About 0.75–2% of pregnant women need non-obstetric surgery every year [1]. Middle pregnancy (namely, second trimester) is considered to be a safe period for surgical anesthesia. With the development of fetal surgery and laparoscopic technology, an increasing number of pregnant women undergo surgery under general anesthesia during middle pregnancy [1]. However, recently, studies have shown that middle pregnancy is a “busy period of neural development” in which neurons proliferate, migrate, and differentiate [2]. Anesthetics commonly used in clinical practice have been found to mainly affect two major neurotransmitter receptors, N-methyl-D-aspartate (NMDA) receptors and gamma-aminobutyric acid (GABA) receptors, during the development of the central nervous system (CNS). Exposure to general anesthetics before and after birth could affect brain development. A warning against prolonged and repeated exposure to general anesthetics during pregnancy was proposed by the American Food and Drug Administration. Sevoflurane, as a GABA type A (GABA${}_A$) receptor agonist/enhancer [3] is a common general anesthetic used in anesthesia for pregnant women. Hence, prenatal sevoflurane exposure during middle pregnancy could affect the neural development of offspring.

As an excitatory transmitter in the early stage of neural development, GABA is a key neurotransmitter in the development of the brain that plays an important role in the proliferation and migration of neurons [4, 5]. When neurons migrate to the target cortex, the process of neuronal migration is terminated by contact of GABA receptor. With the development of the brain, the GABA${}_A$ receptor changes from excitability to inhibition and depolarization to hyperpolarization [4, 5]. As a GABA${}_A$ receptor agonist, sevoflurane could play an important role in the proliferation and migration of neurons.

The “vulnerability window” of neurotoxicity of general anesthetics is mainly at the peak of brain nerve cell proliferation, migration, or synaptic development (i.e., during the middle or late stage of pregnancy) [2]. In middle pregnancy, the fetus’ brain is extremely sensitive to changes in the environment. Middle pregnancy is a critical period for brain nerve development (neuron proliferation, migration, and the formation of neural connections), and most of these events occur in middle pregnancy [6]. In addition to unavoidable emergency operations in middle pregnancy, selective operations are also mostly carried out in this period. Prenatal sevoflurane exposure in middle pregnancy may affect fetal neural development and cause neural dysfunction in offspring.

Interneurons, located with GABA${}_A$ receptors, can control and regulate the pyramidal neurons in the entorhinal cortex (ECT). The layers II and III (LII and LIII) pyramidal neurons of the ECT send out perforating fibers to transmit information to the hippocampus, and the hippocampus forms memories by editing and storing information [7]. The hippocampus is the key center of brain association, learning, and memory. Thus, interference with interneurons may lead to learning and memory dysfunction. However, previous related studies have been mainly focused on ultrastructural and functional impairments of pyramidal neurons after sevoflurane exposure [8, 9, 10, 11]. Relatively little attention has been paid to interneurons.

Neuregulin-1 (NRG1), a member of the epidermal growth factor family, plays an important role in promoting the phosphorylation of epidermal growth factor receptor-4 (ErbB4) [12, 13]. ErbB4 receptors have a variety of isomers, including ErbB (1–4). ErbB4 is highly expressed in the brain and has a strong affinity for NRG1. Most ErbB4 receptors are expressed on parvalbumin (PV) interneurons, which account for 40% of interneurons [12, 13, 14]. The NRG1–ErbB4 pathway can regulate and interact with the secretion function of GABA${}_A$ receptors [15]. The NRG1–ErbB4 pathway plays a great role in the occurrence, migration, and synaptic plasticity of neurons [16, 17, 18] by regulating the composition of interneurons’ subunit and the activity of receptor [19, 20, 21]. Furthermore, alterations of the NRG1–ErbB4 pathway could regulate key receptors connected to learning and memory ability, such as NMDA receptors [22, 23]. Moreover, exposure to general anesthetics can interfere with the key processes of dendritic growth and the development of pyramidal neurons [24]. Li et al. [25] found that disruption of NRG1–ErbB4 signaling in the PV-positive interneurons caused cognitive impairment in rats after exposure to isoflurane. Hence, it is possible that sevoflurane can induce neural dysfunction in offspring by regulating GABA${}_A$ receptors in the PV interneurons of the ECT through the NRG1–ErbB4 pathway. Consequently, the influenced NRG1–ErbB4 pathway could further regulate interneurons and pyramidal neurons and affect the information transmission function of the ECT.

Herein, we hypothesize that sevoflurane could regulate interneurons via the NRG1–ErbB4 pathway in the ECT of offspring during the middle pregnancy. Gestational day 14.5 in pregnant rats is similar to middle pregnancy in humans [26]. In this study, pregnant mice were exposed to sevoflurane in middle pregnancy. The status of the NRG1–ErbB4 pathway, interneurons, NMDA receptors, and the dendrite morphology of pyramidal neurons were examined to test the hypothesis. Discovering the mechanism of sevoflurane-induced neurotoxicity is of great significance for guiding the standardized clinical use of general anesthetics and research into toxicity prevention and treatment.

We previously speculated that prenatal sevoflurane, as an exogenous GABA${}_A$ receptor agonist, down regulated the NRG1–ErbB4 signaling pathway and that this change could lead to the disturbance of interneurons and the abnormal development of the dendrites of pyramidal neurons. The results of this study basically confirmed our hypothesis. This could be helpful for the standardization of the clinical use of sevoflurane and its toxicity prevention and treatment.

With the development of surgical technology and fetal surgery, an increasing number of pregnant women need to be exposed to general anesthetics. Sevoflurane mainly acts as GABA${}_A$ receptor agonist/enhancer and has sedative, analgesic, and muscle relaxant effects. Bartolini et al. [29] found that the inhalation of sevoflurane in 2MAC can inhibit uterine muscle contraction in a dose-dependent manner. Thus, based on its inhibition of uterine contraction, which can prevent premature delivery, sevoflurane is the most widely used inhalation anesthetic during pregnancy. Sevoflurane easily impacts fetuses though the placenta [30]. Zheng et al. [9] found that with the inhalation of 2.5% sevoflurane for 2 hours and 4.1% sevoflurane for 6 hours, the offspring of pregnant rat showed decreased learning and memory ability, accompanied by the release of inflammatory agents in the hippocampus and abnormal synaptic development. In recent years, a large number of studies have shown that sevoflurane it neurotoxic to the developing brain, which can lead to increased neuronal apoptosis, the inhibition of proliferation, neuronal development disorders, and long-term neurobehavioral abnormalities [31, 32]. However, most of the authors of these studies were focus on the hippocampus, which is related to learning and memory, and mainly studies the ultrastructural and functional impairment of projection neurons. Little attention has been paid to LII/LIII of the ECT, which is the key area of hippocampal information input. In LII/LIII of the ECT, interneurons, characterized as GABAergic neurons, could be regulated by sevoflurane. Impacts on the NRG1–ErbB4 pathway in interneurons may further regulate interneurons themselves and pyramidal neurons and affect the information transmission functions of the ECT.

In this study, we simulated clinical practice. The pregnant rats were anesthetized with 4% sevoflurane for 3 hours at 14.5 days of gestation to reach a MAC of 1.8 (lower than normal rats, considering that pregnant rats are more sensitive to sevoflurane), which was equivalent to that of most surgical operations for pregnant females. We found that sevoflurane, as an exogenous GABA${}_A$ receptor agonist, could down regulate the NRG1–ErbB4 signaling pathway (Fig. 2). NRG1 is highly expressed in mammalian embryos and decreases with age. It involves many aspects of neural development, including neuronal migration, survival, axon projection, myelin sheath development, synaptic formation and the regulation of neurotransmitter receptor expression [13, 16, 33, 34, 35, 36, 37]. Previous studies have revealed that the NRG1–ErbB4 pathway is necessary for the generation and migration of intermediate neurons originating from the medial ganglionic eminence (MGE) region [16, 38, 39, 40]. The impairment of the NRG1–ErbB4 pathway can affect the migration of PV interneurons from the MGE region. PV interneurons account for 40% of interneurons, and there was a decrease of 30–50% of interneurons in ErbB4 gene knockout rats. We administered sevoflurane anesthesia during middle pregnancy, which is the critical period of the development of the generation and migration of the embryonic interneurons. In this period, the MGE area produces PV interneurons, which migrate to the edge of the cortex and subventricular area and then radially to the target cortex [29]. Our results also showed that the NRG1–ErbB4 pathway was down regulated along with a decrease in the number of interneurons. Therefore, the decrease in the number of interneurons in LII/LIII of the ECT in the SeV group could have been due to the inhibition of the formation of interneurons by the decreased NRG1–ErbB4 level. At the same time, the decreased NRG1–ErbB4 blocked interneurons’ migration to LII/LIII of the ECT.

The NMDA receptor (NMDAR) is mainly composed of two NR1 and two NR2 subunits. The distribution of NMDAR subunits (NR1, NR2, NR3) also changes with the process of neurodevelopment. NR1 begins to increase after birth until puberty and reaches a peak level in the third week after birth. NR2B and NR2D are the main subunits of NR2 in the embryonic stage. The expression of NR2 changes significantly in the two weeks after birth. NR2A begins to partially replace NR2B until the brain matures. The development of NMDAR subunits marks the transition from immature to mature neurons and is closely related to the development of learning and memory ability. In our experiment, exposure to sevoflurane was found to decrease the expression of NR2A and increase the expression of NR2B in offspring, which is consistent with the findings of previous studies [31, 32]. NRG1–ErbB4 changes could increase the phosphorylation of NR2B by Fyn and, thus, reduce the internalization of NR2B and delay the transition from NR2B to NR2A. NRG1–ErbB4 can regulate not only the release and activity of GABAergic neurotransmitters but also the excitatory synapses on inhibitory neurons (interneurons). The intracellular segment of the ErbB4 protein contains post-synaptic postsynaptic dens 95 (PSD-95)/DiscsLarge/zonula occludens protein-1 (ZO1) (PDZ) domains, which are anchored to the postsynaptic membrane by interacting with other proteins (such as PSD-95) that also contain PDZ [41, 42, 43]. Thus, ErbB4 regulates the function of the NMDAR by the connection of PDZ [41, 42, 43]. The NRG1–ErbB4 pathway phosphorylates the phosphorylation site of NR2B through a member of the Src family of Fyn (SRC/Fyn), which blocks the internalization of NR2B by protein AP-2 [44, 45]. In this way, the NRG1–ErbB4 pathway increases the expression of NR2B on the postsynaptic membrane [44, 45]. Thus, the decreased NRG1–ErbB4 level results in increased NR2B and decreased NR2A levels (Fig. 3). This alteration may lead to the abnormal development of learning and memory ability.

Pyramidal neurons are responsible for information transmission, and their function is very important. Therefore, it has always been a research hotspot in the topic of the neurotoxicity of general anesthetics during development. Interneurons play an important role in regulating pyramidal neurons. We found that prenatal sevoflurane exposure in middle pregnancy resulted in a significant increase in the number of dendritic spines, a significant decrease in the total length of dendrites, and an abnormal spatial distribution of dendrites in offspring (Fig. 4). There could be three possible mechanism behind the increased number of dendritic spines. One explanation is that the decrease in the number of interneurons leads to the weakening of their regulation of pyramidal neurons, resulting in abnormal numbers of dendritic spines. The second explanation is that the change of the NRG1–ErbB4 pathway could regulate the dendritic spines of pyramidal neurons. Barros et al. [35] found that the number of pyramidal dendritic spines decreased after ErbB4 and ErbB2 knockout in the nervous system. However, the number of pyramidal dendritic spines did not decrease in the experiment of ErbB4 knockout by pyramidal cells [44, 46]. It is suggested that ErbB2 may play a role of functional compensation, and there is an over-compensation in the number of pyramidal dendritic spines [37, 47]. The third explanation is that the NR2B subunit of NMDAR combines with PSD-95 and calmodulin-dependent kinase II (CaMKII), which activates a series of downstream signaling pathways and leads to an increase in the number of dendritic spines [48]. The NR2B/PSD95/kalirin-7 pathway is very important in the development of neuronal dendritic spines. Recent studies have shown that PSD-95, kalirin-7, and NR2B form a complex with a postsynaptic membrane through the PDZ domain. NR2B can activate kalirin-7, and then activate Rac1, a downstream RhoGTPase family member [49]. In this way, NR2B could dynamically regulate actin cytoskeleton rearrangement, promote the growth of dendritic spines, induce the formation of spinous structures in neuronal bodies, and cause the excessive formation of dendritic spines in pyramidal neurons and interneurons [49]. The decrease of total length and branches of dendrites in offspring may be explained by the change of the NRG1–ErbB4 signaling pathway and NR2B subunit, which mainly involve the growth and pruning of dendrites [36, 37]. Further, the plasticity of dendritic spines affects learning and memory function, especially the formation of long-term memory [50, 51].

The initial experimental hypothesis has been verified in this study. However, to test the hypothesis, this study focused on the study of the ECT with a concentration on the memory transfer station. Thus, this study did not involve an investigation of the dentate gyrus area of the hippocampus, which was directly projected by pyramidal neurons in the ECT. The time (3 h) for sevoflurane exposure may be too long in actual clinical situation. However, despite the hard exploratory work we’ve done, there is a few results we assumed except for that sevoflurane exposed to fetal brain for at least 3 hours or neonatal brains for more than 6 hours were neurotoxic. These results suggested that the safe concentration and exposure time in most clinical practice is safe.

Prenatal sevoflurane exposure in middle pregnancy could lead to the disorder of the interneurons by activating the GABA${}_A$ receptor and its NRG1–ErbB4 pathway. In this way, prenatal sevoflurane leads to the abnormal dendrite development of the pyramidal neurons in the ECT of offspring rats and, thus, may interfere with the process of information transmission from the ECT to the hippocampus. This study indicated a possible novel neurotoxic pathway in the influence of sevoflurane on ECT of rat offspring. In clinical practice, the concentration of sevoflurane is much lower than the concentration in this study because of the addition of other auxiliary drugs such as opioid sedatives. Secondly, because of the advancement of surgical procedures, most non-obstetric procedures during pregnancy do not require 3 hours. Therefore, the current clinical use of sevoflurane is safe. However, prolonged exposure to high concentrations of sevoflurane still needs to be alert to neurotoxicity.

NMDA, N-methyl-D-aspartate; GABA, gamma-aminobutyric acid; CNS, central nervous system; GABA${}_A$, GABA type A; LII and LIII, layers II and III; NRG1, Neuregulin-1; ErbB4, epidermal growth factor receptor-4; PV, parvalbumin; SD, Sprague–Dawley; SeV, sevoflurane; MAC, minimum alveolar concentration; P30, the day 30 of postpartum; PBS, phosphate buffered saline; GAD67, glutamic acid decarboxylase 67; NR2A, NMDA receptor subunit 2A; NR2B, NMDA receptor subunit 2B; BSA, bovine serum albumin; MGE, medial ganglionic eminence region; NMDAR, NMDA receptor; PDZ, PSD-95/DiscsLarge/ZO1; PSD-95, postsynaptic dens 95; ZO1, onula occludens protein-1; SRC/Fyn, Src family of Fyn; CaMKII, calmodulin-dependent kinase II.

XS, YG and TZ designed research; YG, TZ, YC, ZS, ZS and JL performed experiments; YC analyzed data; YG wrote the paper; TZ and XS critically revised the paper.

This experiment was approved by the Animal Ethics Committee of Guangzhou Medical University, and the researchers strictly followed the relevant provisions of the “Guidelines for the Care and Use of Laboratory Animals” issued by the National Institutes of Health in 1996 (ethic code: 2016-029).

We thank Xiaolong Zeng for assistance in manuscript revision preparation.

This work was granted by the National Natural Science Foundation of China (Granted no. 81870823).

The authors declare no conflict of interest.