In utero Exposure to Anesthetics Alters Neuronal Migration Pattern in Developing Cerebral Cortex and Causes Postnatal Behavioral Deficits in Rats

Abstract

During fetal development, cerebral cortical neurons are generated in the proliferative zone along the ventricles and then migrate to their final positions. To examine the impact of in utero exposure to anesthetics on neuronal migration, we injected pregnant rats with bromodeoxyuridine to label fetal neurons generated at embryonic Day (E) 17 and then randomized these rats to 9 different groups receiving 3 different means of anesthesia (oxygen/control, propofol, isoflurane) for 3 exposure durations (20, 50, 120 min). Histological analysis of brains from 54 pups revealed that significant number of neurons in anesthetized animals failed to acquire their correct cortical position and remained dispersed within inappropriate cortical layers and/or adjacent white matter. Behavioral testing of 86 littermates pointed to abnormalities that correspond to the aberrations in the brain areas that are specifically developing during the E17. In the second set of experiments, fetal brains exposed to isoflurane at E16 had diminished expression of the reelin and glutamic acid decarboxylase 67, proteins critical for neuronal migration. Together, these results call for cautious use of anesthetics during the neuronal migration period in pregnancy and more comprehensive investigation of neurodevelopmental consequences for the fetus and possible consequences later in life.

anesthesia, behavior, brain development, neuronal migration, somatosensory cortex

Introduction

Each year up to 75 000 pregnant women (approximately 2% of all pregnancies) and 6 million children (including 1.5 million infants) undergo surgery and anesthesia in the US (Hollenbeck et al. 1985, 1986; Crowhurst 2002; Dere et al. 2007; Loepke and Soriano 2008). Given the fact that epidemiologic studies have shown an association between prenatal or early-in-life exposure to general anesthesia and learning disabilities, it is surprising that the effects of anesthetics on the brain development have not been meticulously evaluated (Blair et al. 1984; Mazze and Kallen 1989; Hollenbeck et al. 1985, 1986; Van DeVelde and De Buck 2007; Loepke and Soriano 2008; Sprung et al. 2009, 2012; Wilder et al. 2009; Sun 2010; Stratmann 2011; Ing et al. 2012; Turski and Ikonomidou 2012; Palanisamy 2012; Perna et al. 2015; Sanders et al. 2013; Lee et al. 2014; Wang et al. 2014; Xiong et al. 2014). Moreover, administration of anesthetics to pregnant animals has been known to cause histological aberrations of the brain and behavioral abnormalities in their offspring (Uemura et al. 1985; Yasuda et al. 2002; Jevtović-Todorović et al. 2003; Cattano et al. 2008; Hashimoto-Torii et al. 2008; Bercker et al. 2009; Lu et al. 2010; Palanisamy et al. 2011; Zou et al. 2011; Xiong et al. 2014). Therefore, whether prenatal exposure to the anesthetics interferes with human brain development and if so how is an important public health issue (Hollenbeck et al. 1985, 1986; Wilder et al. 2009; Sprung et al. 2009, 2012; Wang et al. 2014).

The fundamental feature of cerebral cortical organization is correct positioning of neurons into horizontal (laminar) and radial (columnar) arrays that ultimately defines their connectivity and subsequent normal brain function (Schull et al. 1986; Ang et al. 2003, 2006; Janusonis et al. 2004; Metin et al. 2008; Glunčić 2009; Jones and Rakić 2010). As detailed by the landmark work of Paško Rakić, neural precursors proliferate in the zone lining the cerebral ventricle (ventricular zone, VZ) and migrate radially along the radial glia cells or tangentially, guided by neurotropic molecules to the nascent cortex in early brain development (Fig. 1). Regardless of the migratory mode, neurons settle in an appropriate cortical layer with respect to the time of their origin. Successively generated neurons bypass their predecessors that have migrated along the same glial fibers before, eventually settling at the outermost level of the cortical plate (CP) and gradually forming cortical layers and minicolumns (Rakić 1990, 2003; Komuro and Rakić 1998; Casanova et al. 2002a, 2002b; Ang et al. 2003, 2006; Hashimoto-Torii et al. 2008; Jones and Rakić 2010; Inada et al. 2011; Allen and Lyons 2018).

Figure 1

Neuronal migration in the developing cerebral cortex. Schematic representation of high-power magnification of tissue section through the medial somatosensory cortex right above the hippocampal formation (left panel). Middle panel depicts radially migrating neurons (red) adhering to the processes of the radial glial cells (black) on their way from the ventricular zone (VZ) to the cortical plate (CP). An important mediator of their migration is reelin, extracellular matrix glycoprotein secreted by Cajal–Retzius cells (yellow), which sparsely populate the marginal zone (MZ). Each successively generated neuron must bypass predecessors (white “squares”) that have migrated along the same glial fiber, before eventually settling at the outermost layer of the cortical plate. Anesthetics certainly have potential to affect molecular mechanisms that regulate neuronal migration (listed on the right).

Contact interactions between migrating neurons and neighboring cells play a crucial role in formation of migratory pathways and determination of final neuronal positions (Rakić 1990, 2003; Šestan et al. 1999; Ang et al. 2003, 2006; Metin et al. 2008). Molecular mediators of those interactions are expressed early in the developing brain and include reelin, an extracellular matrix glycoprotein secreted by Cajal–Retzius cells, voltage-dependent calcium channels, and neurotransmitters N-methyl-D-aspartate (NMDA) and gamma-aminobutyric acid (GABA). These proteins, neurotransmitters, and the respective receptors act as paracrine signaling molecules regulating intracellular calcium levels required for dynamic changes of cytoskeleton that are, in turn, critical for the neuronal migration (Rabinowicz et al. 1996; Alcántara et al. 1998; Komuro and Rakić 1998; Ikonomidou et al. 1999; Berbel et al. 2001; Ang et al. 2003; Janusonis et al. 2004; Manent and Represa 2007; Hashimoto-Torii et al. 2008; Cai et al. 2009; Wang and Kriegstein 2009; Barber and Pierani 2016).

Due to these complex cellular and molecular interrelationships, neuronal migration is highly sensitive to various biological, physical, and chemical factors, as well as to specific genetic mutations. For instance, repeated exposure of the rodent and primate fetal brain to environmental agents, such as alcohol, cocaine, neurotrophic viruses, ultrasound, or ionizing irradiation can result in aberrant neuronal migration pattern and create neuronal ectopias—a plausible neuropathological basis for behavioral abnormalities later in life (Schull et al. 1986; Rakić 1990, 2003; Komuro and Rakić 1992, 1993, 1998; Lidow 1995; Algan and Rakić 1997; Gleeson and Walsh 2000; Berbel et al. 2001; Janusonis et al. 2004; Ang et al. 2006; Hashimoto-Torii et al. 2008; Metin et al. 2008; Bertoglio and Hendren 2009; Glunčić 2009; Duque and Rakić 2011; Inada et al. 2011; Selemon et al. 2013; Grandjean and Landrigan 2014). In addition, misplacement of neurons and fine structural abnormalities of the neuronal distribution within the cortical minicolumns—which can be acquired only during the brain development—have been seen in numerous neuropsychiatric disorders (Casanova et al. 2002a, 2002b; Rakić 2003; Ang et al. 2006; Jones and Rakić 2010).

In humans, time of the most intense neuronal migration and proliferation corresponds to the second trimester, when most nonobstetric surgeries are performed (Schull et al. 1986; Mazze and Kallen 1989; Rakić 1990; Rodier 1995; Gleeson and Walsh 2000; Clancy et al. 2001; Crowhurst 2002; Ang et al. 2006; Metin et al. 2008; Loepke and Soriano 2008; Stammer et al. 2008; Sun 2010; Palanisamy 2012; Sanders et al. 2013; Selemon et al. 2013; Wang et al. 2014). The mechanisms of anesthetic action involve alterations of the synaptic transmission via GABA and NMDA receptors and voltage-dependent calcium channels, and nonspecific lipid membrane perturbations, which are all essential in the regulation of neuronal migration (Fig. 1) (Franks and Lieb 1994; Martin et al. 1995; Rabinowicz et al. 1996; Alcántara et al. 1998; Komuro and Rakić 1998; Ikonomidou et al. 1999; Berbel et al. 2001; Ang et al. 2003; Campagna et al. 2003; Janusonis et al. 2004; Manent and Represa 2007; Hashimoto-Torii et al. 2008; Cai et al. 2009; Wang and Kriegstein 2009; Olsen and Li 2011; Sanders et al. 2013). In spite of these potent and potentially disastrous alterations, the effects of anesthetics on neuronal migration and development of cortical columnar organization have never been tested.

To test this hypothesis, we injected pregnant rats with the deoxyribonucleic acid (DNA) replication marker bromodeoxyuridine (BrdU) at embryonic day 17 (E17) and exposed them to general anesthesia. E17 is a late stage of pregnancy in rodents, when the development of motoric brain regions is practically completed and neurogenetic processes are characterized by the generation of the neurons destined to the upper cortical layers of the somatosensory cortex and intense development of visual cortex and hippocampal formation (Bayer 1980; Bayer and Altman 1991, 2004; Rodier 1995; Algan and Rakić 1997; Gleeson and Walsh 2000; Berbel et al. 2001; Ang et al. 2006; Loepke and Soriano 2008; Deng et al. 2010; Duque and Rakić 2011; Palanisamy 2012; Selemon et al. 2013; Grandjean and Landrigan 2014). Since BrdU is a synthetic nucleoside that can be incorporated into nascent DNA of cells in the S-phase (but also cells undergoing DNA repair), it selectively labels neurons generated at the time of the injection and allows for detection of their migration patterns and positions within the cerebral wall and cortex (Berbel et al. 2001; Ang et al. 2006). Migratory pathways of neurons toward the somatosensory cortex are the longest and most vulnerable while the expression of the key regulatory molecules, such as reelin and glutamic acid decarboxylase 67 (GAD67) peaks at this time (Alcántara et al. 1998; Pesold et al. 1998, 1999; Ang et al. 2003; Janusonis et al. 2004; Hashimoto-Torii 2008; Cai et al. 2009; Popp et al. 2009). Finally, GABA and NMDA receptors and voltage-gated calcium channels are functional by E17 and play a major role in neurogenesis and neuronal migration (Komuro and Rakić 1992, 1993, 1998; Ikonomidou et al. 1999; White et al. 2005; Manent and Represa 2007; Cai et al. 2009; Wang and Kriegstein 2009; Inada et al. 2011).

We hypothesized that anesthetic exposure at E17 will lead to an altered neuronal migration pattern in the somatosensory cortex of rat pups, paralleled by their poor performance on somatosensory function, memory, and learning behavioral tests, but relatively normal performance on motoric activity tests. To anesthetize the animals we used either propofol, an intravenous general anesthetic, or isoflurane, an inhalational general anesthetic. Both are widely used in clinical practice and animal studies and both drugs have been shown to induce apoptotic neurodegeneration in newborn rodents (Franks and Lieb 1994; Martin et al. 1995; Mazze et al. 1985; Crowhurst 2002; Campagna et al. 2003; Jevtović-Todorović 2003; White et al. 2005; Cattano et al. 2008; Loepke and Soriano 2008; Palanisamy et al., 2011; Olsen and Li 2011; Lee et al. 2014; Xiong et al. 2014).

Materials and Methods

Study Overview

To examine the impact of in utero exposure to anesthetics on developing brain, a group of pregnant rats was anesthetized at E17, and another group at E16 (Fig. 2). Female pups of the rats exposed to anesthesia on E17 were euthanized on their postnatal day 10 (P10), and their brains were used for histological studies: standard analysis (hematoxylin and eosin [H&E] staining), cell proliferation analysis (BrdU staining), and stereological analysis (cresyl violet staining). Male littermates were enrolled into behavioral studies at P49. Fetuses of the rats exposed to anesthesia on E16 were harvested on E20 and used to analyze protein expression of reelin and GAD67 in the brain.

Figure 2

Study design. Subscripts indicate age group: a = pregnant adults, e = embryos, p = female pups, ya = young adult males.

Animals

The experiments were conducted on 22 timed-pregnant Sprague–Dawley rats (Charles River Laboratories, Inc., Wilmington, MA, USA) and their respective offspring, following approval of the Rush University Medical Center Institutional Animal Care and Use Committee (Chicago, IL, USA) and adhering to the NIH guidelines, ensuring that the minimum number of animals needed to obtain valid data is used (Dell et al. 2002; Charan and Kantharia 2013; National Research Council Committee on Guidelines for the Use of Animals in Neuroscience and Behavioral Research 2003).

We chose the E17 gestational age for cerebral cortical histological analysis and behavioral studies (timing of the most intense generation and migration of neurons destined for the upper cortical layers) and E16 for reelin and GAD67 levels assessments (timing slightly ahead of the reelin expression peak on E18) (Alcántara et al. 1998; Janusonis et al. 2004; Hashimoto-Torii et al. 2008). On E17, 18 pregnant rats were randomized to 9 groups of 2 animals per group receiving 3 different means of anesthesia (60% oxygen air mixture [control], propofol and isoflurane group) for 3 different exposure durations (20, 50, and 120 min) (Table 1). To eliminate the influence of the estrus on behavioral experiments, female offspring were used for histological evaluation, while male littermates were used in behavioral tests (Berbel et al. 2001; Wenk 2001; Dudchenko 2004; Ang et al. 2006; Bouet et al. 2009; Ennaceur et al. 2009; Palanisamy et al. 2011). Additionally, 4 pregnant rats were randomized on E16 to 60% oxygen air mixture (n = 2) or isoflurane (n = 2) group, exposed for 120 min, and euthanized at E20 for extraction of fetuses.

Table 1

Numbers and weights of pregnant rats and their offspring used for histological analysis and behavioral testing

Exposure duration (min)	Condition	N of dams	Weight of dams (g)	N of female pups	Weight of female pups on P10 (g)	N of male pups	Weight of male pups (g) on
							P28	P35	P39	P46
20	Oxygen	2	274 ± 12	15	22 ± 2	7	96 ± 6	144 ± 11	175 ± 13	181 ± 20
	Propofol	2	289 ± 26	13	22 ± 3	9	105 ± 5	155 ± 9	183 ± 10	197 ± 10
	Isoflurane	2	292 ± 19	13	21 ± 2	9	102 ± 3	155 ± 7	183 ± 8	203 ± 9
50	Oxygen	2	298 ± 11	10	21 ± 3	12	100 ± 9	152 ± 15	187 ± 13	197 ± 12
	Propofol	2	277 ± 10	11	20 ± 1	10	84 ± 11	128 ± 13	157 ± 12	178 ± 13
	Isoflurane	2	290 ± 5	12	19 ± 2	9	100 ± 4	152 ± 9	181 ± 13	195 ± 11
120	Oxygen	2	286 ± 38	12	21 ± 2	7	94 ± 5	152 ± 13	185 ± 9	216 ± 10
	Propofol	2	290 ± 25	10	20 ± 2	12	94 ± 8	149 ± 8	187 ± 14	211 ± 13
	Isoflurane	2	295 ± 27	12	19 ± 3	11	94 ± 5	152 ± 8	187 ± 9	211 ± 12

Note: Pregnant rats were exposed to oxygen (control), propofol, or isoflurane on E17, each at 3 exposure durations. All female pups were sacrificed at P10. From each litter, 3 female pups were analyzed in histological studies giving total of 54 animals. The average weight of female pups is based on 6 pups: 3 animals randomly chosen from each of 2 litters. All the male pups (total of 86 animals) underwent behavioral testing starting at P49. Data are presented as mean ± SD.

Anesthesia Protocols

All animals were placed in anesthetizing chambers and were breathing spontaneously during the experiments. Control animals were exposed to 60% oxygen air mixture. Animals randomized to propofol group were anesthetized by induction dose of 50 mg/kg and maintenance dose 12 mg/kg as needed while exposed to 60% oxygen air mixture. Animals randomized to isoflurane received 1.5% isoflurane (Baxter Healthcare, Deerfield, IL, USA) in 60% oxygen air mixture, delivered through a calibrated vaporizer (Ohio Isoflurane; Ohio Medical Corporation, Gurnee, IL, USA) as part of the SurgiVet Anesthesia Machine (Smiths Medical, Waukesha, WI, USA). Isoflurane concentration was measured with the Capnomac Ultima analyzer (Datex-Ohmeda, Helsinki, Finland) measuring inspired/expired O₂ and CO₂. Concentrations of isoflurane and propofol dosing were selected because they represent approximately 1 minimum alveolar concentration (MAC) in the pregnant rodent (Mazze et al. 1985; Jevtović-Todorović et al. 2003; Kroin et al. 2006; Palanisamy et al. 2011; Xiong et al. 2014). Following anesthesia, the animals recovered in 60% oxygen air mixture for 20 min after return of their righting reflex (Palanisamy et al. 2011).

Histological Analysis

Each pregnant rat was injected with the DNA-replication marker BrdU (Sigma-Aldrich, St. Louis, MO, USA) to label dividing cells in the proliferative zone. BrdU was administered as a single intraperitoneal injection of BrdU at a dosage of 50 mg/kg (Berbel et al. 2001; Ang et al. 2006).

Brains from 4 randomly chosen female pups from each litter were harvested following deep anesthesia with ketamine (10 mg/kg; Sigma-Aldrich)/pentobarbital (60 mg/kg; Sigma-Aldrich) and transcardiac perfusion with 0.9% saline. The brains were subsequently fixated with a 4% paraformaldehyde in 0 .1M sodium phosphate-buffered saline (pH 7.4). The remaining steps are described with the staining protocols (Berbel et al. 2001; Ang et al. 2006; Kordower et al. 2010).

The brain of 1 female pup from each litter was embedded in paraffin and processed for light microscopy by preparing sections of brain tissues and stained by H&E staining (see Supplementary Methods). The other 3 brains were postfixed for 2 h in 4% paraformaldehyde, sectioned, and stained as described in Supplementary Methods (Berbel et al. 2001; Ang et al. 2006; Cai et al. 2009; Kordower et al. 2010).

Images were collected from the medial somatosensory cortex right above hippocampal formation. Images spanned from the marginal zone (MZ)/layer II border to the layer VI/white matter border, and BrdU positive cells were counted as previously described (Berbel et al. 2001; Ang et al. 2006). Histology data consisted of 3 brains per litter and 3 sections per brain. One image from each section was used for counting.

Each image accommodated 2 standardized grids with 3 equal bins arranged radially, with bin I being at the surface (topmost). BrdU-positive cells within 3 bins of each grid were counted (Fig. 3). Each grid produced 1 dispersion calculation presented as percentage—number of the BrdU-positive (BrdU+) cells in the bin divided by the total count in that grid (i.e. in all 3 bins). As the 3 images were analyzed per each animal they produced a total of 6 dispersion calculations per animal and 18 per each litter. In total, we analyzed 54 animals, 162 sections/images, and counted BrdU-positive cells within 324 grids. All the measurements were performed by a blinded observer.

Figure 3

Rats prenatally exposed to anesthesia have lower percentages of BrdU-positive cells in the upper layers of the somatosensory cortex. Pregnant rats received a single BrdU injection on E17. Ten hours after the injection, the rats were exposed to oxygen (control) (n = 6), propofol (n = 6), or isoflurane (n = 6), for 20, 50, and 120 min for each exposure (n = 2 for each exposure duration). On P10, female pups from each litter were euthanized, their brains harvested and stained immunohistologically for BrdU. Coronal sections across the somatosensory cortex were taken at position marked by the hippocampal formation immediately below and medially. The sections were overlapped by standardized grid divided into 3 equally spaced bins, with bin I starting at cortical MZ layer II border, and bin III ending at cerebral cortex/white matter border. (A) Coronal sections of somatosensory cortex above the hippocampal formation stained immunohistologically for BrdU (black frame; ×2.5 magnification of a representative animal). Immunohistochemistry staining revealed increasing numbers of BrdU-labeled cells in bins II and III and adjacent white matter of animals prenatally exposed to anesthetics, compared with animals exposed to oxygen (O₂). (B) Percentage of BrdU-positive cells in the topmost cortical bin, averaged across animals in each exposure’s duration. The data points are mean ± standard error from the model. *Tukey-adjusted P value versus the oxygen-exposed group <0.05.

Stereological Analysis of Brain Cortical and Neuronal Volumes

Stereological analysis was performed by a blinded observer on the brains of 2 female pups from each litter (total n = 18). The Cavalieri estimator (StereoInvestigator; Micro-Bright Field, VT, USA) was used to estimate the total volume of the somatosensory and motor cortex on the cresyl violet-stained 40 μm sections. In each brain, we sampled 5 randomly selected serial coronal sections that extended from the level of forceps minor corporis callosi to the caudal level of the mammillary body using the fractionator principle. The distance between sections was approximately 0.72 mm. In coronal plane, the somatosensory and motor cortex are located in the dorsal lateral brain, defined ventrally by rhinal fissure and dorsally by a line from cingulum to median plane, and outlined using a ×1.25 objective. The thickness of sections was determined as the top of the section was brought into focus with ×60 objective and the stage was zeroed at the z-axis. The stage was then stepped through the z-axis until the bottom of the section was in focus. The area estimation of was performed by means of a 50 × 50 μm point grid with ×10 objective (Gundersen and Jensen 1987; Chu et al. 2006; Bartus et al. 2011).

The nucleator module of StereoInvestigator was used to estimate neuronal volumes in layers II and III of somatosensory cortex. These can be effectively differentiated from neurons in layers I (smaller Golgi type II cell) and IV (packed smaller stellate cells). The nucleator probe was simultaneously run with the optical fractionator to achieve a systematic random sample. A ×60 apochromat oil lens was used for cellular volume measurement. Five rays were placed emanating from the nucleolus. The intersections of the rays with the boundary of the cell soma were marked for each sampled cell. A cell was defined as a neuron if it contained a large, clear, lightly staining nucleus with a single distinct nucleolus, surrounded by distinct staining of the perikarya, differentiating it from other common cell types in the brain like glia or endothelial cells. A neuron was measured in the sample if the nucleolus came into focus within the counting frame or intersecting the lines of inclusion. For the analysis, 20 randomly selected neurons were used (Gundersen et al. 1988; Shumann and Amaral 2005).

Behavioral Testing

Male pups were acclimated to the experimental environment for a week before the start of behavioral testing (Dudchenko 2004; Palanisamy et al. 2011). The testing was performed by 5 blinded experimenters (1 per test). The behavioral tasks were designed to detect changes in locomotor activity (spontaneous locomotor activity test), motor coordination and balance (rotarod performance test), sensorimotor function (adhesive removal test), exploratory behavior (novel object recognition test), and spatial working memory (radial arm maze test) (Wenk 2001; Dudchenko 2004; Sughrue et al. 2006; Bouet et al. 2009; Ennaceur et al. 2009; Palanisamy et al. 2011; Westin et al. 2010; Grandjean and Landrigan 2014; Xiong et al. 2014) (details in Supplementary Methods).

Protein Expression Analysis

Two randomly chosen embryos from each pregnant animal were extracted (total n = 8; 4 from anesthetized and 4 from control pregnant rats), forebrains dissected, and frozen at −80°C. Subsequently, protein extraction and western blotting were performed and described in Supplementary Methods (Stammer et al. 2008; Penumatsa et al. 2010; Inada et al. 2011).

Statistical Analysis

The primary outcome measure in the study was the fraction of BrdU+ cells in the topmost cortical bin, and that measure was used to calculate effect size (Cohen 1988) and sample size (Dang et al. 2008; Rochon 1998). To estimate the effect size we reparametrized the model with the primary outcome as a dependent variable and assumed a normal distribution with random effects using both anesthetic and duration as separate factors, while maintaining dam (mother) nested within anesthetic as a random effect. With a 5% alpha, 80% power, logit link, and an exchangeable autocorrelation structure with a rho correlation of 0.4, we require 18 pups per anesthetic duration to test difference among drugs, for a total of 54 pups to evaluate an effect size of 40% topmost bin reduction for anesthetics conditions versus control.

The proportion of BrdU+ cells in the topmost cortical bin (bin I) was formally assessed by generalized linear mixed model, parameterized as a logistic regression (binomial link function) predicting the percentage of cells in the topmost bin relative to all the stained cells in the grid. The correlated measure in the model was the individual rat (2 grids per section/image, 3 sections per rat) nested within their mother with a compound symmetry covariance structure. The model included 2 factors: anesthetic exposure (3 levels: oxygen/control, propofol, isoflurane) and exposure duration (3 levels: 20, 50, 120 min) as well as an interaction effect among these 2. Structural factors of grid (2 levels: first grid, second grid) and section (3 levels) were also included. Post hoc testing was performed with adjustment using the Tukey multiple comparison method.

Behavioral data from pups born to anesthetized and control rats were analyzed with a multilevel mixed model in which experimental condition and exposure duration were modeled as fixed effects, with a random effect to account for nonindependence among pups born to the same dam. In experiments in which behavioral testing occurred over several days, the time point (day) was entered as a repeated measures effect with a lagged autoregressive covariance structure. General linear models with tests of interaction were used to evaluate the effects of cortex volume and cell volume. Details on each model are provided in the Supplementary Methods.

All analyses were performed using SAS 9.3 (SAS Inc., Cary, NC, USA) using, among other, PROC MIXED, PROC GENMOD, PROC GLM, GEEsize and GLIMIX macros. The plots were generated in SigmaPlot (version 9.01; Systat Software Inc., San Jose, CA, USA) (Raudenbush and Bryk 1992; Littell et al. 2006).

Results

Dispersion of BrdU+ cells across the somatosensory cortex of rats prenatally exposed to anesthetics

During all anesthetic protocol heart rate, mean arterial blood pressure, oxygen saturation, CO₂, ventilation rate, and temperature of the pregnant rats were stable and well within physiological limits (Table 2). All the pups were viable, nursed by their mothers, and there was no difference in the litter size between the oxygen-exposed and anesthetized dams (Table 3).

Table 2

Heart rate (bpm), mean arterial pressure (mmHg), temperature (°C), ventilation rate (n/min), and hemoglobin oxygen saturation (%) of pregnant rats exposed to oxygen (control), propofol, or isoflurane on E17, each at 3 exposure durations

Exposure duration (min)	Condition	N of dams	HR (bpm)/MAP (mmHg)Temp (°C)VR (n/min)/O2Sat (%)
			15 min	30 min	45 min	60 min	75 min	90 min	105 min	120 min
120	Oxygen	2	413 ± 33/110 ± 10 37.3 ± 0.4 108 ± 11/96 ± 3	418 ± 35/108 ± 13 37.5 ± 0.4 96 ± 11/94 ± 2	407 ± 18/103 ± 11 37.6 ± 0.2 98 ± 10/94 ± 1	401 ± 22/108 ± 11 37.6 ± 0.2 93 ± 9/95 ± 1	398 ± 23/109 ± 13 37.7 ± 0.1 91 ± 11/95 ± 2	379 ± 18/100 ± 9 37.7 ± 0.2 89 ± 8/95 ± 2	369 ± 32/99 ± 7 37.8 ± 0.1 92 ± 8/95 ± 1	362 ± 25/108 ± 7 37.8 ± 0.2 95 ± 8/94 ± 3
	Propofol	2	421 ± 55/90 ± 16 38.3 ± 0.4 82 ± 5/99 ± 2	420 ± 19/82 ± 7 38.0 ± 0.3 84 ± 3/93 ± 1	421 ± 40/87 ± 14 37.5 ± 0.0 88 ± 6/95 ± 2	402 ± 15/77 ± 11 37.4 ± 0.3 78 ± 8/94 ± 3	387 ± 8/81 ± 5 37.2 ± 0.4 78 ± 2/97 ± 2	415 ± 75/88 ± 13 37.1 ± 0.1 79 ± 8/99 ± 2	361 ± 29/86 ± 5 37.5 ± 0.1 75 ± 4/96 ± 3	370 ± 26/86 ± 5 37.2 ± 0.3 91 ± 2/94 ± 6
	Isoflurane	2	413 ± 25/90 ± 12 37.8 ± 1.1 76 ± 5/99 ± 3	446 ± 17/110 ± 8 38.0 ± 0.7 60 ± 4/95 ± 1	445 ± 45/107 ± 9 37.5 ± 0.1 81 ± 5/98 ± 2	401 ± 26/96 ± 5 37.5 ± 0.1 77 ± 5/95 ± 1	372 ± 39/97 ± 7 37.0 ± 0.1 81 ± 9/99 ± 2	353 ± 46/94 ± 6 36.8 ± 0.3 73 ± 14/97 ± 2	358 ± 40/87 ± 4 37.1 ± 0.1 68 ± 7/94 ± 3	369 ± 28/94 ± 12 37.3 ± 0.1 66 ± 4/91 ± 2
50	Oxygen	2	392 ± 32/108 ± 11 37.4 ± 0.4 105 ± 13/95 ± 3	402 ± 26/104 ± 9 37.5 ± 0.2 100 ± 11/96 ± 2	410 ± 35/118 ± 13 37.6 ± 0.2 100 ± 11/93 ± 2
	Propofol	2	408 ± 47/84 ± 9 37.5 ± 0.2 92 ± 6/95 ± 3	368 ± 37/83 ± 6 37.6 ± 0.2 81 ± 6/92 ± 3	373 ± 26/86 ± 8 37.7 ± 0.2 88 ± 11/95 ± 2
	Isoflurane	2	374 ± 46/98 ± 12 37.4 ± 0.4 73 ± 5/94 ± 3	382 ± 32/107 ± 11 37.6 ± 0.2 86 ± 9/93 ± 2	389 ± 29/101 ± 6 37.6 ± 0.3 83 ± 5/94 ± 3
20	Oxygen	2	413 ± 22/106 ± 8 37.3 ± 0.3 88 ± 10/96 ± 3
	Propofol	2	392 ± 42/85 ± 9 37.4 ± 0.3 95 ± 9/93 ± 3
	Isoflurane	2	383 ± 32/98 ± 9 37.2 ± 0.4 78 ± 11/92 ± 2

Note: The measurements were performed noninvasively, at 15-min intervals. Data presented as mean ± SD. HR, heart rate; MAP, mean arterial pressure; temp, temperature; VR, ventilation rate; O₂ Sat, hemoglobin oxygen saturation.

Table 3

Average size of rat litters exposed to oxygen (control), propofol, or isoflurane on E17, each at 3 exposure durations

Exposure duration (min)	N of dams	Average litter size (pups)
		Oxygen	Propofol	Isoflurane
20	2	10 ± 1	11 ± 1	11 ± 0
50	2	11 ± 3	11 ± 1	11 ± 1
120	2	9 ± 2	10 ± 1	10 ± 3

Note: Data presented as mean ± SD.

Female pups were euthanized on P10, and their brains were stained to expose the final positions of cells generated at E17. The count of BrdU+ cells (among total cells) in the topmost histological bin (Fig. 3) revealed that the dispersion of BrdU+ cells, that is, decrease of fraction of BrdU+ cells in bin I, was significantly associated with treatment (drug) (P < 0.001), exposure time (P < 0.001), and treatment by exposure time interaction (P < 0.001). As illustrated by Fig. 3, in rats prenatally exposed to anesthetics, BrdU+ cells were more dispersed: The topmost bin (bin I) contained a smaller number of BrdU+ cells when compared with the oxygen-exposed animals, whereas the deeper layers (i.e. bins II and III) as well as the underlying white matter contained more of such cells. In some cases, ectopic BrdU+ cells formed a distinct band at the white matter/cortical border, resembling white matter neuronal ectopias. In contrast to that, the majority of BrdU+ cells in immunohistochemically stained sections from the oxygen-exposed animals accumulated in the topmost cortical bin (bin I), whereas a small number of BrdU+ cells were scattered in the bottom bin (bin III). It should also be noted that the examination of the H&E histological preparations did not reveal any difference in cytoarchitectonic organization between the anesthetized and oxygen-exposed animals.

Moreover, to quantitatively assess the magnitude of reduction of number of BrdU+ cells in the topmost cortical bin upon exposure to anesthesia, we used an estimator of effect size (Cohen’s d). For the histological data presented in Fig. 3, Cohen’s d was 2.67. We also explored the variation between dams and the interaction of that variation with the primary outcome. The experimental factors were reparametrized as follows: cross of anesthesia type (3 means) by duration of anesthesia (3 durations) created 9 experimental conditions, each with 2 dams (i.e. dams were nested within experimental conditions). Based on an ANOVA model, the mean squares for the experimental conditions was 1.671 and for the dams 0.024. In the other words, the variability in the dams was 67 times smaller than the variability between the experimental conditions. We also tested the nested dam factor to see if there were differences between the dams within the same condition, which yielded a nonsignificant result (P = 0.249).

Subsequent analysis of the exposure time for each of the anesthetic conditions was significant for isoflurane (P < 0.001) and propofol (P < 0.001) but not oxygen (P = 0.181). That indicates that, for groups exposed to anesthetic agents, there was an overall inverse relationship between the duration of anesthesia and the numbers of BrdU+ cells in the topmost bin (bin I).

Considering the individual exposure times, effect of treatment was not significant for 20 min (P = 0.330) but was significant for both 50 min (P < 0.001) and 120 min exposures (P < 0.001), pointing to a difference in dispersion of BrdU+ cells at those 2 durations of anesthesia, depending on the anesthetic agent. Post hoc testing revealed that the proportion of BrdU+ cells in the topmost bin of rats exposed to isoflurane was significantly reduced compared with the oxygen-treated group when anesthesia lasted for 50 min (P = 0.024 vs. oxygen) or 120 min (P = 0.003 vs. oxygen). Moreover, anesthesia with propofol for 120 min was also characterized by decrease in BrdU+ cells in the topmost bin (P < 0.001 vs. the oxygen-exposed group) (Fig. 3).

In addition, smaller numbers of BrdU+ cells situated in the topmost bin (bin I) in pups exposed to isoflurane or propofol were mirrored by greater numbers of BrdU+ cells in the lower bins (bins II and III) and throughout the subcortical white matter, compared with the oxygen-exposed pups (Supplementary Fig. 1). The tendency toward accumulation of BrdU+ cells in the bottom bin (bin III) was particularly emphasized by normalizing the numbers of BrdU+ cells in anesthetized animals to the number of BrdU+ cells in the oxygen-exposed animals (Supplementary Fig. 1). It should be also noted that small differences in absolute magnitudes of BrdU+ cell counts in the lower bins of drug-exposed pups compared with the oxygen-exposed pups, represented large proportional increases (Supplementary Fig. 1).

Furthermore, we performed stereological analysis of motor and somatosensory cortex volumes and neuronal cell volumes in layers II and III to rule out possibility that different experimental conditions affected overall size of the brain and/or neurons. Modeling the effect of both treatment (anesthesia type) and exposure duration on cortical volume did not yield significance (P = 0.382). No differences were found between the drugs (oxygen [mean ± standard deviation {SD}]: 53 ± 4 mm³; propofol: 53 ± 3 mm³; isoflurane: 51 ± 4 mm³; P = 0.459) or between exposures (20 min: 54 ± 3 mm³; 50 min: 51 ± 3 mm³; 120 min: 51 ± 4 mm³; P = 0.271). Similar to that, treatment and duration were not associated with neuronal volume (P = 0.493). Cell volumes were comparable between the drugs (oxygen: 1139 ± 85 μm³; propofol: 1078 ± 103 μm³; isoflurane: 1212 ± 148 μm³) and across exposure durations (20 min: 1148 ± 64 μm³; 50 min: 1156 ± 230 μm³; 120 min: 1127 ± 47 μm³; P = 0.955).

Behavioral Deficits of Young Adult Rats Prenatally Exposed to Anesthesia

To explore whether the increased percentage of ectopic neurons affects normal cortical functions, we conducted behavioral tests on male pups, starting on P49: motor skills and coordination (spontaneous locomotor activity and rotarod tests, respectively), sensorimotor function (adhesive tape removal test), and learning and memory (novel object recognition and radial arm maze tests, respectively).

Spontaneous Locomotor Activity and Rotarod Tests

Performance of rats in utero exposed to isoflurane or propofol on the spontaneous locomotor activity tests was comparable to that of oxygen-exposed rats, as measured by distance traveled, and general movement patterns in the horizontal and vertical planes (P > 0.05; Supplementary Fig. 2). Similarly, rats prenatally exposed to isoflurane or propofol showed no difference in latency to fall from an accelerating rotating rod measured in 3 trials per day when compared with the oxygen-exposed animals, for any of the anesthesia durations (20, 50, and 120 min) (Fig. 4).

Figure 4

Rats prenatally exposed to anesthesia perform comparable to controls on the rotarod test. On E17, pregnant rats were exposed to oxygen (control) (n = 6), propofol (n = 6), or isoflurane (n = 6), for 20, 50, and 120 min (n = 2 for each exposure duration). From P49, all the male rats from their litters (n = 86) underwent rotarod testing and latency to fall from an accelerating rotating rod in 3 trials per day was calculated. Effects were tested by repeated-measures general linear mixed-effects model with a first-order autoregressive covariance structure and adjusted for error inflation using Tukey’s post hoc text. The data points are mean ± standard error.

Novel Object Recognition Test

Novel object recognition test relies on rodents’ inherent preference for novelty and depends on perirhinal and visual cortex and the hippocampus. Our testing showed no difference in the number of novel object visits or total time spent with novel object between the rats in utero exposed to either oxygen, propofol, or isoflurane, for the first 2 exposure times, that is, 20 or 50 min (P > 0.05). However, rats prenatally exposed to isoflurane for 120 min visited a novel object less frequently and spent less time with it than rats exposed to oxygen or propofol (Fig. 5A) (P < 0.05).

Figure 5

Rats prenatally exposed to anesthesia underperform on behavioral tests. On E17, pregnant rats were exposed to oxygen (control) (n = 6), propofol (n = 6), or isoflurane (n = 6), for 20, 50, and 120 min (n = 2 for each exposure duration). From P49, all the male rats from their litters (n = 86) underwent behavioral testing. The data points are mean ± standard error (± standard error from the model for the combined maze results). *Tukey-adjusted P value versus the oxygen-exposed group <0.05, #Tukey-adjusted P value versus the propofol-exposed group <0.05. Novel object recognition test (A). The number of visits to a new object and total time spent with the object within 2 min were recorded. Effect was tested by general linear mixed-effects model. Adhesive tape removal test (B). Three trials on each paw (left and right) were performed and time to remove the tape (limited to a maximum of 30 s) was recorded. Effect was tested by repeated-measures general linear mixed-effects model with a first-order autoregressive covariance structure. Combined radial arm maze test results (C). Testing consisted of 10 min sessions, on 3 consecutive days, in which the rat was placed in the maze with 8 arms with food rewards. Three direct performance measures were recorded (shown in Supplementary Fig. 3): the time needed to complete the task, number of correct choices before the first error, and total number of errors. To adjust for the learning effect over 3 days, a repeated measure general linear mixed-effects model with a first-order autoregressive covariance structure was run, and the results plotted in this figure.

Adhesive Tape Removal Test

Rats in utero exposed to isoflurane were less engaged in tape removal than the oxygen-exposed controls in 3 trials on each paw, irrespective of the anesthesia duration (P < 0.05 for each time point). Time spent trying to remove the tape was also significantly shortened in animals in utero anesthetized with propofol for 120 min (mean ± SD: 21 ± 2 s and 28 ± 0.5 s, in propofol and oxygen-exposed group, respectively; P < 0.05 vs. the oxygen-exposed group) (Fig. 5B).

Radial Arm Maze Test

Finally, we assessed memory and spatial learning of prenatally anesthetized rats by using 3 direct metrics of the radial arm maze test: time to complete the maze (min), number of choices before first error, and the total number of errors. We also used a combined radial arm maze test score, by first rescaling 3 direct measures to a 0–100 scale and then calculating the average across the 3, with larger scores indicating better performance (detailed results in Supplementary Fig. 3).

In addition, we analyzed the combined scores on the radial arm maze test after adjusting for the learning effect over the 3 days. As can be seen in Figure 5C, the overall results paralleled those without the adjustment (Supplementary Fig. 3), that is, the rats prenatally exposed to isoflurane had inferior performance than the oxygen-exposed rats (on the average 20 points on a 0–100 scale) for all the exposure times (P < 0.05 at each time point), and also did worse than the propofol-exposed rats at 20 min and 50 min (P < 0.05 at each time point). The propofol group, on the other hand, achieved scores similar to the oxygen group for the first 2 exposure durations, but did poorly at 120 min, being quite similar to the isoflurane group.

Reduced Reelin and GAD67 Protein Expression in the Brains of Rat Embryos Exposed to Anesthesia

To provide an insight to possible molecular mechanisms causing increased dispersion of migrating neurons and postnatal behavioral deficits, we analyzed levels of reelin and GAD67 in anesthetic and oxygen-exposed embryos. To that end, 4 pregnant rats were exposed to 1 MAC of isoflurane or oxygen at E16 (n = 2 per exposure) and sacrificed at E20. Two randomly chosen embryos were removed from each dam (n = 4 per exposure) and western blots performed on the brain lysates. Reelin isoforms were detected, with the density of the most intensive isoform, at 175 kDa, approximately 30% lower in the brains of isoflurane group, compared with the embryos from the oxygen group (Student’s t-test; P = 0.045) (Fig. 6A). Similarly, the GAD67 band (~67 kDa) densities were also reduced in the pups prenatally exposed to isoflurane (Student’s t-test; P = 0.017) (Fig. 6B).

Figure 6

Embryos exposed to anesthesia have lower expression of reelin and GAD67 proteins in their brains compared with controls. Pregnant rats were exposed to oxygen (control; n = 2) or isoflurane (treated; n = 2) for 120 min on E16. Two randomly chosen embryos from each pregnant rat (total n = 8) euthanized on E20 were extracted and their brains harvested. o: pups exposed to oxygen, i: pups exposed to isoflurane. (A) Western blot of total brain proteins immune probed for reelin (~175 kDa) (upper panel) and densities of the reelin band in control and isoflurane-exposed animals after normalizing for the density of the respective actin bands (lower panel). (B) Western blot of total brain proteins immune probed for GAD67 (~67 kDa) (upper panel) and densities of the GAD67 band in control and isoflurane-exposed animals after normalizing for the density of the respective actin bands (lower panel). Data are presented as mean ± SD and compared by Student’s t-test. *P < 0.05 versus the oxygen group.

Discussion

Development of cerebral cortex is genetically determined, precisely choreographed, and highly regulated. In rodents, it develops in a rostro-caudal direction with frontal areas, including motor cortex, developing early and somatosensory cortex, visual cortex, and hippocampal formation developing later (Bayer 1980; Schull et al. 1986; Rakić 1990; Bayer and Altman 1991, 2004; Komuro and Rakić 1998; Ang et al. 2003, 2004; Metin et al. 2008; Jones and Rakić 2010; Buschbaum and Cappello 2019). The brain is therefore differentially vulnerable to epigenetic insults, with histological and functional outcomes depending on the individual’s susceptibility and the timing of the noxious event (Bayer 1980; Schull et al. 1986; Bayer and Altman 1991, 2004; Rodier 1995; Ang et al. 2003, 2004; Stratmann 2011; Palanisamy 2012; Turski and Ikonomidou 2012). Although effects of anesthetics on neuronal apoptosis and/or synaptogenesis have been described (Mazze et al. 1985; Uemura et al. 1985; Jevtović-Todorović et al. 2003; Cattano et al. 2008; Turski and Ikonomidou 2012), this is the first report on neuronal migration.

Our study demonstrated that exposure of pregnant rats to anesthetics affects neuronal migration in the developing cerebral cortex of their fetuses causing that some of the neurons never reach their proper and predetermined laminar positions. More precisely, the histological phenotype of their offspring was characterized by small but significant number of neurons destined for the upper cortical layers that remained inappropriately scattered within deeper cortical layers and/or in the adjacent white matter of the somatosensory cortex. Considering that somatosensory cortex, visual cortex, and hippocampal formation are developing simultaneously (during E17–22 period), we anticipated developmental aberrations in aforementioned areas due to E17 timing of anesthetic exposure (Bayer 1980; Schull et al. 1986; Bayer and Altman 1991, 2004; Clancy et al. 2001). Consistent with our expectations, the behavioral phenotype of offspring was characterized by normal motor activity, that is, development of motoric cortex was almost completed at the time anesthetics exposure at E17, but impaired somatosensory function, novel object recognition memory, and acquisition of spatial working memory. Our notion that behavioral deficits are due to migratory abnormalities was further reinforced by the observation that changes of behavioral phenotype tend to be more extensive with isoflurane in comparison with propofol and increase with exposure duration. These histological and behavioral aberrations are unlikely to be explained by an indirect adverse effect of anesthetics on maternal well being because maternal systemic physiology was normal, and there were no differences in litter size, viability, and weight between the anesthetized and control animals.

Considering molecular mechanisms, we found reduced levels of reelin and GAD67 in the brains of fetuses exposed to isoflurane. Both molecules are essential for the normal cerebral cortex development (Mazze and Kallen 1989; Rakić 1990; Alcántara et al. 1998; Ang et al. 2003; Hashimoto-Torii et al. 2008; Popp et al. 2009; Jones and Rakić 2010; Inada et al. 2011). Reelin is an extracellular matrix glycoprotein that regulates neuronal migration and positioning in the developing brain by controlling cell–cell interactions. It is synthetized by Cajal–Retzius cells, the earliest generated neurons in the developing cortex, with peak levels achieved during the most intense period of neuronal migration (Rakić 1990, 2003; Alcántara et al. 1998; Pesold et al. 1998; Gleeson and Walsh 2000; Ang et al. 2003; Janusonis et al. 2004; Hashimoto-Torii et al. 2008). Reelin exists in the form of multiple peptides that bind to different receptors and have several functions. In the early stages of corticogenesis, reelin acts as a positional signal to migrating neurons and instructs them to detach from their guiding radial glial fibers and stop their migration. In the late stages of corticogenesis, the cerebral cortex becomes thicker and small reelin fragments diffuse from the MZ toward the VZ and create a top-down gradient, which drives newly generated neurons to migrate toward the MZ. The effects of reelin on cytoskeleton dynamics and on adhesive properties of neurons are likely to be different (permissive or inhibitory) by binding to different receptors during different migration phases which determines proper radial movement of migrating neurons: first go then stop (Zhao and Frotscher 2010). Therefore, disruption of the reelin gradient may theoretically prevent late generated neurons from passing early formed layers and cause their detachment from radial glial cells before their leading processes reach the reelin-rich MZ. In addition, recent studies challenged these concepts and demonstrated that the sheet of reelin-rich Cajal–Retzius cells that covers the neocortical primordium is not essential for cortical layers formation and/or order (Yoshida et al. 2006). On the other hand, GAD67 is a GABA-synthesizing enzyme isoform expressed early in the developing cortex. It is crucial for cortical development, as GABA helps regulate the different properties of neuronal progenitors, including multidirectional tangential migration of GABA-ergic interneurons (Gleeson and Walsh 2000; Ang et al. 2003; Manent and Represa 2007; Cai et al. 2009; Popp et al. 2009; Wang and Kriegstein 2009; Inada et al. 2011; Lim et al. 2018). It is also important to notice that approximately 70% of the neurons that express GAD67 also express reelin (Pesold et al. 1998, 1999). Finally, abnormal levels of reelin and GAD67 have also been described in several neurodevelopmental pediatric neurological disorders—including autism—that have been recently shown to have abnormalities of cortical columnar organization as well (Pesold et al. 1998, 1999; Janusonis et al. 2004; Bertoglio and Hendren 2009; Popp et al. 2009). Therefore, here demonstrated disruptions of cortical organization due to prenatal exposure to anesthetics and consequent postnatal behavioral deficits may be conceivably explained by diminished expression of reelin and GAD67. In addition, interneurons arise extracortically from the medial ganglionic eminence during corticogenesis. During their tangential migration to cerebral cortex, these neurons gradually acquire responsiveness to GABA, which is directed by ambient GABA levels and mediated by GABA receptors (Sun et al. 2015). Recent studies showed that pharmacological or genetic manipulation of GABA levels and/or GABA receptors during corticogenesis results in ectopic cell clusters in upper layers and loss of cortical lamination pattern (Luhmann et al. 2015). This implicates that anesthetics affecting GABA levels and/or acting on GABA receptors may also disturb the tangential migration and distribution of interneurons within cortical columns.

Alternative molecular events may be also considered. Rash et al. (2016) demonstrated that Ca²⁺ fluxes propagate bidirectionally within the radial glial cells and provide communication pathway between the proliferative and postmitotic zones—such as CP—in developing brain. That is especially relevant here, since inhalational anesthetics—such as isoflurane—cause apoptosis in radial glial cells via excessive Ca²⁺ release by directly activating inositol 1,4,5-trisphosphate receptors on endoplasmic reticulum (Zhai et al. 2015). Finally, earlier studies have shown that anesthetics induce widespread apoptotic neuronal degeneration in newborn animals, followed by impairment in learning and memory later on in adulthood (Yasuda et al. 2002; Jevtović-Todorović et al. 2003; Cattano et al. 2008; Loepke and Soriano 2008; Bercker et al. 2009; Lu et al. 2010; Xiong et al. 2014). However, it is important to point out that the early postnatal period in rodents corresponds to the late third trimester in humans with respect to brain development—a period normally characterized by intense apoptosis and synaptogenesis (Bayer and Altman 1991, 2004; Rabinowicz et al. 1996; Clancy et al. 2001). In addition, classic neurodegenerative diseases, such as amyotrophic lateral sclerosis and Parkinson’s disease, show a 50-80% loss of targeted neurons before the clinical onset (Lloyd 1977; Bradley 1987), illustrating remarkable ability of mammalian brain to compensate for neuronal loss by plasticity and increased synaptogenesis (Kolb and Whishaw 1989; Jones and Schallert 1994; Jones et al. 1996). Therefore, increased apoptosis may not be reliable neuropathological correlate of anesthetic neurotoxicity or predictor of abnormal behavior later on in life.

Taking the molecular and histopathological data together, we suggest a plausible explanation of anesthetics’ effect on the developing fetal cerebral cortex cytoarchitecture and subsequent function (Fig. 7). Typically, cells generated in the proliferative VZ on E17 migrate radially during the following 2 days and settle in the prospective layers II and III, to be bypassed by the subsequently generated, unlabeled neurons that settle in the more superficial tiers of layer II (Fig. 7) (Bayer 1980; Rakić 1990; Bayer and Altman 1991, 2004; Ang et al. 2003, 2006). If cells generated on E17 that were en route to the cortex were hampered by exposure to anesthetics, some of them would not arrive to their correct position. Subsequently, these cells become intermixed with the earlier-generated neurons within the deeper cortical layers or remain in the adjacent white matter (Fig. 7). These ectopic cells would then disrupt normal cortical columnar organization and ultimately result in behavioral deficits (Rakić 1990; Gleeson and Walsh 2000; Casanova et al. 2002a, 2002b; Ang et al. 2003, 2006; Metin et al. 2008; Jones and Rakić 2010). Our experimental design further strengthens our results since specific exposure timing at E17 causes specific histological aberrations in the areas of the brain that are intensively developing in that period, which later on results in behavioral deficits that are specific for the involved areas.

Figure 7

Anesthetics’ effect on neuronal migration in the developing cerebral cortex. In a normal rat (control), most cells labeled with BrdU (shaded in red) at E17 arrive in the cortex by E19. By P1, those cells become bypassed by subsequently generated neurons and will hence settle predominantly in cortical layers II and III. However, when cells generated at E17 are exposed to anesthetics they slow down, and eventually remain positioned in the deeper cortical layers or white matter. Radial glial cells spanning from the ventricular zone (VZ) to the cortical plate (CP) and marginal zone (MZ) are depicted in white. I–VI: cortical layers.

An obvious question is whether conclusions based on an animal model are relevant for human cortical development, behavior, and clinical practice. Duration of neuronal proliferation and the migratory phase of cortical neurons in the human fetus lasts approximately 12 times longer, occurring between the 6th and 24th week of gestation, with the peak between the 11th and 15th week, compared with the duration of only 1.5 weeks (between E11 and E21, approximately) in the rat (Bayer 1980; Bayer and Altman 1991, 2004; Rodier 1995; Berbel et al. 2001; Clancy et al. 2001; Ang et al. 2003, 2006; Glunčić 2009; Stratmann 2011; Zou et al. 2011; Palanisamy 2012; Turski and Ikonomidou 2012; Sanders et al. 2013; Selemon et al. 2013; Grandjean and Landrigan 2014). Based on that, it can be assumed that the anesthesia durations of 20, 50, and 120 min in our experiments approximately correspond to 4, 10, and 24 h in humans, respectively (Ang et al. 2006). The anesthetic exposure in the present study may have involved a much larger proportion of the time dedicated to development of the cerebral cortex than it would during a typical surgical procedure in humans and could have had a greater overall effect. There are, however, reasons to assume that anesthetics may have a similar or even greater impact on neuronal migration in the human fetal brain. First, the migratory pathways in the convoluted human cerebrum are curvilinear and at least an order of magnitude longer. That means that the number of neurons migrating along the same radial glial fascicle, particularly at the later stages of cortical neurogenesis, is much larger and their routes are more complex, increasing the chance of cells going astray from their proper migratory course (Schull et al. 1986; Rakić 1990, 2003; Ang et al. 2006; Jones and Rakić 2010; Duque and Rakić 2011). Second, the inside-to-outside settling pattern of isochronously generated neurons in primates is more layer-specific than in rodents and conceivably more sensitive to any malposition of the neurons (Rakić 1990, 2003; Algan and Rakić 1997; Casanova et al. 2002a, 2002b; Janusonis et al. 2004; Ang et al. 2006; Jones and Rakić 2010; Duque and Rakić 2011).

Although our study focused on effects of anesthesia on prenatal neuronal migration, it should be noticed that continuous generation of neurons, oligodendrocytes, and astrocytes was demonstrated in the hippocampus of rodents and—to much lesser extent—in young nonhuman primates (Lois and Alvarez-Buylla 1994; Kornack and Rakić 1999; Deng et al. 2010; Duque and Rakić 2011). Furthermore, it has been suggested that local, short-distance migration or migration of isolated neurons exist in the postnatal primate brain (Lois and Alvarez-Buylla 1994; Kornack and Rakić 1999; Ghashghaei et al. 2007; Deng et al. 2010). We hypothesize that the interference of anesthetics with these migratory events—that are primarily noticed in the hippocampal formation and related with learning and memory—is a morphological basis of behavioral and learning difficulties of toddlers (Sprung et al. 2012) and/or cognitive deficits in adults (Fines and Severn 2006) after prolonged and repeated exposure to anesthetics.

Nevertheless, there are several limitations of our study. First, rodent animal models inherently have limited relevance for clinical practice (Clancy et al. 2001; Ang et al. 2006). Second, proposed mechanism of abnormal behavior is based on assumption that opposite sex animals from the same litter have same histological and consequent behavioral abnormalities; that is, for practical reasons and to eliminate the influence of estrus stage on behavior we investigated behavioral changes in males and cortical histological cortical abnormalities in females. The third methodological limitation is the fact that BrdU by itself may disturb neuronal migration as it inherently compromises DNA transcription. Therefore, BrdU cell labeling as an indicator of neuronal migration pattern should be limited to its usage for comparison of different conditions as done in this study (Duque and Rakić 2011). However, laminar distribution of BrdU-labeled cells generated at E17 in our control animals closely resembles those previously reported by studies that were using the same methodology for the assessment of other epigenetic factors’ impact on neuronal migration (Berbel et al. 2001). The fourth caveat is the assumption that the majority of the ectopic BrdU-positive cells detected in deeper cortical layers and the subjacent white matter are actually neurons. Recent study by the Rakić group that complemented BrdU staining by immunohistochemistry has shown that approximately 85% of cells generated in rodent brain on E17 and labeled by the BrdU using the same methodology were also positive for neuron-specific marker NeuN, including those in layer VI and adjacent white matter. The BrdU-positive cells were negative for FoxP2, a marker of deep-layer neurons but were, at the same time, positive for Brn1, maker for superficial-layer neurons. Together, those findings indicate that cells labeled by BrdU on E17 can be considered neurons destined to superficial layers that were arrested along their migratory pathway (Ang et al. 2006). The fifth potential limitation is that in order to quantify neuronal laminar distribution we used standardized counting method (Berbel et al. 2001; Ang et al. 2006) rather than stereology. We recognized that for this particular question stereology may potentially lead to reduction in power since histopathological deviations in the analyzed brains were topological, that is, not distributed diffusely but rather segmentally, within deeper cortical layers (Russ and Dehoff 1999). The deployed quantification method actually provided an additional level of control by matching the laminar distribution of neurons in the control animals in this study to the previously published results (Berbel et al. 2001). Nevertheless, we did use stereology to confirm that volumes of analyzed cortex and neurons were not affected by different conditions, which further ensured the soundness of our histopathological quantification (Berbel et al. 2001). Therefore, our results do not allow us to approximate the extent to which anesthetics affect migrating neurons in developing human fetal brain. However, due to complexity of human brain, even a small number of ectopic cells might, as a result of specific position and inappropriate connectivity, may be a source of epileptic discharge or abnormal behavior (Schull et al. 1986; Gleeson and Walsh 2000; Casanova et al. 2002a, 2002b; Ang et al. 2003, 2006; Janusonis et al. 2004; Metin et al. 2008; Glunčić 2009; Jones and Rakić 2010). In conjunction with recent behavioral pediatric studies, our study definitely raises reasonable concerns (Blair et al. 1984; Hollenbeck et al. 1985, 1986; Sprung et al. 2009, 2012; Wilder et al. 2009; Sun 2010; Ing et al. 2012; Sanders et al. 2013; Lee et al. 2014; Wang et al. 2014). The final consideration point is the sample size. To ensure sufficient statistical power we performed an a priori power analysis and used the calculated sample size (i.e. 54 pups). It is therefore not surprising that the results were statistically highly significant. Regarding robustness of our results to future replication, the observed effect size for the primary finding (histological impact of anesthesia), as assessed by Cohen’s d, can be interpreted as quite large (Sawilowsky 2009). We also emphasize that the unit of analysis for the histological findings is the pup, not the mother. Moreover, the data did not suggest that the differences between mothers contribute much to the effect of anesthesia. To give a sense of magnitude, the variability attributable to the experimental condition was 67 times greater than the variability attributable to mothers.

In conclusion, our results imply that more sophisticated methods should be deployed for testing possible drugs’ neurotoxicity during pregnancy (Glunčić 2009). Some of the medications, including anesthetics, therapeutics, and/or diagnostic physical agents currently approved for use in pregnancy may potentially have adverse effects on the developing human brain that may not be detected with standard neuropathological examinations (Ang et al. 2006; Glunčić 2009; Turski and Ikonomidou 2012). Considering prevalence of pediatric neurodevelopmental disorders such as autism, dyslexia, or attention deficit hyperactivity disorder, it is important to assess whether prenatal exposure to any of these agents contributes to this increase (Rodier 1995; Ang et al. 2006; Van DeVelde and De Buck 2007; Sprung et al. 2009, 2012; Wilder et al. 2009; Sun 2010; Stratmann 2011; Boyle et al. 2011; Palanisamy 2012; Selemon et al. 2013; Grandjean and Landrigan 2014; Lee et al. 2014; Wang et al. 2014). Furthermore, this study highlights the American Society of Anesthesiology’s recommendations to postpone elective surgeries in pregnancy until after delivery (Crowhurst 2002; Loepke and Soriano 2008; Sun 2010; Stratmann 2011; Palanisamy 2012; Perna et al. 2013; Sanders et al. 2013; Lee et al. 2014). In clinical practice, nonurgent surgeries in pregnancy are typically performed in the second trimester, after organogenesis, and when preterm contractions and spontaneous abortion are less likely. Based upon the present work, this concept should be reevaluated since the most active period of neuronal proliferation and migration in human fetal brain occurs between 12th and 24th week (Mazze and Kallen 1989; Rakić 1990; Clancy et al. 2001; Crowhurst 2002; Ang et al. 2006; Loepke and Soriano 2008; Stratmann 2011; Palanisamy 2012; Selemon et al. 2013). Finally, our findings call for examination of the effects of anesthetics on neuronal migration in nonhuman primates, where the duration of neurogenesis and the size and complexity of migratory pathways are more similar to those in humans (Rakić 1990; Rodier 1995; Ang et al. 2006; Metin et al. 2008; Stratmann 2011; Grandjean and Landrigan 2014).

Funding

Internal Departmental and Rush University Medical Center funding.

Notes

Parts of this work were presented at annual meetings of American Society of Anesthesiologists in 2012, 2013, 2014, and 2017.

Previously published as Mickiewicz AL