Neonatal anesthesia and dysregulation of the epigenome

Abstract

Each year, millions of infants and children are anesthetized for medical and surgical procedures. Yet, a substantial body of preclinical evidence suggests that anesthetics are neurotoxins that cause rapid and widespread apoptotic cell death in the brains of infant rodents and nonhuman primates. These animals have persistent impairments in cognition and behavior many weeks or months after anesthesia exposure, leading us to hypothesize that anesthetics do more than simply kill brain cells. Indeed, anesthetics cause chronic neuropathology in neurons that survive the insult, which then interferes with major aspects of brain development, synaptic plasticity, and neuronal function. Understanding the phenomenon of anesthesia-induced developmental neurotoxicity is of critical public health importance because clinical studies now report that anesthesia in human infancy is associated with cognitive and behavioral deficits. In our search for mechanistic explanations for why a young and pliable brain cannot fully recover from a relatively brief period of anesthesia, we have accumulated evidence that neonatal anesthesia can dysregulate epigenetic tags that influence gene transcription such as histone acetylation and DNA methylation. In this review, we briefly summarize the phenomenon of anesthesia-induced developmental neurotoxicity. We then discuss chronic neuropathology caused by neonatal anesthesia, including disturbances in cognition, socio-affective behavior, neuronal morphology, and synaptic plasticity. Finally, we present evidence of anesthesia-induced genetic and epigenetic dysregulation within the developing brain that may be transmitted intergenerationally to anesthesia-naïve offspring.

Neonatal anesthesia can alter the epigenome, and these neuropathological changes can be inherited by anesthesia-naive offspring.

Introduction

General anesthesia is a life-saving component of modern neonatal medicine, and every year millions of human infants worldwide are anesthetized for medical and surgical procedures. Yet, a substantial body of preclinical evidence positions anesthetic drugs as neurotoxins in the developing brain [1, 2]. Histological studies in rodents and nonhuman primates reveal that general anesthesia administered during critical stages in infancy is sufficient to cause rapid and widespread neuroapoptotic cell death in young brain [1, 3, 4].

Outcomes of this acute injury appear to manifest as long-term cognitive impairments [1, 5], suggesting irreparable brain damage from anesthetics used every day in neonatal and pediatric medicine. If anesthesia-induced developmental neurotoxicity were a phenomenon restricted to animal models, then it would not be a cause of much concern. However, a dozen clinical studies now report cognitive deficits in children that were anesthetized as neonates [6–8].

But why is a young and pliable brain unable to recover from a relatively brief period of anesthesia? This question motivated us to study the functional, morphological, and genetic and epigenetic mechanisms that may contribute to anesthesia-induced long-term cognitive impairments.

We reasoned that, apart from acute neurotoxicity, anesthetics cause chronic neuropathology in neurons that survive the anesthetic insult. Our line of investigation has been fruitful because we have learned that, from the macroscopic to the microscopic, neonatal anesthesia exposure interferes with major aspects of brain development, synaptic plasticity, and neuronal function [1, 9, 10].

Recently, we and others discovered yet another deleterious effect of anesthetics on neurodevelopment: epigenomic dysregulation of DNA methylation and histone acetylation in the brains of rodents exposed to anesthesia as neonates [11–13]. Importantly, aberrant DNA methylation and gene expression persisted in anesthesia-naïve offspring born to parents exposed to anesthesia during the neonatal period [13]. These findings raise the troublesome possibility that necessary and potentially life-saving anesthesia in the context of sound medical practice may be causing intergenerational harm.

In this review, we briefly summarize the phenomenon of anesthesia-induced developmental neurotoxicity. We then discuss chronic neuropathology caused by neonatal anesthesia, including disturbances in cognition, socio-affective behavior, neuronal morphology, and synaptic plasticity. Finally, we present evidence of anesthesia-induced genetic and epigenetic dysregulation within the developing brain that may be transmitted intergenerationally to anesthesia-naïve offspring.

Characterizing anesthesia-induced developmental neurotoxicity

Rapid and widespread apoptotic cell death

Anesthesia-induced developmental neurotoxicity is the phenomenon by which neurons in the fetal or neonatal brain commit to apoptotic cell death during anesthesia [1, 2, 14]. The histological gold standard to assess this toxicity is cleaved activated caspase-3 (AC3), a robust and well-validated marker of apoptosis. AC3 immunohistochemistry readily reveals two striking features of anesthesia-induced developmental neurotoxicity: its rapidity and its widespread nature.

In our standard paradigm modeling human infant anesthesia in neonatal and pediatric settings, we anesthetize neonatal rodent pups for 6 h. Neuroapoptosis commences shortly after initiation of anesthesia and continues until anesthesia cessation. Within 2 h postinitiation of anesthesia, AC3 positive neurons are detected in subiculum and caudate/putamen [15, 16]. By 6 h, as the apoptotic reaction in caudate/putamen subsides, neurons in neocortex, hippocampus, superior and inferior colliculi, and cerebellum become highly AC3 immunoreactive [1]. As the duration of anesthesia progresses, however, neuroapoptosis spreads to deeper subcortical structures such as anterior thalamus [17].

In early phases, AC3-positive developing neurons retain their soma morphology with only subtle indications that they are committed to apoptotic death, e.g., corkscrewed and fragmenting dendrites (Figure 1A). In later phases, neuronal soma become pyknotic, and apoptotic blebs of membrane and intracellular contents are apparent in the extracellular space (Figure 1B). But by 24 h after anesthesia, AC3-positive cellular debris is completely phagocytosed, leaving almost no histological evidence of the massive neurotoxic injury encompassing nearly all cortical and subcortical regions in the developing brain.

Figure 1.

Panoramic view of anesthesia-induced developmental neurotoxicity in subiculum with neurons at various stages of apoptotic degeneration. (A) Activated caspase-3 (AC3) immunohistochemistry reveals ongoing neuroapoptosis in the subiculum of a neonatal rat pup anesthetized with sevoflurane. Many of the pyramidal neurons in this image are in the early stages of cell death. AC3 stains the soma and fine dendritic processes of pyramidal neurons as they succumb to the apoptotic reaction. (B) Examples of early- and late-stage apoptotic degeneration. The AC3-positive pyramidal neuron in the center of the image has intact morphology of its soma. However, its apical dendrite has started to corkscrew and fragment (dashed bracket), suggesting that this neuron is in early-stage apoptotic degeneration. Beneath it are two pyknotic neurons in late-stage apoptotic degeneration. They have lost their pyramidal soma morphology and are shedding AC3-positive blebs of membrane and intracellular contents (arrows).

Critics have challenged the translational relevance of anesthetizing rodent pups. However, the nonhuman primate (NHP) brain—similar in its neurodevelopmental trajectory to that of human infants [18]—is also susceptible to anesthesia-induced developmental neurotoxicity and displays similar patterns of AC3 immunoreactivity to those in rodents [2, 14, 19]. Furthermore, neonatal NHPs also experience oligodendrocyte apoptosis as well as neuronal apoptosis [2, 19, 20, 21], suggesting that aberrant myelination may be neuropathological feature of neonatal anesthesia unique to higher order mammals.

Window of vulnerability to anesthetics

In all animal species studied, peak vulnerability to anesthesia-induced developmental neurotoxicity occurs as immature neurons form synapses. Synaptogenesis is a burst of synapse formation in the developing brain as nascent pre-synaptic axon terminals are assembled to postsynaptic dendrites [22–24]. Establishment of this primitive neural circuitry is the foundation for the brain connectome that ultimately supports behavioral and cognitive complexity throughout life.

Rodent synaptogenesis occurs postnatally and spikes around postnatal day (PND) 7. At this age, the ongoing neurodevelopmental processes approximate those of a newborn human infant [25], and we have used PND7 rat and mouse pups extensively in our standard paradigm to assess anesthesia-induced developmental neurotoxicity [1, 25, 26, 27].

NHPs and humans undergo prolonged synaptogenesis starting mid-gestation, peaking in the third trimester, and continuing at an elevated rate throughout the first few years of life [18]. Thus, the window of vulnerability to the neurotoxic effects of anesthetics is comparatively larger in these species than in rodents, extending at least from the fetal period to early childhood. Indeed, maternal anesthesia is capable of triggering extensive apoptosis in the fetal NHP brain [2, 19, 21]. The newborn NHP is also exquisitely sensitive to anesthesia-induced developmental neurotoxicity, and this vulnerability to apoptosis extends to oligodendrocytes as well as neurons [4, 14, 28, 29].

Young rodent pups and fetal and neonatal NHPs have an aggressive neurotoxic reaction to nearly every anesthetic tested, including propofol [2, 26], isoflurane [1, 14], sevoflurane [28, 30], and ketamine [3, 19, 31]. Although general anesthetics modulate multiple cellular targets, at clinically relevant concentrations, they preferentially act as glutamate N-methyl-D-aspartate (NMDA) receptor antagonists and/or γ-aminobutyric acid (GABA) receptor agonists. Notably, other drugs common to neonatal and pediatric medicine such as sedatives and anti-epileptics share this pharmacology and are similarly neurotoxic to the developing brain when administered during synaptogenesis [32–34].

Early studies were based on a clinically relevant triple cocktail of sedative and anesthetic drugs administered to mimic sedation and general anesthesia in infants. The triple cocktail was a sedative dose of the benzodiazepine midazolam (9 mg/kg) followed by 6 h of 0.75% isoflurane and 75% nitrous oxide. Midazolam and isoflurane activate GABA receptors and are therefore GABA mimetics. Nitrous oxide, however, is an NMDA receptor antagonist [35]. The dual pharmacology of the triple cocktail caused a devastating neurotoxic injury that was greater than that produced by a single sedative or anesthetic agent [1].

In 2016, the US Food and Drug Administration issued a black box warning that clinicians should discuss with parents the potential neurological harm of long duration or repetitive exposures to anesthesia [36]. As discussed below, however, long durations or repetitive exposures are not necessary for anesthesia to cause neuropathological changes within the developing brain. Short exposures, single exposures, and even exposures that are subanesthetic are sufficient to alter the developmental trajectory of the neonatal brain permanently.

Chronic neuropathology caused by neonatal anesthesia

Cognition and socio-affective behavior

Anesthesia exposure in the neonatal period causes behavioral and cognitive impairments that persist well into adulthood [1]. Performance in hippocampus-dependent spatial tasks is deficient in juvenile and young adult rodents that were anesthetized as neonates compared with littermates that were never anesthetized [1]. These animals have difficulty in both reward- and aversion-based learning paradigms [1, 37, 38].

The radial arm maze (RAM) and Morris Water Maze (MWM) are behavioral paradigms that probe hippocampus-dependent learning and memory in rodents. In the RAM, food deprived rodents are rewarded by learning and memorizing which arms of the maze are baited with food rewards. In the MWM, rodents—excellent swimmers but water averse—work hard to learn and remember the location of a submerged platform that provides respite from swimming. In both the RAM and MWM, visual cues on the walls around the maze provide spatial reference points to guide task acquisition and memorization.

In an early study on the effects of neonatal anesthesia on hippocampus-dependent memory, we anesthetized PND7 rat pups with the triple cocktail, then tested learning and memory at later developmental stages [1]. Learning the location of the submerged platform in the MWM was impaired in juvenile animals. When we retested these animals as adults, we found that they did not memorize the location either, suggestive of impaired higher-level learning often referred to as executive function. Similarly, in the RAM, juvenile animals anesthetized as neonates took longer to learn which arms of the maze were baited with food rewards. Others have reported similar results with repetitive, short duration exposures. For example, juvenile rodents anesthetized with 1.8% isoflurane for 2 h on PND7, PND10, and PND13 did not remember the locations of the baited arms in the RAM [37]. Data from hippocampus-dependent tasks suggest that neonatal anesthesia causes persistent deficits in short-term working memory needed to learn a task and long-term reference memory needed to memorize it.

Poor learning and memory capacity appears to be a replicable and stable outcome of neonatal anesthesia exposure that extends to nonspatial memory as well. By taking advantage of the intrinsic affinity of rodents for novelty, behavioral testing has shown decrements in object recognition following repetitive propofol anesthesia from PND7 to PND13 and in social memory after sevoflurane anesthesia on PND7 [39, 40], tasks that rely on thalamic and prefrontal cortical function. Motor anomalies documented in juvenile rodents exposed to 50-mg/kg ketamine in utero from gestation day 15 to birth shifted toward hyperlocomotion compared with control animals [41], while the behavioral phenotype of young adult rodents anesthetized with 2.1% sevoflurane for 6 h encompassed deficient sensorimotor gating [11, 42] and the appearance of anxiety-like behaviors [42].

The cognitive phenotype of young NHPs anesthetized for 24 h with ketamine as neonates is remarkably similar to that of rodents—persistent impairments in learning and memory [5, 43]—but with the addition of more subtle behavioral alterations that manifest as motivational deficits, poor psychomotor speed, and emotional hyperreactivity to stressors [5, 44, 45].

The cognitive and behavioral repertoire of humans is even more sophisticated than that of NHPs; thus, one would expect nuanced deficits in higher order cognition and socio-affective behavior in addition to deficits in learning and memory. Indeed, clinical studies report that neonatal anesthesia is associated with increased risk of learning and language disability, impaired executive function, and deficits in internalizing behavior [6, 8, 46, 47].

Despite anesthesia-induced developmental neurotoxicity being an acute phenomenon, the persistence of cognitive and behavioral abnormalities long after the neonatal period is over has led us to the inescapable conclusion that neonatal anesthesia causes chronic neuropathology which interferes with the function and plasticity of surviving neurons.

Synaptic plasticity and neurotransmission

Synaptic plasticity is the fundamental mechanism by which neurons process, encode, and store activity-dependent information. Although the definition is straightforward, in practice, synaptic plasticity is stunning in its complexity, requiring coordinated chemical, morphological, and genetic signals to ensure that neuronal stimuli are faithfully recorded in molecular memory.

To accomplish this task, activity-dependent neurotransmission finely tunes synaptic morphology by regulating intracellular signaling pathways that influence gene transcription. Synaptic architecture is then modified by the addition or exclusion of scaffolding proteins, neurotransmitter receptors, and neurotrophic factors. This, in turn, influences the responsiveness of neurons to synaptic neurotransmission, thereby governing neuronal function. During neurodevelopment, exposure of the developing brain to agents that interfere with this delicate equilibrium leads to cognitive and behavioral disability [48, 49]. Importantly, anesthetics perturb neuronal biochemistry, morphology, and gene expression at critical developmental stages that coincide with the appearance of cognitive and behavioral impairments.

The fundamental biochemistry of synaptic neurotransmission is disrupted by neonatal anesthesia. Using electrophysiological techniques to probe neuronal function, we documented prolonged and enhanced excitatory synaptic neurotransmission within the thalamus of young rodents anesthetized on PND7 with the triple cocktail for 6 h [9]. There was also a parallel reduction of inhibitory neurotransmission affecting presynaptic GABA release as well as postsynaptic GABA receptor responses, thus shifting thalamic neurons toward a state of hyperexcitability. We have documented a similar pattern of hyperexcitability in subiculum—an important communication hub between hippocampus and cortex—which manifested as less GABAergic synaptic inhibition and increased neuronal firing in response to stimulation [10, 12]. In hippocampus and cortex, other groups using short duration, repetitive sevoflurane exposure have reported increased frequency of both excitatory and inhibitory synaptic transmission following neonatal anesthesia [42, 50, 51].

Cognitive flexibility and behavioral adaptation rely on thalamic, subicular, hippocampal, and cortical circuitry to appropriately process and store information. Long-term potentiation (LTP) is a form of plasticity in which synapses within neural circuits are strengthened based on the coordinated patterns of activity between postsynaptic neurons and presynaptic inputs. For example, in hippocampus, high-frequency stimulation of CA3 axons generates lasting excitatory postsynaptic potentials in target CA1 neurons.

However, when we stimulated the CA3-CA1 neural circuit in juvenile rodents that were anesthetized as neonates with the triple cocktail, we found no induction of LTP weeks after the initial exposure [1]. Stimulation of the CA1-subiculum pathway also failed to induce LTP in juvenile animals that were anesthetized with 40 mg/kg of ketamine on PND7 [52]. These data led us to conclude that neonatal anesthesia exposure alters the plasticity of neural circuits, as well as the biochemistry of synaptic neurotransmission, both of which are critical for normal cognitive function.

Synaptic morphology

Neonatal anesthesia-induced changes in synaptic neurotransmission are accompanied by aberrant neuronal morphology and synaptic architecture. For example, the dendritic trees of pyramidal neurons are densely covered with dendritic spines, bulbous protrusions that receive excitatory input from presynaptic axon terminals.

In juvenile animals anesthetized as neonates, spine density is altered either upward or downward depending on age of exposure and region examined. For instance, Briner and colleagues found that in the late neonatal period (PND16), the inhalational anesthetics isoflurane (1.5%), sevoflurane (2.5%), or desflurane (7%) markedly increased spine density in layer V cortical neurons [53]. In a separate study by the authors, propofol anesthesia maintained over 6 h led to reduced spine density in young neonates at PND5 and PND10, but juvenile animals anesthetized with propofol experienced increases in spine density [54]. In addition, repetitive anesthesia with the NMDA antagonist ketamine (100 mg/kg) or the GABA agonist midazolam (50 mg/kg from PND8 to PND12) caused lower spine density in hippocampal CA1 pyramidal neurons [55]. Similarly, single exposure and multiple exposures of neonatal rodents to sevoflurane reduced spine density in hippocampus as well [56, 57]. Thus, neonatal anesthesia is a powerful modulator of dendritic spine density.

When we drilled down to the subcellular level, we found devastating changes to the ultrastructure of brains in juvenile animals that were anesthetized with the triple cocktail as neonates [10, 38, 58, 59]. Neuropil was vacuous and disorganized. There was a dearth of presynaptic axon terminals, while many other synapses were actively undergoing destruction weeks after anesthesia exposure. Not only was mitochondria density reduced at synaptic terminals, but also the mitochondria that were present were degenerating, with swollen bodies and damage to inner cristae. Anesthesia-induced destruction of mitochondria likely impedes synaptic plasticity by leaving neurons in a low-energy state, a phenomenon which has been observed in neonatal NHPs anesthetized with 2.5% sevoflurane for 9 h [60].

The clear implication of chronic disturbances in synaptic neurotransmission and architecture coupled with ongoing destruction of neuropil, synapses, and mitochondria is that neonatal anesthesia shifts the developing brain toward a more disordered state of information processing. Because synaptic plasticity is modulated by an array of genes that are subject to epigenetic regulation, we reasoned that neonatal anesthesia has deleterious effects on the genome and epigenome.

Neonatal anesthesia-induced genetic and epigenetic dysregulation

The genome

As we attempted to connect persistent impairments in cognition with augmented synaptic plasticity at the cellular and intracellular levels, we first focused our attention on the classic genome by studying synaptic plasticity genes and their byproducts such as messenger RNA (mRNA) and proteins. Our studies, and those of others, soon expanded to included genes that participate in neuronal survival and fundamental functions [61–73].

We restrict most of our discussion below to brain-derived neurotrophic factor (Bdnf) and cellular Finkel–Biskis–Jinkins murine sarcoma virus osteosarcoma oncogene (c-fos), two genes that are critical for neuronal function. For a comprehensive list of genes dysregulated by neonatal anesthesia and the anesthetic regimens that caused such dysregulation, please see Supplemental Tables S1–S4.

Bdnf is essential for neurodevelopment and modulates many aspects of synaptic plasticity and neural circuit function throughout the lifespan [74, 75]. Genetic expression of Bdnf is partially under the control of neuronal activity, especially the exon IV transcript, which is upregulated in response to neuronal stimulation [76]. Furthermore, BDNF protein is synthesized in the proximity of dendritic spines and plays a role in spine stabilization and elimination [77]. Bdnf also modulates synaptic neurotransmission by increasing neurotransmitter release by expanding the available pool of synaptic vesicles while affecting both glutamate and GABA postsynaptic receptor kinetics [78]. The faciliatory effects of Bdnf on these processes are mediated via BDNF binding to tropomyosin-related kinase B (TRKB) receptors, which promote synaptic stability and learning and memory [79, 80]. Conversely, BDNF binding to the pan-neurotrophin 75 receptor (P75^NTR) can initiate apoptosis, eliminate dysfunctional or quiescent synapses, and alter synaptic plasticity [81].

Bdnf in hippocampus is consistently downregulated by a variety of anesthetics and anesthetic regimens when administered during rodent infancy [12, 82, 83]. Anesthesia with the triple cocktail markedly reduced total BDNF protein in neonatal hippocampus extracts [12]. When we quantified transcripts of the Bdnf gene, we found reduced levels of activity-dependent exon IV mRNA. Other investigators have reported similar results using repetitive, short duration sevoflurane for 2 h daily from PND6 to PND8. This regimen caused downregulation of BDNF protein that was evident weeks later in juvenile rodents [82]. Anesthesia with ketamine, another NMDA receptor antagonist, for 6 h on PND7 also led to decreased BDNF protein that lasted into the juvenile developmental stage [83].

The general pattern of neonatal anesthesia-induced Bdnf downregulation holds in thalamus as well. Both triple cocktail and propofol anesthesia decreased BDNF protein expression in thalamic extracts [27, 84], while a subanesthetic dose of propofol (25 mg/kg) was capable of reducing mRNA levels of total Bdnf and of exons IV and VI of the Bdnf gene [84]. An exception to anesthesia-induced downregulation of Bdnf, however, has been in cortex. Although 25–30 mg/kg of propofol reduced mRNA expression of total Bdnf and of exons IV and VI, BDNF protein was increased in cortical extracts.

With respect to the receptors through which Bdnf signals, TRKB protein is decreased in thalamus after a subanesthetic dose of propofol, while expression of P75^NTR protein increased after a similar dose of 20 mg/kg propofol [27, 84]. Changes in the expression of BDNF receptors may be a compensatory mechanism to maintain homeostasis of BDNF signaling at critical periods of neurodevelopment when BDNF is necessary for synaptogenesis and synaptic refinement. This line of reasoning is bolstered by evidence that 6 h of neonatal ketamine anesthesia lowered BDNF protein levels in juvenile mice, which coincided with impaired developmental pruning of axons in hippocampus [83].

Another gene whose expression is controlled by neuronal activity is c-fos, an immediate early gene whose protein product C-FOS couples signal transduction from extracellular stimuli to the transcriptional machinery of target genes. The triple cocktail, sevoflurane, or propofol during the early neonatal period significantly downregulate C-FOS protein expression in cortex, thalamus, and hippocampus [27, 61, 84]. Notably, c-fos—such as Bdnf—is susceptible to long-term dysregulation in hippocampus. After multiple exposures to 3% sevoflurane for 2 h spanning PND6 to PND8, C-FOS protein remained downregulated some six weeks later, well into the juvenile developmental phase. Interestingly, in rat pups anesthetized in the late neonatal period on PND14, C-FOS protein was upregulated in cortex, suggesting complex, developmental stage specific effects on immediate early gene dysregulation.

We have also documented disturbances in the expression of genes that modulate early developmental responses to neurotransmitters (e.g., the cation-chloride cotransporters NKCC1 and KCC2) and that regulate neuronal survival (e.g., AKT and XIAP) [61, 85]. Furthermore, as we discuss below, neonatal anesthesia disrupts other immediate early genes that regulate transcriptional activity and neuronal function [13].

But what mechanism underlies gene dysregulation? Our recent research provides compelling evidence that the epigenome is a target-rich environment for disruption by neonatal anesthesia and that this disruption influences aberrant gene transcription.

The epigenome

Epigenetics is defined as the biochemical processes that influence the readability of the genome by remodeling chromatin structure. Chromatin is packaged into nucleosomes, which are comprised of a short strand of DNA wrapped around octamers of the histone proteins H2A, H2B, H3, and H4. Histones have long amino terminal tails that are subject to acetylation, methylation, phosphorylation, and ubiquitylation [86]. Condensed chromatin is inaccessible to gene transcription machinery, whereas relaxed chromatin permits the reading and transcription of genes. Although several proposed mechanisms regulate chromatin structure, two mechanisms have received the most research attention in the field of anesthesia-induced developmental neurotoxicity: DNA methylation and histone acetylation (Table 1).

Table 1.

Genetic and epigenetic dysregulation in various brain regions caused by neonatal anesthesia exposure. Never anesthetized is abbreviated NA. Animals that were never anesthetized were offspring of parents who were anesthetized as neonates. mRNA products are written in upper- and lower-case italics, e.g., Bdnf. Protein products are written in upper case, e.g., BDNF

Hippocampus
Gene	Anesthetic	Dose and Duration	Age of Exposure	Developmental Age of Testing	Reference
5hmC ↓	Sevoflurane	3% 2 h	PND7–9	Neonates	Fan et al. [90]
5mC ↓↓	Sevoflurane NA	3% 2 h + 2.4% 4 h NA	PND7 NA	Juvenile (Gen0) Juvenile (Gen1)	Chastain et al. [13]
5mC ↑	Sevoflurane	3% 2 h	PND7–9	Neonates	Fan et al. [90]
Arc ↑↑	Sevoflurane NA	3% 2 h + 2.4% 4 h NA	PND7 NA	Juvenile (Gen0) Juvenile (Gen1 males)	Chastain et al. [13]
Bdnf (H3 acetyl exon IV) ↓	Cocktail	6 h	PND7	Neonates	Dalla Massara et al. [12]
Bdnf (% methyl exon II) ↑	Sevoflurane	3% 2 h	PND7–9	Neonates & Juvenile	Ju et al. [82]
c-fos (H3 acetyl) ↓	Cocktail	6 h	PND7	Neonates	Dalla Massara et al. [12]
Cbp (HAT activity) ↓	Cocktail	6 h	PND7	Neonates	Dalla Massara et al. [12]
CBP ↓↓	Cocktail Sevoflurane	6 h 3% 2 h	PND7 PND6–8	Neonates Juvenile	Dalla Massara et al. [12] Jia et al. [92]
CBP (fragmented) ↑	Cocktail	6 h	PND7	Neonates	Dalla Massara et al. [12]
Dnmt1 ↑	Sevoflurane	3% 2 h	PND7–9	Neonates	Fan et al. [90]
Dnmt3a ↑↑	Sevoflurane	3% 2 h 3% 2 h	PND7–9 PND7–9	Neonates Neonates	Fan et al. [90] Ju et al. [82]
Dnmt3b ↑	Sevoflurane	3% 2 h	PND7–9	Neonates & Juvenile	Ju et al. [82]
H3 (acetyl total) ↓	Cocktail	6 h	PND7	Neonates	Dalla Massara et al. [12]
H3K9 (acetyl) ↓	Sevoflurane	3% 2 h	PND6–8	Juvenile	Jia et al. [92]
H3K14 (acetyl) ↓↓	Cocktail Sevoflurane	6 h 3% 2 h	PND7 PND6–8	Neonates Juvenile	Dalla Massara et al. [12] Jia et al. [92]
H4 (acetyl) ↓	Sevoflurane	3% 2 h	PND6–8	Juvenile	Jia et al. [92]
H4K5 (acetyl) ↓	Sevoflurane	3% 2 h	PND6–8	Juvenile	Jia et al. [92]
H4K12 (acetyl) ↓	Sevoflurane	3% 2 h	PND6–8	Juvenile	Jia et al. [92]
HDAC3 ↑	Sevoflurane	3% 2 h	PND6–8	Juvenile	Jia et al. [92]
HDAC8 ↑	Sevoflurane	3% 2 h	PND6–8	Juvenile	Jia et al. [92]
Junb ↑↑	Sevoflurane NA	3% 2 h, 2.4% 4 h NA	PND7 NA	Juvenile (Gen0) Juvenile (Gen1 males)	Chastain et al. [13]
Kcc2 ↓↓	Sevoflurane NA	6% induction +2.1% 6 h NA	PND5 NA	Adult (Gen0 males) Adult (Gen1 males)	Ju et al. [11]
Kcc2 (% methyl) ↑	NA	NA	NA	Neonates	Ju et al. [11]
Mecp2 ↓	Sevoflurane	3% 2 h	PND7–9	Neonates & Juvenile	Ju et al. [82]
MECP2 ↓	Sevoflurane	3% 2 h	PND7–9	Neonates & Juvenile	Ju et al. [82]
Psd95 (% methyl exon II) ↑	Sevoflurane	3% 2 h	PND7–9	Neonates	Fan et al. [90]
Reelin (% methyl promotor) ↑	Sevoflurane	3% 2 h	PND7–9	Juvenile	Ju et al. [82]
Shank2 (% methyl promotor) ↑	Sevoflurane	3% 2 h	PND7–9	Neonates	Fan et al. [90]
Syn1 (% methyl exon I) ↑	Sevoflurane	3% 2 h	PND7–9	Neonates	Fan et al. [90]
Syp (% methyl exon I) ↑	Sevoflurane	3% 2 h	PND7–9	Neonates	Fan et al. [90]
Tet1 ↓	Sevoflurane	3% 2 h	PND7–9	Neonates	Fan et al. [90]
Thalamus & Hypothalamus
Gene	Anesthetic	Dose & Duration	Age of Exposure	Developmental Age of Testing	Reference
Kcc2 ↓↓	Sevoflurane NA	6% induction +2.1% 6 h NA	PND5 NA	Adult (Gen0) Neonate (Gen1)	Ju et al. [11]
Kcc2 (% methyl) ↑	NA	NA	NA	Neonate (Gen1)	Ju et al. [11]
Nkcc1↑↑↑	Sevoflurane NA	6% induction +2.1% 6 h NA	PND5 NA	Adult (Gen0 males & females) Adult Gen1 males)	Ju et al. [11]
Reproductive Tissues
Gene	Anesthetic	Dose & Duration	Age of Exposure	Developmental Age of Testing	Reference
Kcc2 ↑ (% methyl)	Sevoflurane	6% induction, 2.1% 6 h	PND5	Adult (male sperm)	Ju et al. [11]

Abbreviations: 5mhC, 5-hydroxymethycytosine; 5mC, 5-methylcytosine; Arc, activity-regulated cytoskeleton-associated protein; Bdnf, brain-derived neurotrophic factor; c-fos, Finkel-Biskis-Jinkins murine osteogenic sarcoma virus; Cbp, CREB-binding protein; Ccl12, chemokine ligand 12, Ccl13, chemokine ligand 13; CREB, cyclic AMP response element binding protein; Dnmt, DNA methyltransferase; H, histone; HDAC, histone deacetylase; Junb, Junb proto-onocogene; Kcc2, potassium-chloride transporter member 5; Mecp2, methyl-CpG binding protein 2; NKCC1, Na-K-Cl cotransporter; Psd95, postsynaptic density protein 95; Syn1, synapsin 1; Syp, synaptophysin; Tet1, ten-eleven translocation methylcytosine dioxygenase 1

DNA methylation is a chemical tag by which a methyl (-CH3) group is added to a cytosine base at the carbon 5 position (5mC methylation; Figure 2A). This process shields gene promoters from transcriptional machinery, and, in general, silences gene transcription by inhibiting the association of transcription binding factors or by recruiting repressor proteins such as MECP2 to transcription sites. In the mammalian brain, methylation of DNA is mediated by a family of DNA methyltransferases (DNMTs) [87]. DNMT1 maintains the fidelity of epigenetic patterns during DNA replication. DNMT3a and DNMT3b, however, are de novo methyltransferases that add methyl groups to previously unmethylated DNA.

Figure 2.

DNA methylation in normal neurodevelopment versus neurodevelopment augmented by neonatal anesthesia. (A) During normal development, DNA methylation plays the biological role of shielding gene promoters from transcription machinery by adding a methyl group (Me) to a cytosine base at the carbon 5 position (5mC methylation). Methylated DNA is restricted to a condensed chromatin state, thereby repressing the transcription of target genes. (B) Neonatal anesthesia causes global hypomethylation in the subiculum of juvenile rodents that were anesthetized as neonates (Gen0). Juvenile males and females in Gen0 that were anesthetized as neonates had upregulated expression of Junb and Arc mRNA. Aberrant DNA hypomethylation was heritable by Gen1 male and female progeny of mothers that were neonatally exposed to anesthesia. However, only Gen1 males were vulnerable to dysregulated Junb and Arc expression.

Another primary epigenetic tag is histone acetylation (Figure 3A). While DNA methylation compacts chromatin, making it inaccessible to transcriptional machinery, histone acetylation has the opposite effect of relaxing chromatin structure, thereby permitting gene transcription. Histone acetylation is mediated by histone acetyltransferases (HAT), while deacetylation—leading to condensed chromatin and restricted gene transcription—is mediated by histone deacetylases (HDAC).

Figure 3.

Histone acetylation in normal neurodevelopment versus neurodevelopment augmented by neonatal anesthesia. (A) The core of the nucleosome is comprised of octamers of the histone proteins H2A, H2B, H3, and H4. Histones have long amino terminal tails that are subject to acetylation, methylation, phosphorylation, and ubiquitylation. During normal development, histone acetylation (Ac) relaxes chromatin structure and permits gene transcription. Cyclic adenosine monophosphate response element-binding protein (CREB) complexes with CREB binding protein (CBP) to regulate the transcription of hundreds of target genes. CBP is a potent histone acetyltransferase (HAT), which relaxes chromatin structure and facilitates gene transcription. (B) Neonatal anesthesia reduced CREB expression and fragmented CBP in developing hippocampus, thus reducing CBP HAT activity. Total acetylated H3 was substantially downregulated, and there was global hypoacetylation of histone H3 at lysine residue 14 (K14). Neonatal anesthesia also led to H3 hypoacetylation of the Bdnf and c-fos promoters, as well as reduced expression of BDNF and C-FOS protein.

Notably, DNA methylation and histone acetylation are subject to activity-dependent changes that may regulate the differential expression of genes involved in neuronal function [88, 89]. There is compelling evidence that neonatal anesthesia can modify chromatin structure—at least in hippocampus—toward a more repressive transcriptional state via the hypermethylation of DNA and hypoacetylation of histones.

Using a repeated exposure paradigm of sevoflurane anesthesia for 2 h daily from PND7 to PND9, Ju and colleagues [82] reported substantial upregulation of Dnmt3a and Dnmt3b mRNA and protein in both neonatal and juvenile hippocampus. An identical neonatal sevoflurane regimen elicited an increase in Dnmt3a, as well as Dnmt1, in juvenile hippocampus [90]. There was also substantial upregulation of 5mC, the chemical tag of methylated DNA. Furthermore, when the authors examined locus-specific methylation status weeks after neonatal sevoflurane exposure, they found increased methylation of exon II of Psd95 and exon I of Syp (synaptophysin), genes that regulate synaptic plasticity and function [90]. The authors also reported decreased PSD95 and SYP protein, linking hypermethylated gene status directly to altered gene expression.

Deacetylation of histones, and subsequent transcriptional repression, occurs primarily through the enzymatic action of a family of HDACs, which remove acetyl groups from histone tails [91]. When we evaluated histone acetylation status in the hippocampus of neonatal rats anesthetized with the triple cocktail, we found that levels of acetylated H3 were markedly reduced [12]. In addition, acetylated H4 remained downregulated some six weeks after repetitive, short-duration sevoflurane anesthesia from PND6 to PND8 [92]. Similarly, there were long-term decreases in HDAC3 and HDAC8 [92], reinforcing the possibility that global histone hypoacetylation is a neuropathological consequence of neonatal anesthesia exposure.

We recently reported that neonatal anesthesia exposure interferes with other mechanisms of histone acetylation in the developing brain [12]. We anesthetized PND7 rat pups for 6 h with the triple cocktail, then probed HAT activity, residue-specific histone acetylation, and the acetylation status of the synaptic plasticity genes Bdnf and c-fos.

Cyclic adenosine monophosphate response element-binding protein (CREB), a cellular transcription factor that regulates hundreds of target genes, was reduced in hippocampus in the immediate aftermath of the anesthetic insult. The related CREB binding protein (CBP) acts as a scaffolding protein to recruit gene transcription activators, including CREB, to cyclic adenosine monophosphate response elements (CRE), recurring motifs that reside in the promoter sequences of many genes. CBP also possesses inherent HAT activity, thereby relaxing chromatin structure to a transcriptionally permissive state [93].

Yet, in rodent pups that were exposed to the triple cocktail, we found that CREB and CBP protein was markedly reduced [12]. The likely mechanism of CBP fragmentation was cleavage by caspases activated via the anesthetic insult. Given CBP fragmentation, we hypothesized that its HAT activity would be impaired, and this was indeed the case. Within the hippocampal formation, there was global hypoacetylation of histone H3 at lysine residue 14 (H3K14), which persisted hours after cessation of anesthesia. When we investigated the activity of endogenous HDACs that deacetylate histones, we found no change in their enzymatic activity, leading us to conclude that hypoacetylation of H3K14 was due to downregulated CBP HAT activity.

Given that histone hypoacetylation is known to compact chromatin structure, thereby restricting gene transcription, we reasoned that the transcription of target genes regulated by CREB–CBP would be reduced. Both Bdnf and c-fos genes have CRE motifs and CREB binding sites in promoter regions [94], and thus their transcription is influenced by CBP HAT activity. Using loci-specific chromatin immunoprecipitation assays, we discovered that neonatal anesthesia caused hypoacetylation of H3K14 within the promoter regions of the Bdnf and c-fos genes [12]. Hypoacetylation was the likely mechanism that led to a substantial reduction in mRNA of the activity-dependent exon IV of the Bdnf gene, as well as protein expression. Furthermore, anesthesia-induced hypoacetylation in hippocampus also downregulated c-fos mRNA and protein (Figure 3B).

Can aberrant patterns of epigenomic regulation of gene expression have intergenerational consequences?

To date, two studies—both with positive results—have explored whether aberrant epigenomic regulation of gene expression can be passed on from parents exposed to anesthesia in the neonatal period to their anesthesia-naïve progeny.

In a study by Ju and colleagues [11], PND5 rat pups were anesthetized with 2.1% sevoflurane for 6 h. A subset of pups, which we term Generation 0 (Gen0), was raised to adulthood then paired with breeding mates that were either never exposed to neonatal anesthesia or exposed to neonatal anesthesia. In DNA methylation analyses of Gen0 animals, the authors found that Gen0 males exposed to anesthesia as neonates had hypermethylation of the cation-cotransporter gene Kcc2 in sperm, while neonatal anesthesia did not affect methylation status in female ovary. Generation 1 (Gen1) offspring were born to parents exposed to neonatal anesthesia but were never exposed themselves. When the authors probed methylation status in Gen1 adults, there were substantial increases in Kcc2 gene methylation in males but not females. The functional relevance of hypermethylation of Kcc2 remains to be elucidated; however, the authors did report that Gen1 males had a moderate decrement in hippocampus-dependent learning and memory, but it is unclear whether it could be attributed to changes in Kcc2 methylation status.

We used a similar approach to study neonatal anesthesia influences on the epigenome in that we anesthetized PND7 rat pups with 3% sevoflurane for 2 h followed by 2.4% for 4 h then explored DNA methylation status in the subiculum of juvenile animals [13] (Figure 2B). A subset of anesthesia-exposed females was raised to adulthood then paired with males that were not anesthetized as neonates. Their offspring, Gen1, were anesthesia naïve. As discussed above, in the juvenile period, there is ongoing destruction of neuropil, mitochondria, and synapses, as well as substantial dysregulation of synaptic neurotransmission. Thus, it is important to understand the pathological mechanisms at work during this dynamic developmental stage.

In contrast to anesthesia-induced hypermethylation in hippocampus, neonatal anesthesia caused significant decreases in 5mC DNA methylation in the subiculum of Gen0 juvenile rodents. Hypomethylation during this developmental stage was not sex-specific, occurring in both Gen0 males and females. Global hypomethylation in the subiculum is expected to facilitate the expression of neuronal genes that would normally be silenced during juvenile neurodevelopment. Thus, we evaluated the expression of two immediate early genes that may participate in the neuropathological changes to neuronal morphology and function observed after neonatal anesthesia: Junb proto-oncogene (Junb) and activity-regulated cytoskeleton-associated protein (Arc). Members of the JUN family of transcription factors dimerize with FOS transcription factors to form the Activator protein 1 transcriptional complex [95]. The complex then recruits other transcriptional activators such as CREB to CRE motifs on target genes, thereby regulating gene transcription. Arc is involved in synaptic plasticity by regulating the pliability of synaptic membranes [96]. Using molecular analyses, we found substantial increases in subicular Junb and Arc mRNA in Gen0 male and female juvenile rodents.

We next questioned whether anomalies in DNA methylation and gene expression could be passed on to progeny that would not be anesthetized at any developmental stage. In adulthood, females that were anesthetized as neonates were paired with males that were anesthesia naïve. We then allowed their offspring—Gen1, who were also anesthesia naïve—to mature to the juvenile developmental stage before analyzing their brains for DNA methylation status and Junb and Arc mRNA. When we probed 5mC methylation in Gen1, we found substantial DNA hypomethylation in subiculum in both males and females. However, Junb and Arc mRNA were elevated only in Gen1 males, suggesting a sex-specific vulnerability in the expression of these genes that can be inherited by male progeny.

Conclusions and future perspectives

The studies included in this review provide compelling evidence of a connection between neonatal anesthesia and lasting genomic and epigenomic dysregulation. We have summarized the findings of our lines of investigation in Figure 4.

Figure 4.

Schematic representation of epigenomic changes caused by anesthesia during the neonatal period. Neonatal anesthesia can influence gene transcription by downregulating cyclic adenosine monophosphate response element-binding protein (CREB) and fragmenting CREB binding protein (CBP). Fragmentation of CBP also compromises its histone acetyltransferase (HAT) activity. This may lead to H3 hypoacetylation in the promoters of the activity dependent exon IV of the Bdnf gene, as well as c-fos. In parallel, exposure of the developing brain to anesthetics can cause DNA hypomethylation, which may lead to inappropriate transcription of genes that are normally silenced at different stages of development. DNA hypomethylation may be inherited by anesthesia-naïve offspring. These disordered epigenomic changes may be responsible, in part, for the persistent neuropathological outcomes documented in animals and humans anesthetized as neonates.

Briefly, neonatal anesthesia interferes with histone acetylation by fragmenting CBP, thus reducing its HAT activity. This leads to hypomethylation of the Bdnf and c-Fos genes. Aberrant DNA methylation and Junb and Arc expression caused by neonatal anesthesia exposure and observed in the brains of juvenile females persist into adulthood; then, they are passed on intergenerationally to offspring that were never exposed to anesthesia during the neonatal period or any other developmental period.

Epigenomic dysregulation of genes critical to neurodevelopment is likely one of the mechanisms underlying the persistent functional pathology in synaptic neurotransmission, LTP, neuronal morphology, and cognition we and others have observed in animals exposed to anesthetics in the neonatal period. Thus, relatively brief anesthesia within the context of sound neonatal and pediatric medical practice may cause intergenerational harm.

Anesthesia in adults produces minimal side effects, which has led to the misguided notion that anesthetics are innocuous agents in the developing brain as well. But the developing brain is not a miniature adult brain. Dynamic processes such as neurogenesis, synaptogenesis, myelination, and synaptic neurotransmission must continue undisturbed to pattern the developing brain to support normal behavior and cognition throughout the lifespan.

In neonatal rodents and NHPs, anesthesia-induced developmental neurotoxicity causes the untimely death of millions of neurons as they mature and begin forming the brain connectome [1, 2, 14]. While the window of vulnerability to anesthetics is wide, the injury itself is temporally narrow—neurons commit to apoptosis, die, then are dismantled completely within a matter of hours—and rapid enough to escape in vivo detection. However, cognitive and behavioral impairments in rodents, NHPs, and human children are disturbing evidence of the power of anesthetics to alter the long-term function of the brain.

Also, we now know that anesthesia-induced neurotoxicity encompasses more than apoptotic cell death. It involves a spectrum of neuropathology that affects synaptic neurotransmission, neuronal morphology, and gene expression via the epigenomic processes of DNA methylation and histone acetylation [9, 10, 12, 13]. Indeed, anesthetic damage to the developing brain has epigenetic hallmarks that may, in part, account for cognitive impairments linked to neonatal anesthesia exposure. In this review, we have outlined disturbances in activity-dependent transcription factors that regulate synaptic plasticity, aberrant histone hypoacetylation of genes in brain structures important for learning and memory, and heritable dysregulation of DNA methylation that may have chronic, deleterious consequences for global neuronal function, behavior, and cognition.

Despite this bleak outlook, we do find hope in the efficacy of pharmacological interventions to mitigate at least some of the deleterious effects of neonatal anesthesia on the epigenome. Our therapeutic approach has been to target histone acetylation due to the excellent safety profile of HDAC inhibitors such as sodium butyrate and entinostat. Pretreatment with entinostat before midazolam, isoflurane, and nitrous oxide anesthesia reversed H3 hypoacetylation in developing hippocampus [97], while sodium butyrate administered postanesthesia exposure on PND7 also restored histone acetylation status [12]. Furthermore, sodium butyrate abolished pathological changes to dendritic morphology in cultured developing hippocampal neurons challenged with the triple cocktail. With respect to parameters of neuronal function, both sodium butyrate and entinostat were able to completely normalize inhibitory GABA synaptic neurotransmission [12, 97].

Thus, at least some deleterious functional changes caused by neonatal anesthesia exposure can be reversed pharmacologically by targeting the epigenome, and the therapeutic window may be large enough to intervene at timepoints before, during, and after the developing brain is subjected to the neurotoxic effects of anesthetics.

When speculating about what these data mean for human infants, it is worth keeping in mind that, for intergenerational studies, mothers were months removed from anesthesia exposure when they gave birth. Their offspring were never anesthetized yet had persistent alterations in DNA methylation and gene expression well into late developmental stages. It is imperative to understand as comprehensively as possible how these epigenetic changes influence neurodevelopment intergenerationally in anesthesia-naïve offspring. Although we have elucidated several mechanisms by which anesthetics interfere with the epigenome during critical periods of neurodevelopment, there are many outstanding questions that await answers.

First and foremost is functional examination of anesthesia-naïve offspring born to parents that were anesthetized during the neonatal period. There is compelling evidence from animal models that dysregulation of the epigenetic machinery causes cognitive impairments, suggesting that the epigenome plays a critical role in learning and memory. Given that neonatal anesthesia exposure causes profound cognitive impairments that do not resolve in adulthood, functional tests should include electrophysiological characterization of neuronal excitability, as well as changes in synaptic neurotransmission. Behavioral profiling in rodents should include standard tests of hippocampal-dependent memory but also more sophisticated tests of cognition to determine whether nuanced impairments noted in first generation parents exposed to anesthesia as neonates extends to anesthesia-naïve rodent offspring. Neurophysiological and behavioral studies should be complemented by detailed morphometric assessment of neuronal dendritic structure, dendritic spine density, mitochondrial integrity, and synaptic architecture.

Sex must be considered as a biological factor that influences susceptibility to anesthesia-induced epigenomic dysregulation. In the intergenerational studies outlined above, male progeny of parents neonatally exposed to anesthesia was particularly vulnerable to inheriting aberrant DNA methylation, while female progeny was not. The sex-specific mechanisms underlying this phenomenon are unclear and await study.

Future investigative endeavors would also attempt to understand neonatal anesthesia influences across scales of the genome and epigenome. Because of time and fiscal concerns, experimenters—ourselves included—have restricted evaluations to select targets in select brain regions. But given our experience with the widespread nature of anesthesia-induced developmental neurotoxicity and neuropathology, we hypothesize that it dysregulates the expression of many more genes and disturbs many more epigenetic processes across the brain. Hundreds of genes contain CRE motifs and are predicted to be responsive to CBP HAT activity. These include many of the genes that we and others have found to be altered by neonatal anesthesia including Trkb, p75^NTR, Arc, and Akt [98]. It would be of substantial research value to map anesthesia-induced changes in DNA methylation and histone acetylation within these genes.

We also envision extending epigenomic mapping to specific brain regions that have been proven to be sensitive to anesthesia-induced developmental neurotoxicity including cortex, thalamus, hippocampus, and subiculum. Patterns of aberrant epigenetic modifications may be identified across the brain, which then may be targeted for the appropriate pharmacological intervention. Also, investigation of histone methylation, phosphorylation, and ubiquitylation states and their influence on anesthesia-induced cognitive impairments may yield additional mechanistic insights.

These lines of investigation are important for the safety of millions of infants and children that are anesthetized every year. And as we have presented in this review, it may also be important for the safety of their children.

Authors’ contributions

Conceptualization, writing, editing: OHC, VJT.

Figures, editing: NU.

Conflict of interest

The authors declare no conflicts of interest.

Data availability

No new data were generated or analyzed in support of this research.

Funding

This work was supported by the Department of Anesthesiology at the University of Colorado Anschutz Medical Campus, Aurora, CO; Eunice Kennedy Shriver National Institute of Child Health and Human Development, Bethesda, MD [R01HD097990, R01HD044517, R01HD044517-S, R21HD080281; F32HD101357]; National Institute of General Medical Sciences, Bethesda, MD [R01GM118197, R01GM118197-11S1], March of Dimes National Award, Crystal City, VA; and CU Medicine Endowments, Aurora, CO.