Cognitive deficits after perinatal anesthetic exposure are well established outcomes in animal models. This vulnerability is sex-dependent and associated with expression levels of the chloride transporters NKCC1 and KCC2. The hypothesis was that androgen signaling, NKCC1 function, and the age of isoflurane exposure are critical for the manifestation of anesthetic neurotoxicity in male rats.
Flutamide, an androgen receptor antagonist, was administered to male rats on postnatal days 2, 4, and 6 before 6 h of isoflurane on postnatal day 7 (ntotal = 26). Spatial and recognition memory were subsequently tested in adulthood. NKCC1 and KCC2 protein levels were measured from cortical lysates by Western blot on postnatal day 7 (ntotal = 20). Bumetanide, an NKCC1 antagonist, was injected immediately before isoflurane exposure (postnatal day 7) to study the effect of NKCC1 inhibition (ntotal = 48). To determine whether male rats remain vulnerable to anesthetic neurotoxicity as juveniles, postnatal day 14 animals were exposed to isoflurane and assessed as adults (ntotal = 30).
Flutamide-treated male rats exposed to isoflurane successfully navigated the spatial (Barnes maze probe trial F[1, 151] = 78; P < 0.001; mean goal exploration ± SD, 6.4 ± 3.9 s) and recognition memory tasks (mean discrimination index ± SD, 0.09 ± 0.14; P = 0.003), unlike isoflurane-exposed controls. Flutamide changed expression patterns of NKCC1 (mean density ± SD: control, 1.49 ± 0.69; flutamide, 0.47 ± 0.11; P < 0.001) and KCC2 (median density [25th percentile, 75th percentile]: control, 0.23 [0.13, 0.49]; flutamide, 1.47 [1.18,1.62]; P < 0.001). Inhibiting NKCC1 with bumetanide was protective for spatial memory (probe trial F[1, 162] = 6.6; P = 0.011; mean goal time, 4.6 [7.4] s). Delaying isoflurane exposure until postnatal day 14 in males preserved spatial memory (probe trial F[1, 140] = 28; P < 0.001; mean goal time, 6.1 [7.0] s).
Vulnerability to isoflurane neurotoxicity is abolished by blocking the androgen receptor, disrupting the function of NKCC1, or delaying the time of exposure to at least 2 weeks of age in male rats. These results support a dynamic role for androgens and chloride transporter proteins in perinatal anesthetic neurotoxicity.
Experimental data in laboratory animals suggest sex-dependent differences in neurocognitive and behavioral vulnerability to early life anesthesia exposure
Steroid sex hormones play an important role in guiding sex-specific brain development
The relationship between steroid sex hormones and developmental anesthesia neurotoxicity is incompletely understood
Blockade of androgen receptors in 7-day-old male rats protects against isoflurane anesthesia-induced behavioral deficits
Androgen receptor blockade results in a premature transition in the developmental expression profiles of chloride transporters NKCC1 and KCC2
These observations suggest that regulation of specific chloride transporters, determining functional modalities of γ-aminobutyric acid–mediated neurotransmission, by androgens is a critical component for developmental anesthetic neurotoxicity
Animal models of early life anesthetic neurotoxicity have consistently found behavioral deficits after long perinatal exposures in species ranging from mice to rhesus macaques.1,2 The extent to which perinatal anesthetic neurotoxicity occurs in humans remains undefined, though the results of several well conducted studies suggest that short exposures are benign3–6 These results are consistent with preclinical studies showing that short anesthetic exposures do not cause significant behavioral deficits for a variety of species.7,8 Regardless of the clinical effect, neonatal anesthetic exposure in animals is an important modulator of developmental plasticity with a mechanism that remains undefined.
Two important variables, sex and age at exposure, impact neurocognitive changes after anesthesia. Experimental animal models often exhibit worsened outcomes in males relative to females.9,10 Similarly, age plays a critical role in susceptibility as a multitude of studies demonstrate deficits after exposures in utero through the perinatal period, while adult and even some juvenile animals do not develop neurocognitive changes after exposure.1,2
Although the underlying mechanism of age- and sex-dependent susceptibility remains unknown, the developing γ-aminobutyric acid–mediated (GABAergic) system is a prime candidate. There are at least three notable characteristics of this system that could impart vulnerability: volatile anesthetics act on the γ-aminobutyric acid type A (GABAA) receptor, the inhibitory properties of the receptor mature in the early postnatal period, and the developmental time frame is sexually distinct.
The flow of chloride into or out of the cell through the GABAA receptor determines the functional response to activation and is thought to be dependent on the relative expression of two solute carrier family proteins, NKCC1 and KCC2, which traffic chloride into (NKCC1) or out (KCC2) of the cell.11 In the mature state, chloride flows into the cell after GABA activation. This direction is down the gradient established by KCC2 and hyperpolarizes the membrane. In the immature state, NKCC1 predominates, and the chloride gradient is reversed. This results in depolarization as chloride flows out of the cell through the GABA receptor. The transition of GABA from excitatory to inhibitory happens shortly after birth and is determined by the changing expression of NKCC1 and KCC2.12 This developmental transition occurs at different rates and varies by cortical location in males and females with the transition occurring at a younger age in females in specific brain regions.13–15
We previously identified differences in susceptibility of male and female rats exposed to anesthesia at the same postnatal age,9 and recently reported that in female rats, both age of exposure and relative expression levels of NKCC1 and KCC2 were predictive of developing a behavior deficit.15 These expression patterns were different for males and females across age with the ratio changing earlier in females. In a separate study, we found that gonadectomized male rats were protected, similar to females. Building on these data, the current study was designed to (1) define the role of the androgen receptor in anesthetic neurotoxicity and developmental regulation of NKCC1 and KCC2 in cortex; (2) test NKCC1’s functional role in developing cognitive deficits; and (3) explore how age and changes in cortex NKCC1 and KCC2 expression relate to vulnerability to cognitive deficits after anesthesia in male rats.
Materials and Methods
This article adheres to the relevant sections of the ARRIVE guidelines.
Animals and Husbandry
The animal experiments were conducted according to the standards and protocols approved by the University of California, San Francisco Institutional Animal Care and Use Committee. Neonatal all-male or all-female Sprague Dawley rat litters were purchased from Charles River (USA) and delivered with a dam. The animals were housed as single-sex litters in a single vivarium room and exposed to a reverse light–dark cycle (12-h light–dark cycle, 18 to 25°C, 45 to 65% humidity). Food and water were available ad libitum. The animals were weaned on postnatal day 21 or 22, after which they were co-housed in groups of 4 to 6 from the same sex and litter. Animals were further segregated into groups of 2 to 3 on postnatal day 36, 1 week before behavioral testing began. Given the potential for environmental enrichment to influence behavior,16 additional enrichment was limited to social housing and a single plastic tube (8 cm in diameter × 15 cm long).
Isoflurane was administered as previously described.15,17 The animals were randomized to isoflurane exposure or control exposure while maintaining equal representations of each group across litters. For postnatal day 7 exposures, the animals weighed 15 ± 5 g. Briefly, the animals were placed on a heated pad (Thermo Haake, USA) in a custom-built exposure chamber. A humidified mix of air and 40% oxygen was delivered to a Datex–Ohmeda (USA) isoflurane vaporizer. Carbon dioxide–absorbent pellets (Litholyme, Allied Healthcare, USA) were placed in the chamber. Isoflurane, carbon dioxide, and oxygen were monitored with a Datex–Ohmeda gas analyzer. Isoflurane was down titrated from 2 to 1.4%, 0.8%, and 0.0% at 2, 4, and 6 h, respectively. Gas variables and pup temperatures were recorded continuously (Supplemental Digital Content, fig. 1, http://links.lww.com/ALN/C431). Toward the end of exposure, nonexposed control animals were separated from dams for 30 min and placed on a warming pad in an anesthesia exposure chamber with room air. Limiting the control animal maternal separation was done to prevent confounding effects, which can influence animal behavior in adulthood.18,19 We conjecture that the isoflurane-exposed animals do not have awareness of maternal separation except for roughly 30 min after anesthesia emergence. In this and previous studies15,20 our isoflurane exposure and control treatment did not translate into measurable nutritional deficiencies because the weight curves for both groups are overlapping (Supplemental Digital Content, fig. 2, http://links.lww.com/ALN/C432).
A total of 115 animals were exposed to isoflurane with 2 mortalities. Both mortalities occurred during the postnatal day 7 exposure, and there were 0 mortalities in the postnatal day 14 exposure group. These rates are consistent with our previous studies.15,17 In total, 122 nonexposed control animals were utilized. All studies were designed with the ideals of replacement, refinement, and reduction of animal usage. This will allow others using similar methods to make accurate power calculations and potentially reduce the number of animals used overall.
On postnatal day 2, all pups were randomly assigned to receive treatment with either flutamide, a selective antagonist of the androgen receptor (Sigma–Aldrich, USA), or vehicle, sesame oil (Fisher Scientific, USA). On postnatal days 2, 4, and 6, pups were given 250-mg subcutaneous injections (volume, 0.1 ml) on their dorsum. The injection sites were sealed using Vetbond surgical adhesive (3M, USA).
Blood samples were taken from adult animals that were treated with flutamide or vehicle as neonates. Before cardiac perfusion at the time of euthanasia, the animals were briefly anesthetized with isoflurane, and ~1 ml of blood was aspirated from the left ventricle into a heparinized syringe. The blood was then centrifuged for 30 min at 14g at 4°C, after which plasma was separated and stored at −20°C. An enzyme-linked immunosorbent assay (ELISA) to quantify testosterone levels, EIA-1559 (DRG International, Inc., USA), was performed according to the manufacturer’s protocol.
Bumetanide (Hospira, USA; 0.25 mg/ml) was diluted in phosphate-buffered saline. The animals were randomized to be injected with bumetanide (dose of 1.8 mg/kg, 0.2-ml volume) or vehicle intraperitoneally 15 min before isoflurane exposure. This dose has previously been reported to show an effect in neonatal rats.21 The injection sites were sealed with Vetbond.
All behavior testing was performed during the dark cycle between 8:00 am and 5:00 pm. To clean all testing materials, 70% ethanol was used between all procedures and rats. Visual cues were placed on the walls for the Barnes maze task and within the testing boxes for all object recognition tasks. Experimenters were blinded to animal group at the time of the testing. For each cohort, order was initially randomized, and the animals were tested sequentially each time.
Barnes maze testing was performed similarly to our previous description.15,17 In the training phase, testing began on postnatal day 43. The rats were first habituated to an escape box for 2 min and then placed into a circular open field arena with 20 holes cut out along the perimeter. The position of the escape box was randomized for each animal (maintaining equal distribution of “goal” among groups), and the box was placed under the assigned goal hole. Movement and latency to goal was recorded with a camera (Basler aca1280, Basler Inc., USA) and tracking software (Ethovision XT 11.5, Noldus Information Technology, Inc., USA). Barnes maze training took place over the course of 4 days with one trial/day. A probe trial was performed 1 week after memory acquisition, during which the escape box was removed, and animal movements and time spent investigating holes were recorded over 90 s. Our primary outcome was the time spent at the goal hole in the probe trial. In the bumetanide cohort, a second probe trial was conducted 5 weeks after the last day of the acquisition phase. Animals that did not find the position of the escape box by day 4 of the learning phase were not included in the probe trial analysis (in total, six were excluded: one from the flutamide control cohort; one from the bumetanide cohort; one from the control/vehicle cohort, one from the isoflurane/vehicle cohort, and two from the isoflurane postnatal day 14 male cohort).
Recognition Memory Tasks
Novel object recognition was performed in a standard way as previously described.22 Movements were tracked with the aforementioned camera and software. During the exposure, the rats were placed in a boxed arena and exposed to two identical objects for 4 min. After a 2-min delay, the rats were placed back into the arena for the testing phase, where one of the identical objects had been replaced with a different object. Time spent investigating the novel object versus the familiar object was recorded for another 4 min. The discrimination index was calculated from these values (time exploring novel object minus the time exploring the familiar object divided by the total time investigating both objects).
Object place recognition was performed in the same way as the novel object paradigm, with the exception that during the exposure the rats were exposed to two different objects and in the test, one of the objects was switched with an object that is identical to the remaining object.22 This object, although identical, should be identified as the novel target because of its location. For both recognition memory tasks, our primary outcome was the discrimination index.
Fluoro-Jade C Staining
Animals undergoing Fluoro-Jade C staining were cardiac perfused with ice-cold 4% paraformaldehyde.15 The brains were dissected and fixed overnight at 4°C, then sucrose-sunk (30% sucrose), frozen with isopentane/dry ice, and stored at −20°C. They were sectioned at 60-μm thickness on a sliding microtome. The sections were mounted on Superfrost slides (Fisherbrand, USA) and stained with 0.0001% Fluoro-Jade C (Millipore, USA) according to the manufacturer’s protocol.
Stereologic analysis was completed with software (Stereo Investigator 10, MBF Bioscience, USA) utilizing the optical fractionator method by a blinded experimenter. Three or four sections of each brain containing representative sections of hippocampus, laterodorsal thalamus, and mediodorsal thalamus regions were traced bilaterally with 4× objective on an upright fluorescence microscope (Nikon Eclipse 80i, Nikon, USA) equipped with a fluorescent lamp and a C11440 Hamamatsu camera (USA). Positive cells were counted according to standard stereology methods using a 40× objective lens. The counting frame was set to 300 × 300, and the systematic random sampling grid was set to 50 and 75% of the region of interest for isoflurane and control sections, respectively.
Western blots were conducted as previously described.15 Briefly, the animals were killed, and their brains were immediately removed and placed on ice. Samples of frontal cortex were dissected from both hemispheres and placed in radioimmunopreciptation buffer (Boston Bioproducts, USA) with protease inhibitor cocktail (Fisher Scientific). The tissue was immediately homogenized, and concentrations were determined by ND-1000 spectrophotometer (Nanodrop Technologies, USA). Standard protein electrophoresis was performed with 7.5% Tris-glycine polyacrylamide gels (Criterion TGX, Bio-Rad, USA) loading buffer and running (Boston Bioproducts) Precision Plus Protein Standards Dual Color (Bio-Rad). Semidry transfer to polyvinylidene fluoride membrane (Bio-Rad) was performed with a Trans-Blot SD device (Bio-Rad). The blots were blocked with 5% nonfat dry milk (Bio-Rad) and then incubated with primary antibodies overnight: rabbit anti-NKCC1 1:1,000 (catalog No. 14581, Cell Signaling Technologies, USA), rabbit anti-KCC2 1:1,000 (catalog No. 07-432, Millipore, USA), and rabbit anti–glyceraldehyde-3-phosphate dehydrogenase 1:2,000 (catalog No. 1440, Cell Signaling Technology). Secondary antibody goat anti-rabbit HRP 1:1,000 (catalog No. A16104, Life Technologies, USA) was incubated for 1 h at room temperature, and then the blot was cut and processed individually with chemiluminescent substrate Super Signal West Pico Plus or Super Signal West Femto (NKCC1 only) and immediately imaged on ChemiDoc Touch imager (Bio-Rad). Band densities were determined with Image Lab 6.0 (Bio-Rad), and NKCC1 and KCC2 densities were normalized to glyceraldehyde-3-phosphate dehydrogenase loading controls.
Statistical analysis was performed using Prism 8.3 software (GraphPad Software, USA). Group size for behavior experiments was determined from previous studies.15,17 For an α of 0.05 and an effect size ranging from 0.98 to 1.5 depending on the specific task, a group size of 8 to 13 animals is predicted to give a power of 0.8. For the cell death assay, we previously found a very large effect size of 6.5 with an α of 0.05, which gave a power of 0.99 with groups of three animals.15 The animal numbers are reported as “n.” The means and standard deviations are reported as the mean value ± SD for all parametric analyses. For nonparametric data, median values followed by 25th and 75th quartiles in parentheses are given. For all experiments, the data were first subjected to Shapiro–Wilk normality testing, which informed the choice of subsequent parametric or nonparametric statistical tests. For all analyses, α was set to 0.05. No outliers were detected or eliminated from any of the analyses. With the exception of six animals (identified above) that did not meet criteria for learning the position of the goal in the Barnes maze, no other data were excluded from analyses.
In the Barnes maze training analysis, a two-way, repeated measures ANOVA was utilized with individual subjects as a blocking factor. This allowed for testing the effect of treatment, day of training, interaction of treatment and training, and the individual subjects effect on maze performance (time to escape). The choice of repeated measures (subjects blocking) design was supported by subject F tests, which suggest rejecting the null hypothesis (that all subjects are the same) for all of the trials. In the probe trial, the time exploring equidistant positions (e.g. ± 1 hole, ± 2 hole etc.) was averaged and compared with the time spent exploring the goal hole with Dunnett’s multiple comparison test.15,17 To compare these patterns across groups, a curve comparison was made using the extra sum of squares F test to ask whether the pattern fit a one-phase decay function or a straight line. We interpreted a P value of less than 0.05 to favor the one-phase decay function over a straight line and concluded that patterns of exponential decay were different from linear functions that formed the basis of comparisons across groups.
In the recognition memory experiments, unpaired, two-tailed t tests comparing the time investigating novel and familiar objects and a two-tailed one-sample t test were used to compare the difference of the discrimination index from zero. In the Western blot studies, differences between groups were tested with an unpaired, two-tailed t test or two-tailed Mann–Whitney U test for nonnormal data sets. In the cell death experiments, one-way ANOVA was used to test for effect of group and post hoc Sidak’s multiple comparison was used to compare the two isoflurane-exposed groups.
Androgen Receptor Inhibition before Isoflurane Exposure Protects Males from Cognitive Deficit
Male rats exposed to 6 h of isoflurane at postnatal day 7 (n = 13) learned the position of the escape box in the Barnes maze similar to controls (n = 17) with a decrease in latency over the training period by repeated measures ANOVA Fday(3, 84) = 40, P < 0.001; there was no effect of group or interaction. (fig. 1). In the probe trial, delayed 1 week after learning, nonexposed (control) animals spent on average 6.0 (4.7) s exploring the goal, which was statistically more time than any other position, displaying a memory of the escape box location (Dunnett’s test goal vs. positions ± 1, 2, 3, 4, 5, 6, 7, 8, and 9 and opposite). Additional analysis showed these data fit a one-phase decay model (F[1, 140] = 6.5, P = 0.012) compared with a linear function. In contrast, isoflurane-exposed males were unable to differentiate the goal from any other position (mean time at goal, 2.3 [2.5] s), and the corresponding curve is best described by a linear function (F[1, 184] = 0.18, P = 0.675; fig. 1C).
We previously found that unlike males, female rats exposed to 6 h of isoflurane on postnatal day 7 do not develop a spatial memory deficit; however, those exposed on postnatal day 4 are susceptible.15 We hypothesized that androgen signaling plays a role in the prolonged susceptibility of males relative to females. We took a pharmacologic approach by injecting the androgen receptor blocker flutamide on postnatal days 2, 4, and 6. We first completed a control cohort comparing flutamide (n = 16) and vehicle (n = 14) and found no effect on spatial memory or the object recognition tasks (Supplemental Digital Content, fig. 3, http://links.lww.com/ALN/C430). There were no differences in weights between the groups, nor were there any differences in serum testosterone levels in adulthood as measured by ELISA (flutamide, n = 14; control, n = 12; means ± SD: flutamide, 1.90 ± 0.53 ng/ml; vehicle, 1.97 ± 0.45 ng/ml; unpaired t test, P = 0.735; Supplemental Digital Content, fig. 2, http://links.lww.com/ALN/C432). A second cohort of flutamide (n = 14) and control (n = 14) animals was studied with the addition of a 6-h isoflurane exposure at postnatal day 7. In this cohort, both groups learned the position of the goal in the Barnes maze and were not different during this acquisition phase (two-way repeated measures ANOVA Fday[3, 78] = 24, P < 0.001; Fgroup[1, 26] = 0.34, P = 0.567; Finteraction [3, 78] = 2.1, P = 0.112; fig. 1E). Animals treated with isoflurane and vehicle (isoflurane/vehicle) performed poorly in the probe trial, spending a mean time of 2.4 (3.5) s at the goal with no differences between the time at the goal and any other position. A curve analysis of these data favored a linear function (F[1, 151] = 1.5, P = 0.229). In contrast, the flutamide-treated males that were exposed to isoflurane at postnatal day 7 (isoflurane/flutamide) were able to successfully differentiate all positions from the goal hole spending on average 6.4 (3.9) s at the goal (fig. 1F). This pattern was different from the isoflurane/vehicle group because it fit a one-phase decay (F[1, 151] = 79, P < 0.001). This protective effect of flutamide was further demonstrated in recognition memory domains because the isoflurane/vehicle group showed no difference between discrimination of novel and familiar objects in the object place test (mean discrimination index ± SD: 0.07 ± 0.20 one-sample t test, P = 0.181), but the isoflurane/flutamide group was able to successfully discriminate the target (mean discrimination index ± SD: 0.09 ± 0.14; one-sample t test, P = 0.032; fig. 1J). Novel object recognition testing revealed successful discrimination by both groups (mean discrimination index ± SD: isoflurane/vehicle, 0.30 ± 0.24; isoflurane/flutamide, 0.41 ± 0.28; one-sample t test, P < 0.001 for both groups; fig. 1I) as is consistent with previous reports.15,17,22
We previously reported a change in the protein levels of NKCC1 and KCC2 in females that occurs between postnatal days 4 and 7, corresponding to the animal’s loss of susceptibility to the isoflurane induced memory deficit over this time frame. We hypothesized that in males, flutamide would change the protein levels of these chloride transporters to reflect a mature pattern as in females. To test this hypothesis, a third cohort of animals was administered flutamide in the same way as the previous experiments and was euthanized on postnatal day 7 (with no isoflurane exposure; fig. 2). Westerns blots were conducted on lysates from bilateral frontal cortex and demonstrated robust changes in the protein levels of NKCC1 and KCC2 with flutamide-treated animals (n = 10), displaying a more mature pattern (Low NKCC1 and high KCC2) compared with vehicle-treated animals (n = 10; mean density ± SD, NKCC1: control, 1.49 ± 0.69; flutamide, 0.47 ± 0.11; t test, P < 0.001; median density [25th percentile, 75th percentile] KCC2: control, 0.23 [0.13, 0.49]; flutamide, 1.47 [1.18,1.62]; Mann–Whitney U test, P < 0.001).
Inhibition of NKCC1 with Bumetanide Protects against Isoflurane-associated Deficits
To specifically investigate the role of NKCC1 in the manifestation of the behavior phenotype, we blocked function of NKCC1 with bumetanide, an NKCC1 antagonist.23 There were three groups of animals: those exposed to bumetanide (isoflurane/bumetanide n = 15) or vehicle (isoflurane/vehicle n = 15) before isoflurane and those injected with vehicle and not exposed to isoflurane (control/vehicle n = 18). In the Barnes maze task, we found a strong learning effect for day of training during the acquisition phase (two-way repeated measures ANOVA F[3, 135] = 224, P < 0.001; fig. 3B). There was no effect of group or interaction. Performance in the probe trial after 1 week revealed control/vehicle and isoflurane/bumetanide-treated animals fit a curve favoring one-phase decay (Fcontrol/vehicle[1, 184] = 10, P = 0.002; Fisoflurane/bumetanide[1, 162] = 6.6, P = 0.011) compared with the isoflurane/vehicle group whose extra sum of squares F test fit favored a linear function (Fisoflurane/vehicle [1, 151] = 1.9, P = 0.174; fig. 3C). A second probe trial was conducted 5 weeks after the initial learning phase. The curve remained linear for isoflurane/vehicle, whereas the control/vehicle and isoflurane/bumetanide followed a one-phase decay. Mean time exploring the goal was 1.3 (2.6) s for isoflurane/vehicle, 4.3 (5.7) s for control/vehicle, and 4.6 (7.4) s for isoflurane/bumetanide. Control/vehicle and isoflurane/bumetanide also showed differences from goal for positions ±4, 5, 6, 7, 8, and 9 and opposite and ±5, 6, 7, 8, and 9 and opposite, compared with no differences for isoflurane/vehicle in this long delay probe trial (fig. 3D). Animal weight was recorded over the developmental window, and no differences between any of the groups were observed (Supplemental Digital Content, fig. 2, http://links.lww.com/ALN/C432).
Postnatal Day 14 Male Rats Have Lost Vulnerability to Isoflurane-associated Deficits
Previously, we showed that female rats are susceptible on postnatal day 4 but have lost vulnerability by postnatal day 7.15 Here we show that males are susceptible on postnatal day 7 but can be protected with flutamide administered before exposure. However, we do not know when this loss of vulnerability occurs in males. We aimed to define anesthetic susceptibility in males at the time point of postnatal day 14. On postnatal day 14, we exposed male rats to the same 6-h isoflurane exposure (n = 15) or control (n = 15) as previously described. Barnes maze learning showed both groups learned the position of the escape box over 4 days (fig. 4, B and C; two-way repeated measures ANOVA Fgroup[1, 28] = 0.23, P = 0.632; Fday[3, 84] = 19, P < 0.001; Finteraction[3, 84] = 0.67, P = 0.571). In the probe trial after exposure at postnatal day 14, both groups successfully discriminated the goal from other positions (control, ±4, 5, 6, 7, 8, and 9 and opposite; isoflurane, ±1, 2, 3, 4, 5, 6, 7, 8, and 9 and opposite). Mean time spent exploring the goal was 4.3 (5.0) s for controls and 6.1 (7.0) s for isoflurane-exposed animals. Curve comparison found one-phase decay is preferred over straight line for both the control (F[1, 162] = 7.5, P = 0.007) and isoflurane-treated animals (F[1, 140] = 28, P < 0.001). Thus, postnatal day 14 represents a defined point where the susceptibility to a memory deficit after isoflurane in male rats has ceased.
The expression of NKCC1 and KCC2 in the cortex was measured in male (n = 10) and female (n = 10) rats at postnatal day 14 by Western blot analysis and found to be no different (mean density ± SD, NKCC1: male 0.98 ± 0.52, female 1.06 ± 0.15, t test, P = 0.649; median density [25th percentile, 75th percentile] KCC2: male, 1.06 [0.59, 1.28]; female, 0.85 [0.76, 1.17], Mann–Whitney U test, P = 0.912; fig. 4, D–H). These results suggest that by postnatal day 14, the developmental sex difference observed at postnatal day 7 in NKCC1 and KCC2 levels has abated.15
Cell Death Does Not Predict Spatial Memory Performance
Because treatment with both flutamide and bumetanide attenuates the deleterious effects of volatile anesthetics on memory, we next explored the effects of these interventions on cell death. We specifically studied the hippocampus (HC), laterodorsal thalamus (LDT), and mediodorsal thalamus (MDT) because we previously reported cell death in these regions with the same anesthetic exposure.15 We found a group effect by one-way ANOVA for most groups (fig. 5; flutamide: FHC[2, 6] = 5.6, P = 0.043, FLDT[2, 6] = 5.1, P = 0.050, FMDT[2, 6] = 3.0, P = 0.12; bumetanide: FHC[2, 6] = 5.2, P = 0.048, FLDT[2, 6] = 6.2, P = 0.034, FMDT[2, 6] = 5.5, P = 0.044). Comparison between the animals that received treatment versus vehicle that were also exposed to isoflurane, post hoc Sidak’s multiple comparison testing revealed no differences between the vehicle control and treatment (mean cells/mm3 ± SD, flutamide cohort: hippocampus isoflurane/vehicle, 28 ± 17; isoflurane/flutamide, 31 ± 5; P = 0.454; laterodorsal thalamus isoflurane/vehicle, 2,300 ± 1,460; isoflurane/flutamide, 1,540 ± 475; P = 0.335; mediodorsal thalamus isoflurane/vehicle, 654 ± 558; isoflurane/flutamide, 532 ± 192, P = 0.677; bumetanide cohort: hippocampus isoflurane/vehicle, 37 ± 16; isoflurane/bumetanide, 17 ± 9; P = 0.067; laterodorsal thalamus isoflurane/vehicle, 262 ± 2,030; isoflurane/bumetanide, 1,680 ± 856; P = 0.590; mediodorsal thalamus isoflurane/vehicle, 591 ± 262; isoflurane/bumetanide, 426 ± 228; P = 0.590).
We previously reported that isoflurane exposure in female rats at postnatal day 4 but not 7 is followed by a memory deficit and that protein levels of NKCC1 and KKC2 in males and females are associated with a temporally defined susceptibility to anesthesia-induced memory deficits.15 Here we show that blocking androgen receptors through flutamide administration before isoflurane exposure in postnatal male rats protects against this behavioral deficit. Flutamide also alters NKCC1 and KCC2 protein expression patterns in males to be similar to those we reported in adults and in female rats at postnatal day 7 when they are no longer susceptible to the memory deficit. Flutamide alone does not affect adult male rats in terms of weight gain, serum testosterone levels as young adults, or ability to perform spatial and recognition memory tasks. The observed effects of flutamide on NKCC1/KCC2 expression and behavioral outcomes suggest that regulation of these specific chloride transporters by androgens is a critical component for anesthetic susceptibility. We also show that this deficit is linked to chloride transporter function, because acute blockade of NKCC1 with bumetanide during isoflurane exposure prevents the cognitive deficit in adulthood. Finally, we narrow the range of days over which male rats are susceptible because postnatal day 7 exposure leads to a memory deficit, whereas postnatal day 14 exposure does not.
The importance of steroid hormones in guiding sex-specific brain development has long been appreciated.24 For many years it was presumed that sexual dimorphism resulted primarily from exogenous testosterone, aromatized to estrogen in the brain, which then exerts masculinizing effects.25 However, it is now understood that this process is substantially more complex, mediated through local and systemic sex hormone signaling, including both androgens and estrogens, with specific regionalization in the brain.26 In our study, we show a specific role for androgen receptors in the susceptibility of postnatal male rats to isoflurane. This period of susceptibility is associated with an immature expression pattern of NKCC1 and KCC2 in the cortex. Treatment with flutamide is similar to testicular feminization, in which genetically male individuals are born with a nonfunctioning androgen receptor.27 In rodent models, all sexually dimorphic nuclei examined with testicular feminized animals show loss of male characteristics,27 which highlights the importance of androgen receptors in brain patterning. We previously found that neonatally gonadectomized males were also protected from the isoflurane-mediated deficit20 similar to the current flutamide studies. However, in contrast to gonadectomy and testicular feminization models, our flutamide intervention targets a narrower time period encompassing the isoflurane exposure and does not cause irreversible gonadal inhibition because testosterone levels are normal in adulthood after neonatal flutamide treatment. Additionally, flutamide alone was not sufficient to affect spatial memory performance (Supplemental Digital Content, fig. 3, http://links.lww.com/ALN/C430), which is important given the role of the early-life sex hormone effect on spatial memory.28
Anesthetic toxicity is dependent on the developmental stage during exposure, providing critical insight into the underlying mechanism of injury. Our study builds on previous work in female rats, which have a susceptibility window across postnatal day 4 that is closed by postnatal day 7. In our current study, males exposed on postnatal day 7 have a phenotype similar to postnatal day 4 females in that they are susceptible to anesthetic deficit and express immature protein patterns of NKCC1 and KCC2. The administration of flutamide changes the protein levels of NKCC1 and KCC2 such that the males reflect the pattern of postnatal day 7 females; KCC2 increases and NKCC1 decreases, which corresponds with protection from the neurotoxic insult. The postnatal day 14 exposure in males did not result in cognitive deficits, and there was no difference in the specific chloride transporter’s expression between males and females. This suggests that vulnerability after isoflurane exposure is minimized by postnatal day 14 through a similar mechanism that protects the female brain at postnatal day 7.15 These differing outcomes after anesthetic exposure support a dynamic model in which susceptibility is regulated through the developmental effects of endogenous sex steroids, structural dimorphisms, and age.26 Intriguingly, others have also shown that GABAergic anesthetics administered in the second week of life in rodents can result in lifelong neuroanatomic changes29,30 that are often associated with improved cognition. Although not designed to evaluate this directly, our study also shows very high performance in isoflurane-exposed males on postnatal day 14. In contrast, Zhu et al.,31 using different species and exposure paradigms, reported worsened cognitive performance after anesthetic exposure at this age.
As an NKCC1 inhibitor, bumetanide can suppress GABA excitability in the neonatal period.32 It is presumably through this action that it has been used to rescue the effects of early life anesthetics in our study and others.21,33 In this study we find that the single dose of bumetanide before isoflurane exposure changed behavior in the Barnes maze Probe trial at 1 week and 5 weeks after initial training. Taken with the KCC2 expression data, these behavioral data suggest that conditions for anesthetic neurotoxicity are met when KCC2 levels are low and NKCC1 is high. However, blockade of NKCC1 leading to a functionally adult phenotype is sufficient to protect against the deficit at a developmental stage when it would normally occur.
Interestingly, unrelated models of neurologic injury in neonates have also implicated NKCC1 function in their disease pathways. In a completely different neonatal insult model than anesthetic neurotoxicity, maternal deprivation, bumetanide was also found to be protective.18 In another application, the action of bumetanide was posited to be useful in treating refractory neonatal seizures,23 although side effects far overshadowed any potential benefit in a small clinical trial.34 In the field of anesthetic neurotoxicity, volatile anesthetics can produce seizure-like activity in neonatal animals that is dependent on GABA excitability.33 Furthermore, sex differences may underlie susceptibility to neonatal seizures in animals35 and humans.36–38 Although these phenomena are studied by very different fields, the mechanistic commonality of NKCC1 function during a narrow developmental window suggests they are related and may be manifestations of different perturbations of the same underlying developmental program.
How the proposed interaction of anesthetics and GABA development translates into a cognitive deficit remains largely unknown. This question is further complicated by the fact that anesthetic agents themselves can influence the expression patterns of these chloride transporters,39 possibly extending the critical period or irreparably altering normal development in this sensitive window. Volatile anesthetic exposure can also lead to DNA methylation of specific genes including chloride transporters, which can alter expression of these molecules in the next generation.40,41
Cell death was initially thought to play an outsized role in the mechanism of anesthetic toxicity, given the reproducible findings in animal models from worms to nonhuman primates.42,43 However, this assumption has been challenged with a number of preclinical studies showing a dissociation between apoptosis and behavioral deficits. For example, apoptosis can be prevented by inhibiting the P75 neurotrophin receptor after neonatal isoflurane or propofol exposure,44,45 but this does not protect against the lasting cognitive deficit.46 Similarly, in our current study, we found elevated levels of cell death in both flutamide and bumetanide treatments when paired with isoflurane exposure. However, despite having similar levels of apoptosis, there was better cognitive performance in the Barnes maze of the flutamide and bumetanide groups compared with vehicle-treated animals. Instead of simply losing critical cells through apoptosis, the neurotoxic insult may exert its effect by influencing processes like synaptogenesis or cell maturation. In this way, normal development is altered during this critical window, and the effects are later unmasked with behavioral deficiencies in adulthood.
There are several limitations with the studies reported here. The link between GABA function and toxicity remains associative and theorized based on measured changes in the expression pattern of NKCC1 and KCC2. However, pharmacologic blockade of NKCC1 does result in a change in cognitive behavior, strongly supporting our hypothesis. We have also taken a pharmacologic approach instead of a genetic one (in using rats as opposed to transgenic mice). Although a genetic model could be very valuable in testing the specific genetic mechanism, we have prioritized cognitive deficit as the clinically relevant outcome. Because of this, using rats in well-defined cognitive tests has distinct advantages in reproducibility and sensitivity compared with mice. In addition, the pharmacologic manipulations allowed for temporal interventions that were not permanent, unlike genetic manipulations.
The expression of NKCC1 and KCC2 is regulated by many factors including age, sex, brain region, and exposure to anesthetic agents. Interpretation of our studies should therefore be limited as our experiments represent a small subset of conditions defined by these factors. This leaves open the possibility that NKCC1 and KCC2 might play significant roles in cognition after anesthesia exposure, which we have yet to explore. Finally, the exposure to volatile anesthetic in perinatal rats should be extrapolated with caution to humans for several reasons including technical limitations in our inability to mechanically ventilate multiple animals at a time.
In conclusion, these studies identify important factors for susceptibility to anesthetic neurotoxicity including a role for androgen receptors and the chloride transporters NKCC1 and KCC2. These developmental insights also have practical significance for the design and reporting of results in the field of sex-difference and developmental neurotoxicity and may hold clinical relevance for future study given the disproportionate distribution of surgical problems in very young boys and girls.
The authors acknowledge Marina DeCastro Deus, B.S. (University of California, San Francisco, California), for assistance with animal behavior and isoflurane exposure; Yasmine Eichbaum, B.S. (University of California, San Francisco, California), and Grant Harris, B.S. (University of California, San Francisco, California), for assistance with animal behavior; Jason Leong (University of California, San Francisco, California) for assistance with Western blot, dissections, and ELISA.
Supported by Ruth L. Kirschstein National Research service award No. T32 GM08440 (to Dr. Chinn), by the National Center for Advancing Translational Sciences, National Institutes of Health through UCSF-CTSI grant No. TL1 TR001871 (to Dr. Sasaki Russell), and by National Institutes of Health grant No. R01 GM112831 (to Dr. Sall).
Dr. Sall received funding from Tasly Pharmaceutical (Rockville, Maryland) for studies on treatment of acute mountain sickness and from Pfizer (New York, New York) for studies of dexmedetomidine in children. The other authors declare no competing interests.