Background
Fragile X syndrome (FXS) is the most common single-gene cause of autism spectrum disorder with social anxiety and auditory hypersensitivity particularly common [
1‐
4]. Despite rapid growth in knowledge of molecular mechanisms from KO mouse studies, there are no known effective treatments for FXS. Development of translational biomarkers that quantify neocortical correlates of sensory sensitivity in FXS can facilitate drug discovery by identifying non-invasive indicators of disease pathology and track treatment response. EEG/event-related potential (ERP) studies, readily performed across species, are a promising and relatively unexplored direction for this purpose in neurodevelopmental disorders such as FXS.
Reduced local circuit inhibition has been proposed as a neural mechanism for sensory hypersensitivity and neural hyper-excitability in FXS [
5‐
7]. Specifically, gamma band activity has been associated with bottom-up sensory processing of stimulus characteristics [
8] and primarily reflects local circuit GLU/GABA interactions involving excitation onto and inhibition originating from parvalbumin positive (PV+) fast-spiking interneurons [
9,
10]. In
Fmr1 KO mice, prolonged persistent gamma activity or “UP” states have been associated with decreased glutamatergic drive onto fast-spiking GABAergic inhibitory neurons in sensory cortex [
5,
6]. Further, network inhibition during UP states is less synchronous, particularly in gamma frequencies. Broadened high-frequency tuning curves suggestive of increased nonspecific excitability also have been found in
Fmr1 knockout mice [
7]. Prolonged asynchronous UP states suggest that the ability to synchronize gamma power may be specifically reduced, but simultaneously net gamma power may be increased because of increased nonspecific excitability in the gamma range. These findings suggest a pattern of increased total high-frequency (gamma) neural activity but reduced temporally synchronous and spatially focused neural activity that may have broad neurobehavioral implications in addition to its impact on sensory processing [
11]. Since gamma is the primary working frequency range of the human auditory system, it is possible to investigate hypotheses generated by pre-clinical models in a relatively non-invasive fashion in FXS using auditory processing paradigms and EEG [
12].
Our previous findings showed significantly increased nonspecific gamma activity (gamma single-trial power) in FXS that was associated with a decreased ability to transiently synchronize evoked gamma (the “gamma spike” during early stimulus registration) and to habituate the neural response to repeated tones [
13]. Related findings have been reported in
Fmr1 KO mice [
14]. Our findings linking gamma power and sensory hypersensitivities in FXS patients suggest a convergence with FXS animal model findings of altered local inhibitory networks in and highlight the need for more systematic study of gamma-band neural activity in FXS.
One previously unexplored strategy is to drive sensory cortex with stimuli oscillating at gamma frequencies to examine the ability to synchronize neural responses to oscillating frequencies of sensory input. When presented with an amplitude-modulated stimulus oscillating at a certain frequency, neural networks oscillate in time with the stimulus frequency, effectively increasing the signal-to-noise ratio (SNR) for activity at that frequency in the local cortical network and thus in the EEG signal. The ability to “drive” cortical networks at desired frequencies to evaluate their functional integrity is especially important when studying high frequencies such as gamma, as neural oscillatory power is lower at high frequencies [
15]. Rather than studying a single frequency as in steady-state examinations, a “chirp” stimulus that increases linearly in frequency across the stimulus presentation period enables the study of neural synchronization across a broad range of input frequencies in a short amount of time.
The current study utilized a chirp stimulus to evaluate neural synchronization to sensory input across a range of frequencies from 1 to 100 Hz. Although relatively novel in neurophysiological research, chirp stimuli have been used successfully to drive and examine neural activity in the gamma frequency range in humans [
16] and rodents [
17], and they have been demonstrated to be sensitive to pharmacological manipulation [
18]. We predicted that FXS patients would be less able to synchronize neural networks oscillations to match the chirp stimulus that such abnormalities would be most pronounced in the gamma band and that this abnormality would be related to total background gamma power and clinical reports of sensory hypersensitivities.
Discussion
The current study utilized a neural entrainment approach for the first time in FXS to selectively drive networks oscillators in auditory cortex to clarify the extent of hypothesized deficits in local circuit inhibition in FXS. FXS patients demonstrated a marked reduction in the ability to synchronize (phase-lock) high-frequency neural activity to the chirp stimulus as well as the first harmonic (doubling of chirp frequency) suggesting impairments in the synchronization of networks at the primary and secondary levels of sensory processing. Both phase-locking abnormalities were highly associated with increased nonspecific gamma power, providing new evidence for a robust functional link between increased local network excitation reflected in raw gamma power and decreased ability to synchronize high-frequency population-level neural activity in sensory networks. The combination of deficits in entrainment coupled with hyper-excitation was specific to gamma, which coincide with highly selective deficits in neural synchronization related to local network excitation/inhibition balance previously identified in FXS translational models and proposed to underlie core clinical deficits in FXS [
5,
43].
Increased desynchronous high-frequency firing or “noise” (gamma STP) and decreased synchronized gamma activity to a driving stimulus indicate that both factors contribute to an overall worsening of auditory cortex SNR during sensory processing in FXS. FXS patients were also unable to selectively increase gamma power above baseline compared to controls, possibly indicating that ongoing gamma power is saturated. Decreased SNR reflecting heightened neocortical excitability suggests a plausible mechanism for the highly prevalent sensory hypersensitivity found in individuals with FXS. Indeed, increased gamma STP was correlated with parental reports of increased auditory hypersensitivity. Increased gamma STP was also associated with autism-associated social impairment using the SCQ, suggesting that cortical hyper-excitability may have broader clinical impact on behavior rather than selectively impacting sensory systems. We speculate that coherent high-frequency inhibitory network interactions play a crucial role in defining receptive fields that are important for sensory, perceptual, and cognitive processes.
Recent work in
Fmr1 KO rodent models have shown a shift toward lengthened activation states and decreased synchronization of fast-spiking interneuron-to-excitatory cell networks in visual cortex [
44] and reduced cross-frequency gamma coupling in hippocampus [
34], suggesting that fast-spiking interneuron alterations findings may not be restricted to a specific sensory modality but may represent a more widespread neocortical abnormality. Because of the potential for an alteration in bottom-up information processing to cause a wide range of neurobehavioral symptoms, high-frequency neural activity may also represent an important treatment target for alleviating multiple behavioral and sensory abnormalities in FXS.
To investigate whether ongoing network oscillatory properties contributed to decreased sensory system SNR in FXS, we examined amplitude-amplitude and phase-amplitude coupling between low frequencies and gamma activity during the pre-stimulus period before chirp presentation. Similar to findings in
Fmr1 KO mouse cortex [
34], we found a shift toward reduced influence of alpha frequency oscillations on gamma power and synchronization in FXS compared to controls. Cortical networks in healthy controls generally utilize alpha oscillations to control and inhibit local network excitation [
45]; in FXS, these networks may rely more consistently on slower theta waves or even delta waves (see Additional file
1) to modulate gamma frequency, given the lack of deficit in theta-gamma coupling. This strategy may have some success, given the moderate correlation between increased theta-gamma phase-amplitude coupling and higher nonverbal deviation IQ (Table
3) and the moderate correlation between delta-gamma amplitude-amplitude coupling and more normalized gamma power and phase-locking (Additional file
1), but it may be a less successful alternative and one with potential adverse sequelae on higher order neurobehavioral processes dependent on phasic delta/theta modulation, including processes associated with ASD-like behaviors (see Additional file
1: Table S2). This shift may also account for increases in theta power and connectivity seen in resting EEG in FXS [
32,
33,
46], increased gamma abnormalities during the chirp and increased pre-stimulus theta-gamma coupling.
We also identified increases in alpha frequency phase-locking during stimulus onset and offset and a decrease in theta frequency phase-locking synchronous with stimulus offset in FXS compared to controls. The increased alpha phase-locking is consistent with previous findings of increased amplitude early ERP components in response to brief auditory stimuli in FXS [
26,
40,
41]. Decreased theta frequency phase-locking simultaneous to stimulus offset is a novel finding in FXS. Given the role of theta-gamma coupling in network inhibition, this theta burst at stimulus offset for healthy controls may represent activation to prepare for the inhibition required to stop the gamma oscillations when the chirp stimulus stops, a mechanism which may be deficient in FXS.
Preclinical work in
Fmr1 knockout mice has related increased gamma excitability to decreased excitatory drive on fast-spiking inhibitory interneurons, resulting in increased and poorly synchronized pyramidal cell firing in the gamma range [
5]. Decreased activation of fast-spiking inhibitory interneurons, which synchronize gamma activity via projections from inhibitory neurons onto multiple pyramidal neurons, has been proposed as a mechanism for heightened neocortical excitability in FXS [
5]. Poorly organized inhibitory drive onto pyramidal cells in auditory cortex from fast-spiking interneurons that control gamma synchronization could account for the pattern we observed of increased gamma power, suggesting increased high-frequency firing of excitatory pyramidal neurons (noise) but with less coherent organization (phase-locking) in response to sensory stimulation. This pattern of alterations may contribute to previously reported decreases in transient gamma phase-locking [
13]. Given the similar network dynamics observed in
Fmr1 rodent models [
5,
7,
14,
44], findings reported here may not only extend mouse model concepts to FXS patients, but suggest that neurophysiological measures may be useful for tracking this local circuit deficit in both mouse models and patients to foster direct translational drug development for this neurodevelopmental disorder.
Certain study limitations should be considered. First, comparative work is necessary to determine the degree to which our findings occur in other neurodevelopmental disabilities, several of which have associated sensory sensitivities as seen in FXS [
47]. While evidence exists for impaired phase-locking ability, increased power, and alpha-gamma phase-amplitude coupling abnormalities in other disorders such as ASD [
48‐
50], these studies primarily describe broad-band power abnormalities and phase-locking deficits, whereas the most salient abnormalities in FXS were seen in the gamma band (but see [
51] for gamma deficits in very young boys with ASD). Berman et al. [
52] also report increased alpha-gamma phase-amplitude coupling in children with ASD, an effect not replicated in FXS, although whether these differences are due to the clinical population or the age of the participants is unknown. Intellectual disability (ID) associated with FXS may also play a role in group differences, suggesting a role for these findings in other disorders associated with ID. A pattern of low-frequency power enhancement at rest has been primarily reported in ASD only for those with ID [
52]. A shift from alpha to theta band activity has been reported in Down’s syndrome [
53]; however, increased theta activity in children with ADHD is robust in low- and high-IQ individuals [
54]. Decreased or normal gamma power in individuals with ID is more commonly reported than the increased power seen in FXS [
55,
56]. Comparative studies between FXS and other forms of ID will determine the specificity of the high-frequency disruptions reported here and their relation to low-frequency function. Second, FXS participants were taking various psychiatric medications, and their potential impact on the data cannot be ruled out. However, excluding medicated patients necessarily excludes a subsample with more severe behavioral problems limiting study representativeness, and the medications are well studied in psychiatric populations without known effects as we observed in our patients. Third, our failure to detect gender differences should be considered in the context of the limited number of FXS female participants. Analyses of the male participants were largely consistent with the findings in the full sample, but future studies with larger samples of female participants should clarify gender differences in patients with FXS. Lastly, future research is needed to investigate the sensory and perceptual consequences of the identified neurophysiological alterations.
Conclusions
Despite extensive progress in understanding the genetic alteration resulting in FXS and its biochemical and local circuit functional consequences, far more modest advances have occurred in disease understanding at the systems neuroscience level in affected individuals [
57]. Further, no treatments have demonstrated clinical efficacy for FXS, and translation from rodent models to human pharmacology has been constrained by a lack of translational biomarkers to evaluate, predict, and understand drug response. The current study provides novel evidence that alterations in synchronous activity in fast-spiking inhibitory interneurons, widely reported in translational models of FXS, may contribute to important neurophysiological alterations which are measurable at the systems level in humans with FXS and are of clinical relevance. Our previous data showed decreased gamma network inhibition in evoked responses during habituation to repeated stimuli [
13] and increased resting gamma power [
33]. Our current report indicates altered gamma-band neural entrainment to oscillatory chirp stimuli which parallel preclinical findings in
Fmr1 KO rodent slice preparation [
8] and evoked electrophysiology research [
14,
44]. Our results indicate that abnormalities in sensory neurophysiology not only are clinically relevant but may be useful for in vivo studies of FXS patients to track alterations in fast-spiking cell populations known to be altered in KO mouse models as a biomarker for predicting and evaluating response to novel therapies. Although additional investigations of test-retest reliability, task effects such as stimulus and ISI length, age effects, and sex effects are warranted, gamma deficits may serve as biologically grounded and clinically relevant outcome measures in future clinical trials for FXS.
Acknowledgements
The authors would like to thank Rachel Greene, Savanna Sablich, and Melanie Soilleux for aid in data collection.