Desflurane, Isoflurane, and Sevoflurane Provide Limited Neuroprotection against Neonatal Hypoxia-Ischemia in a Delayed Preconditioning Paradigm

Cincinnati Children’s Hospital Medical Center Institutional Animal Care and Use Committee approval was obtained for all procedures involving the use of live animals. A total of 140 129T2 × C57 Bl/6 F1 hybrid mice were used for the long-term outcome experiments. Thirty-five mice from each of four mating pairs were randomized within the pair to one of five treatments. The groups are shown in table 1.

The surgical procedure and hypoxia-ischemia treatment are as described previously.1,8Briefly, right common carotid artery ligation was accomplished under isoflurane anesthesia (5 min) using a double ligation and division technique. Two hours later, the mice were placed in temperature-controlled chambers with 10% O²in 90% N²for 60 min. Preconditioning was performed in incubators set to an ambient temperature of 34.7°C. The anesthetic agents were administered in 40% oxygen. The anesthetic concentrations used are equipotent in neonatal mice based on previous studies conducted in our lab.

The sample size was determined by using a sample size calculator for means with repeated measures (STATA; STATA Corporation, College Station, TX) using data from previous trials to calculate the means and variance parameters. The inputs into the calculator were a μ1 = 9.1 and ς1 = 6.5 and μ2 = 13.3 and ς2 = 10; the correlation between day-to-day measurements was 0.63. The computed sample size was 28 per group if data analysis was performed using analysis of variance/covariance.

A separate group of animals (n = 24) were used to measure blood gases and metabolic parameters at the end of the preconditioning period. The blood samples for two of the Group S animals did not give results due to cartridge errors. Measurements were made by using the I-Stat blood gas analyzer (Abbott Point of Care, East Windsor, NJ) with CG-8 cartridges as previously described.1Blood samples were obtained from the right common carotid artery directly into heparinized syringes.

Another cohort of 20 mice (5 per preconditioning protocol) was preconditioned by using the same protocol as the study animals. The mice in this group were anesthetized and transcardially perfused by using heparinized phosphate-buffered saline followed by 4% paraformaldehyde, 15% glycerol, and 5% sucrose in phosphate-buffered saline. The tissue was postfixed in the same solution overnight before treatment with increasing concentrations of sucrose in phosphate-buffered saline. The brains were cryopreserved in 35% sucrose in phosphate-buffered saline 3 h after the completion of the preconditioning period. Sections (40 μm) were cut from dorsal striatum and dorsal hippocampus and were stained by using primary antibodies 1:100 rabbit anti-activated caspase-3 (Cell Signaling 9661S; Cell Signaling Technology, Beverly, MA) and 1:100 mouse anti-NeuN (Chemicon MAB 377; Millipore, Billerica, MA) in 5% horse and 5% goat serum overnight at room temperature. After wash steps, sections were incubated with 1:200 Alexa-fluor 488 conjugated goat antirabbit IgG (Molecular Probes A11034; Invitrogen, Carlsbad, CA), 1:250 Alexa fluor 555 conjugated goat antimouse IgG (highly cross absorbed, Molecular Probes A21424, Invitrogen) and 1:250 Alexa-fluor 633 conjugated goat antichicken IgG (Molecular Probes A21103; Invitrogen) for 2 h at room temperature. Slides were mounted with antifade mounting media.

Imaging was acquired by using a Leica TSC SP5 confocal microscope (Leica Microsystems Inc., Bannockburn, IL) with a 488-nm Argon laser and a 543- or 633-nm Helium-Neon laser. The bandwidth of the photodetectors was set to minimize acquisition of energy emitted by fluors other than the fluor to be acquired by a specific channel. Sequential scanning with line averaging was used to further reduce crosstalk between channels and to enhance signal-noise ratio.

The animals assigned to the long-term groups were weighed on days of life or postnatal day (P)9, P10, P12, P14, and P16 to track early somatic growth. Weights were also obtained on days P50 and at the time of apomorphine challenge (P94). Whole brain weights were obtained when the brains were removed after perfusion and before postfixation.

Behavioral testing began on the first Monday after P34. The testing sequence for the first week was locomotor activity on day 1, rotorod on day 2, acoustic startle on day 3, and prepulse inhibition on day 4. The remainder of the behavioral tests began the Monday after P49. The testing sequence was novel object recognition during week 1, cued water maze during week 2, hidden water maze during week 3, reduced water maze during week 4, and the 2-week-delayed probe trial and apomorphine challenge during week 6. After all behavioral tests were completed, the mice were deeply anesthetized, and the heart was exposed and perfused with 4% paraformaldehyde. The brains were removed and postfixed in 4% paraformaldehyde overnight and then cryopreserved.

Each behavioral assay is designed to assess the functional state of neural circuits or systems in the brain of the experimental animals. Increased locomotor activity results from lesions to the ventral hippocampus; unilateral lesions created in the neonatal period are sufficient to produce the effect.8,9The rotorod test is used to assess cerebellar function and gross motor function. Gross motor impairments or poor coordination impair the animal’s ability to remain on a rotating rod as angular velocity is increased.

Acoustic startle is a reflex that is dependent on an intact cochlear nucleus and caudal pontine reticular nucleus. A low-intensity acoustic pulse (3–12 dB above background) presented before a startle pulse can reduce the startle response. This effect is prepulse inhibition. The circuitry mediating prepulse inhibition converges on the caudal pontine reticular nucleus and involves the nucleus accumbens, medial prefrontal cortex, dorsal hippocampus, and basolateral amygdala. Damage to any of these structures can result in a diminished prepulse inhibition response, especially in the presence of dopaminergic overstimulation.10,11 

Novel object recognition tests the innate ability of a rodent to remember a familiar object. Rodents will spend more time exploring a nonfamiliar or novel object than a familiar object. Performance on tests of novel-object recognition are dependent on the hippocampus for acquisition of object memory12and on the integrity of the perirhinal cortex for retrieval of those memories.13–16 

The cued water maze assesses visual navigation as well as dorsomedial but not dorsolateral striatal function.17An animal must track to a platform with a visible cue. Both the platform location and the entry point are moved with each trial. Intact motor systems are also required for the animals to swim to the platform. The hidden maze is a test of spatial navigation that uses distant visual clues as reference points. The test animal must find a submerged platform by using distant visual cues. The platform location is fixed, but the entry point moves with each trial. In addition to visual acuity, hidden maze performance relies on an intact dorsal hippocampus,18,19dorsomedial striatum,20and entorhinal cortex. In adult mice, bilateral but not unilateral lesions of dorsal hippocampus21or bilateral dorsomedial striatal lesions17result in prolonged hidden maze latencies compared to control (uninjured) animals. The reduced maze tests the ability of an animal to “unlearn” cues previously learned and learn a new set of cues to a higher degree of precision, as the platform size is one-quarter the size of the hidden maze platform. The 2-week delayed probe trial is a test of long-term spatial memory.

The apomorphine challenge assesses the competence of dopaminergic inputs to motor pathways. Animals with unilateral damage to these dopaminergic circuits in the striatum and structures projecting to the striatum will circle repeatedly to the side of the lesion in response to the dopamine agonist apomorphine.22,23 

The behavioral testing protocols for locomotor activity, cued, hidden and reduced water maze, the 2-week-delayed probe trial, and apomorphine challenge have been previously described.8,24Briefly, the locomotor activity test is conducted in a 41 cm × 41 cm acrylic chamber with photo-beams placed a 2.5-cm interval on both axes. The mice are allowed to move freely in the chambers for a period of 1h.

The rotorod test was performed on the day after locomotor activity testing. This test was performed to assess the presence of gross motor abnormalities that could affect the results of other behavioral tests. The rotorod (San Diego Instruments, San Diego, CA) consists of a rotating cylinder; the angular velocity of the cylinder is controlled by a servomotor driven by a computer. The device is equipped with a set of sensors that marks the time the mice fall off the cylinder. Rotorod performance is a measure of motor coordination. The mice were allowed to acclimate to the device for three trials. After acclamation trials, the mice were placed on the cylinder for 10 s before the first test trial. The initial angular velocity of the cylinder was 1 rotation per minute; the velocity increased 1 rotation per minute every 15 s until the animals fell off or the maximum velocity of 20 rotations per minute was reached. The animals were given a 1 h rest period before the second test trial. The integral of time-rotations per minute was used as the measure of rotorod performance.

The acoustic startle response of the mice was tested at 35 days of age. They were placed in an acrylic cylinder within one of two acoustic chambers (SR-Lab; San Diego Instruments) equipped with a piezoelectric transducer to sense small forces. The chambers were calibrated each day to assure uniformity of output. Background noise level was set to 70-dB white noise via  a speaker placed 24 cm above the acrylic cylinder in which the animal was placed. After a 5-min acclamation period, a series of 11 120-dB white noise pulses (20 ms duration) were administered with a random interpulse interval of between 8 and 16 s. The delivery of all pulses and data recording was controlled by a microcomputer equipped with SR-Lab software. The response was measured by sampling 1,000 data points of 1-ms duration. The maximum voltage, peak amplitude, time to peak, and interpulse voltages were recorded. The peak amplitude for each pulse was used as the measure of startle response; the first value was discarded and the remaining 10 were averaged.

The day after acoustic startle response, the mice were tested for prepulse inhibition response. All acoustic pulses were white noise. The five acoustic pulse sequences were (1) a 120-dB pulse of 20 ms duration, (2) no acoustic pulse, (3) a 20-ms pulse 3 dB above background followed 70 ms later by a 120-dB pulse, (4) a 20-ms pulse 6 dB above background followed 70 ms later by a 120-dB pulse, and (5) a 20-ms pulse 12 dB above background followed 70 ms later by a 120-dB pulse. The pulse sequences were randomized by using a 5 × 5 Latin-squares design25with three repeats of the block. The average of the 15 repetitions of a pulse sequence was used as the response for the sequence. A positive correlation between prepulse intensity and reduction of startle intensity demonstrates intact prepulse inhibition circuitry. The use of low-intensity (above background) prepulses has been shown to be a more sensitive way to detect subtle loss of prepulse inhibition compared to use of high-intensity prepulses.26,27 

The novel object recognition protocol was performed in a circular chamber of 91 cm diameter by using the method of Clark12with minor modifications. Mice were habituated to the testing chamber for 10 min/d for 4 consecutive days with two identical objects in the test chamber (tinted glass cylinders) on days 3 and 4. The chamber and objects were cleaned with alcohol between trials. The animals were tested on the fifth day. Testing consisted of two phases; for the familiar-object phase, mice were placed in the chamber with two identical brown ellipsoidal objects positioned 41 cm apart and 25 cm from the wall and observed until they accumulated a total of 30 s exploring the objects or 5 min of time accumulated. One hour later, the mice were returned to the chamber for the novel-object phase. The left-side object was replaced with a novel object (clear glass cylinder with marbles inside), whereas the other object was an identical copy of the one used during familiar-object phase. Mice were again allowed to explore the objects until a total of 30 s of exploration of the objects was accumulated or 5 min of time accumulated. The total chamber time required for a mouse to acquire a total of 30 s of object exploration time was recorded for each phase as well as the time spent exploring the right and left sided objects for both phases of the test.

The water maze tests were performed in a tank of 122 cm diameter, with the water temperature maintained at 21°C as previously described.28The water was tinted with white tempura paint to obscure the platform. A 10-cm × 10-cm platform was used as the goal for cued and hidden platform testing. A 5-cm × 5-cm platform was used for the reduced maze trials. The platform location and entry point were varied according to a preset randomization scheme for the cued phase. The goal location was marked by a visible cue on the platform, and curtains were closed around the maze to reduce distal cues. For hidden (and reduced) platform trials, curtains were opened, and the platform remained in a fixed location while entry point varied. The visible cue on the platform was removed. Each mouse was allowed four trials per day for six days, with a maximum of 60 s per trial and intertrial interval of 15 s. If the mouse failed to find the goal within the allotted time, it was placed on the platform for a period of 15 s.29 

Latency was recorded by hand for the cued maze. All other testing was performed with automated tracking and data analysis (Smart; San Diego Instruments). The software was used to extract the selected data parameters such as latency to reach the platform, path length, and average swim speed.

For the apomorphine challenge, mice were placed in one of the locomotor chambers for a period of 10 min to count baseline circling activity. After acquisition of the baseline data, 1.2 mg/kg apomorphine was given intraperitoneally. The total number of clockwise circles made by a mouse during a 20-min epoch after apomorphine was tallied. The relationship between the extent of striatal injury and the response to apomorphine challenge was determined by examining the correlation between number of circles made by the animals after apomorphine injection and the striatal area ratio.

After all behavioral tests were completed, the mice were deeply anesthetized and then perfused transcardially as described in the section immediate effects of preconditioning . Sections from striatum, dorsal hippocampus and ventral hippocampus were Nissl stained for examination by a neuropathologist blinded to treatment group. The Nissl-stained sections were used to estimate cell loss by comparing the layers and arrangement of neurons in the different regions of the hippocampus, striatum and cortex of the injured to the noninjured sides.

All data were analyzed using STATA 10.0 (SATA Corp.) for Mac (Apple Computer Corp., Cupertino, CA). Hypothesis testing proceeded as follows. Experiment-wise error = 0.05 divided equally between two null hypotheses; H⁰1: metric for Group Sham = metric for Group D, metric for Group I, metric for Group S; and H⁰2: metric for Group HI Control = metric for Group D, metric for Group I, metric for Group S. If Group Sham and Group HI were compared within a behavior, the critical P  value for this comparison was set to keep experiment-wise error at 0.05. Thus, the values for H⁰1 and H⁰2 were adjusted so that the sum = 1 −P  for Group sham versus  Group C.

Normally distributed continuous variables were analyzed by using analysis of variance with corrections for multiple post hoc  comparisons (Bonferroni if variances equal, Games-Howell otherwise based on Bartlett’s test). The locomotor activity data and the water maze data were transformed by using a zero-skew ln transform before analysis using repeated measures ANOVA to determine significant between-subject variables within a day. Results were accepted only if the data within individual cells were normally distributed and residual analysis proved the residuals to be independent, normally distributed with a mean of zero. No transformation met all criteria. Therefore, repeated measures data (locomotor activity and water mazes) were analyzed by using the rank transformation type 2 method described by Conover and Iman.30The ranks were assigned for each time block, and an F-statistic was calculated according to equation 5.3 of the Conover and Iman reference. The cued maze data were also analyzed by using SAS Ver. 9.1 (SAS Corp., Carey, NC) using the Proc Mixed procedure with group as the fixed effect. Nonnormally distributed continuous data were analyzed by using the Kruskal-Wallis rank sum comparison with Dunn testing for post hoc  comparisons among groups.

The somatic weight data were subjected to linear regression to compare rates of growth among groups. The findings were corroborated by using repeated measures analysis.

Novel object and prepulse inhibition data were analyzed by using the Wilcoxon matched-pairs signed rank test with correction for multiple tests. Linear regression was used to determine if a significant relationship exists between prepulse intensity and prepulse inhibition. The apomorphine challenge data were analyzed as both continuous and binary variables. Binary variables were analyzed using chi-square contingency tables. Statistical significance, on an experiment-wise basis, was set to 0.05.

The Results are presented in two sections: early preconditioning effects and late preconditioning effects. The first deals with the early effects of preconditioning on the animals. These data were obtained from animals sacrificed either at the end of the preconditioning period (metabolic data) or 3 h later (histology data). The data obtained from the large cohort of mice that completed the long-term study are presented in the section on late preconditioning effects.

The pH, Pco², base excess, potassium, ionized calcium, and glucose values at the end of the 3-h preconditioning period are shown in table 2for the 22 mice used to measure the effects of the preconditioning protocol. The isoflurane-preconditioned mice exhibited elevated Pco²compared to the sham-preconditioned and desflurane-treated mice. The base excess and blood glucose values were significantly lower, and potassium concentrations were significantly greater among desflurane-, isoflurane-, and sevoflurane-preconditioned mice compared with sham-preconditioned mice. The pH of the isoflurane-treated mice was significantly lower than the Group Sham mice. Ionized calcium did not differ significantly among the groups.

Caspase-3 activation is assumed to be evidence for commitment to cell death or irreversible cell injury. To rule out such injury in the preconditioned mice, a cohort of preconditioned mice were examined for the presence of activated caspase-3 in vulnerable regions, such as hippocampus, retrosplenial cortex, and subiculum. Representative sections of hippocampus from the mice sacrificed 3 h after preconditioning are shown in figure 1. Panels A and B show low-power views of sections from cingulate cortex (A) and retrosplenial cortex and subiculum (B) from P7 mice exposed to 6 h of 1.5% isoflurane in 30% oxygen. Numerous cells exhibit evidence of activated caspase-3 are seen in both sections. A high-power (63× oil immersion objective) view of retrosplenial cortex is shown in panel C. Caspase is seen in the cell body and nucleus as well as along the dendritic tree. A few caspase-3–positive cells were seen in sections from the preconditioned mice, but at two orders of magnitude lesser frequency. Desfluane produced caspase-3 activation in the subiculum (panel D); some of the caspase signal (green) colocalizes with NeuN (red), indicating that mature neurons are affected. Isoflurane produced caspase-3 activation in the cortex (panel E) and sevoflurane in the corpus callosum (panel F); none of the affected cells were also NeuN positive (no colocalization of caspase-3 signal and NeuN signal) in the sevoflurane-preconditioned mice. The cells had the morphologic characteristics of oligodendrocytes. Only isolated caspase- 3–positive cells were seen in sham-preconditioned P9 mice (data not shown).

No animals in Group Sham or Group D died. There were three deaths in each of Groups C, I, and S; the differences among the groups were not significant. Somatic weights and brain weights are shown in table 3. There was no difference among the groups in body weight at the time of entry into the study, P9 (table 3). The Group Sham mice had significantly more rapid growth from P9 to P16; the linear regression coefficient for growth was 0.34 g/d for the Group Sham mice (99% confidence interval 0.286–0.400) versus  0.174 (99% confidence interval 0.118–0.244) for the HI mice, i.e. , Groups C, D, I, and S. However, the body weights did not differ among groups at P50 (Kruskal-Wallis [K-W] chi-square = 5.761, df = 4, P = 0.218) or at P94 (K-W chi-square = 1.209, df = 4, P = 0.877). The brain weights of the HI groups (C, I, S, and D) were all significantly less than those of the Group Sham mice (K-W chi-square = 61.44, df = 4, P = 0.0001; individual comparisons all P  values < 0.001 using Dunn’s test with α= 0.025). H⁰2 was tested by using K-W test (K-W chi-square = 6.478, df = 3, P = 0.0905); the Group D mice showed a trend toward greater brain weight than the Group HI (P = 0.069 using Dunn’s test for post hoc  test of H⁰2 with α= 0.025).

The total distance traveled by each group of mice during each of the 5-min intervals, making up the 1-h observation period was used as the metric for locomotor activity (data not shown). The data were analyzed by using the rank transformation method (RT-2) of Conover and Iman30with interval as the repeated block. There were no differences among groups for distance traveled (F ^4,67= 1.04, P = 0.393, neither H⁰1 nor H⁰2 were rejected).

The metric used for rotorod performance was the integral of time on the device and the speed in revolutions per minute at which it turned. There were no significant differences among the groups for either test trial (K-W chi-square = 5.32, df = 4, P = 0.257 for the first trial; K-W chi-square = 4.88, df = 4, P = 0.346 for the second trial). There was no difference among the groups in the fraction of mice with improved performance on trial 5 compared to trial 1 (Pearson chi-square = 3.298, df = 4, P = 0.509).

The acoustic startle response and prepulse inhibition data are summarized in table 4. In all cases, the data were distributed normally as assessed by the Shapiro-Wilk W-test (table 4). There were no significant differences among groups for startle response or percentage of startle inhibition at any prepulse intensity.

Collectively, the mice exhibited reduction in startle as the prepulse intensity increased; startle inhibition was linearly related to the prepulse intensity (F ^1,259= 286, P ≤ 0.0001, r^2adj= 0.523). Similarly, linear regression analysis was performed for each group (table 3). In all cases, significant correlations existed between fractional inhibition of startle and prepulse intensity (fractional inhibition of startle = a × prepulse intensity + b). Group Sham had the greatest mean coefficient for the prepulse intensity (a = 0.0376 ± 0.0032 SE), whereas Group S had the smallest (0.0228 ± 0.0041 SE). The slopes for Groups D, I, and S were significantly different from the slope for Group Sham (t = 4.454, P = 0.0001; z = 3.087, P = 0.003; z = 4.237, P < 0.0001, respectively after correction for multiple comparisons), whereas the slope for Group HI was not significantly different from Group Sham (t = 1.982, P = 0.253 after correction for multiple comparisons). Identical among-group comparisons were obtained using the Hollander and the Potthoff distribution-free tests for parallelism of two regression lines.31,32There was no difference among the slopes for Groups HI, D, I, and S (H⁰2 not rejected).

The novel object recognition data are summarized in figure 2. The time spent exploring the left side object during the novel object phase is plotted against the time spent exploring the left side object during the familiarization phase by group. The line of identity appears in each panel. Data points above the identity line represent animals that exhibited novel object recognition. The time with the left-side object during the familiar-object phase and the novel-object phase was compared by using the paired Wilcoxon signed-ranks test. The difference between the familiar-object and novel-object time was significantly different for Groups Sham, D, I, and S (P = 0.0005, 0.0015, 0.0045, and 0.008, respectively) but not for Group HI (P = 0.846) after correction for multiple tests from same data set.

The fraction of time spent exploring the left-side object during the familiar-object phase was compared to the same fraction for the novel-object phase using the paired-sample Wilcoxon sign-rank test. Group Sham (P = 0.005) and Groups D (P = 0.0035), I (P = 0.0195), and S (P = 0.027 (all P  values Bonferroni corrected for multiple tests) exhibited an increase in the fraction of time spent with the object on the left side of the test chamber when the familiar object was replaced with a novel object. The group HI mice did not show a significant increase in fraction of time with the left object when the familiar object was replaced with a novel object (P = 0.720).

The raw data for the cued, hidden, and reduced maze acquisition phase testing as well as the delayed probe trial are shown in figure 3. The Group Sham mice performed better than all other groups on all three mazes during both the acquisition phases and the probe trials (not used in cued maze).

All animals exhibited a learning curve during the cued maze testing (fig. 3A). Nonparametric analysis was performed by using rank transformation of the data on a day-by-day basis. The ranks were then used in a repeated measures ANOVA to calculate the F-statistic according to Conover and Iman.30Group was highly significant (F ^4,123= 28.44, P < 0.001), the difference between Group Sham and all other groups contributed most to the F-statistic. There was no block-treatment interaction using rank (F ^5,635= 0.14), and the data were found to be spherical (Huynh-Feldt ϵ= 0.969, Box’s conservative ϵ= 0.200). The within-block correlation was essentially zero among groups as required by Kepner and Robinson for application of the F-statistic to RT type data.33The SAS Proc Mixed results were similar using rank as the dependent-variable, the group effect was highly significant (F ^4,512= 29.29, P ≤ 0.001), and day was insignificant (F ^1,127= 0.00). Analysis of orthogonal contrasts revealed Groups D, I, and S were significantly different from Group HI (F ^1,512= 9.91, 9.10, and 9.62; P = 0.0017, 0.0027, and 0.0020 for Groups D, I, and S, respectively, vs.  Group HI; the critical P  value for each of these comparisons is P = 0.00625).

The raw latency data for the hidden maze are shown in figure 3B. Neither the latency nor the path length data for the hidden and reduced maze could be transformed with a zero-skew transform to yield normally distributed values. These data were analyzed by nonparametric methods using Kruskal-Wallis tests with Dunn’s post hoc  for multiple comparisons. Group Sham had significantly shorter latencies and path lengths on days 2, 3, 4, and 5 of the hidden maze compared to Groups C, D, I, and S. There was no difference among groups in swim speed on day 5 of the hidden maze (K-W chi-square = 8.361, df = 4, P = 0.0792).

The raw latency data for the reduced maze are shown in figure 3C. Group Sham had significantly shorter latencies on days 3, 4, and 5 of the reduced maze compared to Groups C, D, I, and S (all differences in mean rank > 29.52, which is the critical value for P < 0.05 using Dunn’s test). The path lengths were not different for any day among the groups. This apparent discrepancy is explained by the fact that the swim speed of the Group Sham mice was significantly greater than Group HI for days 2–5, Group D for days 3–5, Group I for days 4–5, and Group S for days 1–5 using Dunn’s test.

All groups had difficulty with the delayed probe trial, as can be seen by examining the distributions of the latency to first platform crossing (fig. 3D). The Group Sham mice performed better than all other groups in terms of latency and number of crossings (K-W chi-square = 17.01, df = 4, P = 0.0019, and K-W chi-square = 17.83, df = 4, P = 0.0013, respectively). There were no significant comparisons among the HI groups.

The desflurane-, isoflurane-, and sevoflurane-preconditioned mice all exhibited evidence of impaired spatial learning and memory, as there was a blunted learning curve on the hidden maze among all HI groups compared to the Group Sham mice. Learning can be estimated using the fractional change in latency and fractional change in path length.1,8The fractional change in latency is (day 1 latency − the day 5 latency)/day 1 latency. This measure partially removes the effect of differences in swim speed among groups. For the hidden maze, the fractional change in latency for all HI mice was 0.22 compared to 0.51 for the Group Sham mice (K-W chi-square = 20.41, P = 0.0004); the fractional change in path length for the HI mice was 0.32 compared to 0.59 (K-W chi-square = 19.45, P = 0.0006) for the Group Sham mice.

The distributions for the number of clockwise rotations after apomorphine injection are shown for each treatment group in figure 4. Clockwise rotations after apomorphine results from damage to right-sided striatal structures.22None of Groups D, I, and S made significantly more circles than Group Sham (K-W chi-square = 5.581, df = 3, P < 0.139, with ties). One or more of Groups D, I, and S made significantly fewer circles than Group HI (K-W chi-square = 11.803, df = 3, P < 0.0081, with ties). Using Dunn’s test, Group D made significantly fewer circles than Group HI (P = 0.021), whereas the differences between Group HI and Groups I and S were marginally nonsignificant (P = 0.075 and 0.062, respectively).

The incidence of circling, defined as more than 16 circles in 20 min (mean + 2 SD for Group Sham) was determined and analyzed by using chi-square tables. There was a highly significant difference among groups (K-W chi-square = 22.22, df = 4, P < 0.001). Comparisons were done between Group Sham and Groups HI, D, I, and S as well as between Group HI and Groups D, I, and S. The incidence of circling was not significantly greater among Groups D, I, and S compared to Group Sham (chi-square = 3.07, P = 0.080; chi-square = 6.12, P = 0.013; chi-square = 3.34, P = 0.067, respectively, P -critical = 0.00625; α= 0.025 with four groups). Group HI mice had a significantly higher incidence of circling compared to Groups Sham, D, and S but not compared to Group I (chi-square = 21.16, P = 4.2E-06; chi-square = 9.74, P = 0.0018; chi-square = 9.09, P = 0.0025; chi-square = 5.60, P = 0.0179, respectively; P -critical = 0.005; α= 0.025 with 5 groups).

The numbers of circles after apomorphine injection were correlated with the day-5 latency on the cued maze (Spearman ρ= 0.4370, P < 0.001; Kendall’s τ-b = 0.321, P < 0.001 with continuity correction), suggesting both reflect the integrity of common striatal circuits.

A total of 35 brains were analyzed for injury score in cortex, striatum, dorsal, and ventral hippocampus. Some sections from individual brains were not usable for analysis, so the number per group differed by brain region (see table 5). Dorsal hippocampus and striatum were examined for evidence of neurogenesis and gliogenesis. The brain injury scores are shown in table 5. As expected, no injury was noted in the Group Sham mice. Cystic degeneration of the dorsal hippocampus (100% injury) after hypoxia-ischemia on P10 was noted in mice from all HI groups (C, D, I, and V) without differences in injury score among the HI animals. The injury scores in the ventral hippocampus were lower than in dorsal hippocampus for Groups I and D. These groups did not differ from Group Sham in CA1 injury score in ventral hippocampus, given the numbers of sections in table 5. Post hoc  power analysis revealed that a total of 50–60 mice would be needed to establish significant differences among the groups if the observed pattern is representative of the population.

Severe injury to the dorsal hippocampus was defined as a score of 100 or greater for the sum of CA1–3 plus dentate gyrus (table 5). Of 34 mice, 14 had a score of 100 or greater. The median day-5 latency for this group was 38.6 s versus  a median day-5 latency of 17.6 s for the nonsevere injury group. This difference is significant (Mann–Whitney z =−2.852, P = 0.0043), suggesting a relationship between hidden maze performance and dorsal hippocampal injury.

The ratio of the area of dorsal striatum on the side ipsilateral to the carotid ligation to the side contralateral to the ligation is shown in table 5for the five groups. The area ratio for Group Sham is significantly greater than that of Groups D, I, and S and Group C. However, the area ratio is not different among the groups that had hypoxia-ischemia on P10 based on the number of brains examined. Post hoc  power analysis indicated that 33 animals per group would be needed to detect a difference between Group D and Group HI (α= 0.01, β= 0.80, two-sided comparison). This suggests that Groups D and HI may be biologically different.

A striatal area ratio of less than 0.7 was defined as significant injury. Of 32 animals, 20 had significant striatal injury by this criterion. The median number of circles after apomorphine for this group was 36; the median number of circles after apomorphine for the group with area ratio greater than 0.7 was 8. The difference was significant (Mann–Whitney z =−2.175, P = 0.0296). The incidence of circling was correlated with the striatal area ratio (Kendall’s τ-a −0.240, P = 0.0192 with continuity correction), and the absolute number of circles was not correlated.

The short-term effects of preconditioning in the mice studied include significant metabolic changes, particularly hypoglycemia and acidosis. These effects are not seen in the human neonate. We postulate that both are the result of inhibition of the electron transport chain with a subsequent increase in glycolysis for adenosine triphosphate production. This would result in a reversible acidosis and hypoglycemia. The brains of mice subjected to the 3-h preconditioning paradigm showed some activation of caspase-3 at 3 h after preconditioning; this was not true after 2 h of isoflurane preconditioning,1suggesting that the 3-h period may have been past optimal.

Previously published studies indicate that severe hypoglycemia will produce caspase-3 activation in brain 3 h after reintroduction of normoglycemia.34,35The activation of caspase-3 seen may be the result of hypoglycemia or the effects of prolonged volatile agent exposure. The same stresses may also participate in the development of preconditioning. Short periods of hypoglycemia induce immediate preconditioning in myocardium.36Zhao and Zuo37previously found that inducible nitric oxide synthase activity was required for the delayed preconditioning effect of isoflurane using infarct size as a metric. Kawano et al.  38have found that both inducible nitric oxide synthase and nox-2 activity are required for ischemic preconditioning in adult mice. Nitric oxide and oxygen free radicals combine to form peroxynitrite, a compound that can contribute to preconditioning as well as cell death39(fig. 5). More recent evidence suggests that induction of metallothioneins may play a role in anesthesia preconditioning40; however, metallothionein deficient mice do not incur significantly greater behavioral deficits than wild-type mice after perinatal HI.24 

The choice of a 3-h preconditioning period versus  the 2-h period in the previous study was based on proteomic data obtained after 3 h of desflurane exposure.41The 3-h time would allow ample time for the mediators of the delayed preconditioning effect to be activated. The current data suggest that the optimal preconditioning period is less than 3 h; as protective genes may be activated early,3,37and we observed early caspase-3 activation with 3 h of anesthetic exposure. The optimal preconditioning time remains to be determined.

The current study extends the results of our previous report showing a durable but selective neuroprotective effect of isoflurane when used in a delayed preconditioning paradigm before moderate-severe neonatal hypoxia-ischemia. Additional behavioral tests were used in this study to examine the effects of preconditioning neural circuits not examined in the previously published study. The novel object recognition paradigm was added to test hippocampal and perirhinal-cortical circuits. This provides a second and different assessment of hippocampal function from the Morris water maze. Acoustic startle and prepulse inhibition test sensory-motor modulation circuits, which engage the basolateral amygdale, and its projections to the nucleus accumbens and the medial prefrontal cortex42–44as well as cholenergic projections from the nucleus basalis magnocellularis.45 

The data indicate that desflurane, isoflurane, and sevoflurane, when used in the delayed preconditioning paradigm before a moderate hypoxic-ischemic insult on P10, improved performance on cued maze performance, novel object recognition testing, and apomorphine challenge compared with sham-preconditioned animals. In addition, mice preconditioned with isoflurane (Group I mice) did not exhibit a significantly reduced prepulse inhibition response compared with sham controls. The comparisons of performance after preconditioning with the three agents for all behavioral tests are shown in table 6.

The lack of differences among groups in performance metrics for the rotorod test and the acoustic startle response provide important internal controls for analyzing the results of other tests such as the cued maze and prepulse inhibition. The lack of gross motor impairment in any group implicates cognitive impairment as the likely cause of long latencies on the water maze tests. Similar results for each group on the acoustic startle response indicates that the startle mechanism is intact and that differences in prepulse inhibition are a result of the function of the inhibition circuitry rather than blunted startle response.

The decreased prepulse inhibition response slope noted for Groups D, I, and S, but not for Group HI, relative to Group Sham is a paradoxical result of improved protection of dopaminergic circuits in Groups D, I, and S relative to HI and not the result of more severe injury during preconditioning. All HI groups (HI, D, I, and S) sustained the same degree of injury to the ipsilateral hippocampus. Neonatal hippocampal lesions have been found to disrupt prepulse inhibition.46,47The effect can be influenced by dopaminergic inputs; Groups D, I, and S mice would be expected to have greater dopaminergic input than Group HI on the ipsilateral side, given the results of the apomorphine challenge, and therefore, a reduced prepulse inhibition effect.11The damage to the hippocampus appears to be dominant; there is no significant difference between the prepulse inhibition slopes of the HI groups (Groups HI, D, I, and S).

As in our previously published study with isoflurane, there was no improvement in spatial reference learning and memory resulting from preconditioning with desflurane, isoflurane, and sevoflurane. Unilateral neonatal hypoxia-ischemia has been shown to produce spatial learning deficits in adult animals equivalent to those seen in animals with bilateral hippocampal lesions made as adults.21This suggests that unilateral injury in the neonatal period has a detrimental effect on the development of the contralateral hippocampus that produces long-term deficits in spatial learning.48,49The results of this study suggest that reconditioning with desflurane, isoflurane, and sevoflurane does not alter the process that results in poor spatial learning after unilateral neonatal hypoxia-ischemia. The hypoxic-ischemic insult used in this study may be too severe for any potential protective effect of preconditioning to be seen in the dorsal hippocampus. In slice cultures models, isoflurane preconditioning has been found to partially protect the hippocampus against milder insults.50 

The improved performance on novel object recognition testing in Groups D, I, and S occurred despite equally poor performance compared to group HI on the hidden maze. Poor performance on day 5 of the hidden maze has been shown previously to be correlated with severe injury to the right (ipsilateral) dorsal hippocampus in our model8,24; this correlation was also seen in the current study. Despite the extensive unilateral injury, desflurane-, isoflurane-, and sevoflurane-delayed preconditioning preserves one aspect of murine hippocampal function (novel object recognition) compared to sham preconditioning after neonatal hypoxia-ischemia.

Performance of novel object recognition is both hippocampal-dependent12for object memory formation as well as perirhinal cortex-dependent13–16for object memory retrieval. The current data suggest that the pathways involved in novel object recognition do not rely on a bilaterally intact dorsal hippocampus and that desflurane, isoflurane, and sevoflurane preconditioning prevents the developmental disruption caused by unilateral neonatal HI of the circuits involved in novel object recognition in the contralateral hemisphere. Alternatively, desflurane, isoflurane, and sevoflurane preconditioning protect perirhinal cortex relative to sham preconditioning.

Previous work in our lab indicates that novel object recognition is preserved after neonatal HI in the absence of preconditioning after 45 min of HI.24The data presented herein show that novel object recognition is lost after 60 min of neonatal HI. For this test, the effect of preconditioning is equivalent to reducing the HI by about 10-15 min. The effect may be greater for the improvement in cued maze performance, and the reduction in circling after apomorphine challenge as 45 min of neonatal HI are sufficient to produce differences in cued maze performance and circling after apomorphine between sham and HI mice. This conclusion is based on data that were obtained from the same hybrids derived from inbred strains maintained in-house. The model system has been very stable in our hands.

In summary, 3 h of preconditioning with isoflurane 24 h before unilateral hypoxia-ischemia on P10 improved performance in adult mice on novel object recognition and cued water maze and reduced the number of circles after apomorphine compared to sham-preconditioned mice without reducing prepulse inhibition response. There was no protection of spatial learning and memory circuits under the conditions of this study. Desflurane and sevoflurane preconditioning also improved performance in adult mice on novel object recognition and cued water maze and reduced the absolute incidence of circling after apomorphine compared to sham-preconditioned mice. The improved protection of ipsilateral dopaminergic circuits resulted in a paradoxical decrease in prepulse inhibition response among desflurane- and sevoflurane-preconditioned mice relative the sham control mice. We estimate the preconditioning effect to be equivalent to a 10- to 15-min reduction in HI time for a nonpreconditioned mouse.

The authors thank Todd Nick, Ph.D., Professor, Division of Epidemiology and Biostatistics, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio (current of the Department of Pediatrics, Arkansas Children’s Hospital, Little Rock, Arkansas) for preparation of SAS output for cued water maze data. They also thank Vicki Bitter, B.A., Applications Specialist, Department of Anesthesia, Cincinnati Children’s Hospital Medical Center, Cincinnati, Ohio, for expert assistance in preparing final graphics.