Alterations in resting-state gamma-activity is adults with autism spectrum disorder: A High-Density EEG study

Autism spectrum disorder (ASD) is a neurodevelopmental disorder with a wide range of symptoms that include deficits in social cognition and difficulties with social interactions. Neural oscillations in the EEG gamma band have been proposed as an important candidate neurobiological marker of higher order cognitive processes and social interactions. We investigated resting-state gamma-activity of patients with ASD (n=23) in order to delineate alterations as compared to typically developing (TD) subjects (n=24). EEG absolute power was examined in the gamma (30-100Hz) frequency band. We found significantly reduced spectral power across the entire gamma range in the ASD group. The decrease was most pronounced over the inferior-frontal and temporo-parietal junction areas. We also found a significant decrease in gamma-activity over the dorsolateral prefrontal cortex, especially in the left side. Since these brain areas have been associated with social functioning, the reduced gamma-activity in ASD may represent a cortical dysfunction that could underlie a diminished capacity to interpret socially important information, thereby interfering with social functioning. The alterations we found may lend support for an improved diagnosis. Furthermore, they


Introduction
Autism spectrum disorder (ASD) is typically starts in infancy or early childhood, and the majority of ndividuals with autism exhibit functional impairments for most of their lives.Even though ASD is characterized by heterogeneous symptoms that range across multiple domains, social cognitive impairments and social interaction problems were shown to play a prominent role in determining the level of disability in ASD.ASD may substantially reduce an individual's ability to function in school, work, and other areas of life.Moreover, ASD comorbidities amplify the burden to the society and afflicted individuals alike (Al-Kindi et al. 2020;Hodges et al. 2020;Leigh and Du, 2015).
The identification of the core neurobiological mechanisms that may underlie the social cognitive impairments and social interaction problems in ASD has been the target of intensive reseach for several decades.Based on accumulating evidence summarized by Westby (Westby, 2014), a promising approach to understand the neurobiological underpinnings of ASD is to use the social neuroscience paradigm, as proposed by Cacioppo and Bertson (Cacioppo and Berntson, 1992) and Baron-Cohen (Baron-Cohen et al. 1997;Petinou and Minaidou, 2017;Swenson, 2013), and interpret the findings in the context of social Theory of Mind (ToM) that encompasses two subdomains, including the cognitive and the affective ToM.So far, research in social neuroscience uncovered certain relationships for various brain areas in terms of their involvement in atypical social behavior as evidenced in ASD etiology and symptom presentation (Dinstein et al. 2011;Tomalski et al. 2009).The potential involvement of the prefrontal cortex, including the inferior frontal gyrus (IFG) as well as temporo-parietal junction (TPJ) in the ASD-related syndromes received support from data obtained in experimental conditions which require neural processing related to people's emotions, thoughts, and beliefs (affective and cognitive ToM, respectively) (Petinou and Minaidou, 2017;Shamay-Tsoory et al. 2009).
Additionally, there has been a great deal of progress in terms of identifying potential neurophysiological correlates of social interactions.An important emerging finding pertains to neural oscillations in the gamma frequency band, which have been shown to be an important candidate neurobiological marker of higher order cognitive processes and social interactions.Gamma oscillations are often defined as electric or magnetic activity in the frequency range between 30 to 80 Hz, although there is increasing awareness of even higher gamma frequencies (80+ Hz) which may be relevant from the perspective of cognitive processes (Uhlhaas and Singer, 2006;Uhlhaas and Singer, 2013).
Potential neurochemical and cellular mechanisms of alteration of gamma-activity in ASD have been put forward in the literature (Rojas, 2014).Previous experimental and theoretical work provided support for the notion that γ-aminobutyric acid (GABA)-ergic neurons play a pivotal role in the generation of high-frequency neural oscillations and their local synchronization (Uhlhaas and Singer, 2013).Reduced (GABA)-ergic interneuron cell numbers, which result in lower gamma-activity, represent a common finding in animal models of autism (Fisahn et al. 2009).Moreover, there is human imaging evidence for reduced GABA concentration in frontal, motor and auditory cortices (Sohal et al. 2009;Tang et al. 2009) in ASD.It is therefore conceivable that GABA-ergic deficits represent a common molecular pathway impacted across multiple autism etiologies (Zhao et al. 2021).Overall, based on these findings an association between GABA-ergic deficits in ASD with decreased gammaactivity can be postulated.
It is also noteworthy that high frequency (gamma) oscillations are intimately related to higher order mental processes including the binding of sensory features into a coherent percept ( (Engel and Singer, 2001;Tallon-Baudry et al. 1996), object representation (Bertrand and Tallon-Baudry, 2000), attention (Fell et al. 2003), memory (Herrmann et al. 2004;Lisman and Idiart, 1995), and consciousness (Llinas and Ribary, 1993).Furthermore, gamma oscillations in the superior temporal sulcus (STS) have been implicated in nonverbal emotional communication (Symons et al. 2016) and the coupling of STS/STG activity may play a key role between pairs of communicators (Stolk et al. 2014).Based on the results of a MEG study, gamma-activity is also playing a critical role in social interaction (Pavlova et al. 2010).
Although gamma-activity is associated with perceptual and social cognitive functions that are compromised in autism, alterations of gamma oscillations in ASD have received little attention in the literature.Most prior studies of gamma-activity in ASD were conducted under varying task conditions.Whereas these studies suggest decreased gamma-activity across multiple brain regions, their findings are difficult to integrate due to their heterogenous methodology (e.g., studies included visual and auditory perception tasks, and various higher order cognitive processes [for a summary of findings, see (Rojas and Wilson, 2014)]).
While decreased task-related gamma-activity was found in the majority of studies in ASD, the applied tasks can place a substantial burden on the patients when we consider a potential application in clinical practice.By contrast, resting-state examinations, which do not require subjects to make a response (Wang et al. 2013), are, by definition, task-free.Therefore, they may be more suitable for studies of low-functioning individuals than task-based experimental designs (Cornew et al. 2012).This aspect is especially promising for studying more severely impaired and/or younger patients, who may not be able to perform tasks accurately because of cognitive, physical, or developmental challenges.It should be noted that some data suggests that task-free activity at rest predicts individual differences in brain activity during task performance, based on a strong correlation between task-specific and resting networks (Tavor et al. 2016).
Yet, we found only six studies of resting-state gamma-activity in ASD (Table 1).Overall, as shown by the Table, the findings were equivocal, with a slightly greater preponderance of the findings of decreased gamma (Maxwell et al. 2015;Neuhaus et al. 2021;Sheikhani et al. 2009;Sheikhani et al. 2012) as compared to gamma increases (Cornew et al. 2012;Orekhova et al. 2007).However, "resting-state" studies also varied considerably in their approaches (1 MEG and 5 EEG studies); in the gamma frequency range they focused on (only 2 of the EEG studies examined high frequency gamma-activity, i.e., in the >50 Hz range); and in the resting-state conditions that they used.In fact, two of the restingstate studies applied an attention task (Neuhaus et al. 2021;Orekhova et al. 2007) during the EEG recording and three additional studies had EEG in eyes-open condition (Maxwell et al. 2015;Sheikhani et al. 2009;Sheikhani et al. 2012); and there was only 1 study with eyes-closed condition, which used MEG, (Cornew et al. 2012).Furthermore, it is noteworthy no data are currently available from adults with ASD.
Our aim was to fill the above knowledge gaps by investigating resting-state gamma oscillations in order to delineate potential alterations in ASD in the intrinsic activity of the neural networks which have been linked to social cognitive functioning.The high temporal resolution of the densearray EEG recordings allowed for the examination of the oscillatory dynamics in the full gamma spectrum, including both the low and the high frequencies, which typically were not investigated previously.This is essential since high gamma oscillations are associated with a variety of cognitive functions, and are crucial for the integration of information across neural networks (Castelhano et al. 2014).The high-density EEG was adopted for a delineation of the group differences in terms of brain areas which may be associated with the pathophysiology of ASD (e.g., gamma-activity over TPJ, IFG and the dorsolateral prefrontal cortex [DLPFC] region).Moreover, to explore the relationship of gamma-activity with ASD symptoms, we also examined the association of gamma-activity with the severity of psychopathology as measured by the Autism-Spectrum Quotient (AQ).

Participants
Participants were 47 adults (23 with ASD and 24 age-matched TD controls, aged 18-65 years (mean age =29.04,SD =10.2 in the ASD group; mean age =32.9, SD =13.7 in the TD group).All adults with ASD were diagnosed by a psychiatrist according to the DSM5 criteria (see Online Supplement, Part A/ Supplement to Participants, Diagnosis).Participants with ASD were recruited through the Autism Advocacy Association, Budapest, Hungary.An IQ score of >85 was required in order to be included in the ASD group.Lack of history of psychiatric disorders was required for the inclusion in the TD group.Subjects with TD were recruited from the local community, and clinical staff at the University.Pairs of ASD and TD subjects were matched individually to each other by age and level of education.All subjects with ASD were medication-free.All participants gave written informed consent.The study was approved by the Institutional Research Ethics Committee of Semmelweis University and was conducted according to the Declaration of Helsinki.

Measures
The 90-item Symptom CheckList Revised (SCL-90R) (Derogatis et al. 1973) was used to select controls with no history of psychiatric disorders.The severity of general psychopathology and psychological distress were characterized by the total score on the SCL-90R scale.ASD specific symptoms were assessed by the Autism-Spectrum Quotient Test (Baron- Cohen et al. 2001a;Baron-Cohen et al. 2001b;Bishop et al. 2004;Woodbury-Smith et al. 2005) and the Toronto Alexithymia Scale (TAS-20; (Bagby et al. 1986;Bagby et al. 1994)).AQ is a diagnostic questionnaire for measuring the expression of autism-spectrum symptoms based on subjective self-assessment.The questionnaire comprises 50 items, grouped into five subscales: Social Skill, Attention Switching, Attention to Detail, Communication, and Imagination.The AQ has been demonstrated to be a reliable and valid tool (Armstrong and Iarocci, 2013;Hoekstra et al. 2008;Woodbury-Smith et al. 2005), with higher scores indicating a higher level of psychopathology with respect to autistic traits (Woodbury-Smith et al. 2005).The Toronto Alexithymia Scale, a 20-item self-report scale (total score range:20-100), is the most frequently used measure of alexithymia in adults in prior research of autism (Vaiouli and Panayiotou, 2021).Higher scores indicate higher severity of alexithymia (Parker et al. 2003).Theory of Mind (ToM) functioning was assessed by the "Reading the Mind in the Eyes Test"(RMET), developed by Baron-Cohen and co-workers (Baron-Cohen et al. 1997;Baron-Cohen et al. 2001a).RMET is a widely used test to measure mentalization, empathy and social sensitivity (Kirkland RA, 2013;Oakley et al. 2016).

EEG recording and pre-processing
During the EEG recording sessions, participants sit in a dimly lit room and they were asked to remain still with their eyes-closed for 3 minutes.EEG data were obtained by a 128-channel active electrode system (Biosemi Inc, Amsterdam, Netherlands) at a sampling rate of 512 Hz.The signal was bandpass filtered of 0.5-100 Hz using the average reference.Off-line data storage and analysis was done by using the Electromagnetic Source Signal Imaging (EMSE) Suite and the Statistical Analysis System (SAS 9.4) software.The EEG was re-referenced off-line to the common average potential and filtered between 0.5 and 100Hz with a zero-phase shiftforward and reverse IIR Butterworth-filter.Our principal analyses focused on absolute power values in the full gamma frequency band, ranging between 30 and 100Hz.The full gamma frequency band was then subdivided into two symmetrical frequency bands, 30-48 Hz (low gamma) and 52-100Hz (high gamma), henceforth low gamma and high gamma.Power spectrum data in these gamma ranges were further processed in second level analyses.For further details regarding EEG recording, artifact rejection and preprocessing see Online Supplement (Part B/ Supplement to EEG recording and pre-processing).

Statistical Analyses
The statistical analyses were based on the random-regression Hierarchical Linear Model (HLM).Repeated measurements of spectral amplitudes (log-transformed µV 2 units) across the gamma frequency range were used as dependent variable in the HLM.Group membership (ASD, TD) served as independent variable.Age and gender were included as covariates in the analyses to adjust for a potential confounding by these variables.We conducted a separate analysis was for each of the 128 EEG recording channels, and used the False Discovery Rate (FDR) procedure (Benjamini et al. 2001) in order to correct for the alpha-error inflation due to multiple testing.A similar approach has been used in prior literature (Wang et al. 2021).The analyses were carried-out according to a hierarchical top-down approach.First, the analyses were conducted for the full-gamma range (30-100Hz).Then, we performed analyses separately for the low gamma (30-48Hz) and the high gamma range (52-100 Hz), respectively.
The results of the analyses were summarized in a tabular format, and visualized in the form of topographical maps, which were generated by using the EEGLab software (Delorme and Makeig, 2004).The topographical maps were created to illustrate the gamma-activity (in original µV 2 units) in both groups separately (i.e., as group averages).In addition, to illustrate the magnitude of the group diffences (i.e., effect-size) in standardized statistical units, we computed the Cohen D values for effects size at all EEG electrodes, which were then portrayed in the form of topographical maps.Specifically, for an interpretation of Cohen D measure we adopted the following standards from the literature: values approx.0.3, 0.5, and >0.7 were considered to be small, medium and large effect sizes, respectively (Cohen 1988).Finally, topographical maps were generated to display the FDR-corrected levels of statistical significance for the group difference at each electrode.
In further HLM analyses, we examined whether alterations in the gamma-activity were associated with the level of psychopathology in the ASD group as measured by the AQ-test.Age and gender were included as covariates in the analyses of the association between gamma-activity and the level of symptom severity.These analyses were conducted for those brain areas of interest (including the TPJ, IFG and the DLPFC) where a statistically significant alteration in gamma-activity was detectable in the ASD group.(For the definition of scalp areas of interest see Online Supplement (Part C/ Statistical Analyses, potential scalp areas of interest).In the respective HLM analyses, the total score in the AQ test was included as an independent variable.To delineate the association, we computed the LSmean estimates for gamma-activity for various levels of psychopathology (i.e., total scores on the AQ-test).

Demographics
Demographic and clinical data on study participants are presented in Table 2.The ASD and TD subjects were similar in age and years of education, with no significant difference on these two variables (p=0.28 for age and p=0.49for years of education).All ASD and all control subjects were having a job, or were at school at the time of the study.Moreover, the two samples were similar in terms of the proportion of having a job vs. currently attending school (61% [14 of 23] and 63% [15 of 24], respectively, in the ASD and TD samples).The difference between the two study groups approached statistical significance regarding gender distribution (p=0.08),reflecting the fact that the ASD group predominantly consisted of male subjects while the TD group had an approximately balanced gender distribution.The ASD subjects manifested significantly higher total score on the AQ-scale.Furthermore, significantly (p<0.05)higher severity was found in the ASD group in terms of the level of alexithymia, as measured by the TAS-scale; and in the level of psychopathology and psychological distress as indexed by the total score on the SCL-90R scale.In addition, a numerically lower performance was detectable in the ASD group as compared to TD regarding the % correct recognition rate on the RMET.

Full gamma spectrum (30-100Hz)
As indicated by Table 3, gamma-activity was significantly (p<0.05)diminished across the entire gamma range at multiple electrode locations in the ASD as compared to the TD subjects.In each panel (LEFT/RIGHT/MIDDLE) of the table, the columns display separately for both groups the numerical values for the mean and the standard error of the log-transformed spectral power values in the full gamma range.The next two columns after the spectral power values indicate the magnitude of the group difference in terms of Cohen D effect-size; and the FDR-corrected pvalues, adjusted for multiple testing.Type I errors (adjusted p-values) reaching statistical significance (p<0.05) are highlighted in the table in bold.As shown by Table 3, 79 (61.7%) of the 128 comparisons were statistically significant after the FDR-correction for multiple testing.Additionally, it should be noted the Cohen D values exceeded the medium effect-size for a high proportion of the recording electrodes.The mean (standard deviation) for the Cohen D estimates across the 128 electrodes was 0.57 (SD=0.25).
As shown by Figure 1 (upper portion/panel A), in order to visually examine the topographical distribution of the diminished gamma-activity, we created respectively the topographical maps of observed gamma-activity in the ASD and TD group in µV 2 units.We also depicted the maps of the group differences in terms of standardized statistical effect-size ('Cohen D') units and the scalp distribution of the adjusted p-values for statistical significance ('FDR p').As indicated by the figure, while the gamma-activity generally exhibited a reduction over the entire scalp in the ASD group, the reduction had a rather distinctive topographical distribution, with a preponderance of the changes in the right hemisphere.Specifically, the most pronounced reduction occurred at electrodes located over the right TPJ, IFG and DLPFC, although significant differences were observable in the left hemisphere at the TPJ, IFG and particularly the left DLPFC.Besides the aforementioned differences, we found marked reductions in gamma-activity bilaterally over the left posterior superior temporal gyrus.

Low gamma (30-48Hz)
The visual summary of the observed data and the results of the statistical analyses for the group comparisons are presented by the middle portion of Figure 1 (panel B).As the Figure indicates, the differences in the lower gamma frequency range showed a topographically specific distributiton over the scalp.As revealed by the scalp map of the Cohen D and FDR-corrected p-value statistics, the most pronounced differences in terms of statistical effect-size and FDR-adjusted p-values were observable over the IFG and TPJ areas especially over the right hemisphere; and the prefrontal regions over the left hemisphere.In the low gamma range, the mean (standard deviation) for the Cohen D estimates across all electrodes was 0.38 (SD=0.21).

High gamma (52-100Hz)
The lower portion of Figure 1 (panel C) provides the summary of the results of the observed data and group comparisons for the high gamma frequency range.As revealed by the Figure, the topographical pattern of group differences in the high gamma range was similar to the pattern in the full gamma range.Specifically, the differences followed a spatial distribution which involved the IFG and TPJ areas with preponderance on the right hemisphere and the involvement of the prefrontal region on the left side.However, as shown by the Cohen D and FDR-corrected pvalue statistics, the differences were observable over large brain areas surrounding the aforementioned regions of interest.In the high gamma range, the mean (standard deviation) for the Cohen D estimates across all electrodes was 0.58 (SD=0.23).

Association of gamma-activity with the severity of psychopathology
In the regions of interest, including the TPJ, IFG and DLPFC, our result revealed significant differences between the ASD and TD groups.While the reduction in gamma-activity (p<0.05) in the TPJ and IFG areas was observed bilaterally, the group differences were more prononounced in right hemisphere for these regions.With respect the DLPFC the association was stronger in left hemisphere.In further analyses, we investigated whether the severity of psychopathology, as measured by the AQ total score, was associated with the reduction in gamma-activity in the above areas of interest.We found that the severity of psychopathology showed an inverse relatioship with the gamma power in the 30-100Hz frequency range: higher total score on the AQ scale was associated with lower gamma power for all areas of interest (Figure 2 and Online Supplement, Part D/ Supplementary table).Moreover, based on the steeper decline of the estimated gamma-activity with increasing symptom severity (shown by the regression line in Figure 2), our results showed that the relationship was markedly stronger in the right hemisphere than in the left for the TPJ and IFG area, whereas for the DLPFC the association was stronger in the left hemisphere.

Discussion
The current study investigated resting-state gamma oscillations in the EEG in order to delineate potential alterations in ASD in the intrinsic activity of the neural networks, which have been linked to social cognitive functioning.We expected that in ASD, where problems of social cognition and interactions feature prominently, alterations in gamma-activity would be detectable in multiple brain areas, including the right TPJ and IFG, and the DLPFC.Moreover, we assumed that the examination of the association of low and high frequency gamma-activity with the severity of symptoms would provide insights into the specific brain mechanisms that are responsible for the social interaction difficulties in ASD.We found reduced spectral power across the entire gamma range in the ASD group.The decreased gamma-activity in ASD exhibited a distinctive topographical pattern.Specifically, the decrease was most pronounced at electrodes over the right TPJ and the right IFG; in addition, significant reduction of gamma-activity was also detectable in the left hemisphere at the TPJ and IFG.Besides these changes, we found significant decrease in gamma-activity over DLPFC, especially in the left side.However, in order to interpret the topographical distribution of our findings in terms specific brain areas, one has to bear in mind that the sources of EEG activity cannot be reliably inferred from scalp topography alone, i.e. the inverse problem is ill-posed (see Nunez et al, 2016).
The strong decrease of gamma-activity in patients with ASD provides converging electrophysiological evidence to the neurochemical and neurophysiological abnormalities found by prior studies in ASD (Uhlhaas and Singer, 2006).In particular, several studies reported impairments at the GABA-ergic interneurons, a finding which is of particular importance since this neurotransmitter system is involved in the generation of cortical gamma oscillations (Zhao et al. 2021).Specifically, the network of GABA-ergic interneurons acts as a pacemaker in the generation of high frequency oscillations, by producing rhythmic inhibitory postsynaptic potentials (IPSPs) in pyramidal neurons; these IPSPs, in turn, are considered to synchronize the firing of a large population of pyramidal neurons.GABA-ergic interneurons that express the Ca 2+ -binding protein parvalbumin (PV) are of particular relevance since (1) these are fast-spiking cells whose activity may be causally related to the generation of gamma oscillations; and (2) among the neurons most affected in ASD are those expressing the Ca 2+ -binding parvalbumin (Sohal et al. 2009;Uhlhaas and Singer, 2013).Overall, considering the close relationship between the GABA-ergic system and gamma oscillations, our findings of reduced gamma-activity in ASD are consistent with several lines of evidence for impairments in ASD.These include reduced GABA concentration in the frontal, motor and auditory cortices (Wang et al. 2013), a significantly reduced glutamic acid decarboxylase expression in various brain regions in postmortem autism samples (Zhao et al. 2021), and reduced interneuron cell numbers in the animal models of autism (Filice et al. 2016;Sohal et al. 2009;Uhlhaas and Singer, 2013).
Our findings highlight the predominance of right-hemisphere alterations in the TPJ and IFG areas in ASD.This is consistent with evidence for the predominantly right-lateralization of social cognitive functions in the human brain.Specifically, with respect to the lateralization of social cognitive functions in TD subjects, most evidence for TPJ's involvement in mentalizing pertains to the right TPJ (Koster-Hale et al. 2017;Lombardo et al. 2011;Van et al. 2015) (although in certain mentalization tasks the TPJ appears to be activated bilaterally (Santiesteban et al. 2015;Ye et al. 2015;Young et al. 2010)).Right-lateralized dysfunctions in the TPJ area have been found in various investigations in ASD (Ahmad et al. 2021).Furthermore, a prior study which used transcranial direct stimulation (tDCS) delivered to the right TPJ revealed an improvement in behavioral problems associated with ASD (Salehinejad et al. 2021).
Importantly, a study by Maxwell and collegues (Maxwell et al. 2015) found in children and adolescents with ASD a reduction in resting-state gamma power, which was specific to the righ lateral hemisphere.Overall, our results are congruent with Maxwell et al.'s finding, which, similar to our study, was based on a high functioning sample with ASD.Our findings extend the previous results to an adult sample, which included individuals from both genders (unlike Maxwells's study which included only males).They also provide evidence for left hemisphere alterations over the IFG, TPJ and DLPFC areas.We note that in Maxwell et al's study a non-significant numerical reduction of the left hemisphere gammaactivity was also observed in ASD.The fact that it did not reach statistical significance could have been due to the lower sample size and poorer spatial resolution in that study.
With respect to the potential neurochemical background of the hemispheric aysmmetry in the gamma-activity in ASD, we are aware of one study that examined animal models of ASD in terms of the inter-hemispheric distribution of parvalbumin containing GABA-ergic cells, which participate in the generation of gamma oscillations (Gogolla et al. 2009).This study revealed hemispheric asymmetry in the number of parvalbumin containing cells.This can underlie, at least in part, an inter-hemispheric difference in gamma-activity, but the data require confirmation in humans.
Our study also revealed a reduction of gamma-activity in ASD at the DLPFC.To interpret this finding, we should consider that the DLPFC plays an important role in executive functions (EFs) and higher cognitive functioning (Powell and Voeller, 2004).EFs include the processes that underlie the conscious control of thought, emotion, and action, which in turn are central to the management of one's day-to-day life.ASD is associated with various impairments across a wide range of EFs, including attention, working memory, and emotional processing.Because gamma-activity was shown to be associated with EFs in several studies, the reduced gamma-activity at DLFPC in ASD may be considered as underlying these impairments.This is further supported by the finding that the decreased gamma-activity in our study was closely related to the severity of autism symptoms, as measured by the AQ scale, which besides dysfunctional social skills, captures deficits in EFs.
Even though the functional significance of the differentiation between low and high frequency gamma-activity is still debated, it is noteworthy that according to Uhlhaas (2011) such a distinction is essential since they reflect different physiological processes.This is supported by evidence that (1) both low and high gamma-band oscillations undergo distinct and non-overlapping frequency modulations (e.g., (Colgin et al. 2009;Crone et al. 1998;Hoogenboom et al. 2006)); and that (2) low and high gamma-band oscillations can correlate with different cognitive processes (Vidal et al. 2006;Wyart and Tallon-Baudry, 2008).In addition, while both low and high gamma-band oscillations can be generated by recurrent inhibition, they differ in their relationship to the spiking activity of PV-interneurons; their pharmacological modulation profiles; and layer specificity (Oke et al. 2010;van der Meer and Redish, 2009).
The alterations in the low and high gamma range showed an overlapping, but at the same time a distinctive pattern.In particular, the pattern of alterations in both gamma bands included the IFG and TPJ regions with right hemisphere predominance, and the involvement of the DLPFC.Nonetheless, it should be noted that the alterations in the high gamma range covered considerably more extended brain areas.This was reflected by the median Cohen D value, which was substantially larger in the high-gamma range than in the low range.Since high gamma-activity is considered to reflect synchronization in local neural cell assemblies in the brain, while lower frequencies are thought to be associated with long-range synchronization (Rojas and Wilson, 2014), our findings regarding high gamma-activity may be interpreted as a widespread dysfunction in local synchronization of neural activity in multiple brain areas.On the other hand, the reduction of low gamma-activity in ASD may reflect a dysfunctional sychronization among more distant brain regions.

Limitations
Our study has several limitations.First, while the findings revealed a pronounced reduction of gamma-activity in subjects with ASD as compared to TD, they were based on a relatively small sample of umedicated subjects which necessitates replication in further studies, using a more representative sample that may be subjected to pharmacotherapy.Second, in the autism spectrum we focused on high functioning subjects.This may limit the generalizability to a more impaired patient population; thus, our study needs to be extended to samples with a broader representation of individuals with ASD.The fact that ASD subjects in our sample, similar to some of the previous studies (Sheikhani 2009(Sheikhani ,2012;;Cornew 2012), were medication-free may further restrict generalizability.Third, male subjects were overrepresented in the ASD sample.While our analyses were adjusted statistically for the inbalance by applying gender as a covariate, we think that the examination of larger samples that would allow for a stratificaton by gender would be necessary in future studies.Fourth, since deficits in social cognition and interactions constitute hallmark symptoms of ASD, and gamma-activity has been shown to be closely associated with social functioning, we focused specifically on the gamma-activity.Thus, further studies should examine the specificity of the associations as compared to other EEG frequency ranges.Fifth, we note that while the Cohen D effect-sizes for several scalp areas are considered high according to accepted benchmarks, these values are established by statistical criteria; therefore, their potential clinical significance (e.g., diagnostic sensitivity and specificity) need to be evaluated in further studies.Finally, the myogenic contamination in the gamma band, could bias the findings and lead to a group difference in this frequency-range.This issue typically arises in studies with an increased gamma-activity in ASD (e.g., Orekhova (Orekhova et al. 2007)).We think, however, that since our results showed reduced gamma-activity, it is unlikely that the group differences in our study were due to a contamination by myogenic artefacts in the ASD sample.

Conclusions
Our findings identified reduced gamma-activity in ASD in several brain areas of interest that have been implicated in social functioning in typically developing individuals.Deficient gamma-activity in ASD may represent a cortical dysfunction which can lead to a reduced capacity to process social information, thereby interfering with social functioning.In terms of potential clinical implications, the alterations we found may provide support in the future for an improved diagnosis, and hold out the promise for designing a targeted therapy involving these brain areas in order to improve social cognitive deficits in ASD.

Figure legends:
Figure 1 The first two columns depict the topographical distribution of the gamma-activity, in raw-score (i.e., µV 2 ) units in each of the two groups (typically developing, ASD), respectively.The third column shows the topographical distribution of the group difference of gamma-activity in standardized statistical effect-size (Cohen D) units.The fourth column depicts the scalp distribution of the FDR-corrected p-values for statistical significance.Panel A, B, and C provide the topographical distribution for the full gamma, and for the low and high gamma range, respectively.Higher levels of gamma-activity, greater group difference in Cohen D units, and greater statistical significance is indicated by warmer colors in the Figure .Figure 2 Association between the severity of psychopathology, as measured by the AQ total score, and the reduction in gamma-activity in three brain areas of interest where a significant group difference was found: the dorsolateral prefrontal cortex (panel A), inferior frontal gyrus (panel B) and temporo-parietal junction (panel C) areas.The severity of psychopathology showed a significant inverse relatioship with the gamma power in the 30-100Hz frequency range for all of the areas of interest (p<0.05 in all analyses): higher total score on the AQ scale was associated with lower gamma power.A steeper decline of the estimated gamma-activity with increasing symptom severity, as indicated by the regression lines, shows a closer relationship for both the inferior frontal gyrus and temporo-parietal junction area in the right hemisphere (in blue) than in the left (in red), whereas the association is stronger at the left hemisphere for the dorsolateral prefrontal cortex .

CRediT authorship contribution statement
Authors BK, PC, IB designed the study.BK, MT conducted the EEG data acquisition and prepocessing.PC and BK analyzed the data.BK and PC wrote the original draft.BSZ and IB contributed to conceptualization, writing, review & editing.KB, PC and MT worked on visualization.

Declaration of Competing Interest
None.Neuhaus, 2021 81 male 61female 70 male 68 female 8-17 y Watching a short movies 128 30-50 Hz Decreased gamma activity among females.Among males, age showed a correlation with the gamma-power.
Notes: a ASD: Autism Spectrum Disorder; b TD: Typical development; c gender composition not provided; d Age range: From childhood to late adolescence.Age effects on the resting state EEG in ASD have not been investigated in an adult (>18 years) sample.Only 1 study (Neuhaus et al), which included subjects with <17 years, investigated the effects of age in ASD and TD.The study found that higher age was associated with a decrease in spectral power across low and high EEG frequency bands both in ASD and TD, without an interaction with diagnostic group.Changes with age at higher frequencies including the gamma range were of relatively smaller magnitude compared to lower EEG frequencies.

Table 2
Basic Demographic and Clinical Characteristics of the Study Sample a a : Chi-square test for categorical, ANOVA for continuous variables b : AQ score=Autism-Spectrum Quotient Scale, total score c : TAS-20=Toronto Alexithymia Scale, total score d : SCL-90R= The Symptom Checklist-90R, total score e : RMET =Reading the Mind in the Eyes Test (recognition, % correct, SD)

Table 3 .
Spectral power in the gamma range (30-100Hz) in ASD and TD group, and statistical test results by electrode