Dynamics of Ketamine-induced Loss and Return of Consciousness across Primate Neocortex

Anesthetic-induced altered states of consciousness are thought to be associated with highly structured oscillations in the brain.^1–3  Many general anesthetics are known to act by potentiating the transmission of γ-aminobutyric acid, leading to depression of neuronal function and conscious processing.⁴  Ketamine, however, does not bind with high affinity to the γ-aminobutyric acid type A receptor. Ketamine is primarily a noncompetitive N-methyl-d-aspartate (NMDA) antagonist and has several unique clinical features. The electroencephalography (EEG) studies have shown its distinctive electrophysiologic profiles, including an increase of high beta and gamma oscillations,^5,6  unlike common γ-aminobutyric acid–mediated (GABAergic) anesthetic drugs, which decrease high-frequency activity and produce typical slow oscillations. Ketamine’s unique excitatory effects are attributed to its inhibitory action at the NMDA channels on inhibitory interneurons. Blocking these active inhibitory interneurons at a low dose, ketamine allows downstream excitatory neurons to become disinhibited without proper modulation by the interneurons.^7,8  At a larger dose of ketamine, the NMDA receptors on the excitatory glutamatergic neurons are also blocked, resulting in more cortical suppression. These findings together suggest that ketamine may change neuronal dynamics in stages as its effect deepens.

Ketamine is a well-established analgesic. Ketamine-induced analgesia is thought to be multimodal, including diverse physiologic processes.⁹  Potent analgesic effects with subanesthetic ketamine suggest that nociceptive stimulation is not perceived as pain, which is thought to be attributed to ketamine’s blockade of the NMDA receptor activity and consequent inhibition of central sensitization.^9,10  However, whether the sensory signal is processed in the cortex but is not perceived under ketamine, or the signal itself is altered during transmission before it reaches to the higher cortical area, is not clearly understood. Altered cortico-cortical connectivity is shown during ketamine anesthesia^11–14  and suggests that sensory information may be altered during its cortico-cortical transfer. Although ketamine’s effects on cortico-cortical connectivity seem to be mainly inhibition, its region-selective effects have been reported, including an increase of functional connectivity to the dorsolateral prefrontal cortex.¹³  It is thus crucial to understand sensory signal processing under ketamine in an interconnecting cortical network while the cortical dynamics are simultaneously measured.

Here we investigated how neural dynamics change during ketamine-induced loss of consciousness and return of consciousness. Anesthetic-induced state transitions are suggested to correlate with distinct neural changes with propofol¹⁵  and isoflurane.¹⁶  We hypothesized that ketamine-induced loss of consciousness and return of consciousness are associated with distinct neural changes in a primate cortical network and that the cortical regions are functionally disconnected. We further hypothesized that sensory information is disrupted in the hierarchical network during cortico-cortical transfer. In order to investigate these hypotheses, we developed a nonhuman primate model and performed direct neural recording from a functionally and anatomically interconnecting somatosensory (primary somatosensory cortex and secondary somatosensory cortex) and ventral premotor area network.^17–19  The ventral premotor area is known to link sensation and decision-making as well as to integrate multisensory modalities.^17,20–25  We investigated how the network-specific somatosensory stimulation and cross-modal auditory stimulation are processed under ketamine-induced oscillatory dynamics change. Further, we have successfully determined behavioral endpoints during ketamine anesthesia and recovery, such as loss of consciousness and return of consciousness, as well as full task performance recovery defined as return of preanesthetic performance level (performance return) based on the animal’s task response behaviors.

All animals were handled according to the institutional standards of the National Institutes of Health (Bethesda, Maryland) and according to an animal protocol approved by the institutional animal care and use committee at the Massachusetts General Hospital (Boston, Massachusetts). We used two adult male monkeys (Macaca mulatta, 10 to 12 kg). Before starting the study, a titanium head post was surgically implanted on each animal. A vascular access port was also surgically implanted in the internal jugular vein (Model CP6, Access Technologies, USA). Once the animals had mastered the following task, before the recording studies, extracellular microelectrode arrays (Floating Microelectrode Arrays, MicroProbes, USA) were implanted into the primary somatosensory cortex, the secondary somatosensory cortex, and the ventral premotor area through a craniotomy (fig. 1A). Each array (1.95 × 2.5 mm) contained 16 platinum-iridium recording microelectrodes (approximately 0.5 MΩ, 1.5 to 4.5 mm staggered length) separated by 400 µm. The placement of arrays was guided by the landmarks on the cortical surface (fig. 1A) and stereotaxic coordinates.²⁶  A total of five arrays were implanted in monkey 1 (two arrays in the primary somatosensory cortex, one in secondary somatosensory cortex, and two in the ventral premotor area in the left hemisphere) and four arrays in monkey 2 (two arrays in the primary somatosensory cortex, one in the secondary somatosensory cortex, and one in the ventral premotor area in the right hemisphere). A secondary somatosensory cortex array in monkey 2 did not provide stable signals due to unknown damage. The responsiveness to the somatosensory (air puff) stimulation to the face was tested in single neuron spikes recorded from all arrays after implantation. The recording experiments were performed after 2 weeks of recovery after the array surgery. All experiments were conducted in the radio frequency–shielded recording enclosures.

The animals were trained in the behavioral task shown in figure 1B. After a start tone (1,000 Hz, 100 ms), the animals were required to initiate each trial by holding the button located in front of the primate chair using the hand ipsilateral to the recording hemisphere. They were required to keep holding the button until the task end in order to receive a liquid reward. The animals were trained to perform correct response greater than 90% of the trials consistently for longer than approximately 1.5 h in an alert condition. The animal’s performance during the session was monitored and simultaneously recorded using a MATLAB- based behavior control system.^27,28  The trial-by-trial behavioral response was analyzed as a correct response (button holding until the trial end and release), failed attempt (early release, late touch, or no release of the button), or no response (figs. 1C and 2A). Loss of consciousness was defined as the time at which the probability of task engagement, including correct responses and all failed attempts, falls under less than 0.3, and return of consciousness was defined as the first time, since being unconscious, at which the probability of the engagement recovers greater than 0.3.^29,30  We also defined return of preanesthetic performance level (performance return) as the first time, since being unconscious, at which the probability of a correct response was greater than 0.9. Additionally, we tested arousability in two separate recording sessions in one animal. We applied a series of stimuli (nonaversive ear-pulling, a loud white noise at 100 dB of sound pressure level for 5 s, and hand claps three times at 10 cm from face) at 3 min, 10 min, and 30 min from the loss of consciousness, at the end of the anesthetic infusion, and at 10 min after the end of infusion.

During each trial, one of the four sensory stimulus sets (air puff, sound, simultaneous air puff and sound, no stimulus, fig. 1B) was delivered to the animal at a random delay regardless of their button responses. Air puffs were delivered at 12 psi to the lower part of the face contralateral to the recording hemisphere, via a computer-controlled regulator with a solenoid valve (AirStim, San Diego Instruments, USA). The eye area was avoided from the puff stimulation. Sound stimuli (pure tone at 4,000 Hz at 80 dB of sound pressure level) were generated by a computer and delivered using two speakers at 40 cm from the animal. White noise (50 dB of sound pressure level) was applied throughout the trial to mask the air puff and mechanical noises. All the stimulus sets were randomly presented to the animal regardless of their behavioral response throughout the recording session.

After the 30 min of awake performance, ketamine was infused for 60 min at a fixed rate (100 µg · kg^–1 · min^–1) through a vascular access port. The infusion rate of ketamine was determined to induce loss of consciousness in approximately 10 min for each animal. No other sedatives or anesthetics were used during the experiment. The animal’s heart rate and oxygen saturation were continuously monitored throughout the session (CANL-425SV-A Pulse Oximeter, Med Associates, USA). The animals maintained greater than 94% of oxygen saturation throughout the experiments.

Neural activity was recorded continuously and simultaneously from the primary somatosensory cortex, secondary somatosensory cortex, and ventral premotor area through the microelectrode arrays while the animals were alert and participating in the task and throughout anesthesia and recovery. Analog data were amplified, band-pass filtered between 0.5Hz and 8 kHz, and sampled at 40 kHz (OmniPlex, Plexon). Local field potentials were separated by low-pass filtering at 200 Hz and down-sampled at 1 kHz. The spiking activity was obtained by high-pass filtering at 300 Hz, and a minimum threshold of 3 SDs was applied to exclude background noise from the raw voltage tracings on each channel.

The recordings under ketamine anesthesia were performed five times in monkey 1 and seven times in monkey 2. In separate sessions, the recordings were performed under ketamine in the blindfolded animals that were performing the task (one session in monkey 1 and two sessions in monkey 2) and in the animals that were required no task performance (two sessions in monkey 2). In addition, multiple recordings were performed in alert performing animals without anesthesia.

Local field potential data analysis. All local field potential analyses were performed using existing and custom-written functions in MATLAB (Mathworks, Inc., USA). The local field potential frequency domain was segmented into slow-delta (0.5 to 4 Hz), theta (4 to 7 Hz), alpha (7 to 12 Hz), low-beta (12 to 18 Hz), high-beta (18 to 30 Hz), and gamma (30 to 65 Hz) bands. Multitaper spectral analysis of the local field potential, using the Chronux toolbox, was used to generate spectrograms (fig. 2, B–D), power changes over time (fig. 2E), and spectra (fig. 2, F–H). For the full session analysis, 30-s nonoverlapping time windows were used when generating spectrograms or calculating power changes in a particular band over time. Statistical analyses on the power change of frequency bands were performed by comparing the average preanesthetic power to the consecutive timepoints, using ANOVA test and post hoc Bonferroni multiple comparisons with a significance level of 0.05. Channel-to-channel coherence calculations were performed using the Chronux toolbox function, applying 2-s nonoverlapping windows (fig. 3A). Coherence for each frequency band was computed for 1 min for each epoch. We then analyzed the change of local and interregional coherence between all pairs of channels and compared all channel-to-channel changes between awake versus loss of consciousness, anesthesia, or return of consciousness, using the ANOVA test and post hoc Bonferroni multiple comparisons test with a significance level fixed at 0.05 (fig. 3B). Evoked potentials were analyzed using the raw, unfiltered local field potential, and normalized using the baseline local field potential over the 100 ms preceding each stimulus (fig. 4). For each comparison, the evoked responses’ maximum amplitude and latency to peak were statistically tested using a two-tailed unpaired t test, assuming equal variances, with a significance level of 0.01. Time frequency plots for evoked power change during trial were analyzed using 0.5-s nonoverlapping windows (fig. 5).

All single-unit analyses were performed using custom written software in the MATLAB programming environment using standard signal processing and statistical toolboxes. Briefly, we used template matching and principal component analysis based on waveform parameters to isolate single units.³¹  We analyzed only well-isolated units with identifiable waveform shapes and adequate refractory periods. All these units from two animals across different recording sessions were analyzed as they were independent. Floating microelectrode arrays placed on the cortical surface naturally pulsate with the brain and drift over time and are thought to provide new data sampling in each recording session. Peristimulus time histograms were constructed for each unit by binning the spike activity data into 1-ms bins and convolving a Gaussian function (SD = 50 ms) during a 4-s time window centered on the stimulus delivery. Single units were categorized into three groups (bimodal puff and sound responsive, unimodal puff responsive, or unimodal sound responsive units) based on their response to a puff or sound stimulus during wakefulness. On each trial, average prestimulus firing rate (for 500 ms before the stimulus) was compared against average poststimulus firing rate (for 500 ms immediately after the stimulus) using a two-tailed paired t test (α = 0.01). Thus, each unit was exclusively classified as being puff, sound, or bimodally responsive. For population level peristimulus time histograms, each unit was normalized to its baseline activity using Z-scores, and then averaged across units for each subgroup for 100 trials in each epoch (fig. 6). Average firing rates over the course of entire session were calculated by normalizing each unit’s firing rate to its preanesthetic values using Z-scores. We ran a 1-min Gaussian window with 59-s overlap to obtain a smooth estimate, and then averaged across units. ANOVA and post hoc Bonferroni multiple comparisons were performed to compare the average preanesthetic firing rate and each following time point, using a significance level of 0.05.

We have successfully identified the behavioral endpoints, such as loss of consciousness, return of consciousness, and performance return, in this primate model (fig. 1C). Return of consciousness was identified most of the recording sessions (10/11 sessions in two animals); however, performance return was detected in 7 out of 11 sessions. We have also tested arousability in one animal. A series of brief nonaversive stimuli (ear-pulling, a loud white noise, and hand claps, as described in Materials and Methods) was applied after loss of consciousness at 3 min, 10 min, and 30 min, at the end of anesthetic infusion, and at 10 min after the end of infusion. There were no task attempts at any of these time points in the animal, suggesting no arousability after loss of consciousness.

We first analyzed the dynamics in spectrograms in the primary somatosensory cortex, secondary somatosensory cortex, and ventral premotor area. Beta oscillations were dominant in all regions during wakefulness. After the start of ketamine infusion, while the animal was still performing the task, the beta oscillations were abruptly disrupted (fig. 2, A–D, high beta in fig. 2E). We then found a brief increase in the alpha oscillations in the primary somatosensory cortex (fig. 2, B and E), and the animal’s loss of consciousness was identified when the wide-band high-beta gamma oscillations were gradually emerging (fig. 2, B–E). The slow oscillations did not increase at loss of consciousness, contrary to their immediate increase after loss of consciousness during propofol anesthesia.¹⁵  The slow-delta, theta, and alpha oscillations, however, appeared to increase significantly as ketamine infusion continued (fig. 2, B–E). Return of consciousness was observed during a gradual decrease of the gamma band in its power and frequency. On this continuous drift of the gamma oscillations, performance return was identified with the oscillations approaching to their awake beta frequency range. However, the frequency and power of the beta oscillations were not fully returned to the awake level at the time of performance return (fig. 2, B–D, F–H). Loss of consciousness, return of consciousness, and performance return did not appear to correspond with distinctive neural changes.

We next investigated how ketamine affects communication across these brain regions by examining both local and interregional coherence changes. Corresponding to rather abrupt disruption of the beta oscillations shown in the spectrograms in the primary somatosensory cortex, secondary somatosensory cortex, and ventral premotor area (fig. 2, B–D), local and interregional high-beta coherence (18 to 30 Hz) between the primary somatosensory cortex, secondary somatosensory cortex, and ventral premotor area were all significantly decreased at loss of consciousness, under ketamine and at return of consciousness (high beta in fig. 3, A and B). In the slow-delta frequencies, coherence significantly increased within the primary somatosensory cortex and between the primary somatosensory cortex and higher regions (secondary somatosensory cortex, ventral premotor area) and decreased within secondary somatosensory cortex and ventral premotor area and between secondary somatosensory cortex and ventral premotor area during anesthesia (slow-delta in fig. 3, A and B). Consistent changes were found in theta and alpha coherence. Coherence of the gamma oscillations significantly increased both locally and interregionally in all regions during ketamine anesthesia (gamma in fig. 3, A and B). Overall ketamine-induced coherence changes appeared to be region- and frequency-specific. Additionally, the data indicate that local and interregional coherence, except for high-beta coherence, significantly change after loss of consciousness during the course of ketamine anesthesia.

We further investigated sensory processing during these oscillatory dynamics under ketamine. We first examined evoked local field potential responses to puff and sound, analyzing the average amplitude and latency to the peak of the evoked potentials, as summarized in figure 4. In the primary somatosensory cortex, the amplitude of puff-evoked response was statistically significantly decreased during ketamine anesthesia (t₆₈ = –5.40, P = 9.04^-7), but the evoked responses remained. The puff-evoked responses recovered by the time of performance return, not being significantly different from the awake (t₁₀₉ = –2.34, P = 0.021, fig. 4A). In addition, the latency of the puff-evoked response was significantly prolonged in the primary somatosensory cortex during ketamine anesthesia when compared to wakefulness (t₆₈ = –8.38, P = 4.38^-12). This latency was still significantly delayed at performance return (t₁₀₉ = –6.54, P = 2.06^-9). There were no significant effects of ketamine-induced altered states on the sound-evoked responses in the primary somatosensory cortex (amplitude: t₆₇ = –0.29, P = 0.77; latency to peak: t₆₇ = –2.53, P = 0.013, fig. 4B). In ventral premotor area, there was no statistically significant effect of ketamine on puff-evoked response amplitude (t₆₈ = 1.36, P = 0.18). However, the latency to peak after puff stimulation was significantly prolonged during ketamine anesthesia (t₆₈ = 2.88, P = 0.005), and it was still prolonged at performance return (t₁₀₉ = 3.75, P = 2.86^-4, fig. 4D). The evoked response amplitude to sound stimulation was not significantly different in ventral premotor area (t₆₇ = –0.53, P = 0.60), but the latency was prolonged under ketamine (t₆₇ = 2.96, P = 0.004, fig. 4E). The time-frequency evoked spectrograms demonstrated that the power responses to puff and sound both appeared to become dispersed and prolonged during the pre–loss of consciousness to loss of consciousness period (fig. 5). The puff-evoked power responses recovered at performance return in the primary somatosensory cortex. However, the puff-evoked response appeared to recover in the ventral premotor area before return of consciousness while the animal was still under ketamine infusion (fig. 5, A and D). Sound-evoked power responses seemed to delay in recovery (fig. 5, B and E), suggesting prolonged inhibition of multisensory processing under ketamine. Together, while puff- and sound-evoked potentials appeared to be preserved with small but statistically significant effects on amplitude and latency of the response, evoked power responses additionally demonstrated that power components were dispersed in frequency beginning in the pre–loss of consciousness period.

Last, we focused on single-unit responses during ketamine anesthesia and recovery. We recorded 233 well-isolated single units from these cortical regions in two monkeys, of which 112 (48%) were responsive to either somatosensory or auditory stimulation (n = 74 in primary somatosensory cortex; n = 38 in ventral premotor area). Secondary somatosensory cortex units were recorded only from one monkey and were not included in the analyses. We found that ketamine did not change the average firing rate in the primary somatosensory cortex units or in the ventral premotor area units (fig. 6B). We then identified distinct subpopulations of units on the basis of their responsiveness to the sensory stimulation during wakefulness: units with enhanced firing response to puff, units with suppressed response to puff, bimodal units with enhanced response to puff and sound, and bimodal units with suppressed response to puff and sound. Puff responses appeared to maintain their response magnitude during ketamine anesthesia in unimodally puff-responding units in the primary somatosensory cortex and ventral premotor area (fig. 6C). In the bimodally responding units in the primary somatosensory cortex, sound responses were diminished during ketamine anesthesia, suggesting selective inhibition by ketamine. In contrast to the enhanced puff responses, suppressed puff responses in the primary somatosensory cortex and ventral premotor area and bimodally enhanced responses in ventral premotor area appeared to be diminished under ketamine and delayed in recovery (fig. 6D). There was not a sufficient number of neurons at performance return for the comparison.

Our results demonstrate that the nonhuman primate model can be successfully used in determining anesthetic-induced behavioral endpoints that are relevant to the ones in humans.²  We determined loss of consciousness and recovery endpoints, such as return of consciousness for earlier detection of the initial moment of recovery and performance return for full task performance recovery, based on the probability of task engagement and task performance. Compared to rather consistent loss of consciousness, we found that recovery phase prolonged and recovery endpoints varied significantly in each animal. The state-space model paradigm was thus valuable to predict the animal’s task response state in variable courses of anesthetic recovery.³⁰  Moreover, full performance recovery after anesthetic-induced unconsciousness has not been demonstrated previously with concurrent intracortical dynamics.

We also demonstrate that ketamine generates unique intracortical dynamics during its altered states of consciousness. The evolution of the wide-band high-beta gamma oscillations was characteristic and followed an abrupt disruption of the coherent beta oscillations in all primary somatosensory cortex, secondary somatosensory cortex, and ventral premotor areas. The beta disruption was the only abrupt change in the dynamics observed during ketamine anesthesia and recovery. This disruption appeared to be associated with the animal’s declining performance, but not with the time of loss of consciousness. The synchronized beta oscillations are thought to bind multiple sensorimotor areas.³²  The beta oscillations are associated with supramodal information and perceptual decision-making in the premotor system.^33,34  Recent data suggest that these beta events emerge locally within neocortex from the integration of synchronous bursts of subthreshold excitatory synaptic activities.³⁵  The observed abrupt interruption of the beta oscillations can be explained by disinhibition of cortical excitatory neurons by preferential inhibition of inhibitory interneurons,^7,8  resulting in the failure of maintaining organized cortical excitation, at a critical ketamine concentration.

In contrast to distinctive neural changes that were associated with loss of consciousness and return of consciousness during propofol anesthesia,¹⁵  our current results demonstrate that ketamine-induced behavioral endpoints, such as loss of consciousness, return of consciousness, and performance return, were all identified during a gradual change in the oscillatory dynamics. Recovery of the coherent beta oscillations was also gradual, slowly shifting down from the gamma frequencies toward the range of awake beta oscillations. It is possible that these animals may not have undergone clear state changes under ketamine but were rather continuously conscious internally after loss of responsiveness that we defined as loss of consciousness. In human studies, subjects who underwent clear loss of responsiveness during ketamine anesthesia reported vivid dreams upon awaking.³⁶  Conscious experiences or dreaming during ketamine anesthesia as well as natural sleep are thought to be disconnected from the environmental input. Recent human EEG data suggest that high-frequency activity in a parieto-occipital zone correlates with dreaming, and an increase of low-frequency activity in the region is associated with the absence of dream experience.³⁷  Our local field potential data are consistent with the global EEG change observed in humans.^5,36  An increase of the slow-delta oscillations at the later stage of ketamine infusion may indicate a state of unconsciousness in these animals, suggesting that the internal state still changes after ketamine-induced loss of responsiveness.

The slow-delta oscillations are observed under various general anesthetics with different receptor mechanisms, including GABAergic propofol and sevoflurane and alpha2 adrenergic agonist dexmedetomidine.^3,15,38  An increase of the slow-delta oscillations suggests the inhibition of thalamocortical pathways or subthalamic regions.^39,40  Previous human studies suggested an increase in the slow-delta power is associated with unconsciousness.^2,41,42  High-dose nitrous oxide has also been shown to induce large-amplitude slow-delta oscillations.⁴³  The effects of nitrous oxide are thought to be mediated through its blockade of NMDA receptors, similar to the ones of ketamine. Contrary to delayed development of the slow-delta oscillations under ketamine, these other anesthetics, including nitrous oxide, generate the slow-delta oscillations when subjects become unconscious or unresponsive, suggesting different neural processes in generating unresponsiveness by ketamine. In addition, our data demonstrate that the gamma power maintained its incline until the end of ketamine anesthesia despite an increase in the slow-delta oscillations. The mechanisms of cogrowing gamma and slow-delta oscillations are not clear and warrant further investigation.

Despite these dramatic oscillatory changes, somatosensory tactile responses appear to be largely preserved in both the primary somatosensory cortex and ventral premotor area throughout ketamine anesthesia and recovery. However, temporal selectivity of the sensory information was decreased before loss of consciousness, likely associated with emerging high-beta and gamma oscillations. In contrast, cross-modal sound responses were selectively diminished in these regions. Bimodally responding single units and units with suppressed firing responses also appear to be inhibited under ketamine and delay in recovery. These changes were all observed in the primary somatosensory cortex along with ventral premotor area, suggesting that the sensory signal is altered within or before the primary somatosensory cortex, before entering to cortico-cortical circuits. These findings are not consistent with our hypothesis on hierarchical inhibition of sensory signals during cortico-cortical transfer by ketamine.

Interestingly, the effect of ketamine on sensory processing seems to be comparable with propofol. Somatosensory responses were preserved but multisensory responses were inhibited during both ketamine and propofol anesthesia despite a vast difference in oscillatory dynamics and single unit firing.¹⁵  Olcese et al. suggested that sensory processing exists across a variety of unconscious states, but a strong degree of synchrony may prevent specific cell ensembles from effectively communicating.⁴⁴  We found that ketamine significantly increased gamma band coherence across regions and slow-delta, theta, and alpha coherence in the primary somatosensory cortex and between the primary somatosensory cortex and other regions. Propofol also generates robust synchrony in the slow-delta and alpha oscillations.¹⁵  These results suggest that strong synchrony under these general anesthetics prevents effective multisensory communication. Our results provide a first demonstration at the single-unit level that ketamine and propofol, two anesthetics with distinct receptor mechanisms, both preserve simple nonaversive somatosensory information, but inhibit multisensory processing in a neocortical network.

In conclusion, ketamine generates unique intracortical dynamics during its altered states of consciousness, including the characteristic evolution of high-beta gamma oscillations preceding loss of consciousness. Ketamine-induced loss of consciousness, return of consciousness, and performance return are identified during a gradual change of the high-beta gamma oscillations, and are not associated with distinctive neural changes, suggesting fundamentally different neuronal processes from propofol. Under these dramatic oscillatory changes, nonaversive tactile processing is preserved in the sensory and premotor network, but multisensory processing appears to be diminished during ketamine anesthesia and recovery, suggesting prolonged multisensory inhibition by ketamine. Future use of this primate model will further allow us to investigate nonaversive versus nociceptive information processing under ketamine and various general anesthetics.

The authors thank Tatsuo Kawai, M.D., Massachusetts General Hospital, Boston, Massachusetts, for performing vascular port surgeries; Anne C. Smith, Ph.D., Evelyn F. McKnight Brain Institute, University of Arizona, Tucson, Arizona, for providing expert assistance in the behavioral analyses; and Warren M. Zapol, M.D., Massachusetts General Hospital, for guiding and supporting the project development.

This work was supported by the Foundation for Anesthesia Education and Research (Schaumburg, Illinois), the National Institutes of Health (Bethesda, Maryland; grant Nos. 5T32GM007592 and 1P01GM118269), and Harvard Medical School (Eleanor and Miles Shore 50th Anniversary Fellowship Scholars in Medicine; Cambridge, Massachusetts).

The authors declare no competing interests.