Abstract

Given the importance of gamma oscillations in normal and disturbed cognition, there has been growing interest in their developmental trajectory. In the current study, age-related changes in sensory cortical gamma were studied using the auditory steady-state response (ASSR), indexing cortical activity entrained to a periodic auditory stimulus. A large sample (n = 188) aged 8–22 years had electroencephalography recording of ASSR during 20-, 30-, and 40-Hz click trains, analyzed for evoked amplitude, phase-locking factor (PLF) and cross-frequency coupling (CFC) with lower frequency oscillations. Both 40-Hz evoked power and PLF increased monotonically from 8 through 16 years, and subsequently decreased toward ages 20–22 years. CFC followed a similar pattern, with strongest age-related modulation of 40-Hz amplitude by the phase of delta oscillations. In contrast, the evoked power, PLF and CFC for the 20- and 30-Hz stimulation were distinct from the 40-Hz condition, with flat or decreasing profiles from childhood to early adulthood. The inverted U-shaped developmental trajectory of gamma oscillations may be consistent with interacting maturational processes—such as increasing fast GABA inhibition that enhances gamma activity and synaptic pruning that decreases gamma activity—that may continue from childhood through to adulthood.

Introduction

Synchronous gamma-band (30–80 Hz) oscillations are an important mechanism for coordinating neural activity in the service of cognitive and sensory processing. Altered gamma oscillations have also been implicated in the pathophysiology of neuropsychiatric disorders such as schizophrenia and autism. Accordingly, investigating the normal developmental trajectory of gamma oscillations is of vital importance to understanding the neurophysiologic underpinnings of normal cognitive development and how pathophysiologic disturbances in gamma activity in neurodevelopmental disorders can give rise to deviations from this trajectory.

Gamma oscillations have been most studied in humans using electroencephalography (EEG), which records at the scalp the summed effect of the synchronous postsynaptic activity in a large number of cortical pyramidal neurons (Nunez and Srinivasan 2005). As various structural and functional processes that support gamma oscillations involve a protracted neurodevelopmental course from childhood through to adulthood, the development of gamma oscillations would be expected to follow a similarly extended trajectory before reaching the mature state. For instance, structural magnetic resonance imaging (MRI) studies find that gray matter thickness in auditory cortical areas (posterior superior temporal gyrus) decrease linearly from childhood into early adulthood (Gogtay et al. 2004). Such macro-level observations of gray matter changes are thought to reflect synaptic pruning over development. Early studies by Huttenlocher and Dabholkar (1997) suggested that auditory and visual regions completed pruning processes in early adolescence while prefrontal cortex had a more extended course into mid-adolescence. However, these interpretations were limited by sparse sampling over the adolescence and adulthood periods, and more recent studies with denser sampling over this range have indicated that synaptic elimination continues beyond adolescence into the third decade of life (Petanjek et al. 2011). Such structural changes are accompanied by extensive changes in components of GABA neurotransmission (Beneyto and Lewis 2011) that could critically change the functional capacity to produce gamma oscillations. Parvalbumin (PV) fast-spiking cells are a subclass of GABA interneurons that, given their nonadapting firing and postsynaptic GABA-A receptors that possess fast kinetics, can support sustained, high-frequency firing. These properties allow PV fast-spiking interneurons to play their critical role in the temporal regulation of gamma oscillations through feedback inhibitory coupling with pyramidal cells (Bartos et al. 2007; Buzsáki and Wang 2012). Interestingly, during development, GABA-A receptors shift their subunit composition from alpha-2 to alpha-1 subtypes, resulting in a shift from slower to faster inhibitory decay kinetics (Hashimoto et al. 2009). Since the frequency of gamma oscillations depends critically on the decay time course for inhibition (Bartos et al. 2007), such a shift could increasingly provide the means to support higher frequency oscillations. This developmental shift also follows a protracted temporal course, beginning postnatally and continuing through to adulthood (Hashimoto et al. 2009). Thus, both structural refinements and molecular changes occur with a protracted course but exactly how these interact and give rise to physiologic changes over development in the form of gamma oscillations requires explicit investigation.

Steady-state responses to trains of periodic stimuli have been extensively used to examine oscillations in auditory, visual, and somatosensory modalities (Colon et al. 2012) and have the advantage of a high signal-to-noise ratio (Vialatte et al. 2010). Auditory steady-state responses (ASSRs) have been used to study the development of gamma oscillations over a wide range of ages (Maurizi et al. 1990; Rojas et al. 2006; Poulsen et al. 2009) and are sensitive to selective gamma oscillatory disturbances in schizophrenia (Kwon et al. 1999; Light et al. 2006; Spencer et al. 2008; Brenner et al. 2009; Krishnan et al. 2009; Kömek et al. 2012) and pharmacologic modulation (Vohs et al. 2010; Kömek et al. 2012). Developmental studies of gamma oscillations have found that infants and children (3 months to 6 years) exhibit poor ASSR (Stapells et al. 1988), whereas older children (5–8 years) sustain better gamma responses (Maurizi et al. 1990), with 10-year olds progressing to show comparatively higher gamma responses at 11.5 years (Poulsen, Picton, and Paus 2009). A study spanning a much broader age range (5–52 years), found increasing ASSR from childhood through adolescence, with plateauing from early adulthood (Rojas et al. 2006). However, the relatively sparse sampling at the younger ages (30 subjects younger than 20 years old) precluded a detailed characterization beyond the first-order trend of increases toward adulthood. In the current study, we employed the ASSR paradigm with a dense sampling of the 8–22 years age range to investigate the precise trajectory by which gamma activity emerges and evolves over this critical period of development.

Prior developmental studies using ASSR have focused on magnitude measures (amplitude/power) of gamma oscillations. Another important measure of oscillatory activity is phase-locking factor (PLF), which is a measure of the consistency with which neural activity is phase locked to the stimulus across trials and is analytically independent of the oscillation magnitude. Another important aspect of gamma oscillatory dynamics that has gained increasing attention is cross-frequency coupling (CFC), commonly indexed as the modulation of gamma amplitude by the phase of slower rhythms. Modulation by theta rhythms is most often reported, with critical implications for cognitive processing (Jensen and Colgin 2007) including working memory (Sauseng et al. 2009; Axmacher et al. 2010; Fujisawa and Buzsáki 2011) and sensory selection (Schroeder and Lakatos 2009). Interestingly, theta–gamma CFC in hippocampus is dependent on fast inhibition onto PV interneurons (Wulff et al. 2009) suggesting that the developmental trajectory CFC may track that of gamma oscillatory power which also depends on fast inhibitory kinetics (Bartos et al. 2007). The phases at other lower frequency oscillations, including at delta (1–3 Hz) and alpha (8–12 Hz) bands, have also been noted to modulate gamma band amplitudes (Canolty and Knight 2010).

The current study investigated the developmental trajectory of gamma oscillations using the ASSR. It is the first study with sufficiently dense sampling to permit a detailed characterization of ASSR gamma over the critical developmental age range. We evaluated ASSR gamma amplitude in 188 subjects spanning the age range 8–22 years old, binned in increments of 3 years [late-childhood; 8–10 years, early-adolescence: 11–13 years, mid-adolescence: 14–16 years, late-adolescence: 17–19 years, early adulthood (The age bin definitions correspond to ranges that are commonly used in the literature. However, the particular age limits are arbitrary and definitions can vary widely, in particular, in distinguishing late-adolescence from early adulthood given the extended maturational course of some cognitive and neural processes.): 20–22 years]. This is also the first study to examine the maturation process of PLF and CFC of gamma oscillations.

Materials and Methods

Participants

One hundred eighty-eight participants (age range 8–22 years, M = 14.5, SD = 4.4) were recruited from the greater Allegheny County to participate in this study. Of these, 7 participants were excluded based on participant withdrawal, ineligibility, or technical issues. Thus, data from 181 participants were entered into the analyses (age range 8–22 years, M = 14.5, SD = 4.4). Written informed consent was obtained prior to testing in accordance with the Institutional Review Board at the University of Pittsburgh. Participants and accompanying legal guardians were monetarily compensated for their participation. Recruitment numbers were well balanced across the age range but with greater sampling for the 8- to 10-year-old subjects (Table 1) in order to facilitate future longitudinal follow-up studies. Age bins in increments of 3 years (8–10, 11–13, 14–16, 17–19, 20–22 years) were matched for race, IQ, and gender (these groups are henceforth referred to as age bins 1 through 5, respectively). Potential participants were excluded using the MINI (Sheehan et al. 1998) for having a history of DSM IV Axis I or mental retardation diagnosis, or a first-degree relative with a history of psychosis. Additional demographic information can be found in Table 1.

Table 1

Demographic and behavioral summary for current sample (N = 188)

Variable Age bin (in years)
 
Count 1 (8–10) 2 (11–13) 3 (14–16) 4 (17–19) 5 (20–22) 
n 44 32 37 36 32 
Female 21 16 19 18 15 
Left-handed 
Ethnicity 
 African American 11 12 
 Asian 
 Caucasian 29 21 23 25 24 
 Other/not reported 
Withdrawn 
Mean (SD)      
WASI IQ 107.5 (13.7) 105.1 (11.7) 98.2 (13.1) 103.3 (11.5) 107.2 (14.2) 
Average SES 41.6 (15.6) 44.4 (12.7) 41.4 (14.0) 44.3 (13.9) 47.2 (13.8) 
Standard Trial False Alarm Rate 0.07 (0.10) 0.03 (0.07) 0.03 (0.04) 0.03 (0.07) 0.04 (0.08) 
Standard Trial Reaction Time 245 (125) 199 (130) 237 (134) 219 (157) 260 (97) 
Trial Segment Count (Hz) 
 40 67.0 (24.7) 87.9 (17.5) 89.5 (18.6) 87.5 (25.8) 85.9 (26.3) 
 30 66.3 (27.4) 81.4 (22.6) 89.8 (20.1) 86.3 (29.5) 90.6 (19.7) 
 20 67.7 (27.5) 85.8 (20.1) 90.9 (16.0) 89.2 (22.2) 89.5 (17.3) 
Variable Age bin (in years)
 
Count 1 (8–10) 2 (11–13) 3 (14–16) 4 (17–19) 5 (20–22) 
n 44 32 37 36 32 
Female 21 16 19 18 15 
Left-handed 
Ethnicity 
 African American 11 12 
 Asian 
 Caucasian 29 21 23 25 24 
 Other/not reported 
Withdrawn 
Mean (SD)      
WASI IQ 107.5 (13.7) 105.1 (11.7) 98.2 (13.1) 103.3 (11.5) 107.2 (14.2) 
Average SES 41.6 (15.6) 44.4 (12.7) 41.4 (14.0) 44.3 (13.9) 47.2 (13.8) 
Standard Trial False Alarm Rate 0.07 (0.10) 0.03 (0.07) 0.03 (0.04) 0.03 (0.07) 0.04 (0.08) 
Standard Trial Reaction Time 245 (125) 199 (130) 237 (134) 219 (157) 260 (97) 
Trial Segment Count (Hz) 
 40 67.0 (24.7) 87.9 (17.5) 89.5 (18.6) 87.5 (25.8) 85.9 (26.3) 
 30 66.3 (27.4) 81.4 (22.6) 89.8 (20.1) 86.3 (29.5) 90.6 (19.7) 
 20 67.7 (27.5) 85.8 (20.1) 90.9 (16.0) 89.2 (22.2) 89.5 (17.3) 

Note: Average SES represents a composite SES score using the Hollingshead Scale. WASI IQ represents an age normalized IQ score derived from the Wechsler Abbreviated Scale for Intelligence test. Participant demographics do not reflect withdrawn participants.

Task

Participants were seated ∼80 cm from an LCD computer monitor used to present visual stimuli and wore ER-3A insert earphones (Etymotic Research, Elks Grove, IL, USA) for auditory stimuli. All stimuli were presented using E-Prime software (Psychological Software Tools, Pittsburgh, PA, USA). Click trains of 500-ms duration were presented binaurally at 65 ± 5 dB. The click train repetition frequencies were 20, 30, or 40 Hz and presented in the context of an auditory oddball paradigm to ensure participant attention to the stimuli. Standard stimuli were click trains with individual clicks being 1-kHz carrier frequency whereas Oddball stimuli were click trains with clicks of 2-kHz carrier frequency. During click train presentation and for 200 ms after click train cessation, the screen remained blank (black). Participants were then prompted by the appearance of a central fixation cross to respond by button press with either their left index finger for Standard stimuli (110 trials per block) or their right index finger for Oddball stimuli (10 trials per block). There was one block for each click repetition frequency (20, 30, and 40 Hz) for a total of 3 blocks, with block orderings counterbalanced across subjects. Behavioral analyses were conducted to confirm attention to task but only correct Standard trials were submitted for EEG analyses. Neither behavioral nor EEG analyses were carried out for oddball trials due to insufficient trial counts.

EEG Acquisition and Preprocessing

EEG sessions were conducted in an electrically shielded, sound-attenuated room lit with a low-level ambient light. EEG data were collected using a 128-channel Geodesic Sensor Net and Netstation software (EGI, Eugene, OR, USA) sampling at 250 Hz and referenced to a common reference (Cz). Online filtering was applied using a 0.01- to 100-Hz elliptical bandpass hardware filter, and electrode impedances were maintained at or below 50 kΩ. Epochs were defined as −500 to 1000 ms relative to stimulus onset adjusted to a −350 to −150-ms prestimulus baseline. Only correct Standard trials were included and epochs containing artifacts were excluded (greater than ±100 µV amplitude within epoch or a consecutive sample difference of 60 µV). Segments identified by these criteria were visually inspected prior to rejection. Ocular and ECG artifacts were removed with ICA-based detection and correction methods (EEGLab; Delorme and Makeig 2004). Data were filtered off-line with a 1- to 100-Hz bandpass short infinite impulse response filter. The resulting data were submitted to final review using the above amplitude criteria and for 60-Hz line noise, with bad channel data replaced by interpolation. Data were re-referenced to average reference. As a final preprocessing step, a principle component analysis (PCA) was performed. This additional “denoising” step was motivated by ASSR being maximal at the frontocentral electrode FCz, but distributed across surrounding electrodes. PCA was run on 14 electrodes centered on FCz (EGI system electrodes 5, 6, 7, 12, 13, 20, 30, 31, 80, 105, 106, 112, 118, 129), with the first principle component consistently capturing the ASSR. Accordingly, the time series derived from back-projecting the first component on to FCz was used for all subsequent spectral analyses.

Spectral Analyses

Data were transformed by a complex Morlet wavelet transform as basis for all spectral analytic measures. Morlet wavelets are a Gaussian-shaped sinusoidal function, thus yielding frequency-specific information in a time-specific manner: w(t,f0)=A×exp(t2/2σt2)×exp(2iπf0t), where A is a normalizing constant, σt gives the spread of the Gaussian, and f0 is the frequency of the sinusoid, with σf = 1/2πσt. The transform of the EEG data, then, was a convolution of the wavelet with the EEG data d(t): W(t, f0) = w(t, f0) × d(t). The modulus and argument of the resulting complex value in polar form are estimates of the amplitude and phase values for the particular time point and frequency band. The ratio f0/σf is constant and defines the frequency versus time resolution (higher values favoring higher frequency resolution). A value of 10 was used for phase-sensitive measures (evoked activity and PLF) at the entrainment frequencies. This value emphasized frequency resolution with minor loss of time resolution at such relatively high frequencies. For measures involving lower frequencies (induced activity and cross-frequency coupling), time resolution was emphasized by using a value of 5.

The evoked amplitudes were calculated on wavelet transforms of the averaged EEG data (see Supplementary Fig. 1 for ERP averages), whereas induced amplitudes were calculated as averages of wavelet transformed data on a per trial basis. PLF, the phase consistency with respect to the click trains, was calculated as the modulus of the vector average of the normalized transformed data: 1/pΣ((a+bi)(a2+b2)1/2) where the summation is over the number of data points p, and a+bi is each wavelet transformed data point. PLF values are bounded from 0 to 1, and being equivalent to 1 minus the circular variance of phases, a value of 1 represents no variance, that is, identical phase across trials. The unit normalization of magnitude ensures that PLF yields an independent, complementary measure to evoked amplitude.

CFC analysis addressed the degree of dependence of amplitude on the phase of lower frequencies, that is, phase-amplitude coupling. This can be quantified by measuring the distance between the phase and amplitude distributions (Tort et al. 2010). By binning amplitude values at a frequency on the basis of the concurrent phase of a lower frequency, the latter can be said to modulate the former if its distribution is not uniform. This approach was refined in order to capture coupling even in fast dynamics during short time windows that could otherwise be compromised by uneven sampling of lower frequency phase values. To this end, both phase and amplitude were binned and coupling measured in terms of their mutual information.

Statistical Analysis

All statistical differences across age bins (i.e., 8–10, 11–13, 14–16, 17–19, 20–22 years) were assessed through one-way analyses of covariance (ANCOVA) controlling for gender, handedness, IQ, ethnicity, and socioeconomic status (SES) as assessed on the Hollingshead scale (Hollingshead 1975). SES data were incomplete for 5 participants; data for these participants were imputed, derived from the mean SES for their respective age bins. ANCOVAs were conducted for Standard trial reaction times and error rates. And for each click frequency, separate ANCOVAs were run for evoked amplitudes, induced amplitudes, PLF and CFC. As CFC analyses identified delta–gamma coupling, the total delta and gamma amplitudes were also evaluated separately. Owing to heteroscedasticity of delta amplitudes, a weighted least squares approach was taken where group weights were calculated as the inverse of the variance of residuals from the unweighted model. Note that 3 participants had no viable trials in 1 of the 3 conditions (20 Hz: one from 20–22 year age bin 5; 30 Hz: one each from age bins 11–13 and 17–19 years); therefore, analyses of those conditions reflect the remaining participants' data. Standard trial error rate was arcsine transformed to correct for significant differences in variance across age bins. For evoked and induced amplitudes and PLF, means were extracted from 225- to 525-ms poststimulus onset to avoid differential influence from the initial mid-latency response (50- to 150-ms postonset). Age-related trajectories were assessed using polynomial trend analyses with post hoc t-tests carried out to assess group differences. Lastly, we characterized relationships between measures with partial correlations controlling for age. Post hoc t-tests and partial correlations were evaluated at a Bonferonni corrected α = 0.05 to address possible inflation of type I error.

Results

Behavioral Performance

Error rates were significantly higher for the youngest age bin, F4,169 = 2.55, P = 0.041, ηp2 = 0.06, and followed a negative quadratic trend with age, F1,169 = 6.41, P = 0.012, ηp2 = 0.04. However, pairwise contrasts only found a trend-level difference between the first and third age bins, t(79) = 2.72, P = 0.058. Reaction times showed no clearly discernable pattern by age, and no significant differences were observed across age bins. See Table 1 for performance summaries.

Evoked Amplitude

Analyses of evoked amplitude revealed significant main effects of age bin for 20- and 40-Hz steady-state responses, F4,168 = 3.32, P = 0.012, ηp2 = 0.07; F4,169 = 5.83, P < 0.0005, ηp2 = 0.12, respectively (Fig. 3 and Supplementary Fig. 5; see Supplementary Fig. 2 for 20- and 30-Hz time–frequency plots). Polynomial trend analyses of 20-Hz responses across age bin revealed a significant negative linear trend reflecting decreases in evoked response with age, F1,168 = 10.71, bLinear = −5.21, SELinear = 1.59, P = 0.001, ηp2 = 0.06. Consistent with this trend, mean amplitude significantly decreased from the first to the last age bin, t(73) = 3.57, P = 0.005.

In contrast, evoked amplitude for the 40-Hz response increased from age bins 1–3 before decreasing again from age bins 3–5 (Figs 1A and 3 and Supplementary Fig. 5). Polynomial contrasts confirmed this pattern to be statistically significant with positive linear and strongly significant negative quadratic trends, linear: F1,169 = 5.39, bLinear = 2.61, SELinear = 1.12, P = 0.021, ηp2 = 0.03; quadratic: F1,169 = 15.35, bQuadratic = −4.51, SEQuadratic = 1.15, P < 0.0005, ηp2 = 0.08. These patterns reflect significant increases between the first age bin and bins 2, 3, and 4: t(74) = −3.01, P = 0.030; t(79) = −4.47, P < 0.0005; and t(78) = −3.63, P = 0.004, respectively. The 40-Hz trajectory is especially noteworthy given that its pattern is distinct from the 20-Hz trajectory.

Figure 1.

Time–frequency plots displaying age-related changes in evoked amplitude (A) and phase-locking factor (B) in response to 40-Hz click trains. Stimulus onset is at 0 ms (green dashed line) and offset is at 500 ms (red dashed line). A marked increase in evoked amplitude and phase-locking centered around 40 Hz is apparent during stimulus presentation for all age bins with the largest response for both measures observed at 14–16 years. Insets above each time–frequency plot in (A) show the scalp topography of gamma amplitude (anterior electrodes in downward direction), demonstrating consistent distributions across the age bins, centered on electrode FCz (outlined by black circle).

Figure 1.

Time–frequency plots displaying age-related changes in evoked amplitude (A) and phase-locking factor (B) in response to 40-Hz click trains. Stimulus onset is at 0 ms (green dashed line) and offset is at 500 ms (red dashed line). A marked increase in evoked amplitude and phase-locking centered around 40 Hz is apparent during stimulus presentation for all age bins with the largest response for both measures observed at 14–16 years. Insets above each time–frequency plot in (A) show the scalp topography of gamma amplitude (anterior electrodes in downward direction), demonstrating consistent distributions across the age bins, centered on electrode FCz (outlined by black circle).

Figure 3.

Summary plots of age-related changes in evoked amplitude, induced amplitude, PLF, and CFC (see Supplementary Fig. 5 for corresponding scatterplots). All measures for the 40-Hz condition follow an inverted-U trajectory except for induced amplitude which follows a mirror opposite trend. These patterns are specific to 40 Hz, with responses to 30- and 20-Hz stimuli showing a flat or decreasing trend with age.

Figure 3.

Summary plots of age-related changes in evoked amplitude, induced amplitude, PLF, and CFC (see Supplementary Fig. 5 for corresponding scatterplots). All measures for the 40-Hz condition follow an inverted-U trajectory except for induced amplitude which follows a mirror opposite trend. These patterns are specific to 40 Hz, with responses to 30- and 20-Hz stimuli showing a flat or decreasing trend with age.

To rule out any developmental trends due to differential build-up or attenuation of the ASSR during the course of the experiment, we compared responses from the first and second halves of the experiment. However, there were no significant differences between the first and second halves (Supplementary Fig. 7).

Induced Amplitude

Only induced activity at 40-Hz differed significantly across age bins, F4,169 = 4.27, P = 0.003, ηp2 = 0.09 (see Fig. 3, Supplementary Figs 4 and 5). Polynomial trends showed the trajectory of the 40-Hz response to be the inverse of the evoked responses. With increases in age, induced 40-Hz amplitudes demonstrated significant negative linear and positive quadratic trends indicating that across development induced 40-Hz activity declined decreased at a decreasing rate with a slight increase for the final age bin, linear: F1,169 = 5.50, bLinear = −1.38, SELinear = 0.59, P = 0.020, ηp2 = 0.03; quadratic: F1,169 = 8.95, bQuadratic = 1.80, SEQuadratic = 0.60, P = 0.003, ηp2 = 0.05. Consistent with an inverse pattern to the evoked response, this trend was driven by decreases from age bin 1 to bins 2, 3, and 4: t(74) = 2.97, P = 0.035; t(79) = 3.26, P = 0.013; and t(78) = 3.59, P = 0.004, respectively. Note that the 40-Hz induced amplitudes were negative indicating a significant reduction from baseline levels of nonphase-locked activity; however, since the absolute magnitudes of the reductions did not quantitatively mirror the increases in evoked amplitudes, the evoked activity could not be entirely explained by a phase realignment of ongoing cortical gamma-band activity.

Phase-Locking Factor

Measures of phase-locking mirrored the results for the evoked 40-Hz response, F4,169 = 10.15, P < 0.0005, ηp2 = 0.19 (see Figs 1B and 3 and Supplementary Fig. 5); however, neither the 20- nor 30-Hz phase locking significantly differed across age bins (Supplementary Fig. 2B, D). Similarly, trend analyses of 40-Hz phase-locking found significant positive linear and negative quadratic trends reflecting a general increase in PLF from age bins 1–3 and a decrease from age bins 3–5, linear: F1,169 = 7.64, bLinear = 0.06, SELinear = 0.02, P = 0.006, ηp2 = 0.04; quadratic: F1,169 = 26.34, bQuadratic = −0.12, SEQuadratic = 0.02, P < 0.0005, ηp2 = 0.13. Post hoc analyses revealed these trends to reflect significant increases from the first age bin to all other age bins, 2 through 5: t(74) = −3.82, P = 0.002; t(79) = −6.22, P < 0.0005; t(78) = −4.06, P = 0.001; and t(74) = −3.03, P = 0.026, respectively, as well as a significant decrease from age bin 3–5, t(74) = 2.94, P = 0.040.

Cross-Frequency Coupling

Analyses of CFC strength revealed a significant phase-amplitude relationship between high delta (2–4 Hz) phase and 40-Hz response amplitude (Fig. 3 and Supplementary Fig. 5). Moreover, this coupling strength differed across age as well, F4,164 = 4.34, P = 0.002, ηp2 = 0.10. Group differences followed similar linear and quadratic trends to other 40-Hz response measures; however, the linear term did not quite achieve statistical significance. Linear: F1,163 = 3.49, bLinear = 0.01, SELinear = 0.005, P = 0.063, ηp2 = 0.02; quadratic: F1,163 = 11.47, bQuadratic = −0.02, SEQuadratic = 0.005, P = 0.001, ηp2 = 0.07. With post hoc testing, only the difference between the first and third age bins was significant, t(76) = −3.88, P = 0.001. Interestingly, this increase in CFC strength arose in the context of decreasing monotonic trend in total delta amplitude, F1,169 = 2.16, P = 0.076, ηp2 = 0.05; linear: F1,169 = 7.69, bLinear = −12.62, SELinear = 4.44, P = 0.006, ηp2 = 0.04. There were no significant effects for the 20- and 30-Hz conditions (Supplementary Fig. 3A, B, respectively).

Correlation Analyses

As evoked oscillations may be due, in part, to phase resetting of ongoing oscillations, we evaluated for negative correlations between evoked and induced amplitudes. Partialling out the linear and quadratic trend effects of age, evoked 40-Hz amplitude correlated negatively with induced 40-Hz amplitude (partial correlation pr = −0.83, df = 170, P < 0.0005). We also evaluated the correlation between 40-Hz evoked amplitude and PLF, finding evidence for a close relationship (pr = 0.87, df = 170, P < 0.0005). Finally, as CFC measures of coupling may be influenced by the amplitudes of the coupled oscillations, total (sum of induced and evoked activity) delta and gamma amplitudes were correlated with CFC during 40-Hz stimulation. There were significant positive correlations between CFC and delta (pr = 0.29, df = 177, P < 0.0005) and gamma (pr = 0.70, df = 177, P < 0.0005) amplitudes.

Discussion

This study investigated age-related changes in multiple aspects of gamma oscillatory activity as indexed by ASSR in healthy subjects 8–22 years old. The common pattern across the measures, including evoked power, PLF and CFC, was an inverted-U trajectory, with monotonic increases from age childhood (8–10 years) to mid-adolescence (14–16 years), with subsequent decreases towards adulthood (20–22 years) (Fig. 3). This trajectory stood in stark contrast to the entrained responses to the 20-Hz and 30-Hz stimuli which were either uniformly monotonic decreasing or showed no appreciable changes across the same age span. While the initial increases in gamma activity are consistent with the findings of prior studies that show similar trends, this is the first study with sufficiently dense sampling of this age range to capture the decreases in gamma activity later in development. The inverted-U trajectory is suggestive of 2 underlying developmental processes with opposing effects, such as increases in alpha-1 GABA receptors promoting increases in gamma activity with age that are later offset by synaptic pruning that decreases pyramidal cell excitation and, as a consequence, gamma activity.

Figure 2.

Age-related changes in phase-amplitude cross-frequency coupling (CFC) in response to 40-Hz click trains. Shown are t-value maps of CFC compared with prestimulus baseline. The x-axis shows the lower modulating frequency used for phase estimation whereas the y-axis shows the higher modulated frequency used for amplitude estimation. The amplitude of the 40-Hz response is coupled to the phase of frequencies between 2 and 6 Hz. The frequency specificity of coupling appears consistent across development but the strength of coupling is greatest at 14–16 years.

Figure 2.

Age-related changes in phase-amplitude cross-frequency coupling (CFC) in response to 40-Hz click trains. Shown are t-value maps of CFC compared with prestimulus baseline. The x-axis shows the lower modulating frequency used for phase estimation whereas the y-axis shows the higher modulated frequency used for amplitude estimation. The amplitude of the 40-Hz response is coupled to the phase of frequencies between 2 and 6 Hz. The frequency specificity of coupling appears consistent across development but the strength of coupling is greatest at 14–16 years.

Prior studies of ASSR gamma oscillations have characterized the general first-order developmental trend towards increased gamma, starting with poor ASSR gamma in infants and children (3 months to 6 years) (Stapells et al. 1988) and increases in childhood (Maurizi et al. 1990; Poulsen et al. 2009) that appear to continue monotonically, plateauing in early adulthood (Rojas et al. 2006). In general agreement with these prior reports, we found an initial frequency-specific pattern of increases in evoked gamma power. This increase is consistent with parallel changes in the subunit composition of cortical PV fast-spiking interneuron GABA-A receptors from alpha-2 to alpha-1 subtype that results in a shift from slower to the faster inhibitory decay kinetics (Hashimoto et al. 2009) that are important to the higher frequency oscillations in the gamma-band range (Bartos et al. 2007). This shift follows a protracted developmental course, beginning postnatally and continuing through to adulthood (Hashimoto et al. 2009). Similar observations have been made in hippocampus where PV neurons undergo developmental shifts toward producing stronger, more precise inhibitory currents with faster kinetics that characterize the more mature state and support high-frequency gamma oscillations (Doischer et al. 2008). Interestingly, an integrated MEG/computational modeling study of the ASSR in schizophrenia (Vierling-Claassen et al. 2008) showed that the decreases in 40-Hz power and increases in 20-Hz power for patients could be explained precisely by slower kinetics in GABA neurotransmission through its shifting of the peak frequency to the lower portion of the spectrum. Future developmental studies could use finer sampling of frequencies (cf., Krishnan et al. 2009) to more explicitly demonstrate such age-related shifts.

If GABA-A-receptor kinetics were the only rate-limiting factor, one would expect gamma oscillatory activity to follow a similar monotonic increasing trajectory, perhaps plateauing by early adulthood. Rojas et al. (2006) reported just such a trend but while the study may have been adequately powered (n = 69, ages 5–52 years) to detect this first-order pattern, the sampling of younger subjects was relatively sparse (n = 30 subjects <20 years old) precluding a more detailed characterization. Our much denser sampling of the same age range permitted a more refined characterization, revealing the decreases in gamma activity beyond mid-adolescence toward early adulthood. (While the Rojas et al. (2006) study demonstrated a first-order monotonic increase toward adulthood, a re-examination of their data (provided courtesy of Dr Rojas) revealed a trend toward an inverted-U trajectory resembling the findings of the current study. See Supplementary Data.) This nonmonotonic pattern suggests that in addition to factors such as GABA-A-receptor kinetics, other important developmental processes help to shape the trajectory of gamma activity.

Given that oscillatory activity measurable at the scalp results largely from postsynaptic potentials in the apical dendrites of pyramidal cells, synaptic pruning is one such developmental process that could act counter to that of evolving GABA-A-receptor kinetics, perhaps eventually enough to cause decreases in measured gamma activity. Interestingly, synaptic pruning occurs more in supragranular than infragranular layers (Rakic et al. 1986; Bourgeois et al. 1994; Anderson et al. 1995), precisely the layers where gamma oscillations are more prominent (Chrobak and Buzsáki 1998; Quilichini et al. 2010; Buffalo et al. 2011; Spaak et al. 2012). Gamma oscillations are also thought to primarily reflect local circuitry dynamics in the superficial layers and so it is notable that the extensive pruning in these layers during adolescence occurs in local intrinsic circuitry as opposed to synapses associated with long-range associational axons (Woo et al. 1997). Interestingly, this pruning during adolescence does not appear to depend on the functional immaturity of synapses (Gonzalez-Burgos et al. 2008), as is the case earlier in development (Mirnics et al. 2001). Further, synaptic elimination occurs beyond adolescence into the third decade of life (Jacobs et al. 1997; Petanjek et al. 2011). Consistent with pruning processes that continue into adulthood, the temporal cortex peaks relatively late in gray-matter thickness (∼17 years) (Giedd et al. 1999) and the posterior superior temporal gyrus undergoes linear decreases in gray-matter thickness that continue through adolescence into early adulthood (Gogtay et al. 2004). Thus, the temporal course of functional refinements due to GABA-A-receptor kinetics and structural refinements mediated by synaptic pruning may together shape the resultant developmental trajectory of gamma oscillatory activity. Of note, both these factors could also explain the monotonic decreasing pattern for the 20-Hz ASSR, since the faster receptor kinetics would be progressively less optimal for responses at this lower frequency (Doischer et al. 2008; Vierling-Claassen et al. 2008) and pruning processes could generally decrease the magnitude of any type of measurable EEG signal, including the 20-Hz ASSR.

In addition to changes in GABA-A-receptor kinetics and synaptic pruning processes, other developmental processes could also impact the development of gamma oscillations. While there may be a relatively early functional maturation of excitatory inputs onto pyramidal cells (Gonzalez-Burgos et al. 2008), there are substantial changes in glutamatergic receptors on fast-spiking interneurons during adolescence (Wang and Gao 2009, 2010). The presence of NMDA receptors on the majority of fast-spiking interneurons prior to adolescence reduces to a minority by adulthood (Bitanihirwe et al. 2009; Wang and Gao 2009) while calcium-permeable AMPA receptors increases over the same period (Wang and Gao 2010). The decreases in NMDA-mediated excitation on fast-spiking interneurons could result in a relative disinhibition of pyramidal cells thereby contributing to increases in gamma activity, as would be suggested by increases in ASSR gamma after acute administration of ketamine, an NMDA receptor antagonist (Plourde et al. 1997) and findings that excessive NMDA conductance in fast-spiking cells can decrease gamma power (Rotaru et al. 2011). The sustaining of AMPA-mediated excitation on the other hand could provide the necessary much faster kinetics to support high-frequency oscillations (Johnston and Wu 1997; Compte et al. 2000; Rotaru et al. 2011; Buzsáki and Wang 2012).

Gamma oscillatory development could also be affected by another aspect of GABA neurotransmission, namely, the cannabinoid system, which undergoes significant changes during adolescence. Stimulation of type 1 cannabinoid receptors (CB1R) on cholecystokinin (CCK)-containing interneurons suppress their output (Katona and Freund 2008) and modulation of these receptors can substantially decrease gamma oscillations (Holderith et al. 2011), although the particular effect can depend on the specific circuit (Morgan et al. 2008). CB1R density significantly increases in layers deep 3 and 4 during adolescence (Eggan et al. 2010), possibly contributing to development of gamma oscillations (which predominantly arise from layer 3 (Maier et al. 2010; Buffalo et al. 2011; Spaak et al. 2012) as well as to the particular vulnerability to cannabis use during this period (Ehrenreich et al. 1999; Pope et al. 2003; Moore et al. 2007). Interestingly, a recent rodent study showed that adolescent but not adult exposure to cannabis decreased gamma activity in association with working memory impairments (Raver et al. 2013).

The maturation of cognitive processes such as response inhibition and cognitive control [which is associated with gamma activity (Cho et al. 2006; Kieffaber and Cho 2010; Minzenberg et al. 2010)], reaches adult levels during mid-late-adolescence (Luna et al. 2004), paralleling our findings of increases in evoked gamma until mid-adolescence. The later decreases in gamma activity toward adulthood may be consistent with structural refinements via synaptic pruning that serve to enhance cortical efficiency, as activity in the gamma band is more metabolically demanding than lower frequency activity (Mukamel et al. 2005; Niessing et al. 2005). In fact, fMRI studies using BOLD imaging, which indirectly indexes metabolic demand and correlates strongly with gamma activity (Niessing et al. 2005; Magri et al. 2012), have shown an inverted-U pattern over development in association with working memory and cognitive control tasks, with higher activation levels in executive control regions during adolescence compared with childhood and adulthood (Luna et al. 2001; Ciesielski et al. 2006; Scherf et al. 2006).

Induced gamma activity followed a mirror opposite trend to that of evoked gamma, with values decreasing below baseline, consistent with ongoing prestimulus gamma oscillations becoming phase-aligned to the stimulus upon its presentation. Also supportive of this idea are the high negative correlations between induced and evoked activity, even after partialling out the effects of age. However, the absolute values of evoked gamma were higher than induced gamma suggesting that over and above any such phase realignment, a portion of evoked activity was also due to de novo additional excitation of pyramidal cells due to the auditory stimulus.

The decreases in induced gamma activity in the current study during steady-state evoked responses can be contrasted with the increases typically found in association with tasks that engage gamma oscillations in association with processes that are not generally phase-locked to stimulus onset, such as perceptual decision-making. In a developmental study of gamma oscillations using a Gestalt perception task, Uhlhaas et al. (2009) found that induced gamma in parietal scalp regions had a generally monotonic increasing trend from childhood to adulthood that was interrupted by a relative dip during late-adolescence. These age-related changes more closely resemble that observed for evoked gamma responses in the current study. This suggests that from childhood to mid-adolescence, there may a generic increased ability of cortical circuits to engage in gamma oscillatory activity that can be indexed through different means, including evoked activity as elicited by periodic stimuli or induced activity as elicited by perceptual decision-making. During late-adolescence through to adulthood, the specific pattern of age-related changes may depend on the specific task or region investigated. Future studies could further elucidate the dependencies of age-related changes on the specific paradigm or anatomic region examined to index gamma oscillations.

In contrast with induced and evoked averages which both depend on the amplitude of oscillations, PLF is an index of phase consistency across trials that analytically do not have any dependency on amplitude. PLF showed a similar profile to that of evoked activity, displaying an inverted-U trajectory with a peak at mid-adolescence (14–16 years). The earlier increases in PLF could be understood in terms of enhanced temporal precision of action potentials in more mature PV interneurons, with 4-fold reductions in the coefficient of variation of basket cell conduction velocities observed for mature versus young rodents (Doischer et al. 2008). This enhanced precision, however, is inconsistent with the decreases in PLF observed beyond mid-adolescence. Rather, these later decreases may be attributable to the PLF estimates being affected by signal-to-noise issues. While the analytic determination of PLF depends solely on the phase of the oscillation and not the amplitude, the very estimate of phase may be affected by the relative amplitude of the oscillation compared with the background noise. Thus, assuming constant noise levels, higher amplitudes will tend to yield higher PLF estimates even if the actual PLF remains constant (as opposed to ideal situation of a complete absence of noise in which case amplitude would not affect PLF estimates). Consistent with this idea, we found strong correlations between PLF and evoked power. In contrast to the 40-Hz condition, the evoked amplitude for 20 Hz showed a significant downward trend with age while 20-Hz PLF showed no effect of age, demonstrating that, while correlated, amplitude and phase estimates are not simply redundant indices. Thus, the observed age-related changes in PLF are likely a function of both underlying enhancements in the temporal precision of neuronal activity as well as measurement properties of PLF estimates in non-noise-free time series.

CFC analyses showed significant delta phase modulation of gamma amplitudes in the 40-Hz condition which followed a similar age-related inverted-U trajectory to 40-Hz evoked power and PLF, increasing from childhood, peaking in mid-adolescence, and decreasing toward adulthood. The initial CFC increases may also be attributable to developmental changes in GABA neurotransmission. Wulff et al. (2009) reported on the critical role of fast synaptic inhibition of PV fast-spiking interneurons in hippocampal theta–gamma CFC using experimental studies in rodents and a computational model, showing that ablations of inhibition of PV cells reduced both theta and its coupling to gamma. Delta–gamma coupling has also been observed in the primary visual cortex (Whittingstall and Logothetis 2009; Ito et al. 2013) and modeled computationally as slow shifts (delta frequency) in cortical excitability that modulate the engagement of excitatory–inhibitory loops that produce gamma oscillations (Mazzoni et al. 2010). However, it is not clear how this putative mechanism may evolve over development. Changes in the magnitude of delta band activity could affect the degree of engagement of excitatory–inhibitory loops that produce gamma, as well as affecting the precision of phase estimates in the calculation of CFC due to SNR issues (see preceding discussion of similar issues with PLF measure). Delta amplitude progressively decreased with age, however, indicating that the increases in CFC from childhood to mid-adolescence did not simply result from increases in delta band activity (for further discussion relating to measurement issues, see Supplementary Data).

This study had a number of limitations, including the age range not extending beyond early adulthood (sampled in this study as 20–22-year olds). As such, while the dense sampling of the current study allowed a detailed characterization of gamma activity over a broad, developmentally critical age range (8–22 years), the precise trajectory by which higher levels of gamma activity reached again later in adulthood (Rojas et al. 2006) needs to be investigated. Also, although longitudinal designs offer the strongest inferences regarding developmental trajectory, the current data were collected cross-sectionally. Accordingly, while the age subgroups were matched demographically, cohort effects cannot be completely ruled out. Another limitation is that while the ASSR paradigm has the advantages of excellent SNR and a robust ability to detect gamma-band-specific disturbances in clinical populations, in its standard implementation (as in the current study), it does not directly link the oscillatory activity to a cognitive process. Future investigations could vary task parameters to produce meaningful behavioral responses that enhance the direct cognitive relevance of the EEG findings. Finally, while much finer sampling of frequencies was possible (Krishnan et al. 2009), just 3 frequency bands were selected from those commonly examined in auditory steady-state paradigms in the interest of making the experiment time tolerable (especially to the youngest of subjects). Refinements to the ASSR paradigm that allow more refined sampling of the frequency spectrum while maintaining reasonable experimental duration are currently under development.

In summary, the current study investigated the development of gamma oscillations in 8- to 22-year-old individuals performing the ASSR paradigm, finding that evoked power, PLF and CFC all followed an inverted-U trajectory, with increases from childhood to mid-adolescence with subsequent decreases toward early adulthood. This detailed characterization of multiple aspects of gamma oscillatory dynamics could inform our understanding of the neurodevelopmental basis of sensory and cognitive processing abilities during this developmental period of dynamic changes, as well pathophysiologic mechanisms of disorders that arise in this neurodevelopmental context, such as schizophrenia. Disturbances in gamma oscillations, thought to be a core pathophysiologic mechanism in this disorder, have been extensively characterized using the ASSR paradigm (Kwon et al. 1999; Light et al. 2006; Spencer et al. 2008; Krishnan et al. 2009; Kömek et al. 2012). Schizophrenia has also been hypothesized to arise from dysregulation of synaptic pruning and has well characterized pathological changes in layer III pyramidal cells and PV fast-spiking interneurons across cortical regions including the auditory cortex and PFC (Lewis and Sweet 2009). The findings of the current study could provide important insights regarding the timing and physiologic nature of pathological changes that interact to give rise to disturbances in gamma oscillations in the illness. A particularly clinically useful outcome would be the ability to detect physiologic disturbances prior to psychosis onset. Prior studies have examined gamma-band responses to auditory stimuli in genetic high-risk subjects, finding that evoked power and PLF were reduced (Hong et al. 2004; Leicht et al. 2010). However, given that the subjects were older than the typical age of psychosis onset and were not followed longitudinally, future studies using longitudinal designs in younger clinical or genetic high-risk subjects could better evaluate the potential for detection and prognostic value of identifying disturbances in gamma activity earlier in the developmental trajectory.

Supplementary Material

Supplementary can be found at: http://www.cercor.oxfordjournals.org/.

Funding

This work was supported by the National Institute of Mental Health at the National Institutes of Health (K08 MH080329 to R.Y.C. and P50 MH084053 to D.A.L.).

Notes

The authors acknowledge Debra Montrose and Alicia Thomas for their contributions to subject recruitment and RyAnna Verbiest, Megan Carl, Polina Radchenkova, Tanisha Hill-Jarrett, Annette Richard for their contributions to data collection. They also thank Dr Don Rojas for providing the Rojas et al. (2006) data to permit re-analysis. This work was supported by the National Institute of Mental Health at the National Institutes of Health (K08 MH080329 to R.Y.C. and P50 MH084053 to D.A.L.). These data were presented, in part, at the 2011 Annual Meeting of the Society of Biological Psychiatry, San Francisco, CA, USA. Conflict of Interest: None declared.

References

Anderson
SA
Classey
JD
Condé
F
Lund
JS
Lewis
DA
.
1995
.
Synchronous development of pyramidal neuron dendritic spines and parvalbumin-immunoreactive chandelier neuron axon terminals in layer III of monkey prefrontal cortex
.
Neuroscience
 .
67
:
7
22
.
Axmacher
N
Henseler
MM
Jensen
O
Weinreich
I
Elger
CE
Fell
J
.
2010
.
Cross-frequency coupling supports multi-item working memory in the human hippocampus
.
Proc Natl Acad Sci USA
 .
107
:
3228
3233
.
Bartos
M
Vida
I
Jonas
P
.
2007
.
Synaptic mechanisms of synchronized gamma oscillations in inhibitory interneuron networks
.
Nat Rev Neurosci
 .
8
:
45
56
.
Beneyto
M
Lewis
DA
.
2011
.
Insights into the neurodevelopmental origin of schizophrenia from postmortem studies of prefrontal cortical circuitry
.
Int J Dev Neurosci
 .
29
:
295
304
.
Bitanihirwe
B
Lim
M
Kelley
J
Kaneko
T
Woo
T
.
2009
.
Glutamatergic deficits and parvalbumin-containing inhibitory neurons in the prefrontal cortex in schizophrenia
.
BMC Psychiatry
 .
9
:
71
.
Bourgeois
JP
Goldman-Rakic
PS
Rakic
P
.
1994
.
Synaptogenesis in the prefrontal cortex of rhesus monkeys
.
Cereb Cortex
 .
4
:
78
96
.
Brenner
CA
Krishnan
GP
Vohs
JL
Ahn
W-Y
Hetrick
WP
Morzorati
SL
O'Donnell
BF
.
2009
.
Steady state responses: electrophysiological assessment of sensory function in schizophrenia
.
Schizophr Bull
 .
35
:
1065
1077
.
Buffalo
EA
Fries
P
Landman
R
Buschman
TJ
Desimone
R
.
2011
.
Laminar differences in gamma and alpha coherence in the ventral stream
.
Proc Natl Acad Sci USA
 .
108
:
11262
11267
.
Buzsáki
G
Wang
X-J
.
2012
.
Mechanisms of gamma oscillations
.
Annu Rev Neurosci
 .
35
:
203
225
.
Canolty
RT
Knight
RT
.
2010
.
The functional role of cross-frequency coupling
.
Trends Cogn Sci
 .
14
:
506
515
.
Cho
RY
Konecky
RO
Carter
CS
.
2006
.
Impairments in frontal cortical γ synchrony and cognitive control in schizophrenia
.
Proc Natl Acad Sci USA
 .
103
:
19878
19883
.
Chrobak
JJ
Buzsáki
G
.
1998
.
Gamma oscillations in the entorhinal cortex of the freely behaving rat
.
J Neurosci
 .
18
:
388
398
.
Ciesielski
KT
Lesnik
PG
Savoy
RL
Grant
EP
Ahlfors
SP
.
2006
.
Developmental neural networks in children performing a Categorical N-Back Task
.
Neuroimage
 .
33
:
980
990
.
Colon
E
Legrain
V
Mouraux
A
.
2012
.
Steady-state evoked potentials to study the processing of tactile and nociceptive somatosensory input in the human brain
.
Neurophysiol Clin Neurophysiol
 .
42
:
315
323
.
Compte
A
Brunel
N
Goldman-Rakic
PS
Wang
X-J
.
2000
.
Synaptic mechanisms and network dynamics underlying spatial working memory in a cortical network model
.
Cereb Cortex
 .
10
:
910
923
.
Delorme
A
Makeig
S
.
2004
.
EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis
.
J Neurosci Methods
 .
134
:
9
21
.
Doischer
D
Hosp
JA
Yanagawa
Y
Obata
K
Jonas
P
Vida
I
Bartos
M
.
2008
.
Postnatal differentiation of basket cells from slow to fast signaling devices
.
J Neurosci
 .
28
:
12956
12968
.
Eggan
SM
Mizoguchi
Y
Stoyak
SR
Lewis
DA
.
2010
.
Development of cannabinoid 1 receptor protein and messenger RNA in monkey dorsolateral prefrontal cortex
.
Cereb Cortex
 .
20
:
1164
1174
.
Ehrenreich
H
Rinn
T
Kunert
HJ
Moeller
MR
Poser
W
Schilling
L
Gigerenzer
G
Hoehe
MR
.
1999
.
Specific attentional dysfunction in adults following early start of cannabis use
.
Psychopharmacology
 .
142
:
295
301
.
Fujisawa
S
Buzsáki
G
.
2011
.
A 4 Hz oscillation adaptively synchronizes prefrontal, VTA, and hippocampal activities
.
Neuron
 .
72
:
153
165
.
Giedd
JN
Blumenthal
J
Jeffries
NO
Castellanos
FX
Liu
H
Zijdenbos
A
Paus
T
Evans
AC
Rapoport
JL
.
1999
.
Brain development during childhood and adolescence: a longitudinal MRI study
.
Nat Neurosci
 .
2
:
861
863
.
Gogtay
N
Giedd
JN
Lusk
L
Hayashi
KM
Greenstein
D
Vaituzis
AC
Nugent
TF
3rd
Herman
DH
Clasen
LS
Toga
AW
et al
2004
.
Dynamic mapping of human cortical development during childhood through early adulthood
.
Proc Natl Acad Sci USA
 .
101
:
8174
8179
.
Gonzalez-Burgos
G
Kroener
S
Zaitsev
AV
Povysheva
NV
Krimer
LS
Barrionuevo
G
Lewis
DA
.
2008
.
Functional maturation of excitatory synapses in layer 3 pyramidal neurons during postnatal development of the primate prefrontal cortex
.
Cereb Cortex
 .
18
:
626
637
.
Hashimoto
T
Nguyen
QL
Rotaru
D
Keenan
T
Arion
D
Beneyto
M
Gonzalez-Burgos
G
Lewis
DA
.
2009
.
Protracted developmental trajectories of GABAA receptor α1 and α2 subunit expression in primate prefrontal cortex
.
Biol Psychiatry
 .
65
:
1015
1023
.
Holderith
N
Németh
B
Papp
OI
Veres
JM
Nagy
GA
Hájos
N
.
2011
.
Cannabinoids attenuate hippocampal gamma oscillations by suppressing excitatory synaptic input onto CA3 pyramidal neurons and fast spiking basket cells
.
J Physiol
 .
589
:
4921
4934
.
Hollingshead
AB
.
1975
.
A four-factor index of social status
 .
New Haven
(
CT
):
Yale University
.
Hong
LE
Summerfelt
A
McMahon
R
Adami
H
Francis
G
Elliott
A
Buchanan
RW
Thaker
GK
.
2004
.
Evoked gamma band synchronization and the liability for schizophrenia
.
Schizophr Res
 .
70
:
293
302
.
Huttenlocher
PR
Dabholkar
AS
.
1997
.
Regional differences in synaptogenesis in human cerebral cortex
.
J Comp Neurol
 .
387
:
167
178
.
Ito
J
Maldonado
P
Grün
S
.
2013
.
Cross-frequency interaction of the eye-movement related LFP signals in V1 of freely viewing monkeys
.
Front Syst Neurosci
 .
7
:
1
.
Jacobs
B
Driscoll
L
Schall
M
.
1997
.
Life-span dendritic and spine changes in areas 10 and 18 of human cortex: a quantitative Golgi study
.
J Comp Neurol
 .
386
:
661
680
.
Jensen
O
Colgin
LL
.
2007
.
Cross-frequency coupling between neuronal oscillations
.
Trends Cogn Sci
 .
11
:
267
269
.
Johnston
D
Wu
S
.
1997
.
Foundations of cellular neurophysiology
 .
Cambridge
(
MA
):
MIT Press
.
Katona
I
Freund
TF
.
2008
.
Endocannabinoid signaling as a synaptic circuit breaker in neurological disease
.
Nat Med
 .
14
:
923
930
.
Kieffaber
PD
Cho
RY
.
2010
.
Induced cortical gamma-band oscillations reflect cognitive control elicited by implicit probability cues in the preparing-to-overcome-prepotency (POP) task
.
Cogn Affect Behav Neurosci
 .
10
:
431
440
.
Kömek
K
Bard Ermentrout
G
Walker
CP
Cho
RY
.
2012
.
Dopamine and gamma band synchrony in schizophrenia—insights from computational and empirical studies
.
Eur J Neurosci
 .
36
:
2146
2155
.
Krishnan
GP
Hetrick
WP
Brenner
CA
Shekhar
A
Steffen
AN
O'Donnell
BF
.
2009
.
Steady state and induced auditory gamma deficits in schizophrenia
.
Neuroimage
 .
47
:
1711
1719
.
Kwon
JS
O'Donnell
BF
Wallenstein
GV
Greene
RW
Hirayasu
Y
Nestor
PG
Hasselmo
ME
Potts
GF
Shenton
ME
McCarley
RW
.
1999
.
Gamma frequency-range abnormalities to auditory stimulation in schizophrenia
.
Arch Gen Psychiatry
 .
56
:
1001
1005
.
Leicht
G
Kirsch
V
Giegling
I
Karch
S
Hantschk
I
Möller
H-J
Pogarell
O
Hegerl
U
Rujescu
D
Mulert
C
.
2010
.
Reduced early auditory evoked gamma-band response in patients with schizophrenia
.
Biol Psychiatry
 .
67
:
224
231
.
Lewis
DA
Sweet
RA
.
2009
.
Schizophrenia from a neural circuitry perspective: advancing toward rational pharmacological therapies
.
J Clin Invest
 .
119
:
706
716
.
Light
GA
Hsu
JL
Hsieh
MH
Meyer-Gomes
K
Sprock
J
Swerdlow
NR
Braff
DL
.
2006
.
Gamma band oscillations reveal neural network cortical coherence dysfunction in schizophrenia patients
.
Biol Psychiatry
 .
60
:
1231
1240
.
Luna
B
Garver
KE
Urban
TA
Lazar
NA
Sweeney
JA
.
2004
.
Maturation of cognitive processes from late childhood to adulthood
.
Child Dev
 .
75
:
1357
1372
.
Luna
B
Thulborn
KR
Munoz
DP
Merriam
EP
Garver
KE
Minshew
NJ
Keshavan
MS
Genovese
CR
Eddy
WF
Sweeney
JA
.
2001
.
Maturation of widely distributed brain function subserves cognitive development
.
Neuroimage
 .
13
:
786
793
.
Magri
C
Schridde
U
Murayama
Y
Panzeri
S
Logothetis
NK
.
2012
.
The amplitude and timing of the BOLD signal reflects the relationship between local field potential power at different frequencies
.
J Neurosci
 .
32
:
1395
1407
.
Maier
A
Adams
GK
Aura
C
Leopold
DA
.
2010
.
Distinct superficial and deep laminar domains of activity in the visual cortex during rest and stimulation
.
Front Syst Neurosci
 .
4
:
31
.
Maurizi
M
Almadori
G
Paludetti
G
Ottaviani
F
Rosignoli
M
Luciano
R
.
1990
.
40-Hz steady-state responses in newborns and in children
.
Audiology
 .
29
:
322
328
.
Mazzoni
A
Whittingstall
K
Brunel
N
Logothetis
NK
Panzeri
S
.
2010
.
Understanding the relationships between spike rate and delta/gamma frequency bands of LFPs and EEGs using a local cortical network model
.
NeuroImage
 .
52
:
956
972
.
Minzenberg
MJ
Firl
AJ
Yoon
JH
Gomes
GC
Reinking
C
Carter
CS
.
2010
.
Gamma oscillatory power is impaired during cognitive control independent of medication status in first-episode schizophrenia
.
Neuropsychopharmacology
 .
35
:
2590
2599
.
Mirnics
K
Middleton
FA
Lewis
DA
Levitt
P
.
2001
.
Analysis of complex brain disorders with gene expression microarrays: schizophrenia as a disease of the synapse
.
Trends Neurosci
 .
24
:
479
486
.
Moore
TH
Zammit
S
Lingford-Hughes
A
Barnes
TR
Jones
PB
Burke
M
Lewis
G
.
2007
.
Cannabis use and risk of psychotic or affective mental health outcomes: a systematic review
.
Lancet
 .
370
:
319
328
.
Morgan
NH
Stanford
IM
Woodhall
GL
.
2008
.
Modulation of network oscillatory activity and GABAergic synaptic transmission by CB1 cannabinoid receptors in the rat medial entorhinal cortex
.
Neural Plast
 .
[Internet]
2008
.
Available from: http://www.hindawi.com/journals/np/2008/808564/abs/. Last accessed date: 14 October 2013
.
Mukamel
R
Gelbard
H
Arieli
A
Hasson
U
Fried
I
Malach
R
.
2005
.
Coupling between neuronal firing, field potentials, and FMRI in human auditory cortex
.
Science
 .
309
:
951
954
.
Niessing
J
Ebisch
B
Schmidt
KE
Niessing
M
Singer
W
Galuske
RAW
.
2005
.
Hemodynamic signals correlate tightly with synchronized gamma oscillations
.
Science
 .
309
:
948
951
.
Nunez
PL
Srinivasan
R
.
2005
.
Electric fields of the brain: the neurophysics of EEG
 .
2nd edn
.
New York
:
Oxford University Press
.
Petanjek
Z
Judaš
M
Šimić
G
Rašin
MR
Uylings
HBM
Rakic
P
Kostović
I
.
2011
.
Extraordinary neoteny of synaptic spines in the human prefrontal cortex
.
Proc Natl Acad Sci USA
 .
108
:
13281
13286
.
Plourde
G
Baribeau
J
Bonhomme
V
.
1997
.
Ketamine increases the amplitude of the 40-Hz auditory steady-state response in humans
.
Br J Anaesth
 .
78
:
524
529
.
Pope
HG
Jr
Gruber
AJ
Hudson
JI
Cohane
G
Huestis
MA
Yurgelun-Todd
D
.
2003
.
Early-onset cannabis use and cognitive deficits: what is the nature of the association?
Drug Alcohol Depend
 .
69
:
303
310
.
Poulsen
C
Picton
TW
Paus
T
.
2009
.
Age-related changes in transient and oscillatory brain responses to auditory stimulation during early adolescence
.
Dev Sci
 .
12
:
220
235
.
Quilichini
P
Sirota
A
Buzsáki
G
.
2010
.
Intrinsic circuit organization and theta–gamma oscillation dynamics in the entorhinal cortex of the rat
.
J Neurosci
 .
30
:
11128
11142
.
Rakic
P
Bourgeois
JP
Eckenhoff
MF
Zecevic
N
Goldman-Rakic
PS
.
1986
.
Concurrent overproduction of synapses in diverse regions of the primate cerebral cortex
.
Science
 .
232
:
232
235
.
Raver
SM
Haughwout
SP
Keller
A
.
2013
.
Adolescent cannabinoid exposure permanently suppresses cortical oscillations in adult mice
.
Neuropsychopharmacology
 .
38
:
2338
2347
.
Rojas
DC
Maharajh
K
Teale
PD
Kleman
MR
Benkers
TL
Carlson
JP
Reite
ML
.
2006
.
Development of the 40 Hz steady state auditory evoked magnetic field from ages 5 to 52
.
Clin Neurophysiol
 .
117
:
110
117
.
Rotaru
DC
Yoshino
H
Lewis
DA
Ermentrout
GB
Gonzalez-Burgos
G
.
2011
.
Glutamate receptor subtypes mediating synaptic activation of prefrontal cortex neurons: relevance for schizophrenia
.
J Neurosci
 .
31
:
142
156
.
Sauseng
P
Klimesch
W
Heise
KF
Gruber
WR
Holz
E
Karim
AA
Glennon
M
Gerloff
C
Birbaumer
N
Hummel
FC
.
2009
.
Brain oscillatory substrates of visual short-term memory capacity
.
Curr Biol
 .
19
:
1846
1852
.
Scherf
KS
Sweeney
JA
Luna
B
.
2006
.
Brain basis of developmental change in visuospatial working memory
.
J Cogn Neurosci
 .
18
:
1045
1058
.
Schroeder
CE
Lakatos
P
.
2009
.
Low-frequency neuronal oscillations as instruments of sensory selection
.
Trends Neurosci
 .
32
:
9
18
.
Sheehan
DV
Lecrubier
Y
Sheehan
KH
Amorim
P
Janavs
J
Weiller
E
Hergueta
T
Baker
R
Dunbar
GC
.
1998
.
The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10
.
J Clin Psychiatry
 .
59
(Suppl 20)
:
22
33
;
quiz 34–57
.
Spaak
E
Bonnefond
M
Maier
A
Leopold
DA
Jensen
O
.
2012
.
Layer-specific entrainment of gamma-band neural activity by the alpha rhythm in monkey visual cortex
.
Curr Biol
 .
22
:
2313
2318
.
Spencer
KM
Salisbury
DF
Shenton
ME
McCarley
RW
.
2008
.
Gamma-band auditory steady-state responses are impaired in first episode psychosis
.
Biol Psychiatry
 .
64
:
369
375
.
Stapells
DR
Galambos
R
Costello
JA
Makeig
S
.
1988
.
Inconsistency of auditory middle latency and steady-state responses in infants
.
Electroencephalogr Clin Neurophysiol
 .
71
:
289
295
.
Tort
ABL
Komorowski
R
Eichenbaum
H
Kopell
N
.
2010
.
Measuring phase-amplitude coupling between neuronal oscillations of different frequencies
.
J Neurophysiol
 .
104
:
1195
1210
.
Uhlhaas
PJ
Roux
F
Singer
W
Haenschel
C
Sireteanu
R
Rodriguez
E
.
2009
.
The development of neural synchrony reflects late maturation and restructuring of functional networks in humans
.
Proc Natl Acad Sci USA
 .
106
:
9866
9871
.
Vialatte
F-B
Maurice
M
Dauwels
J
Cichocki
A
.
2010
.
Steady-state visually evoked potentials: focus on essential paradigms and future perspectives
.
Prog Neurobiol
 .
90
:
418
438
.
Vierling-Claassen
D
Siekmeier
P
Stufflebeam
S
Kopell
N
.
2008
.
Modeling GABA alterations in schizophrenia: a link between impaired inhibition and altered gamma and beta range auditory entrainment
.
J Neurophysiol
 .
99
:
2656
2671
.
Vohs
JL
Chambers
RA
Krishnan
GP
O'Donnell
BF
Berg
S
Morzorati
SL
.
2010
.
GABAergic modulation of the 40 Hz auditory steady-state response in a rat model of schizophrenia
.
Int J Neuropsychopharmacol
 .
13
:
487
497
.
Wang
H-X
Gao
W-J
.
2009
.
Cell type-specific development of NMDA receptors in the interneurons of rat prefrontal cortex
.
Neuropsychopharmacology
 .
34
:
2028
2040
.
Wang
H-X
Gao
W-J
.
2010
.
Development of calcium-permeable AMPA receptors and their correlation with NMDA receptors in fast-spiking interneurons of rat prefrontal cortex
.
J Physiol
 .
588
:
2823
2838
.
Whittingstall
K
Logothetis
NK
.
2009
.
Frequency-band coupling in surface EEG reflects spiking activity in monkey visual cortex
.
Neuron
 .
64
:
281
289
.
Woo
T-U
Pucak
ML
Kye
CH
Matus
CV
Lewis
DA
.
1997
.
Peripubertal refinement of the intrinsic and associational circuitry in monkey prefrontal cortex
.
Neuroscience
 .
80
:
1149
1158
.
Wulff
P
Ponomarenko
AA
Bartos
M
Korotkova
TM
Fuchs
EC
Bähner
F
Both
M
Tort
ABL
Kopell
NJ
Wisden
W
et al
2009
.
Hippocampal theta rhythm and its coupling with gamma oscillations require fast inhibition onto parvalbumin-positive interneurons
.
Proc Natl Acad Sci USA
 .
106
:
3561
3566
.

Author notes

Christopher P. Walker and Nicola R. Polizzotto contributed equally to the manuscript.