Excitation of Putative Glutamatergic Neurons in the Rat Parabrachial Nucleus Region Reduces Delta Power during Dexmedetomidine but not Ketamine Anesthesia

All surgical and experimental protocols were approved by the Massachusetts General Hospital Subcommittee on Research Animal Care, and the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines were followed. The numbers of subjects for experiments were chosen based on previously published studies of parabrachial nucleus stimulation during anesthesia.^5,8,10  Thirteen adult male Sprague–Dawley rats (Charles River Laboratories, USA) were fed ad libitum and housed in rooms with a 12-h light–dark cycle, with lights on at 7 am and off at 7 pm. All experiments without anesthesia and all anesthesia infusions were completed during the lights-on hours (although some recordings after anesthesia infusions continued past 7 pm). Because previous studies used male rodents to examine the effects of parabrachial nucleus stimulation during anesthesia,^2,5,8,10  we also used male rats.

Male rats (~10 weeks old, 312.5 to 409.5 g) were anesthetized using 3% isoflurane followed by maintenance at 2.5% isoflurane (Henry Schein, USA) and placed in a stereotaxic frame (model 942, David Kopf Instruments, USA). The animal’s body temperature was maintained at 37°C using a heating pad (50-7220F, Harvard Apparatus, USA). After opening the skin and clearing connective tissue from the top of the skull, craniotomies were made bilaterally over the parabrachial nucleus using a microdrill (Patterson Dental Supply Inc., USA; see fig. 1A for a schematic of craniotomies). A glass micropipette (5-000-1010, Drummond, USA), adapter (MPH6S10, World Precision Instruments, USA), and Hamilton syringe (80301, Hamilton Company, USA) containing the virus (2.8 × 10¹² viral genomes/ml, diluted from 2.8 × 10¹³ viral genomes/ml using saline; pAAV-CaMKIIa-hM3D(Gq)-mCherry; Addgene, USA) for DREADDs transfection were lowered down to the parabrachial nucleus (stereotaxic coordinates, relative to Bregma: −8.0 to −9.06 mm posterior; ±1.96 to ±2.0 mm lateral; −7.2 mm ventral), and 500 nl of virus per side was delivered via microinjection pump (53311, Stoelting Co., USA). Three craniotomies were also made over the left prefrontal cortex (2.2 mm posterior; −2.2 mm lateral), left parietal cortex (−3.84 mm posterior; −2.2 mm lateral), and cerebellum (in the middle of the interparietal bone for a reference EEG), and stainless steel EEG electrodes (791400, A-M Systems, USA) were hooked just under the skull at each craniotomy. Finally, holes were drilled for a ground screw over the cerebellum and bone screws (51457, Stoelting Co.) to support the implant. Stainless steel electromyography (EMG) electrodes (793500, A-M Systems) were inserted for recording nuchal muscle activity. EEG and EMG electrodes were all connected to a Neuralynx electrode interface board (EIB-8, Neuralynx Inc., USA), and the electrodes and electrode interface board were secured to the skull and bone screws with dental cement. After surgery, the rats were given analgesics (ketoprofen, 4 mg/kg; Ketofen, Zoetis, USA) and allowed to rest for at least 14 days before beginning experiments. This allowed for recovery from surgery and expression of DREADDs. In addition to 10 rats that received the full surgery, three rats were given viral injections only (500 nl or less per side; no EEG/EMG electrode implantation) for histological verification of DREADDs effectiveness and specificity of expression.

The excitatory DREADD hM3Dq was used to selectively excite calcium/calmodulin-dependent protein kinase II–positive (CaMKIIa⁺) neurons. hM3Dq is a muscarinic receptor modified to be sensitive to the exogenous ligand clozapine-N-oxide, instead of acetylcholine.²⁹  We used an adeno-associated virus (AAV8 serotype) that contained a plasmid with a promoter specific to CaMKIIa⁺ (putative glutamatergic) neurons, followed by the genes for hM3Dq and mCherry, a fluorescent tag (pAAV-CaMKIIa-hM3D(Gq)-mCherry; Addgene). Using this virus enabled us to constrain expression of hM3Dq to CaMKIIa⁺ neurons transfected by the virus.

EEG (filtered 0.1 to 500 Hz) and EMG (filtered 10 to 500 Hz) signals were acquired using a Neuralynx Digital Lynx SX recording system at a sampling rate of 500 Hz (Neuralynx Inc.). Overhead video recordings were also captured to assess behavioral states.

The purpose of this study was to examine the effects of stimulation of the parabrachial nucleus on arousal during dexmedetomidine and ketamine anesthesia. Five anesthesia conditions were paired with saline and clozapine-N-oxide injections (for a total of 10 treatment groups): (1) no anesthesia, (2) low-dose dexmedetomidine (0.3 µg · kg⁻¹ · min⁻¹ for 45 min), (3) high-dose dexmedetomidine (4.5 µg · kg⁻¹ · min⁻¹ for 10 min), (4) low-dose ketamine (2 mg · kg⁻¹ · min⁻¹ for 30 min), and (5) high-dose ketamine (4 mg · kg⁻¹ · min⁻¹ for 15 min). Each of the subjects of a given anesthesia condition received both clozapine-N-oxide and saline injections, in separate experiments, allowing us to compare clozapine-N-oxide and saline conditions for each rat. Randomization methods were not used to determine which treatments each rat received or the order in which the treatments (anesthetics or clozapine-N-oxide/saline) were received, but the order of experiments was varied between rats (see Supplemental Digital Content 1, https://links.lww.com/ALN/C647, for experiment orders for each rat).

For experiments without anesthesia, intraperitoneal injections of clozapine-N-oxide (HY-17366, MedChem Express LLC, USA; and 6329/50, R&D Systems, USA; 1 mg/kg) or saline were given within the half-hour range between 11:23 and 11:53 am. EEG and EMG were then recorded for 5 h after the injection. For these experiments without anesthesia, the rats were allowed to rest for at least 3 days between experiments if clozapine-N-oxide was given in the first of the two experiments and at least 1 day if saline was given first.

For experiments with anesthesia, before starting recordings, the rats were anesthetized with isoflurane for ~15 min to place an intravenous port in the lateral tail vein. Immediately after cessation of isoflurane, clozapine-N-oxide (or saline) intraperitoneal injections were administered. One hour after the clozapine-N-oxide (or saline) injection, dilutions of dexmedetomidine (2 µg/ml for low dose or 5 µg/ml for high dose; dexmedetomidine hydrochloride injection, Sun Pharmaceutical Industries, India; and dexmedetomidine injection, Accord Healthcare, India) or ketamine (6 mg/ml; ketamine HCl injection, West-Ward, USA) in saline were administered intravenously via a syringe pump (HA3000I, Harvard Apparatus). This time delay allowed for clozapine-N-oxide to reach its maximal effectiveness, which occurs at ~45 min after administration.³⁰  Two exceptions to the 1-h delay between clozapine-N-oxide (or saline) and anesthetic infusion occurred: in one high-dose dexmedetomidine experiment (rat 4), clozapine-N-oxide was injected 140 min before the start of the anesthetic infusion, and in one high-dose ketamine experiment (rat 3), saline was injected 116 min before the start of the anesthetic infusion. Both exceptions were well within the ~9-h period during which clozapine-N-oxide has been shown to be effective.³⁰  Normothermia was maintained with a heating pad under the recording cage while the animals were anesthetized. EEG and EMG recordings were started a minimum of 10 min before beginning anesthesia infusion. The rats were allowed to rest for at least 3 days between all experiments with anesthesia except one high-dose ketamine experiment. Only 2 days of rest were given before that experiment, because it followed an aborted ketamine experiment in which the IV started leaking before the infusion completed. For experiments with anesthesia, recordings continued until the rats recovered from anesthesia and were mobile.

After completion of each subject’s experiments, the rats were euthanized using isoflurane overdose and exsanguination via transcardial perfusion (see “Histology”). Blinding methods were not used in this study.

The EEG and EMG signals from experiments without anesthesia were used to visually score wakefulness, non–rapid-eye-movement (non-REM) sleep, and rapid-eye-movement (REM) sleep in 2-s epochs using Spike2 (CED, United Kingdom). The time spent in waking, non-REM, and REM sleep were then separately totaled over the 5 h after saline or clozapine-N-oxide injection. Each rat’s state was assumed to be awake during the first few minutes after their injection, because it took a few minutes of active handling to plug them into the recording system.

Just after the anesthetic infusion finished, each rat was turned on its side. Recovery time was measured as the time from the end of the anesthetic infusion to when the rat righted itself and had all four paws on the floor. In five of the low-dose dexmedetomidine experiments, the rats never lost the righting reflex; in such cases, return of movement was used to measure recovery time.

The data were analyzed using MATLAB (Mathworks, USA). Before spectral analysis, the mean of each EEG recording was subtracted from the entire signal. Artifacts were then removed by identifying time points where the voltage crossed a threshold of ±1,500 µV. All threshold crossings less than 1 min apart were grouped together as a single artifact, with all time points in between treated as threshold crossings. The signal at and surrounding each of these time points was then multiplied by an inverted Hamming window with a length of 0.15 s, centered at the identified voltage threshold crossing. In this manner, time points identified as threshold crossings were set to 0 µV, and the signal 0.075 s before and after artifacts gradually increased to full magnitude, reducing sharp edges in the signal that would produce harmonics in the spectrogram.

All spectrograms in this study were computed using the multitaper methods described by Prerau et al.³¹  Example spectrograms were computed from 0 to 30 Hz (for experiments without anesthesia) or 0 to 50 Hz (for experiments with anesthesia) using a 30-s window that moved in 3-s steps, a time-half-bandwidth product of 3, and five discrete prolate spheroidal sequence tapers. The signal in each window was linearly detrended before the multitaper spectrogram was computed. For Supplemental Digital Content 2–6 (https://links.lww.com/ALN/C649, https://links.lww.com/ALN/C650, https://links.lww.com/ALN/C651, https://links.lww.com/ALN/C652, https://links.lww.com/ALN/C653), the spectrograms were computed from 0 to 60 Hz (to accommodate gamma band analysis) for both experiments with and without anesthesia. Any windows with 0 µV² total power, such as those comprised solely of removed electrical artifacts, were discarded from the analysis (e.g., see the beginning of the spectrograms in fig. 2 [A and B], in which the artifact time points are given the color of the minimum value in the spectrogram), and power was converted to dB.

All statistical analyses were performed using MATLAB, unless otherwise noted. To measure the power in various frequency bands between conditions over time, the multitaper spectrogram was computed from 0 to 30 Hz (for experiments without anesthesia) or 0 to 50 Hz (for experiments with anesthesia) using a 120-s, nonoverlapping time window, a time-half-bandwidth product of 3, and five discrete prolate spheroidal sequence tapers. The signal in each window was linearly detrended before the multitaper spectrogram was computed. After discarding windows with 0 µV² total power and conversion to dB, the power in each band (delta: 0.5 to 4 Hz; theta: 4 to 8 Hz; alpha: 10 to 15 Hz; beta: 15 to 30 Hz; and gamma: 40 to 60 Hz) was averaged for each time point in the spectrogram. This mean power was aligned in time for all rats, relative to the start of anesthesia delivery.

For comparison of mean power in each frequency band between clozapine-N-oxide and saline conditions, because of the small sample sizes, we used the bootstrapping technique to find the CI values. For each time point, the mean spectral power in a given frequency band from the saline condition was subtracted from the corresponding time point’s power in the clozapine-N-oxide condition of the same rat. The differences from all rats for each time point were sampled with replacement N times (where N = the number of subjects), and the mean was taken. This process was repeated 5,000 times for each time point to form a bootstrapped collection of 5,000 mean power differences. From this bootstrapped collection, the overall median for each time point was taken, and 99% CIs were found by taking the range containing 99% of the bootstrapped mean differences. Thus, when then the CIs do not overlap with 0, it indicates 99% confidence that there is a difference between the mean power in that frequency band under clozapine-N-oxide and saline conditions. P values that are reported for times with peak median bootstrapped mean differences in power were calculated by rerunning the bootstrap analysis for those time points using the “boot.t.test” function in the MKinfer package (version 0.6) in R (version 4.1.0; 99% confidence level, 5,000 bootstrap replicates; reported medians and CIs for peak median bootstrapped differences in power were taken from the MATLAB bootstrap analysis described above). For some figure panels containing bootstrapped mean differences in power (figs. 2C, 3B, 5B, and 6B, and the associated Supplemental Digital Content 2–6 [https://links.lww.com/ALN/C649, https://links.lww.com/ALN/C650, https://links.lww.com/ALN/C651, https://links.lww.com/ALN/C652, https://links.lww.com/ALN/C653]), the number of subjects (N) at each time point varies because of occasional electrical artifacts, so ranges of N are reported. For figure 2C, the first 10 min of measurements after clozapine-N-oxide or saline injections (at time = 0) were not used to calculate the reported range of N, because some rats took longer than others to plug into the EEG recording system.

For statistical comparison of the time spent in each sleep stage (for experiments without anesthesia) or the recovery times (for experiments with anesthesia), we used R (version 4.1.0). The reported median differences for these measurements are Hodges–Lehmann estimates. Because the same animals received both clozapine-N-oxide and saline conditions, we used the two-sided Wilcoxon signed-rank test with an α value of 0.05 to determine statistical significance. The Wilcoxon signed-rank tests, the median differences, and the 95% CIs were calculated using the “wilcox.test” function in R.

After completion of experiments, the rats were euthanized with isoflurane (5%) and transcardially perfused with 1[times] phosphate-buffered saline, followed by 10% formalin. The brains were then sliced at 50 µm using a vibratome (VT1000 S, Leica Biosystems, USA). The slices were mounted on a slide with a 4′,6-diamidino-2-phenylindole–containing mounting medium (H-1200-10, Vector Laboratories, USA). The brain slices were imaged using a fluorescence microscope (Axio Imager.M2, Zeiss, USA) to localize expression of mCherry/DREADDs in the parabrachial nucleus region (fig. 1, B and C).

In addition, immunofluorescence was used in a subset of animals to confirm the effectiveness of DREADDs stimulation and the specificity of DREADDs expression to CaMKIIa⁺ neurons (fig. 1, D through F, respectively). To confirm DREADDs effectiveness, we injected rats with clozapine-N-oxide or saline 60 to 70 min before euthanasia. Perfusion and slicing occurred as mentioned above, followed by immunofluorescence. For the immunofluorescence, the brain slices were first rinsed in phosphate-buffered saline for 5 to 10 min three times. A solution consisting of 0.1% Triton X-100 (Boston BioProducts, USA) in phosphate-buffered saline was used as the bulk solution for a blocking solution containing 10% normal goat serum (ab7481, Abcam, United Kingdom) and 4% bovine serum albumin (001-000-161, Jackson ImmunoResearch, USA). The slices were placed in this blocking solution for 1 h. The slices were then incubated overnight at room temperature in blocking solution with an anti–c-Fos primary antibody (1:2,000; 226003, Synaptic Systems, Germany). After incubation, the slices were rinsed in phosphate-buffered saline or 0.1% Triton X-100 for 5 to 10 min three times and then allowed to incubate for 2 h at room temperature in blocking solution with Alexa Fluor 488 secondary antibodies (1:200; A-11008, Invitrogen, USA). Finally, the slices were rinsed twice in 0.1% Triton X-100 and then twice more with phosphate-buffered saline (without Triton X-100) and mounted on a slide with a 4′,6-diamidino-2-phenylindole–containing mounting medium.

To confirm the specificity of DREADDs expression to CaMKIIa⁺ neurons, the brain slices were first rinsed in 0.3% Triton X-100 in phosphate-buffered saline for 5 to 10 min three times. This solution was used as the bulk solution for a blocking solution containing 10% normal donkey serum (S30-100ML, EMD Millipore, USA). The slices were placed in the blocking solution for 1 h. Then the slices were incubated overnight at 4°C in blocking solution with an anti-CaMKIIa primary antibody (1:50; PA5-19128, Invitrogen). After incubation, the slices were rinsed in 0.3% Triton X-100 for 5 to 20 min three times and then allowed to incubate for 2 h at room temperature in blocking solution with Alexa Fluor 488 secondary antibodies (1:200; 705-545-003, Jackson ImmunoResearch). Finally, the slices were rinsed twice in 0.3% Triton X-100 and then twice more with phosphate-buffered saline (without Triton X-100) and mounted on a slide with a 4′,6-diamidino-2-phenylindole–containing mounting medium.

All washes and incubations for both immunofluorescence protocols were performed with free-floating brain slices in 12- or 24-well plates placed on a rocker. Colocalization of mCherry, indicating DREADDs expression, and Alexa Fluor 488, indicating c-Fos or CaMKIIa expression, was then visualized using the Zeiss microscope mentioned previously. Neurons with and without colocalized fluorescence were counted using Fiji software.³² 

We tested the effects of exciting putative glutamatergic neurons in the parabrachial nucleus by using DREADDs (hM3Dq) and surgically implanted EEG electrodes (fig. 1A). Excitatory DREADDs were expressed in the parabrachial nucleus and in the surrounding region, especially rostral to the parabrachial nucleus (see example expression in fig. 1B and a summary of expression across all 10 rats in two atlas slices in fig. 1C; a summary of mCherry expression across a range of atlas slices is shown in Supplemental Digital Content 7, https://links.lww.com/ALN/C648). Effectiveness of DREADDs stimulation was confirmed by intraperitoneally injecting clozapine-N-oxide, the ligand for DREADDs, and then using immunofluorescence to confirm expression of c-Fos, a marker of neural activity (fig. 1, D and E; 4 of 54 DREADDs⁺ neurons were c-Fos⁺ after a saline control injection, N = 1 male rat; 100 of 124 counted DREADDs⁺ neurons were c-Fos⁺ after clozapine-N-oxide injections, N = 2 male rats). The specificity and extent of DREADDs expression in CaMKIIa⁺ neurons were also confirmed using immunofluorescence (fig. 1F; inside the parabrachial nucleus: 333 of 383 counted DREADDs⁺ neurons were CaMKIIa⁺, and in the same fields of view, 333 of 778 counted CaMKIIa⁺ neurons were DREADDs⁺, N = 3 male rats; outside the parabrachial nucleus: 128 of 173 counted DREADDs⁺ neurons were CaMKIIa⁺, and in the same fields of view, 128 of 956 counted CaMKIIa⁺ neurons were DREADDs⁺, N = 1 male rat). Thus, by intraperitoneally injecting clozapine-N-oxide, the ligand for DREADDs, we were able to selectively excite CaMKIIa⁺ neurons during conditions without anesthesia or with dexmedetomidine or ketamine anesthesia.

Some experimental data were not included in our analyses, despite being collected (see grayed-out experiments in Supplemental Digital Content 1, https://links.lww.com/ALN/C647). These included a low-dose ketamine experiment in which the clozapine-N-oxide never fully dissolved (rat 1) and a high-dose dexmedetomidine experiment in which the rat (rat 3) was clearly aroused by a neighboring rat’s tail flick. In addition, two rats experienced issues that prevented them from completing all the planned experiments: rat 2’s electrode interface board broke during an experiment without anesthesia, and rat 5 suffered from painful limb swelling that confounded the measurement of recovery time during a low-dose dexmedetomidine experiment. These issues led to the euthanasia of these two rats and thus prevented them from completing the low-dose anesthesia experiments. Two other rats, rats 8 and 10, were not included in the high-dose dexmedetomidine and high-dose ketamine treatment groups after we felt we had sufficient data to characterize the effects of parabrachial nucleus stimulation during those anesthetics and doses. Finally, after histological analysis, it was found that one rat that was initially included in experiments did not express DREADDs in the parabrachial nucleus and was thus not included in any of our analyses. One other rat did not survive the initial surgery and was not used for any experimental treatment groups.

Previous research has shown that nonspecific excitation of parabrachial nucleus neurons decreases the delta power associated with the awake and non-REM sleep states (indicating neurophysiologic arousal), without anesthesia.²  Because the parabrachial nucleus is primarily glutamatergic, we first tested whether specific excitation of CaMKIIa⁺ neurons also promotes arousal under these conditions. We measured delta power (0.5 to 4 Hz) in the prefrontal EEG for the first 5 h after an intraperitoneal injection of clozapine-N-oxide (1 mg/kg). In congruence with the effects of nonspecific parabrachial nucleus excitation, we observed reduced delta power relative to control conditions in which intraperitoneal injections of saline were administered to the same rats (fig. 2; trace in fig. 2C summarizes delta power with clozapine-N-oxide injections minus delta power with saline injections, and periods with CIs that do not span zero show statistically significant differences with 99% confidence; peak statistically significant median bootstrapped difference in power [clozapine-N-oxide – saline] = –6.79 dB, 99% CI = –9.05 to –4.16 dB, P = 0.001, N = 9 male rats; for entire trace in fig. 2C, N = 8 to 9 male rats). Whereas the EEG under saline control conditions showed increased delta power, reflecting an increase in sleeping over time, the delta power with clozapine-N-oxide injections remained at awake levels. Power in other frequency bands, including theta (4 to 8 Hz), alpha (10 to 15 Hz), and beta (15 to 30 Hz), likewise decreased with clozapine-N-oxide injections relative to saline injections. Conversely, power in the gamma band (40 to 60 Hz) increased with clozapine-N-oxide injections (see Supplemental Digital Content 2, https://links.lww.com/ALN/C649, which shows the summary power differences [clozapine-N-oxide – saline] for the theta, alpha, beta, and gamma bands).

Moreover, a formal assessment of sleep states during the 5 h after clozapine-N-oxide injections shows decreased time spent in non-REM and REM sleep relative to experiments with saline injections (fig 2D; median difference in time spent waking [clozapine-N-oxide – saline] = 126.31 min, 95% CI = 104.13 to 152.31 min, P = 0.004; median difference in time spent in non-REM sleep [clozapine-N-oxide – saline] = –117.23 min, 95% CI = –137.74 to –99.39 min, P = 0.004; median difference in time spent in REM sleep [clozapine-N-oxide – saline] = –9.07 min, 95% CI = –15.13 to –2.10 min, P = 0.004; N = 9 male rats). These results suggest that excitation of CaMKIIa⁺ neurons alone in the parabrachial nucleus region is sufficient to reduce delta power and sleep in male rats without anesthesia, consistent with increased arousal.

Excitation of the parabrachial nucleus, including selective excitation of glutamatergic neurons,⁸  is effective at reducing both delta oscillations and the time to recovery from isoflurane, sevoflurane, and propofol anesthesia.^5,10  We tested whether excitation of CaMKIIa⁺ neurons in the parabrachial nucleus region has similar effects with dexmedetomidine or ketamine. One hour after injection of clozapine-N-oxide, we administered dexmedetomidine (0.3 µg · kg⁻¹ · min⁻¹ for 45 min). We found that delta power in the prefrontal EEG was reduced relative to control experiments in the same rats with saline injections. This reduction was statistically significant between 2 and 40 min after the start of the infusion and from 44 min until the end of the analyzed period (150 min; fig. 3, A and B; trace in fig. 3B summarizes delta power with clozapine-N-oxide injections minus delta power with saline injections, and periods with CIs that do not span zero show statistically significant differences with 99% confidence; peak statistically significant median bootstrapped difference in power during dexmedetomidine infusion [clozapine-N-oxide – saline] = –6.06 dB, 99% CI = –12.36 to –1.48 dB, P = 0.007, N = 8 male rats; for the entire trace in fig. 3B, N = 7 to 8 male rats). Consistent with experiments without anesthesia (fig. 2), there was also some statistically significant reduction in delta power before the start of anesthesia. Power in the theta, alpha, and beta frequency bands likewise decreased with clozapine-N-oxide injections relative to saline injections. The prevalence of these decreases across the recorded period declined as the frequency of the band increased (i.e., theta > alpha > beta). Conversely, power in the gamma band had slight increases with clozapine-N-oxide injections (see Supplemental Digital Content 3, https://links.lww.com/ALN/C650, which shows the summary power differences [clozapine-N-oxide – saline] for the theta, alpha, beta, and gamma bands).

To measure behavioral arousal, we used the time to return of righting reflex (or time to return of movement if righting reflex was not lost; see “Materials and Methods”) as a surrogate for recovery from anesthesia. In contrast to the reduced delta oscillations, the recovery times between clozapine-N-oxide and saline conditions did not show statistically significant differences (fig. 3C; median difference [clozapine-N-oxide – saline] = 1.63 min, 95% CI = –20.06 to 14.14 min, P = 0.945; N = 8 male rats).

We next sought to understand whether excitation of CaMKIIa⁺ parabrachial nucleus region neurons also reduces delta power effectively at a higher dose of dexmedetomidine. In these experiments, after injection of clozapine-N-oxide, we administered high-dose dexmedetomidine (4.5 µg · kg⁻¹ · min⁻¹ for 10 min). We found that other than one short period with lower delta power from 6 to 8 min, delta power in the prefrontal EEG during the infusion did not show statistically significant differences relative to control experiments in the same rats with saline injections. After the infusion, periods with lower delta power with parabrachial nucleus region stimulation appeared from 60 to 64 min, 78 to 98 min, 108 to 110 min, and 112 to 118 min (fig. 4, A and B; trace in fig. 4B summarizes delta power with clozapine-N-oxide injections minus delta power with saline injections, and periods with CIs that do not span zero show statistically significant differences with 99% confidence; peak statistically significant median bootstrapped difference in power during dexmedetomidine infusion [clozapine-N-oxide – saline] = –1.93 dB, 99% CI = –4.16 to –0.56 dB, P = 0.006, N = 7 male rats; for the entire trace in fig. 4B, N = 7 male rats). Power in the theta, alpha, and beta bands decreased with clozapine-N-oxide injections relative to saline injections, although to a lesser degree than with low-dose dexmedetomidine. Conversely, power in the gamma band increased after the dexmedetomidine infusion was completed (see Supplemental Digital Content 4, https://links.lww.com/ALN/C651, which shows the summary power differences [clozapine-N-oxide – saline] for the theta, alpha, beta, and gamma bands).

As with low-dose dexmedetomidine, recovery after high-dose dexmedetomidine infusion, as measured by the return of righting reflex, did not show statistically significant differences for rats with clozapine-N-oxide injections relative to saline controls (fig. 4C; median difference [clozapine-N-oxide – saline] = 11.01 min, 95% CI = –20.84 to 23.67 min, P = 0.297; N = 7 male rats).

Given that excitation of CaMKIIa⁺ parabrachial nucleus region neurons reduced low-dose dexmedetomidine-induced delta power, we next tested whether stimulation of these same neurons could reduce ketamine-induced delta power. In these experiments, 1 h after injection of clozapine-N-oxide, we administered ketamine (2 mg · kg⁻¹ · min⁻¹ for 30 min). At this dose, all rats lost the righting reflex, although many rats moved when they were placed on their sides (6 of 7 male rats moved with control conditions; 2 of 7 male rats moved with clozapine-N-oxide conditions). In contrast with low-dose dexmedetomidine, for low-dose ketamine, delta power in the prefrontal EEG during the infusion did not show statistically significant differences relative to control experiments in the same rats with saline injections. After the infusion, periods with lower delta power with clozapine-N-oxide injections appeared from 64 to 68 min, 70 to 74 min, 78 to 82 min, 84 to 86 min, 90 to 96 min, 110 to 112 min, and 114 to 116 min. In addition, one brief period with higher delta power appeared from 44 to 46 min (fig. 5, A and B; trace in fig. 5B summarizes delta power with clozapine-N-oxide injections minus delta power with saline injections, and periods with CIs that do not span zero show statistically significant differences with 99% confidence; for the entire trace in fig. 5B, N = 6 to 7 male rats). During the ketamine infusion, there was a brief decrease in theta power but no differences in alpha, beta, and gamma power with clozapine-N-oxide injections relative to saline injections. After the infusion, theta, alpha, and gamma showed brief periods of decreased power, theta showed a brief increase in power, and beta showed no differences (see Supplemental Digital Content 5, https://links.lww.com/ALN/C652, which shows the summary power differences [clozapine-N-oxide – saline] for the theta, alpha, beta, and gamma bands).

Similar to dexmedetomidine, recovery after low-dose ketamine infusion, as measured by the return of righting reflex, did not show statistically significant differences for rats with clozapine-N-oxide injections relative to saline controls (fig. 5C; median difference [clozapine-N-oxide – saline] = 12.82 min, 95% CI = –3.20 to 39.58 min, P = 0.109; N = 7 male rats).

When a higher dose of ketamine (4 mg · kg⁻¹ · min⁻¹ for 15 min) was used, excitation of CaMKIIa⁺ parabrachial nucleus region neurons (via clozapine-N-oxide injections) had little effect on delta power in the prefrontal EEG during the infusion, except for a brief reduction in power from 0 to 2 min. After the infusion, periods with lower delta power with parabrachial nucleus region stimulation appeared from 52 to 54 and 62 to 66 min (fig. 6, A and B; trace in fig. 6B summarizes delta power with clozapine-N-oxide injections minus delta power with saline injections, and periods with CIs that do not span zero show statistically significant differences with 99% confidence; peak statistically significant median bootstrapped difference in power during ketamine infusion [clozapine-N-oxide – saline] = –0.87 dB, 99% CI = –1.62 to –0.18 dB, P = 0.019, N = 8 male rats; for the entire trace in fig. 6B, N = 7 to 8 male rats). During the ketamine infusion, power in the theta, alpha, beta, and gamma bands showed no differences with clozapine-N-oxide injections relative to saline injections. After the infusion, theta and beta showed brief periods of decreased power, and alpha and beta showed some increased power. No differences were seen in gamma power (see Supplemental Digital Content 6, https://links.lww.com/ALN/C653, which shows the summary power differences [clozapine-N-oxide – saline] for the theta, alpha, beta, and gamma bands).

However, contrary to the rest of the anesthesia conditions, recovery after high-dose ketamine infusion, as measured by the return of righting reflex, was longer for rats with clozapine-N-oxide injections relative to saline controls. This elongation of recovery times was statistically significant (fig. 6C; median difference [clozapine-N-oxide – saline] = 11.38 min, 95% CI = 1.81 to 24.67 min, P = 0.016; N = 8 male rats).

In this study, the primary outcome was evidence that excitation of putative glutamatergic neurons in the parabrachial nucleus region is sufficient to reduce delta power during low-dose dexmedetomidine anesthesia but not high-dose dexmedetomidine nor ketamine anesthesia. Furthermore, changes in delta power did not correspond to changes in time to recovery from anesthesia. As a secondary outcome, we found that selectively stimulating putative glutamatergic neurons in the parabrachial nucleus region was sufficient to reduce delta power and time spent in sleep without anesthesia. We also found lesser effects of parabrachial nucleus region stimulation on other frequency bands, including theta, alpha, beta, and gamma, with and without anesthesia. These results advance our understanding of the neural circuit mechanisms underlying EEG patterns under anesthesia and may be useful for improving clinical EEG monitors.

We used male rats in our study so that our results would be comparable to previous studies that used male rodents to examine the effects of parabrachial nucleus stimulation during anesthesia.^2,5,8,10  We cannot rule out, however, that there may be sex differences in the neurophysiologic or behavioral responses to parabrachial nucleus stimulation. Future studies that include both and compare between the sexes are recommended to discern any potential differences.

The extent of colocalization of CaMKIIa expression with glutamatergic neurons in the parabrachial nucleus has not been studied, so it is not known whether all DREADDs⁺ neurons in our study were glutamatergic. However, previous studies have shown a strong association between CaMKIIa⁺ and glutamatergic neurons in the cortex,^33,34  amygdala,³⁵  and thalamus,³⁶  as well as a strong disassociation between CaMKIIa⁺ and GABAergic neurons in the basal ganglia, thalamus, and hypothalamus.³⁷  The parabrachial nucleus has also been shown to be primarily glutamatergic¹⁸  and express CaMKII.^38,39  These previous findings, combined with the large extent of CaMKIIa-driven expression of DREADDs in the parabrachial nucleus in our study, are consistent with the claim that the majority of our DREADDs⁺ neurons were glutamatergic.

In our study, we did not observe any statistically significant reduction in recovery times with clozapine-N-oxide injections relative to saline injections. Contrary to our expectations, we found an increase in time to recovery with excitation of parabrachial nucleus region neurons in the case of high-dose ketamine. This result could indicate a disconnect between neurophysiologic and behavioral indicators of arousal. Indeed, the righting reflex itself does not involve cortical activity⁴⁰  and presumably would not directly affect the EEG. However, there are also other possibilities that should be considered. First, because of the arousable nature of low-dose anesthesia (in particular, dexmedetomidine anesthesia) and the large SD of recovery times even under controlled conditions, the return of righting reflex can be a variable, albeit common measure of behavioral arousal. In addition, during recovery from dexmedetomidine, although a rat may recover from anesthesia enough to return to the upright position, they usually quickly return to an unconscious state, with the recovery time period lasting far beyond the initial arousal. Finally, in the experiments without anesthesia, a subset of rats showed very little movement after clozapine-N-oxide injection, despite lack of sleep (see example in fig. 2B). This stillness with DREADDs stimulation did not correspond in any obvious way to differences in the anatomical regions expressing DREADDs. However, if this behavior was present as anesthesia began to wear off, it could have confounded the time-to-recovery results.

Although not the primary outcome, we also found some differences in the theta, alpha, beta, and gamma bands with parabrachial nucleus region stimulation. In experiments without anesthesia, reduced amounts of sleep with clozapine-N-oxide injections may be responsible for broadband power reductions affecting the theta, alpha, and beta bands, in addition to delta. The reduced amount of sleep with clozapine-N-oxide injections is also consistent with increased gamma power, because gamma power decreases with sleep.⁴¹ 

Power during the dexmedetomidine anesthesia experiments largely follows the same pattern as the sleep experiments. The reductions in theta, alpha, and beta and the increase in gamma suggest a similar effect of parabrachial nucleus region stimulation on dexmedetomidine anesthesia and sleep and may underscore similarities in their mechanisms. For ketamine anesthesia experiments, most changes in power occur after the anesthesia infusion has completed, while the rats are recovering. Thus, the minor changes that we observed may have more to do with the return of sleeping behavior than with an interaction between parabrachial nucleus region stimulation and ketamine.

We found that the ability of CaMKIIa⁺ neuron excitation in the parabrachial nucleus region to promote arousal during concurrent anesthesia depended on the molecular targets of the anesthetics. In particular, excitation of these neurons reduced delta oscillations during dexmedetomidine (fig. 3) but not ketamine anesthesia (figs. 5 and 6). These results may suggest that distinct mechanisms are responsible for delta generation between these two anesthetics.

Anesthetics such as propofol, isoflurane, and sevoflurane enhance or replicate GABAergic neurotransmission in both subcortical and cortical neurons. As a result, the cortex is inhibited both indirectly, because subcortical arousal nuclei provide less excitatory neurotransmission, and directly, because GABAergic neurotransmission onto cortical neurons is increased.^21,26  Like GABAergic anesthetics, dexmedetomidine exerts both indirect and direct effects on the cortex: (1) Dexmedetomidine inhibits norepinephrine release from the locus coeruleus,²⁰  which disinhibits the preoptic area (a sleep-active area), leading to indirect inhibition of other major arousal areas and the cortex. (2) Although less studied, α_2A-adrenergic receptors also exist in the cortex,⁴²  and dexmedetomidine may also exert effects on the cortex directly. Similar to the effects of parabrachial nucleus excitation with GABAergic anesthetics, exciting glutamatergic neurons in the parabrachial nucleus may counteract both the indirect and direct cortical inhibition caused by dexmedetomidine, causing neurophysiologic arousal. This would happen as excitatory, glutamatergic projections from the parabrachial nucleus to other subcortical arousal areas and the cortex are activated. When the inhibition is more profound (as with a high dose of dexmedetomidine), the cortical excitation provided by parabrachial nucleus activation is insufficient to overcome the inhibition (fig. 4).

Although ketamine is like GABAergic anesthetics and dexmedetomidine in its inhibition of subcortical arousal areas, it is distinct in its direct cortical effects.^21,26  Ketamine is thought to disinhibit cortical neurons and produce an overall excitatory effect on pyramidal neurons via inhibition of local interneurons.^27,28,43  Because excitation of the para-brachial nucleus also has an excitatory effect on the cortex, it likely cannot counteract the direct effects of ketamine on the cortex and thus is ineffective at evoking neurophysiologic arousal. In addition, because ketamine antagonizes NMDA receptors, it may blunt the effects of glutamate release that the parabrachial nucleus provides when excited. It is possible that parabrachial nucleus stimulation may lead to arousal with other doses of ketamine; however, movement was not prevented by the low-dose ketamine we used, so decreasing the dose further would cause insufficient sedation to discern differences between clozapine-N-oxide and saline conditions. In addition, because parabrachial nucleus region stimulation did not cause consistent delta power nor recovery time reductions with low-dose or high-dose ketamine, it is unlikely that an even higher dose would be arousable.

Although not examined in this study, it is possible that inhibition of glutamatergic neurons in the parabrachial nucleus increases cortical delta power or the time to recovery from dexmedetomidine or ketamine anesthesia. Indeed, one recent study showed inhibition of glutamatergic parabrachial nucleus neurons increases time to recovery from sevoflurane.⁸  However, other studies of parabrachial nucleus excitation have not reported the effects of inhibition on arousal from anesthesia.^2,5  The inability of DREADDs techniques to completely silence (rather than just dampen) neural activity can make interpreting results a challenge.⁴⁴  For this reason, we chose not to investigate the effects of inhibition on arousal from anesthesia.

Finally, the balance between inhibition and excitation can be crucial for determining the oscillatory state of neural circuits.^45–48  It is therefore possible that the excitation of glutamatergic neurons in the parabrachial nucleus is effective at disrupting that balance with GABAergic anesthetics and low-dose dexmedetomidine, but it does not disrupt that balance with high-dose dexmedetomidine or ketamine. Consistent with this hypothesis, the parabrachial nucleus projects to the basal forebrain, and accordingly, stimulation of the parabrachial nucleus may cause cortical acetylcholine release. Because dexmedetomidine infusions do not change cortical acetylcholine levels,⁴⁹  excitation of the parabrachial nucleus may cause unbalanced cortical excitation via an overabundance of acetylcholine, reducing delta power and leading to neurophysiologic arousal. More research, including measurement of cortical neurotransmitters, will be helpful for explaining the mechanisms responsible for the differing responses to excitation of glutamatergic para-brachial nucleus region neurons with these two anesthetics.

Statistical support was provided by the Anesthesia Research Center within the Department of Anesthesia, Critical Care and Pain Medicine at Massachusetts General Hospital, Boston, Massachusetts.

Supported by grant No. P01-GM118269 from the National Institutes of Health, Bethesda, Maryland, and the Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Boston, Massachusetts.

Dr. Solt is a consultant to Takeda Pharmaceuticals (Tokyo, Japan). The other authors declare no competing interests.