Desflurane Selectively Suppresses Long-latency Cortical Neuronal Response to Flash in the Rat

The experimental procedures and protocols used in this investigation were approved by the Institutional Animal Care and Use Committee of the Medical College of Wisconsin (Milwaukee, Wisconsin). All procedures conformed to the Guiding Principles in the Care and Use of Animals of the American Physiologic Society and were in accordance with the Guide for the Care and Use of Laboratory Animals, National Academy Press, Washington, D.C., 1996.

Adult male Sprague-Dawley rats were housed on 12-h light/dark cycle at a constant temperature of 23 ± 1°C with free access to food and water for 2 weeks before each experiment. Animals were anesthetized for surgery with desflurane, tracheotomized, paralyzed with gallamine triethiodide (80 mg, IV), and artificially ventilated using a rodent ventilator (SAR 830/P; CWE, Ardmore, PA) with the mixture of 30% oxygen in nitrogen plus desflurane. Inspired and expired desflurane, oxygen, and carbon dioxide concentrations were monitored by using a gas analyzer (POET II; Criticare Systems, Inc., Waukesha, WI). Femoral arterial and venous lines were placed for monitoring blood pressure and periodic check of blood gases. Rectal temperature was maintained at 37 ± 1°C with a thermostat-controlled heating apparatus (TC-1000; CWE).

The head was secured in a stereotaxic frame (model 900; Kopf, Tujunga, CA). A craniotomy of approximately 5 × 5 mm over the right occipital cortex was prepared; the dura was left intact. A 7 × 7 rectangular array of electrodes (“Utah” array; length 1.5 mm; spacing 0.4 mm) was inserted into the cerebral cortex by using a pneumatic inserter (Cyberkinetics Neurotechnology Systems, Salt Lake City, UT). The electrodes extended over a 2.4 × 2.4–mm area containing the primary visual cortex with about half of the electrodes in the monocular and the other half in the binocular visual cortex according to the map of Zilles34(fig. 1). The electrode array was initially inserted to a depth of 0.6 mm. The array was then advanced in 0.1-mm increments by using a micromanipulator to a depth of 1.0-1.5 mm to facilitate recording from infragranular neurons. For local analgesia, all surgical sites were infiltrated with bupivacaine. After surgery, the room was darkened and desflurane concentration was kept at 8% for a 1-h equilibration period before neuronal recording was initiated.

Multiunit activity from 49 electrodes was simultaneously recorded by using a 50-channel neural signal acquisition system (Cyberkinetics Neurotechnology Systems). Extracellular spikes were auto-thresholded by using a root mean square multiplier of 2.4 at the end of the equilibration period at 8% desflurane; the settings were then left unchanged for the remainder of the experiment. Desflurane concentration was adjusted sequentially to 4, 2, 6, and 8%, allowing a 30-min equilibration time. The neuronal response to flash was recorded at each level of the anesthetic. Flashes were generated by using a blue (468 nm) light-emitting diode positioned 10 mm from the eye. The light cone emanating from the diode produced diffuse illumination of the entire eye. The diode “on” time was 2 ms at +5V Transistor-Transistor-Logic pulse. Sixty flashes were delivered at 5-s intervals, yielding a total stimulation time of 5 min in each condition. The left eye was stimulated in all experiments and at all anesthetic levels, and the right eye was covered. In addition, in six of the animals at 2% desflurane, the left eye and the right eye were stimulated in alternation at 5-s intervals. In addition to multichannel unit activity, a time marker corresponding to each flash was recorded through an analog channel.

The activities of individual units from the chosen channels were distinguished by using the public domain offline spike sorter PowerNAP (OSTG, Inc., Fremont, CA). The software applies Principal Component Analysis followed by cluster analysis to sort the spikes. Principal Component Analysis defines the linearly dependent factors in the spike data and transforms them to an ordered set of orthogonal basis vectors that capture the directions of largest variation. As the first step, T-distribution was applied to remove synchronous noise artifacts from data. Synchronous artifacts were related to external noise, such as an electrical pulse related to flash generation. This was based on the detection of spike events that occurred on at least half of the units within an interval of 0.5 ms. A scatter plot was then created by using the first two principal components, and K-means clustering was applied to define the cluster boundaries of individual units. Occasional remaining outliers were removed manually. On average, 11% of the traces were rejected as noise. When the spikes were sorted and artifact-free, peristimulus raster plots and peristimulus time-frequency histograms were generated by using the software NeuroExplorer (Version 3.02; Nex Technologies, Littleton, MA). Further analysis was limited to units with an average spike rate of at least 0.3 spikes per second, representing approximately 90% of all active units.

To determine the number of units responding to flash, we tested the statistical significance of the change in spike rate after flash across the trials in each unit at each anesthetic condition. The flash response was divided into an early component (0–100 ms) and a late component (150–1000 ms). We tested for the significance of both response components as compared to the mean firing rate of each unit during a 1-s prestimulus baseline. To this end, individual t  tests with a generalized Bonferroni correction were applied. Since on the average we found approximately 50 units per experiment, a uniform alpha value of 0.001 was used. Statistical testing was done using Matlab Statistical Toolbox (Mathworks, Inc., Natick, MA). Once the responding units were identified, baseline firing rates and magnitudes of early and late responses (i.e. , changes in firing rate from baseline) were averaged across all units within each animal.

The effects of desflurane on baseline and flash-induced firing rates were estimated with linear regression with the anesthetic concentration as continuous and the subject as categorical independent variables. A significant deviation of the slopes from zero was tested with ANOVA. The same procedure was used to test whether the choice of stimulation side (left vs.  right eye) had a significant effect on firing rates. To test for statistical significance of desflurane effect on the number of active and flash responding units, logistic regression was used; again with the anesthetic concentration or stimulation side as continuous and subject as categorical independent variables. The odds ratio of having a late response with versus  without early response was compared by using the chi square test with subject as a blocking variable. Reported means and standard deviations were calculated by averaging within-subject means. Statistical calculations were performed with the software SATA/IC 10.0 for Windows (StataCorp LP, College Station, TX).

The distribution of prestimulus spike rate data was left skewed and could be normalized by logarithmic transformation. Therefore, as appropriate, geometric means and 95% confidence intervals of the original distributions of the prestimulus spike rate were presented.

In eight animals, a total of 578 units from 398 sites were recorded. The average yield was 1.5 ± 0.6 units per electrode when the overall baseline activity was the highest. A few electrodes did not reveal distinguishable spikes. On those electrodes that carried spikes, usually two and occasionally three units could be separated by spike sorting. The average baseline (prestimulus) firing rate at the lightest plane of anesthesia (2% desflurane) was 1.69 (1.60–1.78) spikes/s.

When the concentration of desflurane was increased, there was a small but significant reduction in the number of units active at baseline (table 1, P < 0.001, linear trend). In addition, the baseline firing rate decreased in a gradual manner (fig. 2, P < 0.001).

When the left eye was stimulated with flash, approximately 75% of the units responded with a significant (P < 0.001) increase in firing rate. Although the response pattern varied from unit to unit; a few typical variants could be identified (fig. 3). The majority of responding units at a concentration of 2% showed a distinct increase in firing rate within the first 100 ms, usually peaking at 40 ms. In this paper, we call this the early response. In approximately 60% of the units, the early response was followed by a secondary increase in firing rate that we call late or long-latency response. The late response usually started around 150 ms after flash; it was smaller but more prolonged than the early response, extending up to 1 s poststimulus. It often consisted of multiple bouts. A short period of suppressed firing between the early and late responses was often seen.

A small number of units produced a late response in the absence of an early response. The odds of observing a late response was twice as high with, than without, a preceding early response of the same unit (chi-square, P < 0.001). This proportion was independent of the depth of anesthesia (logistic regression, P = 0.179). The dependence of the late response on anesthetic concentration and stimulation side was similar with or without an early response. Therefore, all units that produced a late response were pooled in the subsequent analysis.

A small fraction (approximately 4%) of the units showed a negative late response that is a decrease in firing rate after flash. They did not seem to be affected by desflurane concentration and, in light of their scarcity, were not analyzed any further.

When desflurane concentration was increased, there was also an increase in the number of units with an early response (table 1, P < 0.001). In contrast, the number of late-responding units was reduced (table 1, P < 0.01).

The effect of desflurane on the magnitude of unit response to flash was calculated separately for the early response (0–100 ms vs.  baseline) and late responses (150–1000 ms vs.  baseline). Only those units that showed a statistically significant increase in firing rate according to the preceding t  test were included. As figure 4shows, the early response in firing rate did not change with anesthetic concentration (P = 0.158). In contrast, the late response was significantly reduced as the anesthetic concentration increased (P < 0.001). The overall reduction in firing rate was approximately 60% from the lowest to the highest desflurane concentration.

The spatial distribution of unit response to flash is illustrated in figure 5. As seen there, the majority of early-responding units to flash are distributed toward the upper left quadrant of the electrode array (same orientation as in fig. 1), corresponding to the monocular visual cortical area. Increasing desflurane concentration augments the early spike response and also recruits units in a wider area. The late-responding units are distributed more widely at shallow anesthesia weighted toward the binocular cortical area. The late response is attenuated, and the number of responding units is reduced with increasing desflurane.

For an additional, model-based analysis of the early and late flash response, we examined the effect of desflurane on the unit population response. Peristimulus spike rate histograms were calculated from all trials at a resolution of 25-ms time bins at the four examined anesthetic concentrations. The spike rate histogram obtained at the lightest anesthetic level (2% desflurane) was taken as a reference and was subjected to Principal Component Analysis. Predicted spike rate histograms (component scores) were calculated by matrix-multiplying the principal components with the original spike rate histograms obtained at 2% desflurane. Predicted spike rate histograms obtained from the first two or three statistically significant principal components, were used as canonical patterns to model the spike rate response of individual units at all four anesthetic concentrations. The latter was accomplished by fitting the canonical histograms to the original peristimulus spike rate histograms by using a general linear model.35For further data reduction, a population response was calculated by averaging the modeled peristimulus spike rate histograms from all units. As illustrated in figure 6, the results confirmed the concentration-dependent suppression of the late population response along with an (transient) augmentation of the early population response.

Finally, in six of the animals we compared the responses to left and right eye stimulation at 2% desflurane. The results are summarized in table 2. Looking at the early response, a smaller number of units responded to right eye stimulation than to left eye stimulation (P < 0.001). The early response in firing rate was also smaller with right eye than left eye stimulation (P < 0.001). In case of the late response, the side of stimulation made no significant difference in the number of responding units (P = 0.529) or in the change in firing rate (P = 0.251).

This study employed flash activation of the rat visual system as a model to test the concentration-dependent effect of desflurane on the reactivity of cortical sensory neurons. The rat is often considered a nonvisual animal; however, ample evidence supports that it can perform cognitive-behavioral tasks demanding complex visual discriminations.36A surprisingly large area of the rat cerebral cortex is devoted to processing visual stimuli – a region that approximates the size of the barrel cortex used for vibrissal exploration.37Furthermore, visual cortex neurons show learning and reward-dependent plasticity.38Our experience with this model shows that the cortical response to flash stimulation in albino rats is highly reproducible and robust.26–28,39Recently, brain imaging studies have adopted flash stimulation in the rat as a useful model to study neurofunctional networks.29 

There are several sources of variance in the data of the current study, some of which have not been fully explored. These include unit-to-unit, trial-to-trial, animal-to-animal, and anesthetic-dependent variations. In our analysis, unit-to-unit variance was reduced by prescreening the units that responded significantly (t  tests) and subsequently categorizing them into subsets for further analysis. Animal-to-animal variance and unit-to-unit variance were explicitly represented in the statistical models used to test the anesthetic effect. Trial-to-trial variation was not considered explicitly; it was averaged over by reducing the data to peristimulus histograms. There are novel ways to address the effect of trial-to-trial systematic variations of spike trains40–this may be considered for further analysis in the future.

First we found, perhaps not unexpectedly, that spontaneous unit activity in visual cortex was reduced by desflurane in a concentration-dependent manner. It has been known that older general anesthetics reduce baseline neuron firing in visual1and other sensory systems.41,42Ogawa et al.  43,44found that volatile anesthetics halothane, isoflurane, enflurane and sevoflurane suppressed multiunit activity in the reticular formation, but they did not study cortical units and only investigated the effect of anesthetic concentrations higher than those that produce loss of consciousness. Villeneuve and Casanova45found that volatile anesthetics suppressed spontaneous cell firing in cat primary visual cortex; isoflurane was more potent than halothane. Again, mainly deeper levels of anesthesia were studied.

Our main question was how desflurane influenced the cortical unit response to visual stimulation. When we stimulated the contralateral eye with light flashes at the lightest anesthetic level, approximately three-quarters of the units responded with an increase in firing rate. Burne et al.  4found in awake animals that 90% of rat visual cortical neurons responded to visual stimuli. They used stationary and moving stimuli, not flash, therefore these figures are not directly comparable. We cannot exclude that the number of responding neurons at 2% desflurane may have been reduced compared to that in waking. Nevertheless, in the range of 2% to 8%, desflurane did not attenuate the early response and the number of responding units was even increased.

Previous studies examined the concentration-dependent effects of various volatile anesthetics, but not of desflurane, on visual cortical neuronal responses. The early investigations by Adams and Forrester revealed that chloroform in small doses enhanced the responsiveness to flash stimulation.1Ikeda and Wright10found that halothane increased from 0.2 to 2.2% reduced the sensitivity of cat striate cortex neurons to stimulus orientation, spatial frequency, and contrast. Halothane suppressed the later components of firing more than the initial component–both within the time frame of our early response. In contrast, Tigwell and Sauter14found reliable neuron responses in monkey striate cortex at isoflurane concentrations 0.5 to 0.9%. Villeneuve and Casanova45compared the concentration-dependent effects of halothane/nitrous oxide and isoflurane/nitrous oxide on cat striate cortex neuron responses and found that isoflurane produced a greater suppression than did halothane. Neither anesthetic influenced the orientation and direction sensitivity and receptive field organization of the neurons.

Our results on the early response appear to be at variance with the general observation that the visual evoked potential (VEP) in humans is sensitive to general anesthesia.46Clearly, we have to be cautious of comparing intracortical unit recordings in rats with scalp-recorded field potentials in humans. However, the most important factor may be the stimulation frequency. We previously showed26,39that the VEP in the rat is preserved if light flashes are presented at long (5 s or longer) interstimulus intervals. These data are relevant because the early unit response peaking at 40 ms in layer 4–5 coincides with the N40 evoked potential with a dominant current sink in the thalamocortical input layer 4C.47,48Likewise, Rabe et al.  49found that VEP in the rat was not depressed by halothane in the concentration range of 0.25–1.0%.49Ogawa et al.  43found that VEP derived from the first 100 ms after flash was not suppressed by isoflurane at a concentration (0.8%) that presumably produces loss of consciousness. Moreover, it was shown that halothane did not attenuate the VEP in human patients when tested with low-frequency (0.5 or 1 Hz) stimuli.50We surmise that the human VEP, commonly recorded with flickering light, is attenuated by general anesthetics because thalamic and cortical neurons increasingly fail to follow high-frequency driving under anesthesia. However, the results from rats and humans appear to be consistent with each other when tested with low-frequency visual stimulation.

The reason for the preserved visual cortical responsiveness in spite of reduced background activity is unclear, but it appears to be true to various sensory modalities.51It may be the result of a transient increase in neuronal excitability as has been seen in the hippocampus in vitro .9Also, it may reflect a preferential anesthetic depression of feed forward inhibition,52leading to a larger response and/or more units being able to respond. Erchova et al.  7suggested that sensory specificity may increase with increasing anesthetic depth because of a reduction of associative inputs; the early response may be augmented by increasing spike synchronization. Others showed that cortical neuronal receptive fields may expand in light anesthesia and contract at deeper anesthesia.42 

Volatile anesthetics may also augment the sensory response by suppressing feedforward inhibition through γ-aminobutyric acid-mediated effects. Orth and Antognini53studied the effect of mesencephalic reticular formation stimulation on electroencephalogram activation during halothane and isoflurane anesthesia. They concluded that cortical neurons remain responsive at anesthetic concentrations associated with unconsciousness.

In contrast to the preserved reactivity of cortical units within 100 ms, the late component of the response was gradually reduced by deepening anesthesia. Very few studies have examined the effect of volatile anesthetics on cortical unit reactivity at latencies longer than 100 ms; as far as we know, none of them have studied desflurane. Early on, Robson13noted that the unit response to flash in cat visual cortex had several late components between 200 and 500 ms, which were more sensitive to anesthesia than was the primary response within 100 ms. Their results obtained with trichloroethylene are difficult to extrapolate to modern anesthetics such as desflurane. Ikeda and Wright11studied the effect of halothane/nitrous oxide on the reactivity of neurons in cat visual cortex and found that, in stage III anesthesia characterized by slow-wave electroencephalogram, the sustained response component (up to 5 s) of certain neurons was suppressed, whereas the primary response was preserved. Likewise, Chapin et al.  5found that 0.75% halothane depressed the long-latency (300 ms) excitatory firing of rat somatosensory cortical neurons after cutaneous stimulation of the forepaw, whereas short-latency responses within 50 ms were little affected. At variance with these findings, Tigwell and Sauter14reported that the sustained response up to 300 ms of monkey striate cortex neurons was preserved under anesthesia with 0.5–0.9% isoflurane + 73% nitrous oxide. Although the exact reason for this discrepancy is unclear, it may have been a result of the presentation of a prolonged visual stimulus as opposed to flash. Recently, Villeneuve and Casanova45reported that halothane and isoflurane reduced in a concentration-dependent manner single-cell responses in primary visual cortex of cats. Isoflurane was more potent, producing a 50% response reduction at 0.8 minimum alveolar concentration—roughly equivalent to our 4% desflurane level. A notable difference from our study was the use of nitrous oxide, which likely potentiated the anesthetic affects. The visual stimuli were complex and temporally extended; therefore, the effect of anesthesia on short and long-latency components of the unit response could not be investigated.

The reason for the differential reduction of the late versus  early response to flash is not entirely clear. General anesthetics are thought to produce a general decrease in neuronal excitability.3,54Several investigators suggested that various anesthetics suppressed the reactivity of thalamocortical neurons.2,6,12Detsch6in particular showed that isoflurane at more than 0.8% concentration suppressed the responsiveness of thalamocortical cells when driven by high-frequency (30–100 Hz) somatosensory stimuli. The initial response to the onset of the stimulus train was present, but the responses to subsequent stimuli were quickly attenuated. We presented the flash stimuli at substantially longer intervals (5 s) to facilitate the recovery of neuronal excitability after each stimulus. None of these findings can be easily reconciled with the anesthetics’ differential effect on the early and late response.

A possible explanation may lie in the difference in the underlying mechanisms of the early and late response. Chapin et al.  5suggested that the cortical long-latency response was brought about by nonspecific spino-reticulo-thalamic pathways. In awake animals, this response was exhibited by cells with large receptive fields, it attenuated rapidly to high stimulus frequencies (more than 2 Hz) and could be elicited by nonspecific, arousing stimuli. The early and late components were separated by an inhibitory period similar to that observed in this study. Long-latency responses could be selectively abolished by cryogenic blockade of centromedian thalamic nuclei.55It is also possible that the selective depression of the late response is a result of its dependence on cortical polysynaptic pathways.46Anesthetic depression of synaptic transmission may result in a cumulative loss of signaling such that the more synapses involved the greater overall suppression is produced. In turn, the most complex information processing would be affected the most.

Flash stimulation of the ipsilateral eye also produced a substantial neuronal response; although the early component was attenuated, the number of late-responding neurons and the magnitude of their firing rate increase were just as large as were obtained with the contralateral eye stimulation. As approximately half of the recording electrodes were targeted to the monocular and the other half to the binocular sensitive area of the primary visual cortex, assuming a nearly even illumination of the retina, we anticipated that approximately twice that many units will respond to contralateral than to ipsilateral stimulation. We saw no such difference; therefore, we surmise that the observed response, especially the late component, may reflect nonspecific or integrative processing that can be elicited from either eye.

The current data suggest that under anesthesia, the visual cortex does not lose its initial reactivity to visual stimuli, but the sustained component of its reactivity is attenuated. What implications may this have for information processing in the anesthetized brain? We know that general anesthetics at increasing doses produce sedation, amnesia, and behavioral unresponsiveness first to verbal and then also to painful stimuli. Unconsciousness ensues at an intermediate dose or concentration, well before nociceptive responses are lost. The concentration of desflurane for unconsciousness falls in the middle of those examined in this work (between 4% and 6%).27At this anesthetic depth, the cortex remains reactive, but is presumably no longer capable of integrating information at the conscious level.56 

Chapin et al.  5suggested that the selective reduction of long-latency response may contribute to the loss of consciousness associated with anesthesia. Whether they are related to so-called nonspecific activation or more complex sensory specific processes is not clear at this time. Although the flash is a simple stimulus, the evoked response likely reflects the brain’s ability to process information because the pattern of cortical population activity evoked by various stimuli is stereotypical, presumably reflecting the wiring of cortical local circuits, and similar to that associated with the performance of complex cognitive functions.57 

It has also been suggested that the long-latency response in primary visual cortex may be mediated by recurrent activity between sensory-specific and higher cortical processing regions and is necessary for conscious perception.58–60Conscious awareness of visual stimuli has been linked to cortical activation seen at 130–280 ms61or 270–300 ms.62,63We previously found that volatile anesthetics reduced feedback information transfer from frontal and parietal cortex to visual cortex at the time unconsciousness ensued.27Fronto-occipital electroencephalogram coherence at rest or after flash stimulation was also reduced.28,64Thus, the attenuation of the late response may also indicate the suppression of recurrent integrative information processing responsible for the anesthetic-induced unconsciousness. Although the mechanism of this effect has not been elucidated, it may, as already mentioned, be due to the accumulation of anesthetic effects at each synapse in a chain so that the complex integrative responses are more depressed because more synapses are involved.

Future extensions of this work could include a more detailed analysis of the temporal details of the flash response and its dependence on various anesthetics. Adams and Forrester1saw more than one component of the primary response, and we noted that the sustained response may consist of several components as well. Also, histologic identification and categorization of the recorded neurons would give further insight; Evarts8found that different populations of neurons may respond to anesthesia different ways.

In summary, our results indicate that the majority of visual cortex neurons do not become “undrivable” by flash stimuli under a range of anesthetic conditions with desflurane that include the transition from consciousness to unconsciousness. At the same time, the long-latency component of the neuronal response to flash is attenuated by desflurane in a concentration-dependent manner. Taken together with previous findings, the attenuation of long-latency neuron response may be related to a suppression of cortical communication and integration associated with the commencement of unconsciousness during general anesthesia.

The authors express their sincere gratitude to James D. Wood, Registered Laboratory Animal Technician, Department of Anesthesiology, Medical College of Wisconsin, Milwaukee, Wisconsin, in memoriam, for his relentless help in performing the experiments. They also thank Kristina M. Ropella, Ph.D., Professor and Chair, Department of Biomedical Engineering, Marquette University, Milwaukee, Wisconsin, for her advice with respect to signal analysis and Richard Rys, B.S., Senior Research Engineer, Department of Anesthesiology, Medical College of Wisconsin, for his contribution in constructing components of the electrophysiological recording system.