Sleep Spindles and Auditory Sensory Gating: Two Measures of Cerebral Inhibition in Preschool-Aged Children are Strongly Correlated

Abstract

Introduction.

Sleep spindles and P50 sensory gating are both reflective of cerebral inhibition, however, are differentially active during different phases of sleep. Assessing whether sleep spindles and P50 sensory gating correlate is a first step to evaluate whether these 2 forms of cerebral inhibition reflect overlapping neural circuits.

Methods.

EEG data were collected between midnight and 6:00 AM on 13 healthy preschool-aged children. P50 sensory gating, calculated during REM sleep, negatively correlated with spindle duration (r=−.715, p=.006) and inter-peak density (r=.744, p=.004). There was a trend toward higher S2/S1 ratios being associated with fewer peaks per spindle (r=−.546, p=.053). In 4-year-olds, 2 established physiological measures of sensory gating and are correlated despite being maximally active during different stages of sleep.

Conclusions.

These results suggest there is an overlap in brain mechanisms underlying each gating mechanism.

At any moment, a vast amount of sensory information is being captured by peripheral modalities and has the potential to reach the cerebral cortex, where it can be processed and acted upon. Much of this information is extraneous, and thus reducing (“gating”) transmission of irrelevant sensory information is necessary in order to avoid overburdening cortical processing capabilities. This is particularly true during sleep, where cortical processing of external stimuli has to be limited to the most critical sensory information. While sensory gating occurs during multiple stages of sleep, there has been little effort to explore correlations between gating across stages: does sensory gating ability at one stage of sleep predict sensory gating during another stage? A correlation between different gating mechanisms would suggest overlapping neural mechanisms of gating and provide support for exploring common etiologic factors. We focus here on 2 sensory gating processes that are both thought to include thalamic interneurons,^1,2 but which are prevalent at different sleep stages: sleep spindle generation, which is limited to Stage 2 non-Rapid Eye Movement (NREM) sleep; and P50 sensory gating, which is maximally effective while awake and during Rapid Eye Movement (REM) sleep. A sample of convenience of overnight electroencephalograms from healthy 4-year-olds was utilized in this initial study.

Sleep spindles are characteristic brief 11-15 Hz waveforms on EEG with a progressively increasing then decreasing amplitude that are specific to the sleeping state.³ They are generated by GABAergic neurons in the thalamic reticular nucleus (TRN)⁴ and are believed to reflect neuronal connectivity patterns in corticothalamic and thalamocortical circuits.^3,5 Ninety percent of TRN neuronal projections are to thalamocortical neurons; activation of TRN neurons inhibits signal transmission to the cerebral cortex.⁶ This thalamic spindle production by cortex-stimulated reticular cells gates sensory input while sleeping, allowing the brain to resist waking in the face of disruptive external sensory stimuli.^{7, 8}

P50 auditory sensory gating refers to a reduction in an early evoked response in the face of repetition of an auditory stimulus. In the most commonly utilized form of the task, an individual is exposed to 2 identical auditory stimuli occurring 500 ms apart. The amplitude of early components of the evoked response is reduced in response to the second stimulus relative to the first.⁹ One common quantification of this effect is to measure the ratio of the amplitudes of the P1 wave, which in adults occurs approximately 50 ms after both the first (S1) and second (S2) stimulus. Intact auditory sensory gating is indicated by a reduction in the amplitude of the evoked P1 wave to the second stimulus that yields a ratio significantly less than 1. P50 sensory gating (ratios closer to 0) occurs during REM sleep but is absent (ratios closer to 1) in NREM sleep.¹⁰ P50 sensory gating can be identified in newborns¹¹ and appears to be fully developed within a few months after birth.¹² P50 sensory gating has been linked to GABAergic neurons^{13, 14} and involves a circuit which includes the thalamus, hippocampus, and prefrontal cortex.²

Sleep spindles have been found to mediate sleep maintenance,¹⁵ memory consolidation,^4,8,16-19 and cortical development during sleep,⁵ among other functions, and reduced spindle activity is thought to be a reflection of inherent thalamic dysfunction as well as thalamic responsiveness to cortical stimulation. Poor sleep spindle generation in 5-year-olds predicts more externalizing behavior and more peer problems 1 year later.²⁰ In a similar fashion, poor auditory sensory gating in infants has been associated with later problems in attention, anxiety, and externalizing symptoms at 40 months of age,²¹ and it has been postulated that diminished sensory gating in infants may reflect an increased risk for later psychopathology.^22-24 fMRI studies also suggest that the thalamus is involved in sensory gating.^2,25 For both sleep spindles and P50 sensory gating, the thalamic reticular nucleus is thought to be responsible for the thalamic response during the gating tasks,^2,6,19 where dysfunction in GABAergic neurotransmission leads to impaired sensory inhibition.^4,13,14

Sleep spindles and P50 sensory gating are both thought to reflect activity of GABAergic neurons and are also both thought to involve the thalamus, yet the 2 measures reflect activity during different sleep stages (Stage 2 for sleep spindles and REM for P50 gating). This raises the question of whether or not the 2 inhibitory processes are correlated—we hypothesize that they are. We have previously reported on overnight P50 sensory gating scores in 4-year-old children as part of a process to determine stability between infant and 4-year-old performance during REM sleep. That study suggested that P50 sensory gating matured to adult levels within a few months after birth and that analysis of 4-year-old results may be generalizable to other ages.²⁶ The current study adds an analysis of sleep spindles during the same overnight recording with the goal of investigating the relationship between 2 measures of cerebral inhibition, sleep spindles and P50 auditory sensory gating. This is the first study to investigate such a relationship and would contribute to the effort to understand the relationship between sensory gating mechanisms.

Method

Participants

Fourteen preschool-aged children (9 females) who were part of a longitudinal study on early child development in a large metropolitan area and who were within 2 weeks of their fourth birthday were recruited for an overnight sleep study aimed at testing P50 auditory sensory gating stability from birth to 4 years of age. Artifact-free data with minimums of 30 minutes of REM and 3.5 hours of total sleep was considered the minimum amount necessary for analysis. Data from 1 participant (1 female) is excluded because there was insufficient artifact-free data. The mean age for the remaining participants in this study is 47.15 (SD=0.99) months. Additional demographic information is summarized in Table 1. This was a healthy pediatric sample, which had been followed since infancy.

Table 1.

Demographics, Spindle, and P50 Characteristics

Demographics	N (%)
Female:male	8:5 (62:38)
Caucasian Hispanic	2 (15)
Caucasian Non-Hispanic	11 (85)
	Mean (SD)
Maternal Socioeconomic Status (SEI)a	55.38 (26.79)
Age (months)	47.15 (.99)
Spindle Characteristics	Mean (SD)
Spindle Duration	838 (144) msec
Number of peaks per spindle	11.5 (1.75)
Inter-peak interval	72.65 (4.19) msec
P50 Characteristics	Mean (SD)
S2 (test) amplitude	1.39 (1.04) μV
S1 (conditioning) amplitude	2.58 (1.36) μV
S2/S1 ratio	0.534 (0.297)
S2 latency	67.31 (7.94) msec
S1 latency	64.2 (8.03) msec

^aThe Socio-economic Index (SEI) of Occupations³⁶ includes 503 occupations scored in a potential range of 0-100. Managerial and professional occupations generally have scores above 60; technical, sales, and administrative support occupations generally score between 35 and 60; service, agricultural, and labor occupations generally have scores below 35; never employed single individuals are assigned a score of 0. Values reported are for the highest occupation value achieved across an individual’s life.

Procedure

All procedures involving human subjects were approved by the Colorado Multiple Institute Review Board (COMIRB), and parents of the participants gave written informed consent. Participants were admitted for an overnight stay to a pediatric clinical research center at a local children’s hospital. All participants were screened prior to admission for acute illness, and once admitting procedures were complete, were provided dinner and access to entertainment until bedtime. Parents of participants were encouraged to follow the child’s typical bedtime routine. One parent remained in the room with the child overnight.

Electroencephalographic Recordings

Ag/AgCl electrodes (Grass; West Warwick, Rhode Island, USA) filled with Ten20 conductive paste (DO Weaver; Aurora, Colorado, USA) were attached to the sleeping child with adhesive medical tape. EEG and auditory-evoked potentials were recorded from the vertex of the scalp (Cz). For aid in sleep staging, bipolar electrooculogram (EOG) was recorded from electrodes directly superior and lateral to either the left or right eye; submental electromyogram (EMG) was also recorded. Times of movement and environmental events were also noted. Signals were recorded using NuAmps (Neuroscan Labs, Sterling, Virginia, USA). EEG signals were amplified 5000 times and filtered between 0.05 and 100 Hz; EOG signals were amplified 1000 times and filtered between 1 and 200 Hz; and EMG signals were amplified 10,000 times and filtered between 1 and 200 Hz. Sampling rate occurred at 1000 Hz. Stimulus presentation (for assessment of P50 auditory sensory gating) and recording began when the electrode impedances were below 10 kΩ. The longest periods of REM (the stage for the reliable assessment of P50 sensory gating) and Stage 2 (the period most associated with sleep spindle activity) sleep occur in the early morning hours; thus, data collection did not begin until 11:00 PM and continued until approximately 6:00 AM.

The data were converted from the Scan 4.1 software (Neuroscan Labs; Sterling, Virginia, USA) format to ASCII format so that further analysis using MatLab (Mathworks; Natick, Massachusetts, USA) software could be conducted. A visual representation of EOG, EMG, and EEG activity was generated and provided a global view of each participant’s sleep cycles (Figure 1A). Sleep staging was completed based on the criteria of Anders et al.²⁶ REM sleep was identified by the presence of rapid eye movements obtained on EOG, low amplitude in the EMG, and low amplitude high frequency in the EEG. Stage 2 NREM sleep was identified by the absence of rapid eye movements as obtained by EOG, increased amplitude in the EMG, and the presence of sleep spindles. Sleep state was then verified by visual inspection of the continuous recording in 20-s epochs.

Figure 1.

(A) Graphical representation of a 5-hour sample drawn from overnight EEG, EOG, and EMG activity recorded from a participant. A period of REM has been highlighted to show increased EOG activity and decreased EMG and EEG spindle activity indicative of this stage of sleep. A period of Stage 2 NREM activity has been highlighted to demonstrate increased spindle activity in the EEG accompanied by decreased EOG activity. (B) An example of a sleep spindle recorded from the vertex. This image is representative of spindle activity following the application of an 11-15 Hz bandpass filter. This particular spindle had 6 peaks with a mean inter-peak interval of 73.33 ms. (C) P50-evoked potential tracings. Two stimuli are presented 500 ms apart (noted by the 2 time 0s in any horizontal pair of panels). The P50 response is delineated by hash marks. Intact sensory gating is demonstrated on the top half of this figure. Note the size of the evoked response to the first stimulus, S1 (1.86 μv) and the corresponding response to the second stimulus, S2 (.11 μv). The resulting S2/S1 ratio is 0.06. Poor P50 sensory gating is demonstrated on the bottom half of this figure. The amplitude of the response to the first stimulus, S1 was 1.64 μv while that for S2 was 1.05 μv. The resulting S2/S1 ratio is 0.64.

Sleep Spindle Analysis

Spindle detection was performed by applying a bandpass filter between 11-15 Hz. A baseline average amplitude was calculated from spindle-free EEG and a threshold value of 2.5 times the baseline average was used for automated identification of spindle cycles. Likewise, the beginning and end of each spindle was determined by an increasing or decreasing peak amplitude that fell below 2.5 times the baseline average. Automated selection was confirmed by visual inspection of the EEG record (Figure 1B). Measurements of spindle duration in milliseconds and the number of individual peaks per spindle were obtained. Mean inter-peak interval (calculated as number of peaks per spindle divided by spindle duration) was used as a measure of spindle density. Spindles occur as a series of electroencephalographic peaks; more peaks per spindle, a longer spindle duration, and a lower inter-peak interval (greater density of peaks) are evidence of increased spindle activity and reflect increased sensory gating.

P50 Auditory Sensory Gating Analysis

Methods for P50 sensory gating assessment in children have been previously described in detail,¹² and will briefly be reviewed here. Paired clicks were presented through 2 speakers positioned on either side of the bed at a distance of .50 m from each ear. Volume was adjusted so that each click was at 85-dB sound pressure level at the ear. The clicks were delivered continuously throughout the overnight study for each subject. The first 20 minutes of the longest-recorded REM cycle was used for analyses. This length of time was selected because it yields an adequate number of stimuli for analysis and reduces variability caused by individual differences in sleep.²⁷

Single-trial-evoked potentials were extracted from 100 ms before each click to 200 ms following each click. Trials were excluded in which the signal on the recording of identified periods exceeded ±75 mV. The average waveforms from single trials were band pass filtered between 10 and 50 Hz to accentuate middle latency components. For each subject, the largest positive peak between 50 and 100 ms after an auditory click (P50) preceded by a negative trough was identified and measured, peak to trough, by a computer algorithm.

For each child, a mean response latency and amplitude of the P1-evoked potential wave to each stimulus (S1 and S2) were calculated. In addition, their ratio was calculated by dividing the amplitude of the P1-evoked response to S2 by the amplitude of the P1-evoked response evoked by S1. A ratio closer to 0 is indicative of robust sensory gating, while a ratio closer to 1 is indicative of diminished sensory gating (Figure 1C).

Statistical Approach

Descriptive data was calculated for spindle measures (duration and density) and P50 auditory sensory measures (amplitudes, latencies, and gating ratio). Bivariate correlational analyses were used to assess the relationship between spindle duration, number of peaks per spindle, and spindle density and measures of P50 auditory sensory gating. IBM SPSS Statistics for Windows, Version 22 (Released 2013, Armonk, NY: IBM Corp.) was used for all analyses.

Results

The mean length of artifact-free sleep data was 4.76 (SD=.64) hours (Range: 4.06–5.96 hours). The mean length of REM sleep data was 57.77 (SD 8.96) minutes (Range 44-69 minutes). Table 1 summarizes the primary electrophysiological measures of interest.

P50 gating ratios were positively correlated with spindle density (r=.744, p=.004), negatively correlated with spindle duration (r=−.715, p=.006) and trended towards a negative correlation with mean inter-peak interval (r=−.546, p=.053) (Table 2 and Figure 2).

Table 2.

Correlations between P50 components and spindle characteristics S1, First stimulus in P50 measure; S2, Second stimulus in P50 measure p<.05

	Spindle Duration		Number of peaks per spindle cycle		Inter-peak interval (Intra-spindle density)
	Pearson Correlation (r)	Significance (p)	Pearson Correlation (r)	Significance (p)	Pearson Correlation (r)	Significance (p)
S1 amplitude	−.044	.885	−.134	.662	.195	.524
S2 amplitude	−.484	.094	−.424	.149	−.364	.221
S1 latency	−.201	.511	−.286	.343	.147	.631
S2 latency	.226	.458	.177	.562	.162	.596
P50 ratio (S2/S1)	−.715	.006a	−.546	.053	−.728	.005a

Figure 2.

Scatterplot showing correlation of P50 ratio (S2/S1) and (A) mean spindle duration (r=−.715, p=.006), (B) number of peaks per spindle cycle (r=−.546, p=.053), and (C) mean inter-spindle interval (intra-spindle density) (r=−.728, p=.005).

Discussion

One of the brain’s most important functions is to inhibit its own signals in order to filter out irrelevant sensory information. It utilizes a number of mechanisms to achieve this goal including 2 processes active during sleep: sleep spindle generation and P50 sensory gating. While thalamic GABAergic inhibition has been postulated to contribute to both inhibitory mechanisms, there has been little investigation of whether performance of one process correlates to the other. In this study, we examined the relationship, in preschool-age children, between 2 measures of cerebral inhibitory functioning that had not been compared before, P50 sensory gating and sleep spindles. Elevated P50 gating ratios, corresponding to impaired auditory gating, were associated with shorter spindle duration as well as increased intra-spindle density (lower mean inter-peak interval).

P50 sensory gating and sleep spindles are maximally active at different stages of sleep, suggesting, on the surface, that they are the product of different neurological circuits. However, our results demonstrating the correlation between the processes are more consistent with the hypothesis that GABAergic thalamic interneurons contribute to both processes. As the TRN functions as the spindle pacemaker, with spindle rhythm persisting in decorticated animals but disappearing after thalamic destruction,⁵ both intrinsic thalamic cells and their connections to and from the cortex are implicated in diminished gating.

An alternative method for examining interrelationships between these 2 sleep-associated gating mechanisms would have been to focus on individuals with deficiencies in cognitive functions that have previously been associated with abnormal sleep spindles and P50 measurements, including attention, executive functioning, and working memory. These 2 sensory gating deficits are present in a range of neurodevelopmental disorders including schizophrenia,^{7,9,16,19,28-30} autism spectrum disorders,^31-33 and ADHD.³⁴ However, even if a correlation between these 2 physiologic measures had previously been elucidated in neuropsychiatrically-ill subjects, it would be difficult to interpret the results. It would be unclear whether these measurements would correlate due to overlapping functional circuits or whether the correlation was a result of similar effects on both processes due to medication treatment or cognitive sequelae of the disease. P50 sensory gating matures to adult levels within a few months after birth²⁷ and is stable after that point.²⁶ Spindle activity also is relatively stable, at least from late preschool years through mid-adolescence.³⁵ Thus, although the pediatric population investigated in this study was a sample of convenience, generalizability to older populations may be reasonable. In addition, an advantage of studying young children is the ability to study processes prior to onset of treatment and prior to the effects of years of having to live with the disease. Establishing that these 2 measures represent the same underlying pathology may guide future studies in clarifying the mechanism behind these faulty inhibitory brain processes.

While not the primary purpose of this report, an important goal in the investigation of inhibitory processes could be to contribute to the identification of biophysiological risk factors for neurodevelopmental disorders. Lower spindle activity and poor P50 sensory gating each are predictive of both later neurocognitive and behavioral difficulties.^20,21 If sleep spindles and P50 auditory gating are indeed correlated and represent the same underlying pathology as our results suggest, concurrent evaluations of both processes may provide better predictive ability than either assessment alone. Future research would be required to assess this possibility. In addition, as this study was done only in young children, future research may also investigate the effect of age on the correlation between sleep spindles and auditory gating.

Conclusion

P50 sensory gating and sleep spindle both function to gate sensory information; however, during sleep, spindle generation is strongest during Stage 2 slow wave sleep, while P50 sensory gating is most active during REM sleep. The correlation between P50 gating ratios and sleep spindles, 2 physiological measures of sensory gating, suggest that, despite the fact that the 2 measures occur during different stages of sleep, these 2 processes may share some underlying neurocircuitry. GABAergic thalamic neurons are one possible contributor to both processes.