Genetic lesioning of histamine neurons increases sleep–wake fragmentation and reveals their contribution to modafinil-induced wakefulness

Abstract Acute chemogenetic inhibition of histamine (HA) neurons in adult mice induced nonrapid eye movement (NREM) sleep with an increased delta power. By contrast, selective genetic lesioning of HA neurons with caspase in adult mice exhibited a normal sleep–wake cycle overall, except at the diurnal start of the lights-off period, when they remained sleepier. The amount of time spent in NREM sleep and in the wake state in mice with lesioned HA neurons was unchanged over 24 hr, but the sleep–wake cycle was more fragmented. Both the delayed increase in wakefulness at the start of the night and the sleep–wake fragmentation are similar phenotypes to histidine decarboxylase knockout mice, which cannot synthesize HA. Chronic loss of HA neurons did not affect sleep homeostasis after sleep deprivation. However, the chronic loss of HA neurons or chemogenetic inhibition of HA neurons did notably reduce the ability of the wake-promoting compound modafinil to sustain wakefulness. Thus, part of modafinil’s wake-promoting actions arise through the HA system.

Another compound that enhances wakefulness, but whose mechanism of action is less clear, is modafinil. Modafinil, as first recognized by Jouvet and colleagues [20,21], is an intensively wake-promoting substance with surprisingly few adverse effects [22]. Under controlled laboratory conditions, for example, human subjects given modafinil can stay continuously awake for 2 days and one night (40 hr) [23]. The drug is approved by the US Food and Drug Administration for counteracting daytime sleepiness during narcolepsy [22,24], shift-work sleep disorder, and obstructive sleep apnea/hypopnea syndrome [22]. Modafinil is sometimes used by service personnel, and unofficially taken (e.g. by students) for cognitive enhancement [25].
We still do not have a full understanding of how modafinil promotes wakefulness. Because modafinil cannot promote wakefulness in dopamine transporter (DAT) knockout (ko) mice [26], this seems convincing evidence that the dopamine transporter (DAT) [26][27][28] is critical for modafinil's action. Modafinil antagonizes the DAT transporter and promotes a rise in dopamine levels in the basal ganglia and noradrenaline levels in the prefrontal cortex [27]. Mice with disruptions of their D1 and D2 receptor genes are insensitive to modafinil [29]. Further evidence that modafinil requires dopamine for its effects comes from humans who are homozygotes for a version of the geneencoding catechol-O-methyltransferase that is less effective in degrading dopamine [23]. In these subjects, modafinil does not enhance wakefulness, presumably because these subjects already have high levels of dopamine [23]. Additionally, modafinil increases serotonin and HA levels in the neocortex [22] and HA levels in the anterior hypothalamus [30]. However, with all these changes, and even if the initial changes are produced by modafinil acting at the dopamine transporter, there is a complicated circle of cause and effect, and it is still unclear how modafinil works at the circuit level [27].
As seen by c-Fos expression, modafinil causes widespread excitation throughout the rodent brain, although certain nuclei such as the preoptic hypothalamic area do not show increases in c-Fos expression [31]. In cats, modafinil, immediately after administration, produced c-Fos expression mainly restricted to the anterior hypothalamus, with little expression anywhere else in the brain [32]. A few attempts at lesioning discrete brain regions have been used to try and locate a specific brain nucleus involved in modafinil's actions. From this approach, modafinil does not appear to work by counteracting the sleeppromoting circuitry of the preoptic hypothalamus. Lesions of the ventrolateral preoptic area, whilst increasing wakefulness, have no effect on modafinil's ability to further promote wakefulness [33]. It was also suggested that modafinil acts through the nucleus accumbens core [34]. Nucleus accumbens core lesions substantially increased the amount of time mice were awake, but modafinil did not increase wakefulness above this higher baseline level [34]. However, it could well be that capacity for arousal was already saturated in these accumbens-lesioned mice, so it is unclear if modafinil could have increased it further.
Giving modafinil systemically to mice excites orexin and HA neurons as evidenced by c-Fos expression [35,36]. Thus, there is a possibility that some of modafinil's effects are through the HA or orexin system, in addition to the dopamine system. Lesions of orexin neurons, however, actually increased the sensitivity of mice to modafinil [31]. This still leaves open the possibility of HA's involvement. However, low doses of modafinil can still promote wakefulness in mice with no HA production (i.e. hdc ko mice) [37], suggesting no direct involvement by HA. Nevertheless, some HA neurons also corelease GABA and possibly dopamine [16,38]. Thus, the hdc ko mice will leave still functional "histamine" neurons that could release other substances.
In this paper, we use mice with specific genetic lesions of their HA neurons to examine first how this lesion affects baseline (chronic) sleep-wake behavior, and second how it affects modafinil's arousal-promoting abilities.

Mice
Experiments were performed in accordance with the UK Home Office Animal Procedures Act (1986); all procedures were approved by the Imperial College Ethical Review Committee. The mouse line used, HDC-ires-Cre (JAX labs Stock 021198), predominantly a C57/BL6J background was generated in our laboratory and described previously [39]. All mice used in the experiments were adult male. Mice were maintained on a reversed 12:12 hr light:dark cycle at constant temperature and humidity with ad libitum food and water.

EEG analysis and sleep-wake behavior
Two days before the recording, mice were attached with mock Neurologgers and then fitted with Neurologger 2A devices [43]. Two electroencephalograph (EEG) and two electromyograph (EMG) channels for each mouse were recorded. Spike2 (version 7.10) was used to analyze the sleep (EEG/EMG) data. The sampling rate was set up to 200 Hz. EMG was filtered by band pass between 5 and 45 Hz. EEG frequency was high-pass filtered at 0.5 Hz. The sleep states (wake, W; nonrapid eye movement, NREM, N; rapid eye movement, REM, R) were scored automatically and manually corrected. For the power spectrum analysis of control and HDC-Casp3 mice, delta power (0.5-4 Hz) or theta power (4-8 Hz) was calculated during wakefulness, NREM sleep or REM sleep, respectively, during the 12 hr "lights on" (the "sleep" period) or 12 hr "lights off" (the "wake" period). To analyze the EEG power spectrum for NREM sleep, NREM sleep was assessed for 1 hr beginning with the first NREM bout that occurred after CNO injection, modafinil injection, or sleep deprivation. The fast Fourier transform (FFT) size for the power analysis was 512. EEG power was normalized to total power.

Locomotion activity: open-field assay
All experiments were performed during the "lights off" (active phase). The locomotion activity was detected in an activity test chamber (Med Associates, Inc.) with an ANY-maze video tracking system (Stoelting Co., United States) using a camera (FUJIFILM co).

Immunohistochemistry
Mice were anesthetized and transcardially perfused with 4% paraformaldehyde (Thermo scientific) in phosphate buffered saline (PBS) (Sigma). Brains were removed and 35 μm thick coronal sections were cut. Free-floating sections were washed in PBS three times for 5 min, permeabilized in PBS plus 0.4% Triton X-100 for 30 min, blocked by incubation in PBS plus 5% normal goat serum (NGS) (Vector), 0.2% Triton X-100 for 1 hr, and incubated with primary antibody diluted in PBS plus 2% NGS overnight at 4°C in a shaker. Incubated slices were washed three times in PBS for 10 min and incubated for 2 hr with a 1:1000 dilution of a secondary antibody (Molecular Probes) in PBS and subsequently washed three times in PBS for 10 min (all at room temperature). Primary antibodies used were rat monoclonal mCherry (1:2000, Invitrogen) and rabbit polyclonal histidine decarboxylase (HDC) (1:1000, PROGEN Biotechnik GmbH). Secondary antibodies were Alexa Fluor 488 goat anti-rabbit IgG and Alexa Fluor 594 goat anti-rat IgG (1:1000, Invitrogen Molecular Probes, United Kingdom). Slices were mounted on slides, embedded in Mowiol mounting medium (with DAPI), cover-slipped, and analyzed using a Zeiss LSM 510 inverted confocal microscope (Facility for Imaging by Light Microscopy, FILM, Imperial College). Images were acquired using Z-scan.

Quantification and statistics
All statistical tests were performed in "Origin 2015" (Origin Lab). We used the Kolmogorov-Smirnov test for normality. We did not use statistical methods to predetermine sample sizes but our sample sizes are similar to those reported in previous publications. The individual tests are given in the figure legends. All data are given as mean ± SEM and stated in the figure legends. The data met the assumptions of the statistical tests used. All t-tests were two-tailed. We provided the t values for t-tests and the F values for ANOVA and the p value in the figure legends. HDC-Cre mice were assigned randomly to the experimental and control groups. HDC-Cre mice received saline, CNO, vehicle, or modafinil injections in random order. All experimental data analysis was blinded, including the analysis of EEG data and animal behavior.

Selective chemogenetic inhibition of HA neurons promotes NREM sleep
Previously we found that acute chemogenetic (metabotropic) activation of HA neurons promoted arousal [16]. Here we tested how selective acute chemogenetic-metabotropic inhibition of those neurons influenced sleep-wake states. For metabotropic inhibition of HA neurons, we bilaterally injected AAV-DIO-hM4Di-mCherry into the TMN area of HDC-ires-Cre mice to generate HDC-hM4Di mice ( Figure 1A). The hM4Di-mCherry expressed in the TMN area, including ventral part of the VTA (VTM) and dorsal part of  Figure 1B). The Cre recombinase ensures that the expression of the cassette encoding the hM4Di-mCherry receptor is restricted to HA neurons. We confirmed this by double-staining sections from the TMN area using antisera against mCherry (which detects the hM4Di-mCherry receptor) and HDC, the unique marker of HA neurons ( Figure 1B). The hM4Di-mCherry receptor was found abundantly expressed on axons and processes, as well as the soma of HDC-positive neurons ( Figure 1B).
To examine the specificity of CNO's actions, we injected AAV-DIO-mCherry into the TMN of HDC-Cre mice. CNO injection into these HDC-mCherry mice had no effect on the amounts of sleep or wakefulness ( Figure 1G) or EEG power spectrum ( Figure 1H) compared with saline-injected mice. Note: we found that CNO given at higher doses of 5 or 1 mg/kg to a variety of other control mouse lines did not alter locomotion or change the amounts of sleep-wake compared with saline injections [44,45].

Ablation of HA neurons does not affect the overt sleep-wake cycle but induces more fragmented wakefulness and NREM sleep
We next conducted chronic lesioning experiments to access the function of HA neurons in regulating sleep and wakefulness. To selectively lesion HA neurons, AAV-DIO-taCasp3-TEV was bilaterally delivered into the TMN area of adult HDC-ires-Cre mice to generate HDC-Casp3 mice (Figure 2A). For the control group, AAV-DIO-taCasp3-TEV was injected into the TMN area of Cre-negative littermates. Six weeks after the AAV injections, the efficiency of the lesion was assessed with immunocytochemistry using an HDC antibody. Compared with AAV-injected Crenegative control animals, the number of HDC-positive cells was substantially reduced in HDC-Casp3 mice ( Figure 2B). We mapped HDC expression in both control and HDC-Casp3 mice throughout the entire TMN area (Figure 2, B and C). Nearly 85% of the HDC-cells were killed in HDC-Casp3 animals (1097 ± 75 vs.173 ± 40, p = 7.4E-7) ( Figure 2D).
We performed sleep-wake recordings of control and HDC-Casp3 mice over the 24 hr cycle. These recordings took place 6 weeks after the AAV-DIO-taCasp3-TEV injections. As found for hdc knockout mice [46], the 24 hr spontaneous sleep-wake pattern was similar between HDC-Casp3 mice and control mice (Figure 3, A-C). Over 24 hr, the amount of wakefulness, NREM, or REM sleep of HDC-Casp3 mice did not differ from control littermates (AAV injected Cre-negative mice). Although the amount of wakefulness was slightly decreased and the amount of NREM sleep slightly increased in HDC-Casp3 mice during the 12 hr "lights off" period, these changes did not reach significance (wake: 7.1 ± 0.25 vs. 6.4 ± 0.26 hr, p = 0.1; NREM: 4.5 ± 0.22 vs. 5.1 ± 0.22 hr, p = 0.08) (Figure 3, A and B). Of note, the HDC-Casp3 mice became aroused more slowly than control mice after the start of the "lights off" period (from time 14:00 to 17:00).
We looked into the sleep microarchitecture of the HDC-Casp3 mice. The episode duration of wakefulness and NREM sleep decreased in HDC-Casp3 mice, particularly during the "lights off" active period (Figure 3, D and E) (wake: 6.18 ± 0.74 vs. 3.7 ± 0.35 min, p = 0.02; NREM: 3.98 ± 0.25 vs. 2.94 ± 0.2 min, p = 0.01). The REM sleep episode duration did not differ between HDC-Casp3 mice and control mice (1.21 ± 0.09 vs. 1.13 ± 0.06 min, p = 0.52) ( Figure 3F). Looking in more detail at the sleep-wake transitions, the HDC-Casp3 mice had more NREM to wake transitions during both the "lights on" and "lights off" periods (Figure 3, F and G) (50 ± 4 vs. 80 ± 6, p = 0.0009) and more wake to NREM sleep transitions during the "lights off" period ( Figure  2F) (64 ± 5 vs. 97 ± 7, p = 0.002). Finally, we assessed the EEG power spectrum of control and HDC-Casp3 mice during each vigilance state during the 12 hr "lights on" period or 12 hr "lights off" period. Both delta (0.5-4 Hz) and theta power (4-8 Hz) of control mice did not differ from HDC-Casp3 mice ( Figure  4). The above results suggest that the loss of HA neurons has been compensated for-the mice were not overtly sleepy, there were no obvious effects on the amounts of sleep and wake in the spontaneous sleep-wake cycle, but HA neurons are needed for consolidating wakefulness, otherwise NREM sleep intrudes.

Modafinil promotes wakefulness partially through HA neurons
Using HDC-Casp3 mice, we next examined whether the wakepromoting effect of modafinil depends on the HA system. We and the graph on the right the total time (5 hr) of wake, NREM, and REM sleep of HDC-hM4Di mice that had received saline (n = 5 mice) or CNO (n = 5 mice) injections.

Ablation of HA neurons does not affect sleep homeostasis after modafinil-induced wakefulness or sleep deprivation
Finally, we tested whether HA neurons are involved in regulating sleep homeostasis after a prolonged wakefulness by examining the delta power of NREM sleep during the starting period (1 hr) of recovery sleep. We first looked at the EEG power spectrum of control and HDC-Casp3 mice after vehicle or modafinil injection ( Figure 7A). After modafinil injection, the delta power in both    Figure 7C). Similar to the modafinil injections, both control mice and HDC-Casp3 mice had an increased EEG delta power during the first hour of recovery NREM sleep compared with their spontaneous baseline NREM sleep (Figure 7, C and D) (control: 31 ± 1% vs. 39 ± 1%, p = 0.001; HDC-Casp3: 28 ± 1% vs. 36 ± 1%, p = 0.0005). These results suggest that the HA system did not participate in the regulation of sleep homeostasis.

Discussion
Our work has shown several aspects of interest for the neuronal HA system. First, selective chemogenetic inhibition of HA neurons produced NREM sleep. But, by contrast, genetic lesioning of HA neurons chronically induced in the adult resembles the phenotype of hdc ko mice (i.e. mice without a functional hdc gene from conception) [46]. This suggests that a similar compensatory mechanism occurs in the adult as in the developing hdc knockout mice. Second, the HA system is not required for sleep homeostasis (defined as the increase in NREM delta power seen in the immediate part of NREM recovery sleep after sleep deprivation), as also concluded in an independent study [9]. Third, HA neurons are required for part of modafinil's actions in promoting wakefulness.
The results obtained with the acute pharmacogenetic and optogenetic manipulations of the HA system that induce NREM sleep (see Introduction) contrast with the results of lesioning of the same cells. It is often noted that Von Economo identified flu-induced lesions in the human posterior hypothalamus, more specifically in the posterior wall of the third ventricle, as producing excessive somnolence (encephalitis lethargica) [47]. Hence, Von Economo suggested that the posterior hypothalamus contained a wake-promoting area [47] and so founded the modern concept that there are wake and sleep-promoting centers in the brain [1]. Ironically, this result (at least on the basis of lesioning the TMN) is not borne out in work on rodents. Killing of neurons in the rat TMN area using saporin-orexin ligand produced no effect on sleep amounts [48,49]. Multiple cell types are lesioned in the TMN in this model, because orexinsaporin kills all neurons that express the orexin receptors, and this expression of the orexin receptor is not restricted to HA neurons, but also occurs in other TMN neurons.
A similar lack of effect of the HA system on sleep-wake was found from lesioning at the gene level. Hdc gene knockout mice, which lack the capability to synthesize HA, are not dramatically impaired in their sleep-wake profile [46]. The mice do have a more limited enthusiasm to investigate novel objects, possibly because they are less aroused, and they have more sleep-wake fragmentation. Additionally, hdc ko mice are more sleepy at the light-to-dark transition that marks the start of their active period. This lack of strong phenotype in hdc ko mice compared with the results obtained with acute experiments antagonizing or stimulating the HA system suggests some form of compensation in the hdc knockout mice, which perhaps occurs during brain development. A similar situation pertains to H1 receptor ko mice which have only mild increases in NREM sleep bouts, possibly because of an upregulated cholinergic system [50]. On the other hand, mice with a permanent upregulation of hdc gene expression also have sleep-wake fragmentation  [51]. Yet, mice with no H3 receptors show reduced wakefulness in nonstressful situations [52], contrasting with the wakepromoting effects of H3 inverse agonists. Perhaps these results show the extraordinary unpredictability of compensatory systems (sometimes there is compensation, sometimes not, sometimes partial) in the brain. In any case, the HA-lesioned mice in our study reproduce the sleep-wake phenotype of hdc knockout mice to a remarkable degree [46]. HA-lesioned mice have more sleep-wake fragmentation, less arousal at the start of lights-off, the active period of the mice. Thus, HA-lesioned mice are likely to have the same compensatory mechanism present in the hdc ko mice, and this effect can emerge in the adult and not only during development. Previously, we found some HA neurons corelease GABA in the neocortex [16]. In the hdc ko mice, these neurons can still corelease GABA, whereas in our HA-lesioned mice, both GABA and HA release will be abolished. Yet the phenotypic effects of hdc gene and HA cell lesions are the same. Knockdown of the vesicular GABA transporter gene expression from HA neurons produced hyperactive animals. We are not sure why we do not get a larger phenotype in HA cell-lesioned mice compared with hdc ko mice. Given the unpredictability of compensations, perhaps the loss of GABA signaling from HA cells has been more compensated than loss of HA signaling.
In spite of compensations at the behavioral level, lesioning studies can be useful, whether cellular or genetic, to reveal drug targets. For example, DAT, D1, and D2 knockout mice are largely insensitive to modafinil [26,29], and H3 knockout mice are insensitive to the wake-enhancing effects of H3 inverse agonists [37,52]. Along similar lines, our cellular lesioning and inhibitory chemogenetic studies suggest that HA neurons contribute to the mechanism of modafinil-induced wakefulness. There is, however, likely to be complex positive feedback in the way that modafinil promotes wakefulness.
VTA dopamine neurons promote wakefulness, in part, via the nucleus accumbens [53,54]. As modafinil antagonizes DAT and raises dopamine levels, and indeed requires DAT for its wake-promoting actions, it seems that raised dopamine in the nucleus accumbens is the obvious way that modafinil promotes wakefulness. But HA neurons also express dopamine receptors and can be excited by dopamine agonists [38]; HA neurons may also be able to synthesize and release dopamine [38]. Thus, the raised dopamine levels, sensed either by the HA soma in the TMN or by HA axons in, for example, neocortex or nucleus accumbens will promote HA release, and possibly even dopamine release. HA probably promotes direct wakefulness in the neocortex via postsynaptic H1 receptors, but also HA directly stimulates the firing of dopamine neurons in the VTA and other amine neurons [18], further promoting wakefulness. Consequently, there will be a number of wake-promoting pathways operating in parallel, some involving dopamine acting through, e.g. the nucleus accumbens on D2 receptors and others where the HA neurons are probably excited by dopamine and then the released HA produces wakefulness via many targets, e.g. the basal forebrain cholinergic neurons [9] or the neocortex or nucleus accumbens.
A further point is that in hdc ko mice, modafinil still produces as much wakefulness as in mice with an intact hdc gene [37]. Possible reasons for this discrepancy are that the earlier study used a lower dose of modafinil, or that dopamine or GABA released from the HA axons in the neocortex, rather than HA itself, mediate some of the effects of modafinil. Another interpretation is that HA neurons are needed for maintaining some of modafinil's actions, but not for initiating them. Background strains of mice could also influence the sensitivity to modafinil, as clearly there is considerable variation in the sensitivity of humans to the drug, which is likely to be based on genetic differences [23].
In conclusion, selective genetic lesioning of adult HA neurons shows their requirement for consolidating wakefulness and for sustaining some of the wake-promoting effects of modafinil. The effects on vigilance state produced by acute inhibition of HA neurons compared with those produced by their long-term removal are much larger.