Impact of anesthesia exposure in early development on learning and sensory functions

Abstract

Each year millions of children undergo anesthesia, and both human and animal studies have indicated that exposure to anesthesia at an early age can lead to neuronal damage and learning deficiency. However, disorders of sensory functions were not reported in children or animals exposed to anesthesia during infancy, which is surprising, given the significant amount of damage to brain tissue reported in many animal studies.

In this review we discuss the relationship between the systems in the brain that mediate sensory input, spatial learning and classical conditioning, and how these systems could be affected during anesthesia exposure. Based on previous reports, we conclude that anesthesia can induce structural, functional and compensatory changes in both sensory and learning systems. Changes in myelination following anesthesia exposure were observed as well as the neurodegeneration in the gray matter across variety of brain regions. Disproportionate cell death between excitatory and inhibitory cells induced by anesthesia exposure can lead to a long-term shift in the excitatory/inhibitory balance, which affects both learning-specific networks and sensory systems. Anesthesia may directly affect synaptic plasticity which is especially critical to learning acquisition. However, sensory systems appear to have better ability to compensate for damage than learning-specific networks.

1. Introduction

Each year approximately 6 million children in the USA undergo anesthesia (DeFrances, Cullen, & Kozak, 2007) during the course of surgeries, imaging studies and other diagnostic procedures (L. Sun, 2010), and there is an increasing concern about the potential pathogenic effects of anesthesia on the developing brain (Crosby & Davis, 2013; Lee, Zhang, Wei, & Yu, 2015; Olsen & Brambrink, 2013; Rothstein, Simkins, & Nunez, 2008; L. Sun, 2010; Taylor, 2009; Vutskits & Culley, 2019). Such effects do not necessarily occur in all cases, and likely depend upon the length of anesthesia, frequency of exposure and other factors related to anesthesia protocol. However, a growing body of literature from human and animal studies indicates that exposure to anesthesia, especially at an early age, can affect a variety of aspects of neuronal development, leading to learning and behavioral deficits.

Most of these findings were based upon retrospective studies of children and adolescents who had been exposed to anesthesia early in development (Crosby & Davis, 2013; Lee et al., 2015; Olsen & Brambrink, 2013; L. Sun, 2010; Wilder et al., 2009). For example, Wilder (Wilder et al., 2009) analyzed 5357 children, 593 of which received general anesthesia before age 4 years. Learning disabilities were evaluated in children and adolescents up to 19 years old and it was found that the risk of developing learning impairment increased with the number of anesthesia exposures and cumulative duration of anesthesia. Strikingly, one study reported that young children exposed to anesthesia were more than twice as likely to exhibit behavioral disorders or developmental deficits, including mental retardation, autism and language or speech problems, in later years (DiMaggio, Sun, Kakavouli, Byrne, & Li, 2009).

These reported pathological effects in children are supported by a variety of animal studies that have sought to uncover potential mechanisms that could account for anesthesia-related deficits in learning and behavior. Typically, these studies involve exposure of immature rodents to anesthesia, followed by histological and immunohistochemical analysis of the brain or behavioral studies in adolescent or adult animals. For example, it was shown that commonly used anesthetics lead to widespread apoptotic neurodegeneration in the developing brain (Jevtovic-Todorovic et al., 2003). The most vulnerable brain regions were thalamic nuclei and parietal cortex where a sparsely scattered pattern of cell death was detected. Other regions, including subiculum, globus pallidus, retrosplenial cortex, cingulate cortex and hippocampus, were also affected. Table 1 summarizes the effects of anesthesia that have been reported in different brain structures.

Table 1.

Effects of anesthesia exposure on the learning and sensory structures

Structure	Effect	Drug	Protocol	Animal/Age	References
Thalamus	Elevated levels of neuroapoptosis in cells and nuclei	Isoflurane (ISF)	1.5%, 3 hours, 3 times	Mice, P3, P5, P7	(Maloney et al., 2019)
	Altered synaptic transmission	ISF/N2O	6 hours, once	Rat brain slices, P6	(Woodward, Timic Stamenic, & Todorovic, 2019)
	Apoptotic degeneration of both neurons and glia was less severe than in cerebral cortex	Propofol	5 hours, once	Rhesus macaques, P6	(Creeley et al., 2013)
	DNA fragmentation and neuronal death	ISF,Sevoflurane (SVF) or desflurane (DSF)	2% ISF or 3% SVF, DSF 4 or 8%, 6 hours, once	Mice, P6	(Kodama et al., 2011)
Sensory cortex	DNA fragmentation and neuronal death in layer IV of sensory cortex	ISF, SVF, DSF	2% ISF or 3% SVF, DSF 4 or 8%, 6 hours, once	Mice, P6	(Kodama et al., 2011)
	Apoptosis in layers II and V of visual cortex	ISF	1.5%, 5 hours, once	Rhesus macaques, P6	(Brambrink et al., 2010)
	Neuronal and oligodendrocyte apoptosis in V1, somatosensory and auditory cortex	ISF	1 MAC, 3 hours, once	Macaques, P6	(Noguchi et al., 2017)
Parietal cortex	The level of neuronal degeneration (layer II) depended on the number of involved drugs	ISF/N2O and optional midazolam	1.5%, 6 hours, once	Rats, P7	(Jevtovic-Todorovic et al., 2003)
Cingulate cortex	Neuronal and oligodendrocyte apoptosis	ISF	1 MAC, 3 or 5 hours, once	Macaques, P6	(Noguchi et al., 2017)
Retrosplenial cortex	DNA fragmentation in neurons and neuronal death	ISF, SVF, DSF	2% ISF or 3% SVF, DSF 4 or 8%, 6 hours, once	Mice, P6	(Kodama et al., 2011)
Prefrontal cortex	Neuronal and oligodendrocyte apoptosis	ISF	1 MAC, 3 or 5 hours, once	Macaques, P6	(Noguchi et al., 2017)
Entorhinal cortex	Neuronal and oligodendrocyte apoptosis	ISF	1 MAC, 3 or 5 hours, once	Macaques, P6	(Noguchi et al., 2017)
Hippocampus	Neuronal and oligodendrocyte apoptosis, reduction in astrogliosis, and alterations in gene expression	ISF	2%, 6 hours, once	Pigs, P1	(Broad et al., 2016)
	Apoptosis in all regions of hippocampus	ISF	1.5%, 3 hours, 3 times	Mice, P3, P5, P7	(Maloney et al., 2019)
	Decrease in hippocampal volume, increased Apolipoprotein E expression	SVF	2.6%, 2 hours, 3 times	Rats, P7, P14, P21	(Jiang et al., 2018)
	Increased cellular degeneration in dentate gyrus, CA1, and CA2/CA3 pyramidal cell layer.	ISF	1.5%, 6 hours, once	Mice, P7	(Loepke et al., 2009)
	Decrease in the hippocampal stem cell pool	ISF	1.7%, 35 min., 4 times	Rats, P14	(Zhu et al., 2010)
	Reduction in neuronal density in CA1, morphological changes in the pyramidal cells	Propofol	75 mg/kg, once every day for 7 days	Rats, P7–P13	(Yu, Jiang, Gao, Liu, & Chen, 2013)
	Neuronal apoptosis in CA1, CA3, and dentate gyrus.	ISF	1.1%, 4 hours, once	Rats, P7	(Li et al., 2013)
Amygdala	Reduction in glutamatergic neurons	SVF	3%, 6 hours, once	Mice, P6	(Satomoto et al., 2018)
	Widespread apoptosis of both neurons and oligodendrocytes	Propofol	5 hours, once	Rhesus macaques, P6	(Creeley et al., 2013)
	DNA fragmentation and neuronal death	ISF, SVF, DSF	2% ISF or 3% SVF, DSF 4 or 8%, 6 hours, once	Mice, P6	(Kodama et al., 2011)
Cerebellum	Neural apoptosis in the granule layer	ISF	1.5%, 6 hours, once	Mice, P7	(Deng et al., 2014)
	Impairments in Bergmann glia development, reduction in Purkinje cell density, and lower dendritic length.	propofol	60 mg/kg, once	Mice, P7	(R. Xiao et al., 2017)
	Neurones were relatively spared	Propofol	5 hours, once	Rhesus macaques, P6	(Creeley et al., 2013)

Moreover, the extent of brain damage depends on the duration and frequency of anesthesia exposure. For example, it has been shown that multiple exposures to anesthesia (i.e., three exposures for 2 hours each) or prolonged exposure (i.e., 6 hours) produced a different and typically greater neurotoxic effect than a single, relatively short exposure (i.e., once for 2 hours) (Amrock, Starner, Murphy, & Baxter, 2015). On the other hand, three recent human studies, general anesthesia spinal (GAS), Pediatric Anesthesia Neurodevelopment Assessment (PANDA) and Mayo Anesthesia Safety in Kids (MASK), which examined shorter exposure to general anesthesia does not find detectable neurodevelopmental impairment (Davidson et al., 2016; McCann et al., 2019; L. S. Sun et al., 2016; Warner et al., 2018). Thus, in general it appears that in order to produce visible neurodevelopmental effects longer and/or multiple exposures to anesthesia may be necessary.

The mechanisms underlying anesthesia-related pathology are not well understood, although several possible explanations have been proposed. Yang et al. have suggested that anesthetics may induce cell damage through abnormal calcium release from endoplasmic reticulum (H. Yang et al., 2008). Istaphanous et al. (Istaphanous et al., 2013) proposed that isoflurane can downregulate central GABA-synthesizing enzymes GAD65 and GAD67, constituting a possible toxic effect of anesthetics on the inhibitory system. Wu et al. hypothesized that exposure of the developing brain to anesthesia triggers epigenetic modification, involving the enhanced interaction among transcription factors, which inhibits brain-derived neurotrophic factor expression (Wu, Bie, & Naguib, 2016). Furthermore, anesthesia-related hyperoxia related to the use of supplemental oxygen may also play a role. Anesthesia is often delivered in combination with higher-than-air concentrations of oxygen (typically 30%−100%) in order to avoid hypoxia (Bang, 2015; Dikmen & Onur, 2017; Sola, 2008), and may lead to additional damage to the brain due to oxidative stress (Felderhoff-Mueser et al., 2004). Such damage can occur when increased levels of reactive oxygen species (ROS) overwhelm the brain’s antioxidant capacity (Macri et al., 2010). We have shown previously that isoflurane delivered at 1 MAC can greatly increase brain tissue partial oxygen pressure (PO₂) in neonates, especially when combined with supplemental oxygen (Aksenov, Dmitriev, Miller, Wyrwicz, & Linsenmeier, 2018). Isoflurane delivered in 80% oxygen increased brain tissue PO₂ by up to 300%. Overall, animal studies point to neuropathological changes in a variety of brain regions that could lead to the learning and memory deficits observed in children (Loepke & Soriano, 2008).

Associative learning requires the presence of two or more stimuli or events to establish the association between them, and thus is vulnerable to damage to sensory pathways as well as to the non-sensory networks that mediate learning-specific plasticity. Notably, however, disorders of sensory functions has not been reported in children exposed to anesthesia during infancy. It was shown that patients exposed to anesthesia during first 3 years of life did not develop visual acuity deficiency by age 20 (Yazar et al., 2016). Moreover, full neurological examination revealed no sensory deficiency at the age of 8 years in children exposed to anesthesia during infancy (Bellinger et al., 2003). Considering the significant extent of brain damage found in animal studies, some degree of sensory deficiency would be expected, and thus the absence of reports of sensory deficits in children with learning and memory impairments is surprising.

In this review we will analyze the key mechanisms that could account for the differential impact of early anesthesia exposure upon learning vs. sensory systems, with a particular emphasis on hippocampus-dependent learning. For this purpose we will focus mostly upon animal studies, as they can better employ appropriate control groups, they are conducted in a standardized manner, and they are often accompanied by histological analysis of the brain tissue. First, we will compare the relationship between sensory and learning systems involved in the most common memory tests. Then we will discuss the impact of anesthesia on critical factors that can account for major structural and functional differences between sensory and learning systems, namely, the level of myelination, the excitatory/inhibitory balance, synaptic plasticity and developmental trajectory. Finally we will address the implications of these conclusions for future work involving the effects of anesthesia.

2. The involvement of sensory and learning-related structures in behavioral tests

Assessing the effects of anesthesia exposure on learning and memory depends heavily upon the specific tests that are used to measure the deficits, as each one may engage widely different neuronal substrates. Figure 1 shows a schematic that illustrates key regions of the sensory and learning circuitry. The most common approach to evaluate anesthesia-induced deficiency in animal subjects is to conduct spatial learning and memory tests in rodents. These tests involve acquiring and retaining information about the subject’s orientation and spatial position relative to the environment. Spatial navigation primarily depends on the function of the hippocampus and entorhinal cortex (Vorhees & Williams, 2014). Place cells in the hippocampus form a representation of the environment and their cross-talk with grid cells in the medial entorhinal cortex (Buzsaki & Moser, 2013; Igarashi, Lu, Colgin, Moser, & Moser, 2014; Moser, Moser, & McNaughton, 2017) enables remapping of the environment according to changes in the body’s position.

Figure 1.

A simplified schematic of sensory (red outline) and learning-related circuitry (blue outline). For the purpose of comparison the key structures for three learning paradigms (spatial learning, fear conditioning and eyeblink conditioning) were combined with sensory pathways. The nature of the anesthesia-induced damage reported in these structures is described in details in Table 1. Amy: Amygdala, Cb: Cerebellum, Cg: Cingulate cortex, EC: Entorhinal cortex, Hpc: Hippocampus, PfC: Prefrontal cortex, RsC: Restrosplenial cortex, Th: Thalamus.

A recent paper (Y. Wang, Han, Han, Su, & Li, 2017) reported that mice exposed to daily intraperitoneal injections of propofol from day 6 postnatal (P6) to P11 required more time to locate the hidden platform in the Morris water maze compared to controls at age P28 to P35. Another study (Liu et al., 2016) reported that rats exposed to daily 3% sevoflurane from P4 to P6 for 2 hours exhibited a longer escape latency in water maze testing and reduced platform crossing time as compared to the control group at age P60-P66. Shen et al. (Shen et al., 2013) reported that mice exposed to 3% sevoflurane in 60% oxygen at P6–P8 for 2 hours had a longer escape latency and decreased the platform crossing times in Morris water maze testing compared to the control group at age P30-P37. It is important to note that no sensory or motor deficiencies of any sort were reported in these studies. Rodents which were exposed to anesthesia as neonates had the same swimming speed as animals in control groups; and a battery of sensorimotor tests, including walking initiation, ledge walking, inclined plane performance, and elevated platform performance did not reveal any deficits (Jevtovic-Todorovic et al., 2003).

However, spatial learning tests do not all rely equally on the various sensory functions. Adequate swimming performance, for example, requires fully functional proprioception but the primary sensory system (i.e., visual) is less critical as long as the animal can see cues and make decisions based on this external information. Indeed, it was reported that the acuity of a still-developing visual system is sufficient to perform spatial navigation (Carman, Booze, & Mactutus, 2002; Carman & Mactutus, 2001) even when the task is performed with one eye (Rudy & Stadler-Morris, 1987). Thus, the primary sensory system for spatial memory can be impacted to some degree but spatial memory tests will not necessarily reveal these effects.

Like spatial learning, classical conditioning (CC) depends on primary sensory systems, but it does not require animals to make decisions about how to respond. CC is a typical example of simple associative learning which pairs a neutral conditioned stimulus (CS) with a physiologically salient unconditioned stimulus (US). Unlike spatial learning, CC is sensitive even to minor functional deficits in the primary sensory systems involved in the perception of the CS and US. It has been shown, for example, that a difference in the auditory tone can severely affect the learning rate of conditioned responses (Grice & Hunter, 1964) and it is expected that damage to sensory pathways can affect the learning rate.

As is the case for spatial learning, a wide variety of CC paradigms can be employed to engage different aspects of brain function. The two most common forms of CC are fear and eyeblink conditioning, which can be distinguished by key structural as well as methodological differences. Both of these tests have been used to evaluate anesthesia-induced learning deficiency (Aksenov, Miller, Li, & Wyrwicz, 2016; Satomoto et al., 2009) but they differ in terms of the structures involved. For example, whereas the amygdala is essential for acquiring fear CC (Curzon, Rustay, & Browman, 2009) it plays a non-essential modulatory role in eyeblink CC (Y. Yang, Lei, Feng, & Sui, 2015). Fully functional nociception is critical for fear conditioning (Finn et al., 2004; Fujii, Koshidaka, Adachi, & Takao, 2019) and auditory function must be sufficient to hear the auditory CS clearly (Weinberger, 2011). In eyeblink conditioning the function of the trigeminal nerve should be preserved to sense the aversive US to the eye, and diminished function of the trigeminal nerve can result in slow acquisition of CRs (Christian & Thompson, 2003). The CS, typically an auditory, visual or somatosensory stimulus, must have sufficient intensity and frequency to ensure optimal learning. Functional deficiency in any of these systems can affect the learning rate as quantified by the percentage of CRs.

Furthermore, CC typically is performed either with a delay or trace paradigm. In the delay conditioning the CS and US overlap, whereas in trace conditioning the CS and US are separated by a stimulus-free interval (i.e., the trace interval). Each of these paradigms is characterized by the involvement of different circuitry to support the formation of the learned association. Whereas involvement of the cerebellum is common to both delay and trace eyeblink conditioning paradigms, the hippocampus, for example, is critical for the trace fear (Curzon et al., 2009) and eyeblink CC (Moyer, Deyo, & Disterhoft, 1990) but not for delay CC.

Due to the fact that eyeblink conditioning typically requires immobilization of the head or body, it is difficult to perform this cognitive test in awake rodents (although some investigators have done so, for example, (Galvez, Weiss, Cua, & Disterhoft, 2009)), and thus the majority of CC studies exploring the effects of anesthesia have used fear CC paradigms. In an anesthesia study at an early developmental stage involving delay fear conditioning (Satomoto et al., 2009), P6 mice were exposed to 3% sevoflurane for 6 hours and then CC was performed during maturity, either at 8 weeks or 14–17 weeks. Subjects were presented with an auditory CS (80 db white noise, 20 s) followed by a foot shock US (1 mA, 1 s), and it was found that the freezing response was reduced significantly in the anesthesia-exposed group compared to the control group after conditioning. In this case learning can depend both on sensory systems that mediate nociception as well as the processing of the received information to form the learned association. If the sensitivity to a painful stimulus is even slightly decreased it can affect learning.

Changes in the function of primary sensory systems can be evaluated with electrophysiological recording or neuroimaging methods in order to assess their integrity. In our previous work we found that rabbits exposed to isoflurane anesthesia during infancy exhibited a significantly lower CR percentage following training with a trace eyeblink CC paradigm (50 Hz whisker vibration CS with duration 250 ms, and 150 ms US as airpuff to an eye with 500 ms trace interval) as compared to controls (Aksenov et al., 2016). Blood oxygenation level dependent (BOLD) functional magnetic resonance imaging (fMRI) was performed in the somatosensory system during whisker stimulation prior to trace CC, and no difference was found either in the area of temporal behavior of the BOLD for rabbits exposed to anesthesia during infancy vs. the control group. These findings support the results obtained from human patients (Bellinger et al., 2003; Yazar et al., 2016) and a battery of sensorimotor tests (Jevtovic-Todorovic et al., 2003), which directly indicate that the function of the primary sensory system was preserved in humans or animals that underwent anesthesia exposure in infancy. Thus, learning-related differences could not be attributed to deficiency in sensory perception.

It is therefore important to consider the differences in sensory vs. learning systems with regard to their sensitivity to the various effects of anesthesia. Such differences can be structural as well as functional. Neurons and glia in learning vs. sensory systems may exhibit different inherent levels of susceptibility to anesthesia-induced damage. It is possible that changes in the proportion between excitatory cells and interneurons as a result of cell loss, and the resulting shift in the excitatory/inhibitory balance, affects learning-specific networks more than sensory systems. Furthermore, anesthesia may directly affect synaptic plasticity which is especially critical to learning acquisition. In the sections that follow we will consider the impact of each of these factors.

3. Effects of anesthesia on sensory and learning-related structures

The damage caused by neonatal anesthesia can either result in structural changes (e.g., apoptosis of different degree) or in functional changes (e.g., altered synaptic plasticity), and can depend directly on the timing of exposure in relation to the critical developmental period of each structure. Moreover, different cell types can be affected differently by anesthesia.

There are two major types of cells in the brain: glial cells (including oligodendrocytes responsible for myelination) and neurons. Currently no consensus exists regarding the extent of demyelination induced by early anesthesia exposure. Brambrink et al. (Brambrink et al., 2012) reported that both white and grey matter were affected in the neonatal primate brain following isoflurane exposure for 5 hours, and that the defects in myelination occurred due to widespread apoptosis of oligodendrocytes. It was also reported a three-hour exposure to isoflurane in P6 macaques resulted in the apoptosis of neurons and oligodendrocytes (Noguchi et al., 2017). In contrast, other studies have described more localized effects which were limited to cell loss in grey matter in infant mice and rats exposed to isoflurane for 6 hours (Istaphanous et al., 2011; Jevtovic-Todorovic et al., 2003). Such differences in the injury to white and gray matter could depend heavily upon the specific developmental period in which anesthesia was presented in monkeys and rodents. Developmental periods can be characterized by different levels of oligodendrocyte vulnerability to the direct apoptotic effects of anesthesia exposure.

A number of mechanisms have been proposed to account for this neurotoxicity, such as abnormal calcium release (H. Yang et al., 2008). Another potential factor can be the use of supplemental oxygen. Oligodendrocytes, which are responsible for myelination, are highly sensitive to oxidative stress due to a diminished capacity for antioxidant defense (Smith, Kapoor, & Felts, 1999). Brambrink et al. (Brambrink et al., 2012) reported that venous oxygen tension increased by a factor of 2 from the pre-anesthesia level after 4.5 hours of anesthesia, which is an indirect indication of increased brain oxygen level.

Proper myelination is critical for normal function of sensory pathways (Kaur & Bennett, 2007; You et al., 2019) and learning-related circuits (Nickel & Gu, 2018). However, the absence of reported sensory deficiencies following anesthesia exposure suggests that demyelination occurs to a limited extent under these circumstances, and thus damage of the grey matter may be a key factor in learning deficiency.

3.1. Sensory system

The sensory systems comprise several key nodes, including the sensory nerves, the nuclei of the thalamus, and the cerebral sensory cortex. Anesthesia-induced damage on sensory nerves has not been well studied, but a substantial difference in the effects of anesthesia would be expected depending on the time of exposure and the species of the subject. For example, the sensory systems exhibit differential development in humans vs. rodents. Rodents are born with closed eyes and ears, and sensory stimulation is the major factor in the postnatal development of sensory systems. In humans, for example, most structural development of the visual system occurs prenatally and it is functionally poor at birth (for review see (Johnson, 2013). Acuity develops during infancy due to migration of receptor cells in the retina, myelination of the optic nerve, etc. All fibers of the optic nerve, for example, are myelinated by 7 months of age (Magoon & Robb, 1981). Similar patterns occur in the development of the auditory system, where most changes happen by the time of term birth. Auditory neurons in the brainstem increase their size, reaching 50–60% of adult size, and these changes are combined with rapid growth of dendrites (Moore, Guan, & Shi, 1998). The myelin density in the cochlear nerve and brainstem pathways develops to an adult-like stage by 6–12 months of age (Moore & Linthicum, 2007; Moore, Perazzo, & Braun, 1995). Thus, the structural development of the sensory systems occurs mostly prenatally whereas functional development occurs postnatally.

3.1.1. Thalamus

The thalamus is an important relay structure, transmitting sensory information to the cerebral cortex and participating in alertness/consciousness (for review, see (Moustafa, McMullan, Rostron, Hewedi, & Haladjian, 2017)). There are multiple critical points in thalamic development depending on the type of cells and connections. For example, in mice the postnatal development of synapses between retinal ganglion cell axons and thalamocortical relay neurons involves 3–4 phases (Kano & Watanabe, 2019): segregating into eye-specific projection zones before eye opening at P12, decreasing the number of retinal ganglion cell axons (up to P20), maintaining of retinogeniculate synapses in a visual experience–dependent manner (P20-P30) and fine-scale refinement (up to P60). Retinal spontaneous activity (i.e., not related to visual experience) plays a critical role during the first two phases of postnatal development (Hooks & Chen, 2006). For comparison, neurons of the rat auditory thalamus (i.e., medial geniculate body) obtain mature electrophysiological properties after P14 (Tennigkeit, Schwarz, & Puil, 1998).

Widespread damage in thalamic neurons due to neonatal anesthesia was reported by several studies (Creeley et al., 2013; Kodama et al., 2011; Maloney et al., 2019) where anesthesia was presented before eye opening. Thus, only the first phases of postnatal visual thalamic development could be directly affected. Isoflurane exposure in mice at P3 significantly increased acute neuroapoptotic response in the thalamus, but exposure at P5 increased acute neuroapoptotic response in all brain regions (Maloney et al., 2019). Creeley et al. (Creeley et al., 2013) showed that exposure to propofol at P6 in rhesus macaques resulted in less severe damage to thalamus compared with cerebral cortex or exposure during fetal period of development. These data point to a possible critical period in the development of thalamus which starts earlier than in other brain structures.

Although the thalamus is an essential part of the sensory system, damage to this structure, which results in decreased output, can affect learning as well since the information about stimuli is relayed through the thalamus. Indeed, it was shown by the injection of GABA-agonist muscimol into the somatosensory relay thalamic nucleus that S1 cortical responses to sensory stimulation were greatly decreased (Castejon, Barros-Zulaica, & Nunez, 2016). In another study, the inactivation of the medial auditory thalamus with muscimol also greatly decreased auditory trace eyeblink conditioned responses (Hoffmann, Zara, DeLord, & Mauk, 2018). However, the preservation of sensory functions in patients who were exposed to anesthesia in early life suggests that in spite of possible neuronal apoptosis the function of the thalamus is mainly preserved.

3.1.2. Sensory cortex

The development of the sensory cortex is different in human and rodents. Electrophysiological data based on early cortical activity patterns indicates that the most intense period of neuronal growth and synaptogenesis, sometimes called “brain spurt,” occurs during the fetal period in humans and during the postnatal period in rodents (Khazipov & Milh, 2018). Specifically, in rats and mice the neonatal period between P0 and P10 approximately corresponds to the second half of gestation in humans in terms of sensory cortex development. Moreover, the formation of topographic thalamocortical maps, for example, for the somatosensory cortex, occurs during the first postnatal week in rodents and during the fetal period in humans (Khazipov & Milh, 2018; O’Leary, Ruff, & Dyck, 1994). On the other hand, the time courses of spine and synapse development in visual cortex are similar between monkeys and humans (Oga, Aoi, Sasaki, Fujita, & Ichinohe, 2013). Moreover, unlike rodents, monkeys can have earlier prenatal development of the brain than humans, in terms of brain growth spurt (Brambrink et al., 2010).

Studies in macaques showed that 3 hours after exposure to isoflurane at P6, massive apoptosis was observed in various divisions of the cerebral cortex, including layers II and V of the visual cortex (Brambrink et al., 2010). P6 in rhesus macaques approximately corresponds to six month old human in terms of brain growth spurt (Brambrink et al., 2010). Another study showed that following a 3 hour exposure to isoflurane apoptosis was observed in the cerebral cortex, including the somatosensory, visual and auditory cortices (Noguchi et al., 2017). Notably, glial (i.e., oligodendrocyte) apoptosis was evenly distributed throughout the white matter whereas neuroapoptosis occurred primarily in the cortex. Kodama at al. (Kodama et al., 2011) described neuronal death in layer IV of the sensory cortex after exposure to 4% or 8% desflurane for 6 hours in P6 mice. 2% isoflurane and 3% sevoflurane also caused neuroapoptosis, but to a lesser degree than desflurane.

Specific types of cortical neurons may be more susceptible to anesthetics which act upon glutamate-ergic or GABA-ergic receptors. It has been shown that certain excitatory pyramidal neurons in the piriform cortex are highly sensitive to N-methyl-d-aspartate (NMDA) antagonists and NMDA blockage can result in the selective death of these neurons (L. Zhou, Welsh, Chen, & Koliatsos, 2007). The selective toxic effect of propofol on GABAergic neurons in developing telencephalon was demonstrated previously (Honegger & Matthieu, 1996). The selective death of specific cortical neurons can affect the excitatory/inhibitory balance which plays very important role in the function of cerebral cortex. For example, decreased inhibition in the cerebral cortex typically causes epileptic activity or even seizures (Rossi, Wykes, Kullmann, & Carandini, 2017) whereas increased inhibition can cause altered perception of stimulation, as has been shown in BOLD fMRI and electrophysiology studies (Aksenov, Li, Miller, & Wyrwicz, 2019).

3.2. Learning-related system

3.2.1. Hippocampus

In humans almost all important milestones in hippocampal development are reached during the prenatal period, although the volume of hippocampus increases rapidly up to 2 years of age and continues to increase more slowly thereafter (Utsunomiya, Takano, Okazaki, & Mitsudome, 1999). In rats or mice the peak of dentate gyrus growth, for example, occurs at P7 and P1, respectively; and the tertiary dentate matrix evolves into the sub granular zone through P30 (Li, Mu, & Gage, 2009) and P14, respectively. The most significant difference in hippocampal development between humans and rodents is that approximately 80% of human dentate gyrus cells are formed at the prenatal stage, whereas approximately 85% of rodent dentate gyrus cells develop postnatally (Fan, Sun, & Liu, 2018).

Anesthesia-induced damage of the hippocampus has been well studied because this structure is important for learning and memory. For example, apoptosis in all regions of the hippocampus was shown in mice after 3 exposures to isoflurane at P3, P5 and P7 for 3 hours each (Maloney et al., 2019). Neuronal and oligodendrocyte apoptosis, reduction in astrogliosis, and alterations in gene expression were reported in pigs after isoflurane exposure for 6 hours at P1 (Broad et al., 2016). In some cases hippocampal damage due to neonatal exposure to sevoflurane was found so be extensive that it resulted in a decrease of hippocampal volume (Jiang et al., 2018). The pyramidal cell layer in CA3 was selectively analyzed after mice were exposed to isoflurane at P7 and apoptosis was found (Loepke et al., 2009). There is also evidence that isoflurane caused excitotoxic injury in immature (5 day culture) hippocampal rat pyramidal neurons (Zhao et al., 2011). It was shown that after isoflurane exposure in 7P rats, 54% and 14% of apoptotic cells in the CA1 subarea of hippocampus were GABAergic and glutamatergic neurons, respectively (Z. W. Zhou et al., 2011). This result indicates that, in addition to loss of neurons, changes in the proportion of preserved excitatory vs. inhibitory cells potentially can lead to localized seizure-like activity (Katsumori, Minabe, Osawa, & Ashby, 1998) and lasting hippocampal damage due to chronic seizure activity (Cheng & Zhang, 2014).

The difference in hippocampal development can explain why animal models typically show a greater effect of neonatal anesthesia exposure on learning than human studies. While undergoing significant post-natal development, the rodent hippocampus can potentially be more vulnerable to the effects of anesthesia as compared to the human hippocampus, which is much more extensively developed by the time of birth. Thus, in order to study the effects of neonatal anesthesia exposure on hippocampus-dependent learning in a manner consistent with human development, anesthesia in animal models should be delivered closer to the end of hippocampus development. Earlier exposures may better serve as a model of a human prenatal exposure.

3.2.2. Amygdala

The amygdala in humans rapidly develops early in postnatal period since neuroanatomical architecture of the amygdala already exists by birth (Tottenham, 2012; Ulfig, Setzer, & Bohl, 2003), but the structural growth can last up to 18 years old in males (Giedd et al., 1996). In rodents the volume of the amygdala increases between P20 and P35 with, a subsequent decrease between P35 and P90 and changes in density of the fibers between the amygdala and medial prefrontal cortex continuing into early adulthood (Ganella & Kim, 2014).

Neonatal anesthesia not only decreased glutamatergic neurons in the amygdala after sevoflurane exposure for 6 hours in P6 mice (Satomoto, Sun, Adachi, & Makita, 2018) but caused widespread apoptosis in P6 monkeys after 5 hours of propofol infusion (Creeley et al., 2013). Isoflurane and desflurane also caused neuronal damage in the amygdala (Kodama et al., 2011). Interestingly, GABA-ergic neurons in the amygdala were not affected by neonatal sevoflurane anesthesia (Satomoto et al., 2018). Anesthesia-induced damage of the amygdala can result in the suppression of neuronal activity and would affect the acquisition and expression of fear conditioning (Muller, Corodimas, Fridel, & LeDoux, 1997).

3.2.3. Prefrontal cortex

The development of the human PFC shares many similar features with the sensory cortex, but the development of dendrites is occurs within a different tie frame. PFC synaptic density reaches its peak around at the age of 5 years whereas sensory cortices typically have peak synapse density at approximately 1 year (Kolb et al., 2012; Petanjek et al., 2011). The PFC is also characterized by the greatest overproduction of synapses, compared to other sensory cortical regions, the slowest elimination speed in late childhood, which can reflect its involvement in processing the environment during puberty (Kolb et al., 2012). The same principle can be applied to all laboratory mammals. For example, in rats even by P18 layers of the PFC are still immature (Van Eden & Uylings, 1985) but rapid changes occur in the mPFC during adolescence, including reorganization of synapses (Ganella & Kim, 2014).

Both neuronal and oligodendrocyte apoptosis was reported in PFC following a 3 hour exposure to isoflurane in P6 macaques (Noguchi et al., 2017). Since the PFC is actively developing postnatally in both humans and animals, this structure is particularly vulnerable to anesthesia toxicity, and animal models of anesthesia-induced PFC damage have clear translational value.

3.2.4. Cerebellum

Postnatal development of the cerebellum is different in humans and rodents. In humans the peak of cerebellar neuronal proliferation is around 30 post conceptual weeks and the cerebellum is almost completely formed by the time of birth. Final maturity of cerebellar circuits can be reached by the end of the second postnatal year (Haldipur & Millen, 2019). In mice, for example, the peak of external and internal granule layer expansion, foliation and Purkinje cell maturation occurs during the first two postnatal weeks (Haldipur & Millen, 2019; Marzoll, Saygi, & Dinse, 2018). Moreover, in neonatal and early postnatal mice each Purkinje cell is innervated by several climbing fibers and only later a pruning process eliminates excessive fibers to achieve a the adult mono-innervation pattern (Beckinghausen & Sillitoe, 2019).

Anesthesia-induced damage of the cerebellum varies according to the type of anesthetic used. It was shown that isoflurane caused neuronal apoptosis in the excitatory granule layer of the cerebellar cortex (Deng et al., 2014) whereas propofol induced apoptosis of GABA-ergic Purkinje cells (R. Xiao et al., 2017) in mice. Interestingly, ketamine infusion had no effect on the cerebellum (Slikker et al., 2007). Due to the differences in postnatal development between rodents and humans, it can be assumed that the anesthesia-induced cerebellar damage is greater in rodent animal models than in children. A study, which showed that cerebellar neurons are relatively spared in rhesus macaques after propofol anesthesia at P6 (Creeley et al., 2013), also suggests that rodent models cannot be used as an indicator of human postnatal anesthesia-induced cerebellar toxicity.

Effects of anesthesia on synaptic plasticity

Synaptic plasticity in the brain depends upon changes in dendritic trees as well as in the synapses themselves. For proper dendritic tree development (Jan & Jan, 2010), dendrites must cover the area that encompasses their sensory inputs, the branching pattern and density of dendrites must be suitable for sampling and processing these inputs, and dendrites must maintain the flexibility for experience-related adjustment. It was shown that under normal conditions the number of neuronal connections greatly increases in the early postnatal period throughout all brain regions and then then slowly declines to adult levels during childhood and adolescence (Stiles & Jernigan, 2010). Studies of the effects of anesthesia on the development of dendritic trees are limited, but it was shown that anesthesia-induced impairment of synaptogenesis is strongly age-dependent (Jevtovic-Todorovic, 2012). At a younger age synaptic densities decreased, but in later development they were characterized by excessively up-regulated synapse formation. Furthermore, anesthesia exposure alters the formation of dendritic spines (Vutskits L, De Roo M, Kalauser P, Adrian B, & D., 2008) which are the primary location of excitatory synapses. It is possible that anesthesia can affect developing spines, whereas existing spines are tolerant to the damaging effect of anesthetics. However, since dendrites are important for both sensory and learning structures, both system should exhibit deficits in the case of massive dendritic pathology.

Another major component of synaptic plasticity is long-term potentiation (LTP) of synaptic response following a brief, high-frequency stimulation or similar method of LTP induction. LTP can manifest in multiple forms (i.e., presynaptic or postsynaptic) and can vary not only across brain regions but even within the same synapse (Baltaci, Mogulkoc, & Baltaci, 2019). The majority of studies about this phenomenon have focused on the mechanisms and role of NMDA receptor-dependent LTP because it is believed that this type of LFP is important for associative learning. It has been shown in a number of studies that different hippocampal synapses are selectively modified in strength during the acquisition of classical conditioned responses involving NMDA and other mediators (Gruart, Leal-Campanario, Lopez-Ramos, & Delgado-Garcia, 2015).

LTP related to sensory stimulation has also been described, though not as extensively as LTP related to associative learning. For example, sensory whisker stimulation was able to induce synaptic LTP in pyramidal cells of layers 2–3 in the somatosensory cortex and LTP induction depended on the occurrence of mediated long-lasting depolarizations associated with NMDA receptors (Gambino et al., 2014). However, the functional role of LTP in sensory systems is not entirely clear. It is possible that this phenomenon is necessary to prevent cortical neurons from weakening synaptic input due to spike-timing-dependent long-term depression (Gambino et al., 2014), because the spontaneous and evoked activity of pyramidal cells in somatosensory cortex is low (Aksenov, Li, Miller, Iordanescu, & Wyrwicz, 2015). It has also been proposed that sensory-related LTP can improve tactile acuity or sensitivity (Ragert, Kalisch, Bliem, Franzkowiak, & Dinse, 2008) or improve discrimination performance in case of visual stimulation (Marzoll et al., 2018). In this case the disturbances in sensory LTP may lead to a deficiency in the improvement and adaptation of perception to the changed environment, which can be undetectable via standard neurological evaluation at the clinical level.

A study by Wang et al. showed that the postnatal application of propofol impairs hippocampal LTP in adulthood (Y. L. Wang, Chen, & Wang, 2015). Specifically, the response to stimulation of the field excitatory postsynaptic potentials (EPSPs) decreased after propofol application as well as LTP. It was suggested that propofol may down-regulate PI3K/AKT pathway, which regulates spike genesis and affects hippocampal synaptic development, LTP and memory. Another study found that propofol injected during development decreased the expression of Ca2+/calmodulin-dependent protein kinase II α (CaMKIIα) as well as its activated state (pCaMKIIα), reduced the pCaMKIIα/CaMKIIα ratio, as measured by immunochemistry and Western blotting, and decreased LTP in the hippocampus (Gao, Peng, Xiang, Huang, & Chen, 2014). CaMKII is the main component of postsynaptic density and together with pCaMKIIα plays an important role in LTP, synaptic plasticity and the regulation of learning and memory. It also was shown that neonatal exposure to sevoflurane inhibits LTP in the hippocampus during P35–42 in rats by decreasing the EPSP slope (H. Xiao, Liu, Chen, & Zhang, 2016). Thus, synaptic plasticity can play an important role in the development of anesthesia-induced pathology in learning-related networks and in sensory circuits.

Compensatory abilities of sensory and learning systems

In this context we consider compensation to be the ability of structures or pathways to resist damage or adapt to injury in order to prevent or minimize functional loss. In the previous sections we showed that sensory circuitry was affected by anesthesia along with learning circuitry in light of histological findings that have shown widespread apoptotic neurodegeneration in the developing brain after anesthesia (Table 1), as well as changes in excitatory-inhibitory balance and synaptic plasticity. The preservation of sensory system functionality under these circumstances is surprising. However, it is possible that sensory systems can compensate for moderate damage and maintain normal function even if some portion of the neurons are lost or damaged.

Indeed, it was shown that normal visual acuity requires no more than 44% of the normal quantity of fovelar, neuro-retinal channels (Frisen & Frisen, 1979). Moreover, the ability to compensate for damage of the visual system is even greater in children than in adults, including the ability to differentiate functional tissue within a larger dysplastic cortex during its formation, or to develop new thalamo-cortical connections (Guzzetta et al., 2010). Another study in the auditory system indicated that tone detection behavior was preserved even if 95% of cochlear nerve afferent synapses are lesioned (Chambers et al., 2016). A study of the somatosensory system showed that lesions of the somatosensory dorsal column pathway (up to 98% of the input) in monkeys resulted in near-complete recovery in terms of somatotopy and responsiveness to stimulation (Qi et al., 2016). The olfactory system is also characterized by extensive regenerative abilities after injury (Chehrehasa, Jacques, St John, & Ekberg, 2018; Graziadei & Graziadei, 1979). It is reasonable to assume that taste and nociceptive systems have similar abilities to compensate for a decreased number of functioning neurons up to a critical level.

On the other hand, even minor damage to the key learning structures appears to induce learning deficiency. For example, a study compared patients with mild cognitive impairment (MCI) to healthy subjects and found that in the MCI group the neuronal loss was 12–13% more in the entorhinal cortex compared to the control group (Arendt, Bruckner, Morawski, Jager, & Gertz, 2015). Another study reported that an unbiased stereologic sampling electron microscopy study did not show a significant difference in the total number of synapses within the dentate gyrus between individuals with MCI compared to healthy subjects (Scheff, Price, Schmitt, & Mufson, 2006), although the overall group mean was 13% lower in the MCI group. A similar study compared the number of synapses in CA1 and found no significant difference between MCI and control groups, although the overall group mean was 18% lower in the MCI group (Scheff, Price, Schmitt, DeKosky, & Mufson, 2007). Overall, these studies indicate that even small loss in the number of synapses in learning-related circuitry can lead to clinically visible cognitive impairment.

4. Conclusions

In summary, understanding the mechanisms of anesthesia-related damage in the systems that mediate sensory input vs. learning-related networks is an important aspect to building a more complete understanding of the effects of anesthesia in neonates. We conclude that anesthesia can induce structural, functional and compensatory changes in both systems, which can be influenced by a wide variety of factors. Changes in myelination induced by anesthesia exposure appear less significant compared to the neurodegeneration observed in the gray matter in a variety of brain regions. Disproportionate cell death between excitatory and inhibitory cells induced by anesthesia exposure can lead to a long-term shift in the excitatory/inhibitory balance, which affects both learning-specific networks and sensory systems. Furthermore, anesthesia may directly affect not only synaptic plasticity, which is especially critical to learning acquisition, but also sensory adaptation to stimulation. Sensory systems appear to have better ability to compensate for damage than learning-specific networks. However, as it has been shown that severity of the brain damage depends on the anesthesia protocol, it is possible that sensory dysfunction could be produced by a combination of damaging factors such a longer anesthesia duration, a greater number of anesthesia exposures, the use of multiple drugs, the addition of supplemental oxygen, etc.

Notably, the postnatal development of the learning and sensory systems are different, and the timeline of development is not necessarily equivalent in humans and laboratory animals. Future studies of early anesthesia exposure in humans should account for these developmental differences by examining only subpopulations that were exposed to anesthesia before 6 months of age in order to draw conclusions about the impact on sensory systems. Likewise, future studies involving animal subjects should consider the structure-dependent developmental differences vs. humans in order to achieve a more consistent comparison.