzed the buildup of delta power (relative delta power) upon transition into consolidated NREM sleep as described earlier [22]. To this end, we identified all consolidated NREM episodes lasting longer than 6 min. This value was determined based on established criteria (7 min [23]) and adapted empirically to ensure a significant amount of consolidated sleep bouts in the analysis (6 min time frame).

Electrode placement confirmation
Upon completion of the experiments, we sacrificed the rats via intracardiac perfusion as previously described [24]. We verified correct electrode placement and visualized potential tissue damage due to the stimulation by hematoxylin/eosin stainings in coronal 40 µm fixed brain sections (Figure 2b-g) [18].

Statistical analyses
We expressed the light, dark and 24 h proportions of vigilance states, and the logarithm of delta power as medians and quartiles with 95% confidence intervals (CI), while fragmentation and data presented in 1 h intervals were plotted as mean ± S.E.M. The full spectrum EEG data were reported as absolute values. Bivariate comparison of delta power and sleep fragmentation ON vs OFF stimulation was done by Wilcoxon's signed-rank tests and Bonferroni corrections (n = 4 tests). A two-way analysis of variance (ANOVA) with repeated measures (factor band -4 levels: alpha, beta/gamma, delta, theta; factor DBS -2 levels: OFF, ON) was used for the statistical assessment of the full spectrum EEG data for the three vigilance states and for the analysis of delta-buildup in time, followed by Bonferroni post-hoc tests as appropriate. One-way ANOVA with repeated measures (factor: hour -24 levels: 12 h OFF vs ON) was additionally applied within each light and dark period, for the detailed analysis of the time course of the vigilance states and delta power, followed by Student-Newman-Keuls post-hoc tests as appropriate. We used a threshold for statistical significance of P ≤ 0.05. We performed all statistical analyses using StatView ® (SAS Institute Inc., USA) and R (Team).

No distinct side effects or tissue damage upon DBS
Importantly, we did not observe side effects with respect to body weight, body temperature and/or overall home cage behavior (data not shown) in association with our interventions. Evaluation of hematoxylin/eosin staining revealed no significant tissue damage, besides the electrode tracks, in or around the target structures ( Figure 2).

Distribution of the electrical field in the VLPO
To determine the optimal pulse amplitude to be applied during the VLPO electrical stimulation, we investigated the distribution of the electrical field in the brain tissue pre-experimentally, based on computational simulations. We modeled DBS electrodes either at the top or the center of the VLPO to cover the main anticipated configurations of electrode placements. The final outcome of the simulations revealed distributions of the electrical fields beyond the target region at the three intensities of interest for both top and center configurations (Figure 3). Indeed, starting at 40 µA, we observed a leakage outside the VLPO borders reaching up to 20% of the amplitude of the stimulus in the central configuration of the electrodes with, yet, a restricted stimulation area of the VLPO. Electrical field distribution in the top configuration at either selected intensities (40, 90 or 150 µA) showed an important leakage toward the dorsal region above the VLPO. Stimulation at 90 and 150 µA would target 50 to 100% of the VLPO volume, depending on the position of the electrodes (i.e., top or center). In comparison to 90 µA, however, at 150 µA, the boundaries of the main surrounding structures (och and mfbb) would largely be affected by stimulation. Still, even at lower intensities of stimulation, the modeling results indicate that we cannot avoid affecting, at different degrees, the surrounding structures of VLPO, predicting that targeting a specific region without any current leakage is unlikely. Thus, we chose to stimulate at 90 µA to minimize the main leakage effects while substantially targeting the VLPO. We also chose this amplitude of stimulation for the PeFLH area, a relatively bigger area than VLPO, assuming we will be further limiting leakage effects.

DBS modulation effect on the sleep intensity
Following no changes in the sleep-wake behavioral patterns and along the hypothesis that DBS could alternatively have specifically modulated the sleep intensity, we further analyzed temporal changes of delta power (i.e., density in the delta frequency band during high-amplitude low-frequency oscillatory activity, which mirrors the  Figure 6b).
To get more insights into the 24 h time course of delta power changes at HFS, we analyzed the measure in 1 h intervals during both light and dark periods. No specific time windows during the light or dark periods of neither VLPO (Figure 6c) nor PeFLH (Figure 6d) DBS effects were detected (one-way repeated measures ANOVAs followed by nonsignificant Student-Newman-Keuls post-hoc comparisons as appropriate; P >0.05).
To further explore the effect of VLPO and PeFLH HFS on the sleep intensity dynamics [23], we calculated the buildup of delta power during consolidated NREM episodes (≥6 min; Figure 6e and f). We observed no significant effect from neither VLPO nor PeFLH stimulation on delta power buildup over consolidated sleep (two-way repeated measures ANOVA, P >0.05), indicating that changes in delta power were global and not related to its dynamics.
We further explored alterations in the 24 h average full EEG spectrum upon HFS in both target regions, for which we calculated EEG power spectra OFF vs ON stimulation for all 3 vigilance states (Figure 7). Despite some variability observed in the EEG spectrum, notably in the delta activity of WAKE, multiple bivariate comparisons for all frequency bands (delta, theta, alpha, beta/gamma) did not reveal significant band-specific changes upon HFS (two-way ANOVA with repeated measures followed by Bonferroni post-hoc comparisons as appropriate, P >0.05).
Overall, the HFS of VLPO increased the delta power, whereas the HFS of PeFLH decreased the delta power over light and dark periods of recordings.

Discussion
In this report, we investigated the direct effects of sleepwake behavior and characteristics of high-frequency DBS on sleep-and wake-controlling centers in healthy animals. Our findings suggest that 24 h DBS of the VLPO at 125 Hz modulates sleep-wake characteristics mainly by deepening slow-wave sleep (SWS) as measured by an increase in the delta power (36%), while sleep architecture and fragmentation remain unaffected. In other words, the intensity but not the amount of sleep or its stability is enhanced. Similarly, the proportion of vigilance states remained unchanged during 24 h DBS of the wake-promoting PeFLH region but, in contrast to VLPO, stimulation at 125 Hz depressed the delta power (30%), weakening SWS. Furthermore, the sleep microstructure, as measured by the delta buildup over time, was unchanged upon DBS in both targets. Moreover, based on the marked, yet the nonsignificant effect on delta activity upon VLPO DBS, we cannot rule out a behavioral state-specific effect on SWA selectively in wakefulness. This would indicate that VLPO DBS might have a sleep-inducing rather than a sleepenhancing effect by acting more on the sleep-inducing permissive neurons rather than on the sleep-maintenance executive neurons [26]. However, due to the small effect and sample size, these conclusions remain speculative. The negative results of the detailed spectral power analysis in each behavioral state and the buildup analysis of delta power during consolidated SWS (or NREM) episodes, demonstrated that DBS-elicited changes were global in nature and did not affect the sleep intensity dynamics.

Electrical neuromodulation of VLPO and PeFLH areas
The VLPO has been reported to present a low-frequency intrinsic firing pattern of ∼10 Hz [27]. Thus, detecting increased delta power during NREM sleep with HFS in the VLPO was rather unexpected, given that HFS is empirically known to produce a functional inhibition of the target nucleus. However, this initial oversimplification has been actively disputed, and HFS appears to rely on more complex and multifactorial mechanisms [15,28,29]. DBS may notably reduce cellular activity while concurrently exciting the axons of the stimulated neurons, which was shown in both humans and animals [30][31][32]. Interestingly, excitatory stimulation (i.e., increased output) in parts of the basal ganglia in nonhuman primates at HFS has shown to excite both glutamatergic [33] and GABAergic [34] efferent neurons. Therefore, it is conceivable that the cell bodies of the VLPO neurons were inhibited by DBS while the output from this GABAergic nucleus increased simultaneously, inhibiting the downstream arousal systems. VLPO sleep-active neurons have also been shown to progressively increase their firing rate along with sleep depth in rats [35]. The increased intensity of SWS of about 36% upon HFS revealed in our experiments may thus rely on this neurophysiological property. Indeed, HFS could further increase the discharge rate of VLPO sleep-active neurons, based on the synchronization of the target neuronal firing to the stimulus frequency as shown in other nuclei [33], and consequently, increase the depth of sleep, as observed in this study.
PeFLH HFS, on the other hand, decreased the delta power during NREM sleep by approximately 30%, suggesting an excitatory effect of HFS on this wake-promoting region. Once more, the apparent activation of the PeFLH region at, supposedly inhibitory, HFS is unexpected. Nevertheless, activation of the tuberomammillary nucleus (TMN), a neighboring wake-promoting target with similar firing rates to PeFLH arousal-related neurons, upon HFS (100 Hz) has been already demonstrated in rodents [36]. Moreover, the PeFLH is crossed by the mfbb, carrying projections from groups of neurons critical for the sleep-wake control [37], which unintended stimulation could additionally compound in the observed effects. Although the exact neurophysiological mechanisms sustaining these results remain speculative, comparable processes (i.e., activation of neuronal processes rather than cell soma) in VLPO and PeFLH regions may have been involved, increasing the output of these simplistically regarded sleep-and wake-promoting regions by excitation of efferent neurons.
Noteworthily, however, a decrease in SWA has been previously observed by electrical stimulation of the PeFLH neurons [8]. The delivery of trains of electrical stimuli at 50 Hz in the PeFLH area of anesthetized rats increased the mean firing rate of LC neurons and induced activation of the EEG shown by a decrease in the proportion of delta waves together with an increased percentage of faster (>4 Hz) waves. This result, although obtained under anesthesia as opposed to our awake, freely moving rats, reveals a direct role of the PeFLH neurons in the modulation of delta power, in line with our findings.

Lack of behavioral modulation with VLPO and PeFLH DBS: potential compensatory mechanisms
Although VLPO is an important sleep-promoting center and PeFLH is an organizer of wakefulness/sleep stages, the electrical modulation of these regions did not change the proportions of sleep-wake stages or the fragmentation of behavioral states in response to HFS. In this line, we cannot exclude a possible DBS effect on regions neighboring VLPO, predominantly populated by wakeactive cells in the lateral preoptic/anterior hypothalamic area [35], and/or on the main monoaminergic arousal projections from LC, raphe nuclei and TMN to VLPO, which could counteract the effect of stimulation on VLPO sleepactive neurons. Similarly, we cannot exclude the activation of the main inhibitory afferents (i.e., VLPO and median preoptic nucleus) to the PeFLH region upon DBS that could have counterbalanced the stimulatory effects of this conceptualized wakepromoting area. Also, stimulation of a subpopulation within the heterogeneous PeFLH, the melanin-concentrating hormone (MCH) neurons, known to promote sleep [38], might have played a role. Indeed, these neuronselectrically silent in the absence of synaptic activityshow pronounced firing at 100 Hz upon brief repeated current injections in rodent brain slices [39,40]. Strikingly, the application of high levels of orexin peptides excited some MCH neurons in vitro [41]. In this context, possible overexcitation of orexin neurons by electrical stimulation may have in turn stimulated MCH neurons as a feedback mechanism to prevent hyperarousal and, consequently, overall changes in sleepwake amounts.

Limitations
The most challenging part of any DBS study is identifying the optimal combination of parameters producing a targeted effect while limiting side effects, for which the lack of conclusive dose-response assessments for each parameter analyzed (frequency, intensity and duration) is one of the main limitations of our study. Also, despite the use of bipolar electrodes producing a concentrated current around the tip of the electrodes, our study expectedly suffers from a fundamental DBS limitation which, unlike other highly specific approaches such as optogenetics [42], lacks selectivity between the activation of local cells and that of passing axons using conventional symmetrical biphasic pulses, as illustrated in computational models with monopolar electrodes [43]. Additionally, cell heterogeneity within a target constitutes an intrinsic biological limitation of DBS. For instance, electrical stimulation of PeFLH, whose activation has been shown to produce changes at feeding and motor activity [44], could have additionally affected another subpopulation besides the wake-promoting hypocretin neurons, such as MCH neurons, known to discharge in a reciprocal manner to orexin neurons across the sleep-wake cycle [45,46] and promote sleep [38]. Nevertheless, DBS remains to be a preferred tool of clinical choice, as opposed to approaches involving invasive mutagenic and viral strategies. For this, its exploration remains remarkably important in the context of the search for new therapeutic targets.
A limitation of technical nature is the use of relatively short pulse width in our study. Longer pulse widths (in the range of milliseconds) could be used in the future to preferentially stimulate the cell bodies of our targets [47]. However, increasing the pulse width would also increase the charge density (i.e., the amount of electrical charge per surface area) accumulated per pulse, thus potentially reaching safety limits and causing tissue lesions, which we avoided in our study.

Conclusion
Our investigation has provided new insights into DBS modulation of important centers of the sleep-wake-regulating network. In summary, our unexpected results suggest that tight compensatory mechanisms counteracted the intended changes in the sleep-wake behavior in our healthy rats. However, escaping the tight regulatory controls, sleep intensity is specificallybut contrary to predictedmodulated by hypothalamic DBS, shedding light on nonacute neuromodulation outputs of hypothalamic high-frequency electrical stimulation. Overall, although our results indicate that clinical implementation of HFS in hypothalamic centers for the treatment of disabling sleepwake behavior disorders should eventually be explored with the utmost caution, they encourage new avenues of research intended to further explore and understand the findings, for instance, by means of implementation of hypothalamic HFS in animal models of sleep-wake disorders. Such models would benefit from future study designs, in which DBS outputs of prolonged stimulation and additional relevant frequencies of stimulation can be explored in an intrinsically deficient system.