Early postnatal development of sensory gating

Abstract

Sensory gating represents the nervous system’s ability to inhibit responding to irrelevant environmental stimuli. In order to characterize the early development of acoustic sensory gating, suppression of auditory evoked potential component P1 (i.e. P50) in response to paired clicks was measured during REM sleep in healthy infants (1–4 months) that were without genetic risk for disrupted sensory gating function (i.e. having a relative with schizophrenia). As a group, the subjects exhibited significant response suppression. A correlation between increasing age and stronger response suppression was uncovered, even within this restricted age range. Parallel changes in sleep physiology could not be ruled out as the explanation for this change. Nevertheless, these results demonstrate that the neural circuits underlying sensory gating are functional very early in postnatal development.

INTRODUCTION

Evoked and event-related potentials are increasingly utilized to investigate early postnatal brain development in humans. With these methods, sensory and perceptual capacities of an infant’s brain can be probed without reliance on behavioral measures [1–3]. Recently it has been demonstrated that certain neurophysiological measures can predict a baby’s relative risk for subsequent neurodevelopmental disorders, including dyslexia [4,5].

Sensory gating, manifest electrophysiologically as the reduction in amplitude of particular pre-attentive auditory evoked potential components (notably P1) in response to stimulus-repetition, is impaired in people with schizophrenia (reviewed in [6]), schizotypal personality disorder [7] and in some otherwise healthy individuals with inherited risk for the illness [8]. Because schizophrenia results from abnormal neurodevelopment that begins in the perinatal period (reviewed in [9]), sensory gating measures might be useful in this context for tracking early development of relevant brain systems and identifying individuals at increased risk. However, normative data on infant sensory gating are lacking.

P1 sensory gating function in healthy adults is relatively robust across most experimental situations examined, with the exception of acute stress [12,13]. Disruption of auditory gating under this condition is believed to result from the associated increase in central norepinephrine release [14,15].

Previous studies in healthy pediatric populations have demonstrated that the neural circuitry underlying sensory gating develops before a child reaches 10 years of age [10]. A more precise developmental timeline, however, has not been uncovered. Before 10 years of age, and as young as 18 months old, sensory gating measures are extremely variable from child to child [11]. Observed variability in this population, particularly in children prone to react with stress to novel situations, might be explained by the state-dependence of sensory gating measures mentioned above [10].

We recently demonstrated that P1 sensory gating can be measured in adults during REM sleep, a state characterized by reduced central norepinephrine, and that gating impairments associated with schizophrenia persist into this state [16,17]. Because of these properties, and because neonates spend a proportionately large amount of time in this state [18], REM sleep appears to be an ideal time to assess sensory gating in young infants without potential confound from acute stress.

The present study was designed to provide an initial characterization of sensory gating during REM sleep in the early postnatal period. We employed a traditional paired-click paradigm, and measured the amplitude of auditory evoked potential component P1 (i.e. component P50 in the adult literature).

MATERIALS AND METHODS

All procedures involving human subjects were approved by the Colorado Multiple Institute Review Board, and mothers gave written informed consent. Healthy infants (1–5 months post-term) without neurological or hearing problems were recruited. Because genetic risk for schizophrenia affects sensory gating measures [8], infants with a known family history of schizophrenia, and infants whose mothers admitted to experiencing symptoms of psychosis in the past, were excluded.

Evoked potentials were recorded during a daytime nap from 21 infants. A total of 10 infants were excluded from analysis because of insufficient REM sleep during the recording (n = 9), or because evoked potentials were not larger than background noise (n = 1). The remaining 11 infants were between 45 and 55 weeks conceptual age (mean 51.1) at time of recording. All were born without complications between 36 and 41 weeks conceptual age (mean 38.2), and birth weights between 2.27 and 3.63 kg (mean 3.14 kg). Seven of the 11 were female, five were bottle-fed at time of recording, and the mothers of six infants had continued smoking throughout all three trimesters of pregnancy. None of the mothers drank more than 1 alcoholic beverage per week once they were aware of the pregnancy. According to the mother’s choice, nine of 11 recordings were performed in the infants’ homes, the remainder in a perinatal laboratory at the University of Colorado Health Sciences Center.

Ongoing electroencephalogram (EEG) and auditory evoked potentials were recorded from vertex (Cz) referenced to right mastoid, with a ground electrode at left mastoid. Bipolar electrooculogram (EOG) was recorded between electrodes directly superior and lateral to left eye. Submental electromyogram (EMG) was recorded with a bipolar configuration. Electrodes were gold-plated discs (Grass; West Warwick, RI, USA) attached with Ten20 conductive past (DO Weaver; Aurora, CO, USA) and paper tape. All impedances were maintained below 10 k[OMEGA]. Respiration was measured by breathing-effort strap (Grass).

After attaching electrodes and respiration strap, the infant was fed by mother and laid down to sleep. Throughout the subsequent recording session, paired-clicks (0.5 s between clicks, 10 s between pairs) were played continuously through two speakers positioned at either side of the infant’s head, each 0.5 m from the corresponding ear. Volume was adjusted so that each click, ~1 ms in duration, exhibited a peak loudness of 85 dB sound pressure level at the ear.

All physiological signals were recorded with a portable Quicktrace amplifier system (Neuroscan Labs; Sterling, VA, USA). For acquisition, EEG signals were amplified 5000 times and filtered between 0.05 and 100 Hz, EOG amplified 1000 times and filtered between 0.05 and 100 Hz, EMG amplified 10 000 times and filtered between 1 and 200 Hz, and output of the breathing sensor was amplified 100 times and filtered between 0.05 and 30 Hz. All channels were sampled at 1000 Hz. Continuously recorded data were converted from Neuroscan’s Scan 4.1 software format to ASCII format and imported into the Matlab software package (Mathworks; Natick, MA, USA) for further analysis with custom programs.

Sleep state was determined by visual inspection of 20 s epochs. Because sleep architecture changes significantly at 3 months post-term (daytime naps [19,20]), sleep state for infants under 3 months was determined by the criteria of Anders et al. [21], whereas sleep stage for infants over 3 months was determined by the adult criteria of Rechtschaffen and Kales [22]. Although definition and staging of non-REM sleep differs between these publications, in both instances REM (i.e. active) sleep is identified by the presence of rapid eye movements on the EOG channel, suppressed EMG activity corresponding to reduced muscle tone, low-amplitude high-frequency EEG activity, and irregular breathing.

Average evoked potentials were computed from the EEG activity evoked by the paired-clicks during the initial 15 min (corresponding to ~85 click-pairs) of the first REM period that lasted >= 15 min. A REM period was considered to begin when at least two consecutive 20 s epochs were scored as REM sleep, and end when at least three consecutive epochs were scored as non-REM sleep or waking. Single-trial evoked potentials, defined from 100 ms before to 200 ms after each click, were de-trended (linear) and subject to artifact rejection. In particular, if the signal on any recording channel exceeded +/− 75 [mu]V during an evoked potential, that trial was excluded from further analysis. Average waveforms computed from these single trials were band-pass filtered between 10 and 50 Hz to accentuate middle latency components. This filter was applied both forward and reverse to eliminate phase distortion (Matlab’s filtfilt function).

Sensory gating was measured by comparing the amplitude of wave P1 evoked by the second (test) click of the pair to P1 amplitude evoked by the first (conditioning) click. Specifically, a ratio of magnitudes, the test/conditioning or T/C ratio, was computed to quantify sensory gating. A T/C ratio close to 0 indicates robust suppression (very small test response compared to conditioning response) and a T/C ratio of 1 indicates essentially no sensory gating (test and conditioning responses were comparable in magnitude). In adults, T/C ratios for component P1 range from 0 to well over 1. Ratios < 0.4 are generally interpreted as intact sensory gating, whereas T/C ratios > 0.5 correspond to impaired sensory gating [8].

For each subject, component P1 in response to the conditioning click was defined as the largest positive peak between 50 and 100 ms (P1 latency is generally tens of milliseconds longer in infants than adults [23]). To maintain consistency with previous sensory gating literature, P1 amplitude was measured from the preceding negative trough. Component P1 for the test response was defined as the largest positive peak within +/− 10 ms of conditioning P1 latency. If no peak existed within this window, test P1 amplitude was set to zero and test P1 latency was set to the corresponding conditioning P1 latency.

To assess response suppression, mean T/C ratio for the group was compared to 1 by one-distribution t-test. A ratio < 1 indicates that the test response was significantly smaller than the conditioning response. A putative relationship between conceptual age and sensory gating (as quantified by T/C ratio) was assessed by Pearson’s correlation. Post-hoc analyses were conducted to determine whether other variables might explain the observed changes in sensory gating. For example, because sleep physiology is known to change during the developmental period studied, correlation analyses were conducted between REM sleep variables that are known to change within the 0–6 month window [19] and T/C ratio: latency from sleep onset to first REM period [20], percentage of total sleep time occupied by REM sleep [18], and spectral power in the [theta]-band (4–8 Hz) during REM sleep [24].

RESULTS

Eleven of the 21 infants recorded satisfied the criteria for evoked potential analysis (at least one REM period of >= 15 min and conditioning P1 amplitude of >= 0.5 [mu]V). After artifact rejection, the number of click-pairs remaining to compute averages ranged from 65 to 78 for these infants. Figure 1 illustrates grand- and individual-average evoked potentials. For the group, mean (+/− s.d.) conditioning P1 latency was 86.6 +/− 5.5 ms, mean conditioning amplitude 2.22 +/− 1.73 [mu]V, mean test latency 87.6 +/− 8.5 ms, and mean test amplitude 0.90 +/− 0.85 [mu]V. These values are consistent with previous studies in infants [23]. Mean T/C ratio of 0.57 +/− 0.53 was different from 1 (t = −2.67, df = 10, p < 0.05), indicating that significant suppression of test compared to conditioning response does occur in young infants during REM sleep. However, only six infants exhibited T/C ratios that, according to past studies in adults [8], would be considered intact sensory gating (T/C < 0.4).

Fig.1.

Grand average(a) and individual examples (b,c) ofresponses evoked during REM sleep by conditioning (left) and test( right)

Fig. 1 Opens a popup window Opens a popup window Opens a popup window

Note that the two individual infants shown in Fig. 1 differ in conceptual age by ~4 weeks, and only the older infant exhibits suppression of component P1 in response to the test click (panel C). This finding of increased response suppression (i.e. decreased T/C ratio) with increasing age applied generally to the group (Fig. 2a). In particular, T/C ratio was significantly correlated with conceptual age (Pearson r = −0.77, n = 11, p < 0.01). None of the other evoked potential variables (latencies and amplitudes) were significantly correlated with age for either condition or test stimuli.

Fig. 2.

Relationship between age and select electrophysiological variables. (a) T/C ratio generally decreases (i.e. sensory gating improves) with advancing age. (b) EEG spectral power in the y-band (4^ 8 Hz), associated with REM sleep, generally increases with age.

Fig. 2 Opens a popup window Opens a popup window Opens a popup window

Select sleep variables known to change during the course of early infant development (especially 0–6 months) were investigated as possible explanations for the change in sensory gating with age. Latency to first REM period and percentage of total sleep time spent in REM sleep were not correlated with age or with T/C ratio in our sample. Spectral power in the [theta]-band (4–8 Hz) during REM sleep was significantly correlated with age (r =0.70, n = 11, p < 0.05; Fig. 2b), but not with T/C ratio. A significant correlation between conceptual age and T/C ratio persisted even when [theta] power was controlled for (r = −0.66, n = 11, p < 0.05).

A final attempt to discover sleep variables that might explain the change in sensory gating with age was made by dividing infants into those with so-called intact (T/C < 0.4, n = 6) and impaired (T/C > 0.5, n = 5) sensory gating measures. Two-distribution t-tests between these groups for conceptual age and the selected sleep variables described above uncovered a significant difference only for age (t = 2.70, df = 9, p < 0.05). When group division was based on perinatal environment variables (prenatal exposure to nicotine, bottle vs breast feeding, etc.), no differences in sensory gating measures were found.

DISCUSSION

From the present results it is tempting to conclude that the neural circuits responsible for sensory gating are developing rapidly during the first few months of postnatal life. Such a conclusion, however, would be premature. Given that evoked potential analysis was performed only on responses elicited during REM sleep, it is difficult to rule out the idea that alterations in sleep physiology, especially those that occur around 3 months post-term, could be responsible for the observed improvement in sensory gating with advancing age. To investigate this possibility, we focused on those daytime sleep variables which change within the 0–6 month window, and that are most closely associated with REM sleep [19] (see Materials and Methods). Of these, we replicated a significant linear increase in REM sleep [theta]-power with increasing age [24]. This suggests that even though REM sleep was defined (i.e. staged) in the same manner for all infants, some physiological aspect of REM sleep itself might be changing across the developmental window studied in this investigation. Nevertheless, the correlation between sensory gating and conceptual age was significant even when REM [theta]-power was controlled for. Ultimately, disentangling the separate contributions of age and sleep physiology to the expression of sensory gating will be extremely difficult to achieve without measuring evoked potentials across different states of consciousness. Unfortunately, as discussed in the Introduction, REM sleep is probably the only practical state for measuring this brain function in young infants.

Although not tested directly, it is unlikely that changes in catecholamine levels (including norepinephrine) across the age range studied could explain the relationship between sensory gating and conceptual age, since these levels at are generally low and relatively stable until the end of the first year of life [25]. The cholinergic innervation of the hippocampus, which is crucial for intact sensory gating, exhibits extensive remodeling during pre- and early post-natal development [26] that could more readily explain the observed change in sensory gating patterns at around 3 months of age. Although no prenatal and/or postnatal environmental conditions accounted for the differences in sensory gating measures between babies, a larger sample size would be required to rule out such effects. Before this investigation, P50 sensory gating had not been measured in infants. The present results are consistent with the notion that sensory gating circuits are normally functional during childhood, and further suggest that variability previously observed during waking recordings in children under 10 years probably resulted from stress induced by the experimental situation [10,11].

CONCLUSION

The neural system responsible for P1 sensory gating functions very early in postnatal development during paradoxical sleep. Although applicability to the waking state is yet unproven, sensory gating measures in adults taken during REM sleep are comparable to those taken during waking [16,17]. Therefore, we believe that measurements of sensory gating during REM sleep in infants can be useful to assess the relative contributions of genes and environment to the early development of neural circuitry responsible for sensory inhibition, particularly as it relates to risk for development of schizophrenia.