The Concentration of Isoflurane Required to Suppress Learning Depends on the Type of Learning

The Committee on Animal Research of the University of California–San Francisco approved our study of 171 male specific-pathogen-free Sprague-Dawley rats weighing 300–350 g obtained from Charles River Laboratories. All were housed two per cage (the “home cage”) in our animal care facility and had continuous access to standard rat chow and tap water during 12-h cycles of light and dark for 1 week before study. The dose–response effect of isoflurane on learning was measured in six groups for freezing to context (n = 8 per group) and seven groups for freezing to tone (n = 8 per group). Untrained control groups given anesthesia (n = 16 per group) and not given anesthesia (n = 8 per group) were included for each of the two learning preparations. Two additional groups (n = 11 and 8 per group) were studied to compare shock delivery methods used during the training paradigms.

Anesthetic was delivered to each animal in his home cage. We replaced the normal cage cover with a cover containing ports for inflow, outflow, and sampling of isoflurane. This cover also supported an acrylic cylinder (12-cm diameter, 20-cm-high chimney) through which rats could be easily removed from the cage without altering the isoflurane concentration inside the cage.

Training was accomplished in four identical rectangular-shaped chambers (25 × 20 × 17 cm) constructed of clear acrylic and located in a well-lit room. The cover of each training chamber was lined with a rubber gasket providing a gas-tight seal with the chamber rim. Chamber inlet and outlet ports allowed continuous ventilation through the chambers. A circular flow from the four chambers, through a carbon dioxide absorber, and back to the chambers was maintained by a fan producing a background noise of 70 dB (A-scale) (Sound Level Meter; Radio Shack, Ft. Worth, TX). Fresh gas inflow was provided by a 5-l/min oxygen flow through a calibrated isoflurane vaporizer, and overflow gases were scavenged. Gas concentrations were sampled from a port in the circle system. The floor of each training chamber consisted of 14 stainless steel rods (6-mm diameter) spaced 1.8 cm center to center or 31 stainless steel rods (3-mm diameter) spaced 0.8 cm center to center. The closer spacing was used for studies of higher concentrations of isoflurane because at higher concentrations the feet of the rats sometimes fell between bars spaced more widely. The rods were wired to a shock scrambler (Gemini Avoidance System; San Diego Instruments, San Diego, CA). A speaker was mounted on the rear wall of each training chamber. Training chambers were cleaned with 2% ammonium hydroxide before and after each animal occupied it.

Testing to tone took place in chambers differing in appearance and odor from the training chambers. The clear acrylic test chambers had a triangular perforated roof. The base measured 25 × 28 cm and the sides measured 21 × 28 cm, and they had smooth floors and were located in a different room from the training chambers. The test room for tone was lit with a 25-W red light bulb. The test chambers were cleaned with a pine-scented solution, and there was no background noise. A speaker was mounted on the rear wall of each test chamber.

On the training day, animals were brought to the training room and marked on the tail with a permanent felt-tipped pen for identification. After at least a 1-h period of rehabituation to their home cages, isoflurane was delivered for 30 min to these cages at the target anesthetic concentration. Each rat was then removed from its home cage via  the chimney and rapidly (< 5 s) placed in a training chamber previously charged with isoflurane at the target concentration. (We separately determined that the introduction of the rats did not materially change the chamber concentrations.) Animals were allowed to explore the training chamber for 3 min before training began. Isoflurane concentrations were continuously measured by a respiratory gas analyzer (Capnomac II; Datex, Helsinki, Finland), which was calibrated with a standard commercially available gas mixture (Quick cal, Datex). Calibration was confirmed by gas chromatography.

For tone conditioning, animals received three tone-shock pairs consisting of a 30-s tone (90 dB, A-scale, 2,000 Hz) coterminating with a 2-s electric shock (11-Hz bipolar square wave, 2 mA for oxygen [0.00 MAC] groups and 3 mA for groups receiving isoflurane); shock pairs were 90 s apart. Animals were returned to their home cages (purged of isoflurane) within 60 s after the last shock. For context conditioning, the identical procedure was used except that no tone was administered.

The next day, we assessed freezing to context and tone. For context testing, each rat trained to context was returned to the chamber in which it was trained for 8 min. Neither tone nor shock was administered. Four animals were observed simultaneously, one in each of the four chambers. For tone testing, each rat trained to tone was placed in a triangular testing chamber in the different room. After 3 min of exploration, a tone (90 dB, A-scale, 2,000 Hz) was continuously sounded for 8 min. Shocks were not administered. Four animals were observed simultaneously, one in each of the four chambers. Additionally, after a 4-h rehabituation to the home cage after the tone test, each tone-trained animal was tested to context by an 8-min observation period in the identical chamber in which it was trained, and then tested to tone plus context by applying the tone for an additional 4 min in that chamber. Observations of rats during testing were via  a video camera. No personnel were in the testing room during this period.

Two trained observers assessed fear conditioning. Each animal was scored for 2 s once every 8 s. 16Behavior was judged as freezing if there was no visible movement except for breathing. For example, animals that froze did not move for periods ranging from seconds to minutes, sometimes freezing in immobile crouched positions while at other times in unusual postures resembling statues. By comparison, control animals continuously explored the testing chambers, moving about the chambers sniffing and licking the walls, corners, and flooring, or paused to groom themselves or chew. For animals that froze, the initial period of freezing usually began within the first minute of reexposure to the conditioned stimuli and was then followed by alternating periods of exploring–grooming and freezing. The observation periods were video recorded for later reference.

To compare the foot-shock method of conditioning the animals with a tail-shock method, animals were conditioned with shocks delivered to the tail through subcutaneous needles. 20After induction of anesthesia with isoflurane in the home cage, two platinum needle electrodes (type E2; Grass Instruments, Quincy, MA) were placed 3 and 5 cm from the base of the tail. After 30-min equilibration at the target concentration, the animal was transferred to a training chamber, and the needles were connected to the shock scrambler to permit application of a tail shock electrically identical to a shock across two stainless steel rods (the stainless steel rods had previously been removed from the training chamber cage floor). Tone-shock pairs were administered at 3 mA. The impedance between needles was measured (Checktrode; UFI, Morro Bay, CA), and voltage was calculated.

For each group, the isoflurane concentration was calculated as the mean of the concentrations measured in the home cages, in the training cages before, and in the training cages after training of that group. The MAC fractions were calculated by dividing by 1.49 vol%, the value we took for MAC. 21 

The percentage of time an animal froze during the 8-min observation period was calculated as the number of observations judged to be freezing divided by the total number of observations in 8 min, i.e. , 60. 16These freezing scores are probability estimates amenable to analysis with parametric statistics. 7,8,22For each group, the mean and SEM were calculated. Many of the freezing scores were in a range approaching 0%, so an arcsine transformation was applied to the scores in preparation for further analysis. 23Single-factor analysis of variance (Excel; Microsoft Corp., Redmond, WA) was used for overall comparison of the differences among the context-conditioned groups and the tone-conditioned groups. Newman-Keuls testing provided pairwise comparisons between the groups. A group was defined as having learned if the scores were different from the corresponding control (unshocked) group scores; if not, learning was defined as abolished. Learning was defined as impaired if the group score differed from the corresponding unanesthetized (0.00 MAC) trained group. Two-factor analysis of variance was used for overall comparison of the difference between context versus  tone conditioning, and t  tests were used to compare differences between context versus  tone groups at identical concentrations. A P  value < 0.05 was regarded as significant for all comparisons.

Additionally, each animal was dichotamously classified as freezing or nonfreezing, using the percent freezing scores in the control (unshocked) groups to define the thresholds. For the context paradigm, the threshold equaled any percent freezing greater than that of any animal in the 0.75-MAC context control (unshocked) group; the threshold for the tone paradigm was similarly chosen from the scores in the tone control (unshocked) group. Logistic regression analysis was then applied to provide the ED⁵⁰values and SEMs for suppression of learning context–fear and tone–fear associations (BMDP Statistical Software; University of California Press, Berkeley, CA). The statistical significance of the difference between the ED⁵⁰values was calculated using a two-tailed t  test.

We calculated the correlation between the freezing scores of the two observers measured for individual animals.

For the target groups in the context training, termed 0.12, 0.25, 0.38, 0.50, 0.75, and 0.75 Control (unshocked), the measured MAC fractions were 0.13 ± 0.00 (mean ± SE), 0.25 ± 0.01, 0.38 ± 0.01, 0.46 ± 0.01, 0.71 ± 0.02, and 0.73 ± 0.01, respectively. For the target groups in the tone training, termed 0.12, 0.25, 0.38, 0.50, 0.62, 0.75, and 0.75 Control (unshocked), the measured MAC fractions were 0.13 ± 0.01, 0.25 ± 0.00, 0.38 ± 0.00, 0.48 ± 0.02, 0.62 ± 0.01, 0.71 ± 0.01, and 0.73 ± 0.01, respectively.

Table 1shows the percent freezing at each concentration step for rats trained to context with unsignaled shocks (i.e. , without tones). The freezing scores of animals trained to context at 0.00 MAC, when later tested to context, were 78 ± 8% (mean ± SE), a value not different from that at 0.12 MAC (52 ± 12%). By comparison, the scores of 0.75 MAC context control (unshocked) animals were 1 ± 1%. The scores of the 0.12-MAC group differed from those of this control (unshocked) group (P < 0.05), showing that animals learned context–shock associations at 0.12 MAC. In contrast, the scores of animals trained at 0.25 MAC were 24 ± 11% (P < 0.10, NS), and at 0.38 MAC were 4 ± 2% (NS), showing that 0.25–0.38 MAC decreased or abolished learning context–shock associations. The values at 0.25 and 0.38 MAC also differed significantly from the 0.00-MAC value.

Rats trained with shocks signaled by tone showed minimal freezing during the 1 min before the onset of the test tone; the highest baseline group score before the tone was 3 ± 3%. Table 1shows freezing scores for these animals during the test tone. Learning occurred at 0.5 MAC, as evidenced by freezing scores of 32 ± 11% (P < 0.05 compared with the 0.75-MAC tone control [unshocked] group). Learning to tone appeared to be abolished at 0.62 MAC (3 ± 2%, NS); learning at this MAC fraction tended to be depressed relative to learning at 0.00 or 0.12 MAC. Figure 1graphically displays these data and also shows that at 0.25, 0.38, and 0.5 MAC the scores for freezing to tone versus  to context differed significantly; two-factor analysis of variance yielded significant main effects of tone versus  context (F^1,84= 17.0, P < 0.001) and concentration (F^5,84= 21.4, P < 0.001), and significant interaction between the two (F^5,84= 3.8, P = 0.004).

The tone-trained animals were retested after 4 h in their home cages, this time in their original training chambers. Animals trained at 0.12 MAC with shock signaled by the tone showed less freezing (27 ± 9) to context than the animals trained with unsignaled shock (52 ± 12;table 1), an effect (i.e. , that warning the animal of an impending shock produces less conditioning of fear to context than an unsignaled [no warning] shock) reported previously with this fear conditioning preparation. 24The score of 27 ± 9 (P < 0.05 compared with the corresponding 0.75-MAC control [unshocked] group) shows that, as expected, 7,8animals can learn fear to context at the same time that they learn fear to tone. Importantly, testing to context in animals trained to tone at 0.25 MAC shows decreased freezing (10 ± 4, NS) and abolition at 0.38 MAC (1 ± 1, NS), a finding consistent with that associated with the concentrations suppressing learning previously found with the freezing to context with the unsignaled (no tone) shock training (table 1).

Furthermore, when tone-trained animals were retested to tone in their original training chambers, i.e. , trained to tone and tested to tone–context, their freezing scores were similar to those during their first tone test. At 0.5 MAC, percentage freezing was 33 ± 11%, indistinguishable from that during the first tone test, 32 ± 11%, but not significantly different from the corresponding 0.75 MAC control (unshocked) group (10 ± 3%). Freezing was abolished at 0.62 MAC, again consistent with the previous result found for suppression of learning.

For the context paradigm, the freezing scores of the 16 animals in the 0.75-MAC control (unshocked) group ranged from 0 to 5%, and 5% was taken as the threshold: scores greater than 5 were classified as freezing, whereas scores less than or equal to 5 were classified as nonfreezing. Similarly, for the freezing-to-tone paradigm, the freezing scores of the 16 animals in the 0.75-MAC control (unshocked) group presented a tone and tested to tone ranged from 0 to 19%, and 19% was taken as the threshold. Logistic regression analysis applied to the classified (dichotomous) data yielded an ED⁵⁰for the suppression of freezing to context of 0.25 ± 0.03 MAC (value ± SE) and of freezing to tone of 0.47 ± 0.02 MAC. The difference between ED⁵⁰values was significant (P < 0.01). Figure 2shows the logistic regression curves for freezing to context and tone.

The independently obtained real-time scores of the two observers correlated closely (r²= 0.99). In addition, the videotapes were rescored by one of the observers blinded to the experimental conditions; the correlation coefficient of the observer’s original and blinded scores was r²= 0.98. The blinded scores were used to calculate the results shown in the text, table, and figures. There were no significant differences between results calculated with the blinded compared with the real-time scores.

As a check on the effect of shock stimulus and site of shock, two groups in a freezing-to-tone paradigm received tail shocks instead of foot shocks. At 0.5 MAC (n = 11), the freezing scores were 13 ± 8% to tone and 30 ± 10% to tone–context, whereas at 0.75 MAC (n = 8), these values were 5 ± 4% to tone and 4 ± 2% to tone–context. Thus, subcutaneous tail shocks did not provide greater conditioning than foot shocks. The stimulus voltage across the needles was 9.3 ± 0.1 V (mean ± SE), an intensity similar to that of standard MAC stimuli. 20 

Animal dose–response studies using volatile agents show that 0.25 MAC suppresses learning, a result consistent with our data for suppression of learning to context. Using the fear-potentiated startle paradigm, Kandel et al.  5and Sonner et al.  6conditioned rats during desflurane administration by pairing a light (conditioned stimulus) with a foot shock (unconditioned stimulus). To measure the amount of light-shock conditioning, the next day the acoustic startle reflex was measured in the presence of the light. The amplitude and speed of this reflex are increased when an animal is fearful, and this increase (i.e. , this potentiation) indicates the amount of fear conditioning. 25,26Conditioning to the light was suppressed or abolished by 0.25–0.30 MAC desflurane. Additionally, El-Zahaby et al.  4presented paired tones (conditioned stimuli) and electric shocks (unconditioned stimuli) to produce classical conditioning in rabbits receiving isoflurane. As measured by the effect of the tone on nictitating membrane responses, a concentration of 0.2 MAC suppressed and 0.4 MAC abolished conditioning.

In contrast, in the present study we found that more than 0.5 MAC was required to suppress fear conditioning to tone. A relatively high dose was previously suggested by fear conditioning to tone studies using other anesthetics and different methods to measure learning. Weinberger et al.  27and Gold et al.  28found that learning had occurred during an anesthetic of sodium pentobarbital supplemented with chloral hydrate, but only if epinephrine was administered before conditioning. Edeline et al.  29reported that learning had occurred during a ketamine dose sufficient to suppress vibrissa movement but not the corneal reflex of rats. Ghoneim et al.  30found that learning had occurred after a 100-mg/kg subcutaneous dose of ketamine in rabbits. Pang et al.  31reported that learning had occurred during halothane; however, the concentration at the time of learning is uncertain because equilibrated conditions were not described.

Our finding of a greater concentration required to suppress fear conditioning to tone versus  context is consistent with the finding of Melia et al. , 32that ethanol doses of 1.0 and 1.5 g/kg suppress fear conditioning to context but minimally affect fear conditioning to tone.

Why did our animals learn to fear tone yet not context at 0.5 MAC, and why do other dose–response studies find that the concentration abolishes learning? We believe that the neurophysiologic processes underlying conditioning fear to context are more easily depressed than those underlying conditioning fear to tone. Both conditioning paradigms require processing by the amygdala, but the contextual paradigm additionally requires hippocampal processing. 7–14The conditioning tone provides an intense, novel, and discrete cue transmitted directly to the amygdala, where it becomes associated with the shock. In contrast, the conditioning context, i.e. , the chamber shapes, smells, touches, light intensity, and sounds, consist of less intense and more interrelated and complex cues that require processing by both amygdala and hippocampus.

We further speculate that responses using learning paradigms that employ words, pictures, or other targets requiring declarative memory processing within the hippocampus or associated medial temporal lobe structures would be suppressed by doses similar to the dose that we found suppressed fear conditioning to context, i.e. , 0.25 MAC. 33However, this may not explain why 0.25 MAC suppresses fear-potentiated startle yet not freezing to tone. Perhaps it is because of differences in timing and intensities of the condition stimuli and unconditioned stimuli used for the two paradigms, or that light was the conditioned stimuli for one paradigm, whereas sound was the conditioned stimuli for the other. We do not know which of these factors might have resulted in the difference of concentrations required to suppress learning.

We modeled our test approaches after well-defined experimental protocols used previously. Additionally, because our studies required a stable equilibrated concentration of volatile agent, we developed a technique of introducing anesthetic into the rats’ home cages, allowing a 30-min period of equilibration at the target concentration, and then transferring the rats to the anesthetic-charged training chambers 3 min before applying the conditioning shocks. We developed this technique after finding that the more straightforward approach of inducing and equilibrating for 30 min in the training chamber failed to reveal learning to context. With this more straightforward approach, we found that conditioning to context was low, even for animals receiving only oxygen and no anesthetic. The transfer technique decreased the animals’ exposure time to context to 3 min and produced far higher scores for fear conditioning to context. Presumably, the 30-min exposure period in the training chamber allowed animals to habituate to the contextual cues.

Our results emphasize the complexity of learning and memory. Learning takes place at several levels, no doubt many more than tested in the present study. Whether clinical concentrations of potent modern inhaled anesthetics (i.e. , > 0.5 MAC) abolish all learning is not known, but clinical evidence suggests that they do. In contrast, evidence for some older inhaled anesthetics, such as nitrous oxide and diethyl ether, suggest they do not. 2,34 

In summary, we found that isoflurane concentrations that suppress learning of one conditioned stimulus may be insufficient to suppress learning to another. We found that the concentrations required to abolish fear conditioning to both forms of learning exceeds 0.5 MAC.