Brain Maturation in Neonatal Rodents Is Impeded by Sevoflurane Anesthesia

The safety of anesthesia in children has recently become a topic of much scrutiny among anesthesiologists.^1–3  The increasing concern is based on preclinical studies demonstrating increases in the number of cells undergoing apoptosis after anesthesia exposure(s) in the developing rodent brain, associated with long-term behavioral changes.^4–9  Although this has also been seen in young, nonhuman primate brain,^10–12  these important studies fall short of providing direct evidence of similar anesthesia-induced neurotoxicity in neonates and young children. So far, investigations on anesthesia-related neurotoxicity in humans have been limited to retrospective studies correlating the number of anesthesia exposures during early childhood with difficulties in learning and/or abnormal behavior and psychosocial issues.^13–16 

A more direct approach to investigate whether anesthesia exposure(s) in the young human brain causes damage is by means of noninvasive imaging. Proton magnetic resonance spectroscopy (1HMRS), which can be used to detect a variety of metabolites and biomarkers in both the human and animal brains, might be useful for this purpose because it is noninvasive. Small molecular weight moieties such as N- acetylaspartate (NAA; neuronal marker), myo-inositol (myo-Ins; marker of glia cells and inflammation), and choline compounds (marker of cell membrane turnover), in addition to metabolites involved in energetics (i.e. glutamate, glucose, and lactate), can be quantified by noninvasive 1HMRS.^17–20  For example, in clinical trials of patients with Alzheimer disease (AD), NAA is used as a biomarker to track disease progression; total gray matter NAA has been shown to decrease (reflecting neuronal loss) in patients with AD compared with controls and/or nontreated subjects with AD.^21–24  NAA has also been identified as a potential biomarker for traumatic brain injury (TBI), specifically for mild TBI, which is not always associated with obvious structural changes.²⁵  We recently demonstrated feasibility of using in vivo 1HMRS to characterize the metabolic profiles in the brain of children during anesthesia, identifying different cerebral metabolic status with inhalational compared with propofol anesthesia.²⁶  1HMRS can also be used to track normal brain maturation in the very young brain via changes in NAA.^27–31  Specifically, NAA has been shown to increase from 31 to 45 weeks of gestation,³²  and this continues into the first decade of life in children, reflecting normal human brain development.^32–34  Similarly, in rodents, NAA, myo-Ins, and choline compounds are among the most sensitive markers for tracking brain maturation.²⁷ 

NAA is also an important biomarker for brain neurotoxicity. For example, abnormal (lower) levels of NAA have been demonstrated in children undergoing treatment for acute lymphoblastic leukemia³⁵  and/or inborn errors of metabolism.^36,37  The striking evidence of neurotoxicity in neonatal rodents after anesthesia exposure(s) is highly likely to be reflected as derangements of the metabolic profiles including NAA, myo-Ins, or choline. Thus, on the basis of previous reports, we hypothesized that neurotoxic exposure to inhalational anesthesia of neonatal rats, when the brain is most vulnerable to anesthesia toxicity during peak synaptogenesis, would be reflected in changes of metabolic profiles, in particular NAA obtained in vivo by 1HMRS. We further hypothesized that the same exposure scheme would not lead to such changes in older less vulnerable rats.

The local institutional animal care and use committees at Stony Brook University, Stony Brook, New York, and Brookhaven National Laboratory, Upton, New York, approved all animal procedures. Sprague Dawley® male rats were used for the study. Lactating dams-with-litter were ordered using only male pups (Taconic, New York). Table 1 shows the different experimental groups. At the appropriate ages, the rat pups were randomly divided into four groups based on the age and anesthesia exposure regimen. Groups 1 and 3 were exposed to 1 minimal alveolar concentration of sevoflurane for 5 h on postnatal day (PND) 7 and PND15, respectively (designated “exposed” groups). Groups 2 and 4 served as sham-controls and not anesthetized on PND7 and PND15, respectively (designated “unexposed” groups; table 1). All rats were weighed daily during the course of the study.

Group 1 and 3 rats were separated from their mother on PND7 and PND15, respectively. Anesthesia was induced with 5% sevoflurane in 100% oxygen in an induction chamber, and when the righting reflex was gone, anesthesia was maintained at 2.2% sevoflurane delivered in a 1:1 O₂:air mixture. During anesthesia, the animals were placed on a warming pad, and their body temperature was strictly maintained to 37° ± 1°C, while also monitoring heart rate, respiratory rate, and oxygen saturation. The total duration of exposure to sevoflurane for each animal was 5 h. Minor adjustments were made to the anesthetic inhalational regimen as needed to maintain an average respiratory rate between 60 and 70 breath per minutes and average heart rate between 300 and 400 beats per minutes. After the 5 h anesthesia exposure, the rats were allowed to recover (recovery = regaining of righting reflex). On recovery, the rats were returned to their mother.

Group 2 and 4 rats were removed from their mother on PND7 and PND15, respectively, and observed in the induction chamber for the duration of the 5 h but without anesthesia and physiological monitoring; however, skin temperature was checked periodically to be certain that the animals were maintaining normal body temperature. After completion of the 5 h period, unexposed control rats were returned to their mother. Experimenters were not blinded to animal groups.

1HMRS spectra were acquired on a 9.4 Tesla magnetic resonance imaging (MRI) instrument interfaced to a Bruker Advance console and controlled by Paravision 5.0 software (Bruker Bio Spin Corp, USA). A custom-built animal cradle system was used for accurate and consistent positioning of the animal head in relation to the surface coil and the center of the magnetic field. Each rat in groups 1 and 2 was scanned twice, on PND8 and PND9; and each rat in groups 3 and 4 was scanned on PND16 and PND17 (table 1). For 1HMRS scanning, the rats were first induced with 5% sevoflurane and then maintained with 2.2% sevoflurane mixed in a 1:1 O₂:air mixture while continuously monitoring their heart rate, respiratory rate, body temperature, and oxygen saturation using optical monitors compatible with MRI (SA Instruments, Inc., USA). Body temperature was strictly controlled using a computer-assisted heating system (SA Instruments, Inc.) to maintain temperature at 37° ± 0.5°C during acquisition times. For radio-frequency (RF) signal transmission and reception, an 11.2-cm RF volume coil and a custom-built 3-cm surface RF coil, respectively, were used. After positioning in the center of the magnet, anatomical images of the brain were acquired in three orthogonal planes using a Rapid Acquisition with Relaxation Enhancement sequence (Repetition Time = 2,500, Echo Time = 40 ms, Number of Averages = 2, Rapid Acquisition with Relaxation Enhancement factor = 8, number of slices = 25, in plane resolution = 0.117 mm/pixel, slice thickness = 0.9mm, slice gap = 0.1 mm) to identify the location of the thalamic region. A single voxel of 3 × 2 × 3 mm was placed covering the thalamic region of both hemispheres (fig. 1A). B₀ inhomogeneity in the voxel was minimized by applying both first- and second-order shims using MAPSHIM (software package for Paravision). 1HMRS was acquired using point-resolved spectroscopy sequence (Repetition Time = 4,000 ms, Echo Time = 12 ms, Number of Averages = 1,024, spectral width = 8012 Hz, number of acquired complex points = 4,096and scan time ~69 min). Each free induction decay signal was recorded separately and corrected for frequency and phase changes during the scan using custom software written in MATLAB^® (The MathWorks, Inc., USA). A typical 1HMRS spectrum from thalamus of a PND8 rat pup is shown in figure 1B. For the quantification of metabolite concentrations, a water unsuppressed scan was acquired using the same MR parameters as used for the water suppressed scan. Once completed, the animals were removed from the scanner and allowed to recover from the anesthetic before being returned to their litter. This process was repeated for each animal the next day (table 1).

Phase and frequency coherent summed spectra were analyzed using LCModel software (Stephen Provencher Inc., Canada)³⁸  and fitted by using a set of 18 simulated metabolites: alanine, aspartate, creatine (Cr), phosphocreatine (PCr), γ-aminobutyric acid, glucose, glutamine, glutamate, glycerophosphorylcholine (GPC), phosphorylcholine (PCh), glutathione, myo-Ins, scyllo-inositol, lactate, NAA, N- acetylaspartylglutamate, phosphoethanolamine, and taurine. In addition, simulated spectra of mobile lipids (Lip20, Lip1.3a, Lip1.3b, and Lip09) were also included. Spectral profiles of population averaged macromolecules (MMs) from a separate group of juvenile rats³⁹  were also incorporated as described previously.⁴⁰  Because of the higher concentration of water in neonatal rat brains,²⁷  the quantitative data were linearly interpolated, and the water concentrations were estimated as 48.6 mM for PND8, 48.4 for PND9, 46.7 for PND16, and 46.5 for PND17.

The paired 1HMRS spectra obtained over two consecutive days from each animal were carefully examined after processing and LCModel analysis to assure that they matched in quality. Matching of spectral quality was assured by comparing signal-to-noise ratio (SNR) and full width half maximum (FWHM) derived from LCModel analysis from the two spectra. Paired spectra from rats where the SNR and/or FWHM deviated more than “4” and/or 0.020 ppm, respectively, were excluded from analysis. In addition, we excluded spectra with obvious motion artifacts and aberrant baseline fitting. The statistical analysis focused on comparing changes of five key metabolites (NAA, [GPC + PCh], glutamate, myo-Ins, and Cr + PCr) in exposed and unexposed rats. The Cramer–Rao lower bounds, which estimate the accuracy of the fitted concentrations, were less than 15 to 20% for most of these metabolites.

Statistical analyses were performed using SAS software version 9.3 (SAS Institute, Inc., USA) and XLSTAT software, version 2011 (Addinsoft, USA). Normality was examined using the Shapiro–Wilk test. A two-way repeated measures ANOVA was performed to examine whether the preselected metabolite concentration changes from PND8 to PND9 were different between groups 1 and 2. For all preselected quantified metabolites, a two-sided paired t test was performed as post hoc testing to examine how these metabolites changed between day 8 and 9 within each group, respectively. Similar analysis was also applied to group 3 and 4 rats. All data are presented as mean ± SD. We executed a preliminary power analysis based on 1HMRS in vivo data demonstrating that concentration of NAA ([NAA]) in normal neonatal brain increases 0.5 mM/day from PND7 to PND10,²⁷  and the hypothesis that brain growth reflected by [NAA] increases in sevoflurane-exposed neonatal rat brain would be reduced to 0.1 mM or less/day. According to this preliminary data analysis, the assumption was made that the NAA change from PND8 to PND9 would be from 2 to 2.1 mM for the sevoflurane-exposed group 1 rats and 2 to 2.5 mM for the unexposed group 2 rats. The pooled NAA SD for the two groups is assumed to be 0.25, and the SD of NAA change is assumed to be 0.4. Data generated with R function “mvrnorm” (package “MASS”) for 1,000 times, and “aov” was applied to calculate the P value of group-by-time interaction effect for each simulation. Percentage of simulations with a P value of interaction effect less than the significance level 0.05 is regarded as the power. A total sample size of 46 (23 for each group) was found to achieve a meaningful power of 0.85 to detect a significant group by time interaction effect.

A subset of PND7 rats was processed for histology 1 h after exposure to sevoflurane or nonexposure. The rats selected for histologic processing underwent trans-cardiac perfusion fixation accomplished by placing the animal under very deep anesthesia (5% sevoflurane). The left ventricle was cannulated with a 23-gauge blunt-tipped needle (Intramedic^® Luer Stub Adapter, Becton, Dickinson and Company, USA) and directed toward the base of the aorta. After cannulation, the right atrium was slit open to allow for venting, while 1 ml/g of heparinized saline was slowly injected, immediately followed by a slow injection of 1 ml/g of 4% formaldehyde solution. on Completion, the brain was extracted, placed in 4% formaldehyde solution for 24 h, and then placed in phosphate-buffered saline (PBS) until processing.

The harvested brains were processed in the following manner; first the left hemisphere of the brain was sectioned at 80 µm in the sagittal plane using a Vibratome^® 1500 (Leica Microsystems Inc., USA). Brain sections were serially collected and sequentially assigned every sixth section, at 480 μm apart. The brain sections were stored at 4°C in PBS with 0.1% sodium azide. For immunohistochemistry, caspase 3 immunohistochemistry was performed as described previously.⁴¹  Briefly, this was done by first washing the sections with PBS, then incubating the samples in PBS with 5% goat serum and 1% Triton X-100 for 2 h at room temperature. The sections were then washed and incubated overnight at 4°C and additionally for another 2 h at room temperature, with the primary antibodies mixed with PBS, 3% goat serum, and 0.2% Triton X-100. The primary antibodies were rabbit antiactivated caspase 3 (Cell Signaling, USA, #9661S, D175; 1:400 dilution) for degenerating cells, mouse anti-NeuN (Millipore, USA; MAB-377; 1:800 dilution) for neurons, and Hoechst33342 (Life Technologies, USA; 1 μg/ml), a nucleus counter-staining dye. The sections were then incubated for 2 h for the secondary antibody reaction in PBS, 3% goat serum, and 0.2% Triton X-100. Secondary fluorescent dye–conjugated antibodies (Life Technologies, 1:400 dilution) were goat antirabbit IgG-Alexa Fluor-488 for anticaspase 3 and goat antimouse IgG-AF-568 for anti-NeuN. The sections were then washed and mounted onto slides using florescent mounting medium DakoCytomation (Agilent Technologies, Denmark) and stored at 4°C until imaging. The microscope used for image acquisitions was a Zeiss LSM780 confocal laser scanning microscope with a fully automated stage for image tiling, multiple laser lines, and advanced spectral separation (Carl Zeiss Microscopy GmbH, Germany). Images were acquired at both 5X magnification for whole brain imaging and at 20X specifically for the thalamus. Representative image of caspase-positive neurons with nuclear changes was taken under 63X magnification. Zen Image Analysis Software (Carl Zeiss Microscopy GmbH) was used for the creation of stitched whole brain images and stitched images of the thalamus. Image brightness and contrast were enhanced using Apple Aperture software version 3.4.5 (Apple, USA).

The rats in groups 1 and 2 demonstrated significant weight gain, increasing by approximately 20% from PND7 to PND9, as would be expected during this neonatal period (fig. 2A). A two-way repeated measures ANOVA was executed to examine the differences in weight gain between the two groups, and no significant time × group effect was found. Groups 3 and 4 rats also demonstrated an expected increase in weight from PND15 to PND17 (fig. 2B), and again, no significant time × group effect was found between exposed and unexposed groups.

After excluding scans with poor quality as determined by a FWHM of greater than or equal to 0.040, a SNR of less than or equal to 10, and/or scans with poor baselines, 1HMRS spectra from a total of 40 PND7 rats (exposed rats [group 1]: N = 19); unexposed rats [group 2]: N = 21) were of sufficient quality to be included in the final data analysis. Table 2 shows the concentrations of the five metabolites measured on PND8 and PND9 and demonstrates that the [NAA] increases by 16% from 2.08 to 2.41 mM (P < 0.0001) in the unexposed group 2 rats; however, this increase is not evident in the sevoflurane-exposed group 1 rats during the same developmental time period. In addition, the concentration of choline compounds seems to decrease by 25% in the exposed rats from PND8 to PND9; but it is unchanged in the unexposed, control rats. Figure 3 illustrates these profile differences specifically for [NAA] (fig. 3A) and concentration of total choline compounds [GPC + PCh] (fig. 3B); as can be seen, the rate of increase in [NAA] is less in the exposed rats compared with the unexposed rats. Therefore, thalamic [NAA] and [GPC + PCh] is reduced in the PND9 animals, which were previously exposed to sevoflurane. Table 2 also shows the result of the ANOVA analysis demonstrating that the changes over time observed in [NAA] (P = 0.031) and [GPC + PCh] (P = 0.024) from PND8 to PND9 were different between groups 1 and 2, with a significant interaction effect by group and time.

For the older group of animals, 1HMRS spectra from a total of 30 animals (exposed rats [group 3]: N = 16; unexposed rats [group 4]: N = 14) were of sufficient quality to be included in the final data analysis. P values of group-by-time interaction effect from the two-way repeated measures ANOVA in table 3 shows that all the metabolite concentration changes from PND16 and PND17 were not different between the two groups. In other words, the increases from PND16 to PND17 were the same in both groups and not affected by anesthesia exposures as shown for the neonatal rats. As shown in figure 3, the [NAA] increased by 16% from 4.45 to 5.00 mM (P < 0.01) in the exposed rats from group 3 and [NAA] increased by 9% from 4.58 to 4.97 mM (P < 0.01) in the unexposed rats from group 4 (fig. 3C). The [NAA] change between these two groups was not different (P = 0.436). The observed change in [GPC + PCh] from PND16 to PND17 (fig. 3D) also demonstrates no difference between these two groups (P = 0.674).

Because previous studies have demonstrated a possible association of mobile lipids with apoptosis,^42–44  neuronal loss, and/or neuronal stem cell generation,^45,46  we were also interested in exploring whether there were changes in mobile lipids resonating at ≈1.30 and/or 0.9 ppm in the sevoflurane-exposed rat pups. The spectral signature at 1.30 ppm (represented by [Lip13a]) showed that there was a trend toward increase in the exposed group, particularly on PND8; however, this difference was not statistically significant (P > 0.05, results not shown). The LCModel-derived [Lip13a] in the spectra was associated with large errors (as measured by the Cramer–Rao lower bound), probably caused by either very low tissue concentrations of mobile lipids and/or insufficient spectral resolution and SNR. To increase SNR, we summed spectra obtained from each group on PND8 and PND9. When compared in this manner, [Lip13a] was found to be higher in the sevoflurane-exposed animal group on PND8 seen in figure 4A. Furthermore, figure 4B demonstrates that there was no difference in summed [Lip13a] for the older rats between exposed and unexposed on PND16. Although these findings corroborate the hypothesis that mobile lipids are higher in the PND7 sevoflurane-exposed rat pups, this result needs to be interpreted very carefully, because reliability of lipid peaks are influenced by broad and overlapping neighboring peaks of MMs (i.e., MM02).

In a subset of PND7 rats from each of the two groups, the brains were harvested 1 h after exposure or nonexposure and processed for histologic evidence of caspase 3–positive cells, indicating cells undergoing apoptosis. Figure 5 shows representative sagittal sections of whole brain images from each group with a clear difference in the number of caspase 3–positive cells (green cells) throughout the brain exposed to sevoflurane for 5 h on PND7 (fig. 5B) in comparison with the unexposed brain (fig. 5A). Figure 6 demonstrates high-powered views of the thalamic region from an unexposed rat (fig. 6B) and a rat exposed to 5 h of sevoflurane on PND7 (fig. 6, C–E), confirming a larger number of apoptotic cells in the anesthesia-exposed animals (compare fig. 6B with fig. 6, C–E).

We demonstrated differences in the rate of increase of [NAA] from PND8 to PND9 in neonatal rats exposed to 5 h of sevoflurane anesthesia on PND7 when compared with age-matched unexposed rat pups. Furthermore, we showed that the rate of increase of [NAA] from PND16 to PND17 was not affected by the same anesthesia regimen compared with sham controls. These metabolomics findings are important, because it supports histologic data demonstrating that PND7 rats exposed to anesthesia undergo increased apoptosis, whereas older rats (>PND14) are not as sensitive. The increased vulnerability of neonatal PND7 rats to anesthesia-induced neurotoxicity has been correlated with the time of peak of synaptogenesis.^8,47  Because NAA is a marker of neuronal integrity and metabolism,⁴⁸  and [NAA] increases reflect normal brain maturation in rodents²⁷  and in humans,^49–51  we interpreted our findings as indirect evidence of sevoflurane-induced neurotoxicity. This assumption was strongly supported by the demonstration of (1) parallel decreases in choline (likely reflecting less “new” cells), (2) demonstration of more apoptotic cells in the thalamus of sevoflurane-exposed rat pups when compared with age-matched unexposed rats, and (3) lack of a similar change in [NAA] in the older animal exposure group who are not susceptible to the neurotoxic effects of anesthesia.

The main goal of this study was to characterize changes in key metabolites of rat pups from PND8 to PND9 as measured by 1HMRS. Baseline [NAA] at PND8 was approximately2.1 mM; and in unexposed, control rat pups [NAA] increased significantly by 16% over 24 h, which is in agreement with previous reports.²⁷  Specifically, Tkác et al.²⁷  showed that [NAA] increases from 2 to 3.8 mM from PND7 to PND10 or approximately 0.5 mM/day. Because NAA is a known marker for neuronal integrity and metabolism, the NAA increase is inferred to be the result of an increasing number and size of neurons and myelination in the brain during early stages of development. The expected brain growth reflected as an incremental [NAA] increase was not observed in the sevoflurane-exposed rat pups (fig. 3A). The documented differences in the daily rate of increase of [NAA] between the two groups of neonatal rat pups, although small, might represent a real, quantifiable loss of neurons as a result of anesthetic exposure at this age. The fact that an altered [NAA] trajectory was observed only in the PND7 rats (a postnatal time point highly susceptible to neurotoxic effects of anesthesia) and was not observed in the PND15 rats (known to be less susceptible) further supports the claim that this may in fact represent a quantifiable loss of neurons in the PND7 neonatal rats. This possibility is also supported by other published reports. First, NAA is located and synthesized in neurons,⁵²  and [NAA] increases during early postnatal brain development and is involved in myelin synthesis.⁴⁸  Second, there are several recent 1HMRS studies in human demonstrating that low [NAA] is indicative of neonatal brain damage.⁵³  For example, in a recent study, cerebral [NAA] was measured in infants (gestational age, 39 ± 1.8 weeks) with mild or severe encephalopathy and compared with age-matched normal infants and found to be 20 to 50% lower in the infants with encephalopathy.⁵³  Third, 1HMRS has been applied to document changes after mild traumatic brain injury (mTBI) in military personnel and also in professional and semiprofessional athletes; there is often no evidence of direct parenchymal brain injury,^54,55  but specimen and postmortem studies have demonstrated apoptosis in human brains afflicted with mTBI.⁵⁶  For example, in a recent study, looking at athletes who had suffered concussions, [NAA] ratios were found to decrease.⁵⁷  Although we documented neuronal apoptosis in the sevoflurane-exposed rats, we did not observe a decrease in [NAA] as is observed in mTBI.⁵⁷  The main difference is that the [NAA] in the developing brain is constantly increasing, which is not the case in the adult brain. Therefore, neuronal injury via NAA would more likely be reflected as a change in the trajectory of [NAA] rather than a decrease. However, it is important to point out that [NAA] decreases in association with brain injury, such as stroke or TBI, can be transient and several studies have reported near-complete recovery of NAA levels after ischemia.^25,48  In other words, the [NAA] decreases observed from PND8 to PND9 in sevoflurane-exposed rats may be recoverable at later postnatal stages. The key question remains, however, whether this change in [NAA] is indicative of any long-term cognitive deficits. Future studies focused on implementing long-term follow-up, and behavioral testing will answer this question.

The fraction of neurons that undergo apoptosis as result of anesthesia exposure at this critical time point in development, albeit much higher compared with normal, is still relatively small when considering the total number of neurons in the brain. This can explain why a large sample size was needed to have sufficient power to detect a significant interaction effect by our repeated measures ANOVA analysis. This may also give some credence to the theory that the damage caused by anesthesia may not be long lasting and potentially reversible, especially given the plasticity of the brain at that age. Conversely, the fact that we can measure a change in [NAA] given this small fraction of injured neurons may point more toward the opposite conclusion, that there are some underlying metabolic changes taking place that can have some long-lasting, deleterious effects. For this to be better understood, more follow-up testing would be needed to look at changes in [NAA] over the complete developmental time course in exposed animals. Furthermore, there is evidence demonstrating that the neurotoxic effects of anesthetics in neonatal rat pups can be decreased by inhibition of the NKCC1 pathway using bumetanide, as well as by certain anesthetics such as dexmedetomidine.^58–61  Further studies would be needed to determine whether bumetanide, dexmedetomidine, or other agents that reduce anesthesia neurotoxicity might annihilate the observed changes in the [NAA] trajectory resulting from the anesthesia exposure in neonatal PND7 rat pups.

The decrease in the total choline as measured from PND8 to PND9 in the exposed animal group is also noteworthy. The total choline measured by 1HMRS in our study mainly represents phosphocholine and glycerophosphocholine; and these compounds become mobile (and measurable by 1HMRS) when the cell membrane has broken down. A change in choline is interpreted as a change in cell membrane turn over. For example, an increase in total choline is observed in certain neoplastic lesions.^62,63  Therefore, the decrease in choline observed from PND8 to PND9 could signify impeded brain growth within the region investigated, which may have long-term sequelae for cognitive and behavioral development. Further studies characterizing the association between long-term memory impairment and choline deficits during neuronal development would be needed to corroborate such possibilities.

There are several limitations inherent to this study. First, acquiring 1HMRS spectra from neonatal rat pups requires exposure to general anesthesia; therefore, we do not have a complete “nonanesthesia” control of NAA and other metabolite changes from PND8 to PND9. Given that the preponderance of evidence that any anesthesia exposure or multiple exposures during the vulnerability period can cause damage to neurons and cognitive impairment, it is crucial to point out this drawback to the study. It is possible that a larger increment would have been observed in the controls if the studies could be done in an awake setting. There is very limited evidence of 1HMRS metabolites measured in awake animals, given the technique’s sensitivity to motion and the ultimately stressful impact of restraint during long scanning. Some of these technical difficulties might be overcome with appropriate, comfortable, sound-isolated animal holders, particularly given that MRI can be accomplished in neonatal and young human infants without anesthesia.^64–66  Another limitation is the short postnatal developmental period studied and the lack of long-term behavioral data to elucidate the impact of the observed [NAA] changes. We are in the process of developing the means necessary to execute such studies.

In conclusion, our data show that noninvasive 1HMRS is sufficiently sensitive to detect subtle differences in the developmental time trajectory of [NAA] in the neonatal rodent brain. This is potentially clinically relevant because 1HMRS can be applied across species, and therefore, a similar approach can be used to provide more direct evidence of neurotoxicity in the human neonatal and/or young brain from anesthesia exposure(s). Normal time trajectories of NAA changes during development are well documented in human brain^29,67,68  and could be used in future studies focused on assessing human brain injury related to anesthesia exposures.

The authors thank Mei Yu, B.S. (Department of Anesthesiology, Stony Brook Medicine, Stony Brook, New York), for assistance with animal anesthesia.

The work was supported by funds from the Department of Anesthesiology, Stony Brook Medicine (Stony Brook, New York); National Institute of Child Health and Human Development (Bethesda, Maryland) grant R21HD080573 (to Drs. Makaryus, Benveniste, and Nedergaard); National Institutes of Health (Bethesda, Maryland) grant 1S10RR02551401 (to Dr. Benveniste); National Institute of Aging (Bethesda, Maryland) grant R01AG040209 (to Dr. Enikolopov); and National Institute of Mental Health (Bethesda, Maryland) grant R01MH092928 (to Dr. Enikolopov).

The authors declare no conflicts of interest.