Hemispheric Utilization of Alpha Oscillatory Dynamics as a Unique Biomarker of Neural Compensation in Females with Fragile X Syndrome

Abstract

Fragile X Syndrome (FXS) is a neurodevelopmental disorder caused by a trinucleotide expansion on the FMR1 gene and characterized by intellectual disability, sensory hypersensitivity, executive function difficulties, and social anxiety. Recently, efforts to define neural biomarkers for FXS have highlighted disruptions to power in the alpha frequency band, however the dynamic mechanisms supporting these findings are poorly understood. The current study aimed to explore the temporal and hemispheric dynamics supporting alpha phenotypes in FXS and their relationship with neural phenotypes related to auditory processing using electroencephalography during an auditory evoked task. Adolescents and adults (N =36) with FXS and age/sex matched typically developing controls (N = 40) completed an auditory chirp task. Frontal alpha power in the pre-stimulus period was decomposed into “bursts” using percentile thresholding, then assessed for number of bursts per second (burst count) and burst length. Data were compared across left and right hemisphere to assess lateralization of neural activity. Individuals with FXS showed more differences in alpha power compared to TDC primarily in right hemisphere. Notably, alpha hemisphere outcomes in males with FXS were driven by the number of times they entered a dynamically relevant period of alpha (burst count) rather than length of time spent in alpha. Females with FXS showed reduced burst counts but remained in sustained high alpha states for longer periods of time. Length of time spent in alpha may reflect a modulatory or compensatory mechanism capable of recovering sensory processing abilities in females with FXS resulting in a less severe clinical presentation. Right hemisphere abnormalities may impact sensory processing differences between males and females with FXS. The relationship between alpha burst length, count, sex, and hemisphere may shed light on underlying mechanisms for previously observed alpha power abnormalities in FXS and their variation by sex.

Graphical Abstract

1. Introduction

Fragile X Syndrome (FXS) is a rare genetic neurodevelopmental disorder caused by >200 CGG trinucleotide repeats on the Fragile X messenger ribonucleoprotein 1 (FMR1) gene resulting in methylation-based reduction of Fragile X messenger ribonucleoprotein (FMRP) protein output [1–3]. Key neurodevelopmental symptoms associated with FXS include intellectual disability (ID), sensory hypersensitivity, social anxiety, and general difficulty with executive control [4–6]. The need for better treatment options that address underlying biological pathways has led to promising findings in electroencephalography (EEG) biomarkers in FXS. Biomarker development for patient stratification and assessing group-level mechanistic differences remain important goals for FXS research, and may benefit from more sensitive EEG biomarkers that incorporate spatial and temporal dynamics of the EEG [4,7].

1.1. Relationship between gamma, theta, and alpha

Our prior work supports links between altered cortical oscillatory activity and sensory hypersensitivity in FXS, including increased theta and broadband gamma power and decreases in chirp-evoked phase-locking in the gamma range compared to typically-developing controls (TDC) [8–13]. Increased gamma power in FXS is linked to neural hyperexcitability and potentially represents an underlying mechanism of sensory processing deficits. Relative alpha power is reduced in FXS compared to TDC and has been linked to neural hyperexcitability via decreased capacity to thalamocortically modulate gamma-band activity via high-low frequency coupling [11,13]. Power and chirp synchronization deficits have been replicated in the fmr1^−/− knock-out (KO) mouse [14] with intriguing findings related to hemispheric differences in the mouse that have yet to be fully explored in humans with FXS [15]. It may therefore be important to address hemispheric asymmetry to fully utilize neural oscillatory biomarkers in FXS.

1.2. Asymmetry & Alpha Bursts

Frontal asymmetry in the alpha-band (8–13 Hz) has been investigated in relation to emotion, social engagement, motivation, and psychopathology [16–18] and shows high internal consistency and modest stability over short periods of time in both nonclinical and clinical populations [19,20]. Frontal asymmetry reflects a relatively dynamic neural response but as traditionally measured is limited by the inability to capture endogenous recurrent dynamic neural processes. The majority of asymmetry studies use fast Fourier transform (FFT), or other similar metrics averaged over multiple seconds to minutes’ worth of EEG data, assuming stationarity of the contributing neural signal [16]. Neural activity is fundamentally dynamic; fluctuating cortical relationships are lost through data averaging inherent to FFT-based power analyses [21,22]. One method for reducing stationarity assumptions and improving temporal precision is to quantify the transient dynamics of alpha within each hemisphere using endogenous alpha bursts [21].

Alpha bursts (Figure 1) index alpha oscillations at specific points in time that reflect dynamic recurrent alpha states. Where frontal alpha asymmetry is interpreted as generalized lateral inhibition of neural activity reflecting increased cortical activity in the opposite hemisphere of increased alpha power, bursts reflect transient lateral suppression [21]. Bursts add the benefit of fine-tuned temporal and potentially mechanistic exploration of alpha-band utilization, which has shown relevance to hemispheric asymmetry and pathophysiology in disorders such as depression [21].

Figure 1.

Schematic example of one participant to illustrate alpha burst quantification from electrodes F3 and F4. Top row depicts 2.5s of raw data. The next row depicts the same 2.5s of data after filtering in the alpha range (8–13 Hz). Data were then Hilbert transformed and natural log of the squared absolute value were taken to compute power. The final row depicts the continuous difference calculated from positive and negative power signals with threshold lines at the 75th percentile. Everything above (positive/F4) and below (negative/F3) the threshold lines constitutes an alpha burst. Adapted with permission from [21]. Copyright 2010, the authors.

1.3. Sex differences in FXS neural oscillatory dynamics

Sex differences have been found in the TDC population for alpha asymmetry and frontal theta power measures [23–26]; additionally, sex may moderate alpha asymmetry-related treatment outcomes [27]. Sex differences in FXS are to be expected in many cases due to the obligate mosaicism present in females with FXS, which is thought to be tied to individual differences in symptom severity. Sex differences in EEG power have been reported in both chirp-evoked and resting data [10,12].

1.4. Current Study

The current study aims to address underlying neural dynamics in FXS by exploring the extent to which bursting activity and lateralization of that activity in the alpha, theta, and low gamma range can be used to further parse previously reported global assessments of neural activity differences, particularly in alpha power. Additionally, we explored potential sex-differences with the expectation that females with FXS would show different patterns of asymmetry compared to males with FXS due to differences in symptom presentation and severity.

2. Results and Discussion

2.1. Participant Demographics

There were no significant differences in age between males M = 5.72, SD = 8.70) and females M = 22.96, SD = 10.07) with FXS, t(34) = .85, p =.40. Participants were predominately male, consistent with diagnosis rates (CDC; Data and Statistics on Fragile X Syndrome).

2.2. Alpha Burst Counts by Hemisphere

Data were evaluated at all measured thresholds (75^th, 80^th, 90^th, 95^th percentile), however we found that results were generally consistent from the 80^th percentile upwards, whereas group differences became less apparent at the 75^th percentile. Upon visual inspection the 75^th percentile value tended to combine bimodal bursts and may thus be less sensitive to ongoing dynamic fluctuations, therefore results are presented in the main body of the text for the 80^th percentile as most indicative of the lowest threshold that separated bimodal bursts in this cohort, and thus produced group differences consistent with the higher thresholds, and results for the 75^th, 90^th and 95^th percentiles can be found in the Supplemental Materials.

Three-way repeated measure ANCOVAs were run to assess hemisphere by sex by group effects for alpha burst counts per second. At the 80^th percentile threshold level there was a significant main effect of sex, F(1,70) = 7.22, p = .009, ES = .093. Females exhibited reduced numbers of alpha burst counts M = 1.83, SE = .03) compared to males M = 1.92, SE = .02). There was a significant main effect of group, F(1,70) = 21.65, p < .001, ES = .25. Individuals with FXS exhibited a significantly reduced number of bursts per second M = 1.79, SE = .03) compared to TDCs M = 1.96, SE = .03). There was also a significant interaction between group and sex, F(1,70) = 13.94, p < .001, ES = .166. Females with FXS exhibited reduced bursts per second M = 1.67, SE = .04) compared to males with FXS M = 1.91, SE = .03) and both male TDCs M = 1.94, SE = .03) and female TDCs M = 1.98, SE = .04). Lastly, there was an interaction between hemisphere and sex, F(1,70) = 4.12, p = .046, ES = .056. Females exhibited greater alpha burst counts in the left hemisphere M = 1.94, SE = .05) compared to the right hemisphere M = 1.71, SE = .06) and compared to males (left hemisphere: M = 1.91, SE = .04; right hemisphere: M = 1.94, SE = .05).

2.3. Alpha Burst Lengths by Hemisphere

There was a main effect of sex at the 80^th percentile threshold level, with females exhibiting increased alpha burst lengths M = 28.413, SE = .434) compared to males M = 27.223, SE = .343), F(1,68) = 4.577, p =.036, ES = .063. A significant interaction for hemisphere and sex occurred at the 90^th and 95^th percentile levels (supplemental results).

There were significant interactions between group and sex at all percentile threshold levels, 80^th percentile: F(1,68) = 22.267, p < .001, ES = .247. Females with FXS M = 29.15, SE = .65) had the longest burst lengths compared to FXS males M = 26.92, SE= .56) and TDC (Female: M = 25.39, SE = .49; Males: M = 28.34, SE = .56).

2.4. FFT-Based Power by Hemisphere

As a group, individuals with FXS exhibited increased power across all frequency bands except alpha, however females with FXS showed increases in theta and alpha power and decreases in gamma power compared to males with FXS, suggesting both more normative responses (alpha and gamma) combined with potential compensatory power elevations (theta) that may contribute to reduced clinical impact in females with FXS.

2.4.1. Alpha.

A three-way repeated measures ANCOVA assessed hemisphere, sex, and group differences on alpha power showed an interaction between sex and group, F(1, 68) = 14.890, p < .001, ES = .180. Females with FXS exhibited greater alpha M = 51.716, SE = .777) compared to FXS males M = 48.946, SE = .559), TDC females M = 47.907, SE = .667), and TDC males M = 50.147, SE = .579), while the latter three groups did not differ from each other. We also found an interaction between hemisphere and sex F(1, 68) = 4.656, p =.034, ES = .064. Females exhibited greater alpha in the left hemisphere M = 50.162, SE = .525) compared to the right hemisphere M = 49.461, SE = .528) and compared to males (left hemisphere: M = 49.510, SE = .415; right hemisphere: M = 49.583, SE = .417).

2.4.2. Theta.

A three-way repeated measures ANCOVA assessed hemisphere, sex, and group differences on theta power and showed an interaction between sex and group where females with FXS exhibited greater theta power M = 53.234, SE = .651) compared to males with FXS M = 51.849, SE = .494) and compared to TDC females M = 49.778, SE = .559) and TDC males M = 51.369, SE = .485), F(1, 68) = 7.488, p = .008, ES = .099.

2.4.3. Gamma.

A three-way repeated measures ANCOVA assessed hemisphere, sex, and group differences on low gamma power. Age was a significant covariate, F(1, 69) = 6.819, p = .011, ES = .091. There was a significant main effect of group where individuals with FXS exhibited elevated low gamma power M = 40.136, SE = .498) compared to TDC M = 37.669, SE = .457), F(1, 68) = 12.445, p = .001, ES = .155. There was a main effect of sex with females exhibiting reduced low gamma power M = 38.013, SE = .515) compared to males M = 39.793, SE = .406), F(1, 68) = 7.273, p = .009, ES = .097. Lastly, there was a main effect of hemisphere where gamma was reduced in the right hemisphere M = 38.749, SE = .390) compared to the left hemisphere M = 39.056, SE = .350), F(1, 68) = 4.388, p = .04, ES = .061.

2.5. Frontal Asymmetry Ratio

Group differences were not significant between FXS and TDC for frontal alpha asymmetry, frontal theta asymmetry, or frontal low gamma asymmetry. A three-way repeated measures ANCOVA was used to address potential differences between FFT-based alpha asymmetry compared to calculations of alpha asymmetry using alpha burst measures including values for alpha burst counts and alpha burst lengths by group and sex. There was not a main effect of asymmetry suggesting that alpha burst length and count asymmetry measures index alpha asymmetry similarly to FFT-based frontal alpha asymmetry measures.

2.6. ITC Chirp

A three-way repeated measures ANCOVA assessed low gamma chirp phase-locking by hemisphere accounting for sex and FXS diagnosis with covariates of age and trial length. This analysis builds on previous work [6] by accounting for hemispheric differences between groups. There were no significant main effects of hemisphere, group, or sex but a significant interaction of hemisphere and group on ITC emerged, F(1, 68) = 4.67, p = .034, ES = .06. Chirp phase-locking was marginally decreased in FXS in the right hemisphere M = .05, SE = .008) compared to TDC M = .07, SE =.007 ), t(72) = 1.98, p = .051. However, this measure did not differ in the left hemisphere between individuals with FXS and TDC (FXS, M = .07, SE = .01; TDC, M = .06, SE = .009), t(72) = 0.75, p = 0.46, indicating that decreased global chirp phase-locking in FXS primarily reflects underlying sensory processing difficulties in the right hemisphere (Figure 2).

Figure 2.

(Left column) Inter-trial coherence (ITC) for FXS, TDC, and difference maps (FXS minus TDC) for the left hemisphere. (Right column) Inter-trial coherence (ITC) for FXS, TDC, and difference maps (FXS minus TDC) for the right hemisphere.

There was a significant interaction between group and sex for low gamma ITC, F(1, 68) = 5.589, p = .021, ES = .076. Females with FXS had increased low gamma ITC M = .074, SE = .012) compared to males with FXS M = .043, SE = .009) where TDC females had decreased low gamma ITC M = .058, SE = .012) compared to TDC males M = .077, SE = .009). Generally, these results indicate that males with FXS had proportionately increased sensory processing difficulties.

As some of the interaction described above between groups appeared to be driven by sex differences in FXS, differences between hemispheres and sex within the FXS group only were further explored via two-way repeated measures ANOVA. There was a significant main effect of sex, F(1, 31) = 4.85, p = .035, ES = .14. Females with FXS exhibited increased low gamma ITC (M= .07, SE = .010) compared to males with FXS M = .04, SE = .008). A significant interaction between hemisphere and sex emerged, where females with FXS exhibited greater left hemisphere ITC M = .09, SE = .02) compared to the right hemisphere M = .06, SE = .01) and males with FXS were equally affected in both hemispheres (left hemisphere: M = .04, SE = .01; right hemisphere: M = .04, SE = .01), F(1, 33) = 6.06, p = .020, ES = .16. This finding suggests that males with FXS show sensory processing deficits in both hemispheres, whereas females with FXS primarily only show these deficits in the right hemisphere (Figure 3).

Figure 3.

(Left column) Inter-trial coherence (ITC) for FXS males, FXS females, and difference maps (FXS males minus FXS females) for the left hemisphere. (Right column) Inter-trial coherence (ITC) for FXS males, FXS females, and difference maps (FXS males minus FXS females) for the right hemisphere.

2.7. Correlations.

2.7.1. Alpha Burst Count Clinical Correlations.

In males with FXS, increased alpha burst counts were associated with lower communication, language, coping, and play skills but decreased social avoidance (Table 1).

Table 1.

Correlations between Alpha Burst Variables and both EEG and Clinical Variables for Individuals with FXS Including Evaluation by Sex

	FXS Group				FXS Males				FXS Females
	Alpha Burst Length		Alpha Burst Count Per Second		Alpha Burst Length		Alpha Burst Count Per Second		Alpha Burst Length		Alpha Burst Count Per Second
EEG Variables	LH	RH	LH	RH	LH	RH	LH	RH	LH	RH	LH	RH
Alpha Power
Alpha Asymmetry	−.38*	.44**	−.41*	.40*				.44*	−.59*	.75**
Left Hemisphere Alpha	.48**			−.48**					.65*
Right Hemisphere Alpha
Theta Power
Theta Asymmetry	−.65**	.51**	−.44**	.63**		.55**	−.59**	.54**	−.64*	.61*		.60*
Left Hemisphere Theta	.42*			−.42*		−.48*	.59**	−.42*	.58*
Right Hemisphere Theta											−.58*
Low Gamma Power
Low Gamma Asymmetry					−.50*
Left Hemisphere Gamma	−.46**
Right Hemisphere Gamma	−.44**			.35**
Clinical Variables
ABCFXS Social Avoidance		.39*	−.49**				−.46*
ABCFXS Irritability/Aggression
Vineland AE Expressive							−.51*
Vineland AE Written			−.43*			.53*	−.61**
Vineland AE Coping			−.40*				−.52*
Vineland AE Interpersonal Relationship	.45*								.67*
Vineland AE Communication			−.47**			.61**	−.64**
Vineland AE Play/Leisure			−.41*			.49*	−.67**					−.74*
Behavior Variables
Flex Correct								.49*
Distractor Errors										−.77**		.72*

Note. All correlations are Spearman’s rho. All correlations represent the FXS group only. For EEG variables, all variables are included in the table. For clinical/behavior variables, only variables with at least one significant correlation were retained. Blank = N.S.

^*p <0.05

^**p < 0.01

AE: age equivalence, LH= left hemisphere; RH= right hemisphere

2.7.2. Burst Length Clinical Correlations.

Increased alpha burst length in the right but not left hemisphere was significantly correlated with increased social avoidance and increased communication scores. Increased alpha burst length was correlated with interpersonal relationship skills in the left but not right hemisphere. (Table 1).

2.7.3. Frontal Asymmetry Clinical Correlations.

Increased alpha asymmetry was correlated with decreased sensation seeking, indicating that those with increased alpha power in the right hemisphere (i.e., decreased modulatory deficits in the right hemisphere) exhibit more sensation seeking behavior and may have less sensory hypersensitivities. Increased low gamma asymmetry was significantly correlated with decreased lethargy/social withdrawal, potentially reflecting reduced attentiveness noted in FXS populations with increased levels of gamma. (Table 2).

Table 2.

Correlations between Frontal Asymmetry Variables in all Frequency Bands and both EEG and Clinical Variables for individuals with FXS including Evaluation by Sex

	FXS Group			FXS Males			FXS Females
	Frontal Asymmetry			Frontal Asymmetry			Frontal Asymmetry
EEG Variables	Alpha	Theta	Gamma	Alpha	Theta	Gamma	Alpha	Theta	Gamma
Alpha Power
Left Hemisphere Alpha		−.43**			−.43*
Right Hemisphere Alpha
Theta Power
Left Hemisphere Theta		−.56**			−.63**
Right Hemisphere Theta
Low Gamma Power
Left Hemisphere Gamma			−.55**			−.56**
Right Hemisphere Gamma			.34*					.64*
Clinical Variables
ABCFXS Lethargy social withdrawal			−.37*
Vineland AE Communication					.44*			−.71*
Vineland AE Coping					.49*
Vineland AE Play/Leisure					.52*			−.67*
Vineland AE Receptive					.45*
Vineland AE Written					.57**
CSP Seeking	−.48*
CSP Registration		−.47*
Behavior Variables
Distractor Correct							.74*
Distractor Errors	.45**	.39*		.49*				.85**
Flex Correct
Go/NoGo Correct			−.37*						−.65*
Go/NoGo Errors	.38*			.53*					.85**

Note. All correlations are Spearman’s rho. All correlations represent the FXS group only. For EEG variables, all variables are included in the table. For clinical/behavior variables, only variables with at least one significant correlation were retained. Blank = N.S.

^*p <0.05

^**p < 0.01; AE: age equivalence

2.7.4. ITC Chirp Clinical Correlations.

Increased low gamma chirp phase-locking (ITC) in both hemispheres was significantly correlated with decreased SCQ total scores, increased auditory attention, and increased non-verbal raw IQ scores. (Table 3). These correlations between each respective hemisphere were not significantly different (SCQ: z(31) = −.509, p = .31; WJIII: z(28) = −.59, p = .28; IQ: z(32) = −1.161, p = .123), but were driven by FXS males and sex differences in range of values. Increased low gamma ITC was significantly correlated with decreased obsessive/compulsive behavior and increased play skills in the left hemisphere but not the right.

Table 3.

Correlations between ITC in the Low Gamma Range and both EEG and Clinical Variables for Individuals with FXS only

	FXS Group		FXS Males		FXS Females
	Lateralized ITC 40 Hz		Lateralized ITC 40 Hz		Lateralized ITC 40 Hz
EEG Variables	Left hemisphere	Right Hemisphere	Left hemisphere	Right Hemisphere	Left hemisphere	Right Hemisphere
Alpha Power
Alpha Asymmetry
Left Hemisphere Alpha	.56**	.34*	.47*	.46*
Right Hemisphere Alpha	.51**		.55**	.48*
Theta Power
Theta Asymmetry
Left Hemisphere Theta	.45**
Right Hemisphere Theta
Low Gamma Power
Low Gamma Asymmetry
Left Hemisphere Gamma
Right Hemisphere Gamma
Clinical Variables
SCQ	−.46**	−.39*				−.89**
WJIII	.41*	.49**	.45*	.46*
ABCFXS Lethargy/ Social Withdrawal					−.81**
ABCFXS Social Avoidance					−.77**
Vineland AE Social		.37*
Vineland AE Play/Leisure	.43*
ADAMs Obsessive Compulsive	−.57**		−.49*
Nonverbal Raw Score	.37*	.53**		.52*
Nonverbal Z Score		.49**
Behavior Variables
Flex Correct		.55**
Distractor Correct		.38*				.73*
Go-Nogo Correct
Go-Nogo Errors

Note. All correlations are Spearman’s rho. All correlations represent the FXS group only. For EEG variables, all variables are included in the table. For clinical/behavior variables, only variables with at least one significant correlation were retained. Blank = N.S.

^*p <0.05

^**p < 0.01

AE: age equivalence

2.7.5. Physiological Correlations.

Significant physiological correlations between EEG variables are shown in Table 3 and supplemental results. Increased alpha asymmetry were significantly correlated with increased theta asymmetry which may reflect dynamic alpha-theta coupling at the hemispheric level.

2.8. Moderation analyses

Alpha burst lengths and counts did not significantly moderate the relationship between sex and ABCFXS social avoidance scores (Lengths: Left: R² = .08, F(3, 26) = 0.72, p = .55; Right: R² = .09, F(3, 26) = 0.95, p = .43; Counts: Left: R² = .21, F(3, 26) = 2.25, p = .11; Right: R² = .03, F(3, 26) = 0.28, p = .84).

Right and left hemisphere alpha burst lengths significantly moderated the relationship between sex and Vineland Play/Leisure scores with an overall significant model result where the models explained 54.1% and 42.6% of the variability in Vineland Play/Leisure, respectively (left hemisphere: R² = .54, F(3, 27) = 10.61, p < .001; right hemisphere: R² = .43, F(3, 27) = 6.68, p = .002). Right and left hemisphere alpha burst counts per second also significantly moderated the relationship between sex and Vineland Play/Leisure scores with an overall significant model result where the models explained 51.4% and 56.4% of the variability in Vineland Play/Leisure, respectively (left hemisphere: R² = .51, F(3, 27) = 9.51, p < .001; right hemisphere: R² = .56, F(3, 27) = 11.66, p < .001).

Left and right hemisphere alpha burst length significantly moderated the relationship between sex and Vineland interpersonal relationship skills with an overall significant model result where the models explained 25.5% and 34.1% of the variability in Vineland interpersonal relationship skills, respectively (left hemisphere: R² = .26, F(3, 27) = 3.08, p = .044; right hemisphere: R² = .34, F(3, 27) = 4.66, p = .009). However, the path coefficients for the model testing left hemisphere alpha burst length as a moderator were not significant. Left and right hemisphere alpha burst counts per second also significantly moderated the relationship between sex and Vineland interpersonal relationship skills with an overall significant model result where the model explained 43.8% and 34.5% of the variability in Vineland interpersonal relationship skills, respectively ((left hemisphere: R² = .44, F(3, 27) = 7.02, p = .001; right hemisphere: R² = .34, F(3, 27) = 4.74, p = .009).

Left hemisphere alpha burst lengths trended toward significantly moderating the relationship between sex and Vineland coping skills where the model explained 24.1% of the variability in Vineland coping skills, R² = .24, F(3, 26) = 2.75, p = .063. Right hemisphere alpha burst lengths significantly moderated the relationship between sex and Vineland coping skills where the model explained 26.9% of the variability in Vineland coping skills right hemisphere: R² = .27, F(3, 26) = 3.19, p = .040. Left and right hemisphere alpha burst counts per second also significantly moderated the relationship between sex and Vineland coping skills where the model explained 49.9% and 33.3% of the variability in Vineland coping skills, respectively (left hemisphere: R² = .49, F(3, 26) = 8.66, p = .000; right hemisphere: R² = .33, F(3, 26) = 4.32, p = .013) (Table 4).

Table 4.

Moderation Model Path Coefficients

	Results
Model Path	Path Coefficient	SE	t	p-value
LHB CPS ABC Social Avoidance
Intercept	2.87	.56	5.13	<.001
Sex	−0.28	1.21	−0.23	.819
LHB CPS	−5.33	2.07	−2.57	.016
Sex X LHB CPS	−2.42	4.65	−0.52	.608
RHB CPS ABC Social Avoidance
Intercept	2.66	0.68	3.90	.001
Sex	0.12	1.51	0.08	.936
RHB CPS	0.38	2.07	0.18	.855
Sex X RHB CPS	−3.61	4.16	−0.87	.393
LHB Length ABC Social Avoidance
Intercept	2.80	.66	4.26	<.001
Sex	−0.54	1.42	−0.38	.706
LHB Length	0.15	0.14	1.06	.299
Sex X LHB Length	0.11	0.25	0.45	.659
RHB Length ABC Social Avoidance
Intercept	2.94	0.59	4.93	<.001
Sex	−0.00	1.27	−0.00	.997
RHB Length	0.25	0.15	1.65	.111
Sex X RHB Length	−0.01	0.31	−0.04	.971
LHB CPS VPL
Intercept	8.86	0.75	11.78	<.001
Sex	6.01	1.63	3.69	.001
LHB CPS	−5.09	2.82	−1.81	.082
Sex X LHB CPS	19.93	6.35	3.14	.004
RHB CPS VPL
Intercept	6.64	0.80	8.29	<.001
Sex	3.41	1.81	1.88	.071
RHB CPS	1.16	2.25	0.52	.611
Sex X RHB CPS	−20.85	4.68	−4.45	<.001
LHB Length VPL
Intercept	5.43	0.95	5.69	<.001
Sex	4.45	1.76	2.53	.018
LHB Length	−0.60	0.22	−2.78	.009
Sex X LHB Length	1.25	0.29	4.17	<.001
RHB Length VPL
Intercept	6.91	0.98	7.01	<.001
Sex	6.03	1.77	3.41	.001
RHB Length	0.69	0.28	2.48	.019
Sex X RHB Length	−1.22	0.44	−2.78	.009
LHB CPS VIRS
Intercept	5.53	0.77	7.15	<.001
Sex	5.12	1.39	3.66	.001
LHB CPS	−4.36	2.78	−1.57	.129
Sex X LHB CPS	18.13	5.45	3.33	.003
RHB CPS VIRS
Intercept	5.27	0.89	5.88	<.001
Sex	2.13	1.78	1.19	.241
RHB CPS	0.87	2.75	0.32	.754
Sex X RHB CPS	−9.53	4.58	−2.08	.047
LHB Length VIRS
Intercept	6.41	0.82	7.85	<.001
Sex	3.24	1.79	1.81	.082
LHB Length	0.08	0.16	0.51	.616
Sex X LHB Length	0.28	0.30	0.93	.359
RHB Length VIRS
Intercept	7.07	0.69	10.11	<.001
Sex	4.94	1.51	3.27	.003
RHB Length	−0.09	0.19	−0.50	.620
Sex X RHB Length	−0.81	0.38	−2.16	.039
LHB CPS VC
Intercept	8.98	0.61	14.66	<.001
Sex	2.64	1.31	2.02	.054
LHB CPS	−5.86	2.34	−2.51	.019
Sex X LHB CPS	18.67	5.18	3.60	.001
RHB CPS VC
Intercept	7.49	0.79	9.40	<.001
Sex	2.11	1.78	1.19	.246
RHB CPS	3.11	2.22	1.39	.174
Sex X RHB CPS	−13.32	4.62	−2.88	.008
LHB Length VC
Intercept	6.85	1.00	6.84	<.001
Sex	2.78	1.81	1.53	.137
LHB CPS	−0.48	0.22	−2.16	.041
Sex X LHB CPS	0.67	0.31	2.19	.037
RHB Length VC
Intercept	8.07	0.90	8.92	<.001
Sex	2.53	1.58	1.59	.122
RHB CPS	0.67	0.27	2.46	.021
Sex X RHB CPS	−0.92	0.41	−2.26	.033

Note. Abbreviations: RHB = right hemisphere burst; LHB = left hemisphere burst; CPS = counts per second; ABC = aberrant behavior checklist (FXS version); VPL = Vineland play/leisure; VIRS = Vineland interpersonal relationship skills; VC = Vineland coping skills.

3. Discussion

This study extends previous work by addressing spatial and temporal burst characteristics of the EEG signal, and the relation between dynamic analyses and typical neural phenotypes of FXS measured globally by closely characterizing the non-stationary nature of neural activity [10,13]. One novel finding demonstrated more affected right hemispheric cortical activity in FXS compared to TDC, and group differences in alpha power were largely driven by number of alpha events (burst count) rather than length of time spent in an alpha-driven state for males with FXS. Alpha burst length, on the other hand, may modulate compensatory mechanisms in females with FXS, who typically present with less severe or even no clinical deficits.

3.1. Alpha Bursts

Endogenous EEG alpha bursts index periodic times in the alpha state allowing for quantification of alpha dynamics in a manner that is spatially and temporally accurate [21]. Where alpha burst lengths reflect temporal aspects of alpha frequency use in FXS, average alpha burst counts per second provides access to dynamic recurrent alpha states with the capacity to assess lateralization of alpha usage [21]. FXS females had lower average burst counts per second compared to both TDC and males with FXS suggesting they enter alpha states less frequently. Individuals with FXS showed significantly fewer burst counts overall, an effect that was particularly strong in FXS females, which may be a supporting mechanism behind reduced global alpha power previously observed in FXS [11,13]. The interaction for burst counts per second between hemisphere and sex suggests 1.) lateralization of alpha-band frequency use may be more specific to females and 2.) FXS females experience less sensory processing-related deficits in the left hemisphere than the right.

The length of an alpha burst allows exploration of stability and flexibility of alpha-band frequency use [21]. In this study, there were significant sex differences in alpha burst lengths by group. Interestingly, females with FXS had significantly longer burst lengths compared to female TDC and males with FXS, indicating that females may modulate time spent in recurrent alpha states above typical levels. Increasing time spent in individual alpha states in females with FXS may reflect a compensatory mechanism to reduce neural hyperexcitability through high-low frequency coupling. Increased time spent in singular alpha bursts by females with FXS may contribute to potential differences in alpha power between males and females with FXS, where longer burst lengths compensate for decreased alpha burst count by attempting to reduce high frequency neural activity (i.e., gamma), supported by the negative correlation between alpha burst length and gamma power in the FXS group. At higher threshold levels, sex differences were localized to the left hemisphere, supporting the conclusion that burst length reflects a compensatory mechanism for potential deficits in the right hemisphere. In summation, females with FXS exhibit reduced burst counts per second with longer alpha bursts and males with FXS exhibit typical alpha burst counts per second but each burst is shorter than those of FXS females and typically developing males, which may reflect a mechanistic explanation for previous findings related to alpha power differences by group and sex for FXS.

Increased alpha burst length was correlated with greater interpersonal relationship skills for the right hemisphere but increased social avoidance for the left hemisphere, suggesting that hemisphere-specific utilization of alpha may be relevant to social development and approach/withdrawal behavior [28]. Moderation analyses were conducted to assess specific relationships between sex, brain activity, and social function, specifically supporting effects of alpha utilization on sex differences in play and interpersonal skills. Social avoidance scores were unrelated to sex in these analyses, but the relationship between ABCFXS social avoidance and left hemisphere alpha burst CPS remains suggesting dynamic utilization of alpha impacts social avoidance behaviors unrelated to sex in FXS. Increased alpha burst counts in the left hemisphere were correlated with decreased writing, coping, communication, and play/leisure skills, an effect primarily driven by FXS males. This finding suggests that shorter, more frequent bursts, while contributing to overall increases in alpha power measured at the more static FFT level, are not necessarily beneficial to general functional ability over multiple domains. This effect may be most apparent in the left hemisphere for the above domains due to its strong association with language and communication. Indeed, in males with FXS, increased burst count in the right hemisphere is associated with decreases in irritability and aggression and increases in behavioral flexibility, suggesting that alpha effects in this hemisphere may be tied to other domains such as mood regulation and inhibitory control. Alpha burst dynamics across both hemispheres significantly moderate sex-based differences in both play and leisure skills and coping skills, suggesting a mechanism by which females with FXS may engage more successfully with their social environment.

3.2. Asymmetry

FFT-based metrics capture general functional disruption but fail to address the dynamic nature of altered cortical oscillatory activity. In our case, FFT-based power measures accurately reflected previous findings in the literature and highlighted sex differences within FXS. Relative alpha power measured globally is typically decreased in FXS, which is also reflected by a lack of absolute alpha power differences between groups when all other frequency bands show elevated absolute power in FXS [11,13]. Both relative and absolute gamma power is typically elevated in FXS [13]. While our findings support these group relationships, the lack of significant group differences on frontal alpha asymmetry indicate that static power differences between groups do not reflect hemispheric differentiation. There were, however, correlations between alpha burst lengths in the left hemisphere and frontal asymmetry measures across theta, alpha, and low gamma in the FXS group, implying that alpha modulation may retain a more typical functional role in the left hemisphere in FXS. Indeed, opposing correlations between frontal alpha asymmetry and frontal gamma asymmetry with Go/Nogo errors reflect specific inhibitory control deficits related to neural hyperexcitability and thalamocortical modulation in right hemisphere.

3.3. ITC Chirp

Previous lateralization findings in fmr1^−/− KO mice support evaluation of hemispheric differences in chirp-evoked gamma phase-locking [15]. Current results support and extend prior findings of reduced ability to synchronize high-frequency neural activity in FXS suggesting increased sensory processing difficulties [10] differ in lateralization by sex. Males with FXS are equally affected by sensory processing difficulties (i.e., reduced ITC) across hemispheres whereas females with FXS are predominantly affected in the right hemisphere. Sex differences in sensory processing coupled with above average alpha burst lengths in females suggest protective compensatory mechanisms may recover left hemisphere sensory processing capacity, potentially reducing sensory hypersensitivity in females with FXS [11]. Individuals with ASD show gamma and alpha oscillatory deficits in right hemisphere [29,30] which are relevant to social symptoms characteristic of ASD [30]. This finding is supported here by a correlation specific to females with FXS between decreased ITC in the right hemisphere and more severe deficits on the SCQ, which assesses autism-like characteristics.

Chirp phase-locking was also related to behavioral changes. FXS individuals with better cognitive flexibility and fewer errors in response to distractor and no/go stimuli exhibited increased chirp phase-locking in the right hemisphere, suggesting that links between sensory processing fidelity and behavioral inhibition/control may be hemisphere-specific. However, this demarcation does have caveats, in that left hemispheric phase-locking deficits were also associated with increased ADAMS obsessive/compulsive scores.

3.4. Limitations and Recommendations

Bursts were assessed in the alpha range as supported by the literature [17]. The non-continuous state of the dataset limited capacity to apply burst calculation methods to lower frequency bands. However, theta bursts may also provide insight into theta’s role in thalamocortical modulation or compensation [31,32]. Additionally, females with FXS were underrepresented in this study relative to males; future studies should use continuous baseline data to explore lower frequencies and specifically over-recruit females to fully explore compensatory cortical oscillatory activity. Mechanistic claims regarding alpha bursting and its contribution to both overall power differences in the FXS brain, as well as its modulatory effects on other networks supporting sensory and social domains, will necessitate confirmation via translational animal model and human pharmacological investigation.

3.5. Conclusions

The current study addresses analytic approaches to cortical oscillatory activity that adds interpretive capacity to traditional stationary power metrics used previously to define neural biomarkers of FXS with electrophysiology. In particular, the relationship between alpha burst length, count, sex, and hemisphere may shed light on underlying mechanisms for previously observed alpha power abnormalities in FXS and their variation by sex. Alpha burst calculation represents an advancement in the exploration of FXS sensory processing dynamics to further biomarker development and translation to clinical-based interventions and mechanistic research. While many of the findings presented here are hypothesis-generating and will require confirmation both translationally and in larger samples, evidence for differential contributions to alpha power via number of bursts vs. lengths of bursts is of particular interest not only to FXS research, where alpha power plays a strong role in neuro-modulatory deficits [13] but more broadly in understanding how non-stationary dynamics of the EEG signal contribute to stationary power effects.

4. Methods

4.1. Participants

Participants were 36 adolescents and adults with FXS (Age: M = 24.8, SD = 9.1; age range 10–45; 12 females) and 40 age- and sex-matched typically developing controls (Age: M = 27.7, SD = 12.1; age range 12–57; 17 females)]. FXS and methylation status was determined using southern plot and PCR testing. FXS females (N= 12) are by definition all obligate mosaics. FXS males were primarily fully methylated (N = 14) but others were mosaic (N = 9; either due to methylation mosaicism or size mosaicism that includes an unmethylated premutation-range repeat count in some cells). Typically developing controls (TDC) had no current or prior history of neuropsychiatric illness. Task-related data from the auditory chirp task and participant set was previously published, and detailed demographic information is available in Ethridge et al., [10]. Exclusion criteria included history of seizures and current use of anticonvulsant and benzodiazepine medications due to the known EEG effects of both medications.

Clinical assessments were the parent-report measures Social and Communication Questionnaire (SCQ) [33], Anxiety Depression and Mood Scale (ADAMS) [34], Aberrant Behavior Checklist-Community (ABC-C, optimized for FXS) [35], the Vineland Adaptive Behavior Scales [36], and Child Sensory Profile [37]. We also administered the performance-based measures Woodcock-Johnson III Tests of Cognitive Abilities Auditory Attention subscale [38], and the computerized Test of Attentional Performance for Children (KiTAP) [39,40]. IQ was assessed by clinicians for both FXS and TDC participants with the Stanford-Binet Intelligence Scale 5th Ed. Abbreviated IQ [41] using deviation scores for calculating verbal and non-verbal IQ in the lower IQ range [42]. All participants provided written informed consent prior to participation, as approved by the Cincinnati Children’s Hospital Institutional Review Board.

4.2. Procedure

The auditory chirp stimulus consisted of a white noise carrier stimulus amplitude-modulated by a sinusoid linearly increasing in frequency from 0–100 Hz over 2000 ms. Chirp stimuli were presented 200 times separated by an inter-trial interval randomly jittered between 1500–2000 ms. Stimuli were delivered at 65 db SPL through headphones while participants watched a silent movie to facilitate compliance with testing procedures as in prior studies [7,9–10]. EEG data have been previously reported [7] with methods briefly described here for clarity.

4.3. EEG Recording

EEG was continuously recorded and digitized at 1000 Hz, filtered 0.01–200 Hz, referenced to Cz, and amplified 10,000x using a 128-channel Electrical Geodesics system (EGI, Eugene, Oregon) with sensors placed approximately according to the International 10/10 system [43,44].

4.4. EEG Preprocessing

Briefly, no more than 5% of sensors were interpolated for bad data; data were digitally filtered from 0.5 to 120 Hz with 60 Hz notch, submitted to independent components analysis via EEGLAB [45] for artifact removal, average referenced and epoched into 3250 ms trials (−500 to 2750 ms). Number of valid trials retained after artifact correction was higher for TDC compared to FXS for the chirp task (FXS M = 128.6, SD = 35.2; Control M = 154.8, SD = 30.5, t(72) = −3.4, p = 0.001), therefore trial count was evaluated as a covariate for all analyses and retained when significant.

4.5. Asymmetry EEG analysis

Using conventional electrode locations for frontal asymmetry analyses [16], data from F3 (left hemisphere) and F4 (right hemisphere) electrodes were extracted for asymmetry scores. Conventional asymmetry scores for power spectra were calculated on concatenated pre-stimulus periods via fast-Fourier transform (FFT) in the theta (4–7 Hz), alpha (8–13 Hz), and low gamma (30–55 Hz) ranges. Higher scores indicate greater right power. Greater alpha asymmetry scores (greater alpha power in right hemisphere) should be interpreted, however, to reflect greater left hemisphere neural activity due to the general role of alpha oscillations as inhibitory to neural processing, thus driving an inverse relationship between alpha oscillations and neural activity. Normalized difference scores (NDS), operationalized as (R-L)/(R+L), were computed to normalize for overall power, where a positive score reflects more alpha power in right hemisphere (i.e., increased neural activity in left hemisphere). The correlation between NDS and the natural log difference scores [i.e., ln(R)-ln(L)] were checked to confirm linearity. NDS correlated highly (rho > .9, p < .001) with the natural log calculation, as expected [19].

4.6. Exploratory Data Reduction for Alpha Bursts

Alpha burst metric methods reflect those proposed by Allen and Cohen [21], calculated using MATLAB R2018b (The Mathworks, Natick, MA, United States). Concatenated pre-stimulus data were bandpass filtered from 8–13 Hz. A Hilbert transform was applied to filtered data and instantaneous power at each timepoint was calculated as the natural log of the squared absolute value of the complex result. Data representing F3 were subtracted from F4 [power (Right) – power (Left)] to provide continuous asymmetry values. Endogenous lateralized alpha bursts were quantified by identifying the upper 75^th percentile from the absolute value of the asymmetry values and using that value as an individually-determined threshold for each participant. All positive values >= the 75^th percentile threshold for each participant were identified as right alpha bursts (greater relative right alpha power) and all negative values less than or equal to the 75^th percentile threshold values for each participant were identified as left alpha bursts (greater relative left alpha power). Count ratios were calculated by dividing right burst counts by left burst counts. Burst counts per second was calculated to account for differences in data length. All clusters containing 3 or more consecutive values above the threshold level were considered sustained bursts and entered into alpha burst length analyses, with the number of consecutive values quantifying length (Figure 1). Alpha burst NDS were computed to normalize for overall power to match FFT-based frontal alpha asymmetry calculations.

Alpha burst data were visually inspected to confirm burst count, length count, and assess the validity of a 75^th percentile value threshold. Alpha bursts appeared uniform with respect to typical oscillatory pattern where one burst represents one rise and fall in alpha power. However, there were cases where multiple rise/fall patterns occurred in succession without falling below the 75^th percentile threshold value implying that alpha bursts are relatively dynamic and justifying that inclusion of additional threshold values more conservative than the 75^th percentile. The 80^th, 90^th, and 95^th percentile values were calculated and applied to further explore the dynamic range of alpha bursts following the same protocol used for the 75^th percentile (see supplemental results in Supporting Information).

4.7. Chirp ITC

Intertrial coherence (ITC) was calculated in the same manner and using the same parameters to define low gamma phase-locking as Ethridge et al., [10], here applied to each hemisphere at electrodes F3 and F4.

4.8. Statistical Analysis

4.8.1. Group comparisons.

Average alpha burst length, alpha burst counts per second, FFT-based power measures from each hemisphere, and low gamma ITC were evaluated in four separate three-way repeated measures ANCOVAs with between subject factors of group and sex and repeated measure of hemisphere. All effect sizes are reported as partial eta squared. Trial length and age were assessed as covariates and retained when significant. To explore the effect of methylation status (fully methylated vs. partially methylated due to either methylation mosaicism or size mosaicism that includes a premutation-range repeat count in some cells) as a proxy for variation in FMR protein levels that contributes to neurophysiological variation within subgroup, alpha burst count and length were also compared via two-way repeated measures ANCOVA within the FXS male subgroup only. Females with FXS were not included in these analyses as they are obligate mosaics and would thus all fall in the same category, recapitulating sex stratification analyses above. Results for methylation status analyses are presented in the Supporting Information.

4.8.2. Within EEG, Clinical Correlations and Moderation Analyses.

We examined exploratory correlations between global theta, alpha, and gamma power, asymmetry scores (alpha asymmetry, theta asymmetry, and low gamma asymmetry), asymmetry burst count ratio, and asymmetry burst length with clinical variables using Spearman’s rho. EEG variables were also correlated with each other to examine physiological relationships. Correlations were exploratory and hypothesis generating, and thus not corrected for multiple comparisons. The relationships between sex differences and clinical outcomes relevant to social function were further investigated using moderation analyses, split by hemisphere, in which alpha burst metrics (burst count and length) were separately introduced as moderators for each clinical score (ABCFXS social avoidance, Vineland Play/Leisure, Vineland Interpersonal Relationships, and Vineland Coping Skills).

Availability of data and material

The datasets used and/or analyzed during the current study available from the corresponding author on reasonable request.