Autism spectrum disorders (ASD) are clinically defined by impairments in social interaction and communication and by restricted/stereotyped behaviors. The prevalence for ASD, which includes autistic disorder, Asperger's syndrome and pervasive developmental disorder - not otherwise specified, is estimated to be as high as 1 in 110 [CDC, [¹]]. Although diagnosable medical conditions, including genetic syndromes, are estimated to account for as many as 10% of cases, most cases remain idiopathic [²,³]. Family studies indicate that idiopathic ASD is highly heritable [⁴,⁵], with an estimated heritability as high as 90%. Studies of first-degree relatives have shown increased prevalence of anxiety and depression, personality changes, deficits in the social use of language (that is, pragmatics) and significantly higher scores on assessments of autism traits such as social responsiveness [^6-10]. This subclinical expression of the ASD phenotype is termed the broad autism phenotype (BAP) and provides further evidence for the heritability of autism.

Although studies of first-degree relatives have identified a broad range of changes in the behavioral phenotype, few studies aside from genetics have examined the underlying biology of the BAP. However, studies of parents and siblings of people with ASD have reported increased rates of macrocephaly [¹¹,¹²], enlarged hippocampi [¹³], cortical gray-matter changes [¹⁴], altered occulomotor function [¹⁵,¹⁶], increased platelet serotonin levels [¹⁷], reductions in face-specific early visual processing [¹⁸], and reduced γ-band oscillatory phase-locking [¹⁹].

Gamma-band oscillatory activity (that is, 30 to 80 Hz) is of significant interest as a biomarker and/or endophenotype in ASD for two reasons: 1) there is a putative relationship between perceptual binding and/or connectivity and γ [²⁰,²¹], which have been proposed as cognitive deficits in the disorder [^22-25]; and 2) mechanisms for generating γ-band activity in the cerebral cortex and hippocampus are relatively well characterized [²⁶].

Auditory γ-band responses are not unitary, however. Early, obligatory γ-band responses are produced to any auditory stimulus, and usually appear at 30 to 80 milliseconds after stimulus [²⁷]. This early, highly phase-locked response is called the transient γ-band response (tGBR). However, when stimuli are modulated in amplitude, either as part of a train of clicks or by formal amplitude modulation, a later auditory steady-state response (ASSR) is produced, in this case at or near the frequency of modulation, which peaks at rates in the γ-band range [²⁸,²⁹]. Steady-state stimulation produces both types of responses [³⁰]. The mechanisms of generation for these two types of responses, and also whether they are related to cognitive or perceptual processes, may vary. The ASSR may partly reflect a linear superposition of transient mid-latency auditory evoked responses [³¹,³²], although this is not completely accepted by all investigators [²⁸,³⁰,³³,³⁴]. Regardless, the purported association between cognitive functions and the tGBR is not established for the ASSR.

We first reported a significant reduction in MEG-measured evoked or phase-locked ASSR power in children and adolescents with autism compared with control subjects matched for age and gender [³⁵]. Subsequently, we found that adults with ASD and first-degree relatives of people with ASD exhibited reduced tGBR evoked power and increased tGBR induced power, compared with healthy controls [¹⁹]. Across trials evoked responses are consistently phase-locked to the stimulus, whereas induced responses are not. Together, these two types of responses constitute total stimulus-related power. Increases in non-phase-locked γ-band power have also been reported in other studies [^36-38]. We proposed that the deficit in the γ-band electrophysiology may in reduced inter-trial phase-locking to the stimulus, which causes a shift in γ-band power from phase-locked, evoked power to non-phase-locked induced power, while preserving total γ-band power. A recent MEG study replicated reduced auditory tGBR phase-locking in a sample of children with ASD [³⁹], but did not report significant differences in either evoked or induced power. It should be noted that phase-locking factor (PLF), also known as inter-trial coherence, is an amplitude-independent measure, unlike evoked power, so although the two measures may be correlated, phase-locking will tend to be more robust in noisy data, having lower between- and within-subject variance [⁴⁰] (see Additional file 1).

Our previous report on parents of children with ASD measured only the tGBR component [¹⁹], whereas our original finding of reduced evoked γ-band power in children with ASD reflected only the ASSR component [³⁵]. The current study was therefore designed to ascertain whether adult first-degree relatives of people with autism exhibit changes in both the tGBR and ASSR. We hypothesized that phase-locked auditory evoked γ-band activity, as well as being a direct measure of stimulus-related phase locking, would be lower in first-degree relatives of people with ASD for both types of γ-band responses. Three different amplitude modulation rates were used to assess whether relatives of people with ASD would exhibit changes in ASSR-evoked γ-band activity specific to 40 Hz or across a wider γ-band range. Measures associated with the BAP, including the Autism-Spectrum Quotient (AQ) [⁴¹] and the Social Responsiveness Scale (SRS)[⁴²], were included to assess potential relationships between the BAP and γ-band activity, because neither of our earlier studies obtained such measures for correlation with the electrophysiological data.

We found reduced ASSR γ-band evoked power and phase-locking in the pASD group relative to the HC group in the current study. This is consistent with our earlier published results for the auditory 40 Hz ASSR in children with autism, in which we reported reduced evoked power (PLF was not calculated in this earlier report) [³⁵]. However, although mean tGBR evoked power and phase-locking were lower in the pASD group than in the HC group, the results for the tGBR portion of the response were not significant. This contrasts with our earlier findings for the transient γ-band responses, [¹⁹], which showed similar magnitude reductions for subjects with autism and pASD subjects compared to control subjects.

Several factors could contribute to the difference between the tGBR results of the current study and those of our previous study [¹⁹]. First, the stimuli in this study were amplitude-modulated specifically to produce robust ASSR responses, whereas the earlier study used pure tones, which only produce tGBR; however, we are not aware of any studies systematically comparing the effect of AM versus non-AM type stimuli on tGBR, so this is entirely speculative. Second, the pASD group in the earlier study may have had more BAP or relevant underlying physiological abnormalities than the group in the current study, which did not exhibit high ASD trait loading; however, we did not record such measures in the earlier study, preventing us from exploring this question. Finally, it should be noted that a more recent study in children with ASD did not find significant reductions in evoked power in the tGBR, although they did report significantly reduced PLF, which they attributed to the differential effects of noise on PLF versus evoked power [³⁹] (see Additional file 1). It is therefore not clear even in probands with ASD that the finding of reduced evoked power is robust, and further studies with larger numbers of subjects (both probands and parents) will be needed to clarify these issues.

The current study also expanded on the earlier findings by including several modulation conditions, all of which produced some reduction in evoked power in both hemispheres. These reductions appeared stronger for the two highest modulation frequencies, but as we did not examine rates above 48 Hz, we do not know if there are higher rates at which the group differences are more pronounced. The restriction of the ASSR PLF findings to the left hemisphere is more consistent with our earlier published data on the ASSR in children with autism [³⁵]. For tGBR elicited by pure-tone stimuli, our previous work suggested that reduced evoked power and PLF were present in both hemispheres [¹⁹]. This may represent a difference between ASSR and transient type auditory stimuli or γ-band response, although we did not replicate the earlier tGBR finding in this study. Recently, in a study of another set of potential ASD endophenotypes, Mosconi et al. [¹⁶] reported left-lateralized deficits in two measures of occulomotor function: open-loop pursuit gain and procedural learning for rightward saccades. Previous studies have suggested atypical lateralization of language-related brain structures in ASD [^43-46]. Taken together, these findings may suggest abnormal cerebral lateralization and/or stronger left-hemisphere involvement in the disorder. Although group differences were noted primarily in the left-hemisphere data, across groups the tGBR and ASSR evoked power was stronger in the right hemisphere. This finding is consistent with previous EEG and MEG research [^47-50]. Ross et al. [⁴⁹] previously interpreted this in the context of right-hemisphere dominance for pitch perception, because the ASSR is known to closely entrain to the temporal envelope of sounds, and is very sensitive to disruptions in acoustic periodicity [⁵¹]. Previously, we have not found significant differences in total γ-band power (evoked plus induced), but for ASSR in the current study, there was a significant reduction in the left hemisphere of the pASD group. This will be important to parse in future studies, because if γ-band phase-locked power is significantly lower in ASD/pASD participants with no change in total γ-band power, it suggests a shift of γ-band activity from phase-locked to non-phase-locked activity in ASD, rather than a deficit in the generation of γ-band activity. Indeed, several previous studies have shown that induced (non-phase-locked) and spontaneous γ-band activities are increased in ASD [¹⁹,^36-38]. If, however, there is a change in total stimulus-related power - in this case a reduction - it may suggest that γ-band-generating mechanisms themselves are altered, at least for auditory responses.

Gamma-band responses are of significant interest in ASD because their association with well-described cortical circuitry makes them ideal candidates for translational neuroscience. Glutamatergic input to inhibitory interneurons, particularly those expressing the calcium-binding protein parvalbumin (PV), results in the recurrent, phasic inhibitory modulation of pyramidal neurons [²⁶,^52-54]. PV-expressing interneurons play a crucial role in this inhibition via gamma-aminobutyric acid (GABA)_A-receptor mediated mechanisms, the timing of which results in γ-band frequency output from the pyramidal neurons [⁵⁵]. The GABA_A-receptor antagonist bicuculline effectively eliminates γ-band oscillations [²⁶]. Recently, studies of visual γ-band responses using MEG combined with magnetic resonance spectroscopy showed for the first time in human subjects that γ-band response frequency is associated with the cortical concentration level of GABA itself [^56-58]. PV cell deficits are also a common feature across many mouse models of ASD [⁵⁹]. In a potentially interesting parallel of our γ-band findings with two putative mouse models of ASD (prenatal exposure to valproic acid and neuroligin-3 R451C mutants), both models expressed deficits in PV inhibitory neurons in only one hemisphere [⁵⁹].

The potential importance of GABA dysfunction to autism has been repeatedly stressed in the literature [⁶⁰]. Blatt et al. [⁶¹] reported significantly reduced GABA_A-receptor binding in strong binding regions of the hippocampus, with no significant differences noted in binding of serotonergic, cholinergic and glutamateric receptors. This has been extended recently to several areas of cortex and cerebellum [⁶²]. A small study of children with autism aged 5 to 15 years reported increased plasma GABA levels [⁶³], but the authors suggested that because the relationships between plasma, cerebrospinal fluid and brain levels of GABA are unknown, the implication of the finding for a specific central nervous system directionality is unclear. Messenger RNA levels of glutamate decarboxylase (GAD), the enzyme that converts glutamate to GABA and is closely related to intraneuronal GABA, have been reported to be reduced by about 40% in cerebellar Purkinje cells in people with autism [⁶⁴], and up to 50% in parietal and/or cerebellar tissues, depending on the specific isomer (GAD65 or GAD67) [⁶⁵].

A reduction in GAD expression also implies a corresponding increase in cortical glutamate levels. Increased glutamate concentrations have been reported in the hippocampal region of people with autism using proton magnetic resonance spectroscopy [⁶⁶], consistent with previous reports of increased serum levels of glutamate [⁶⁷], giving rise to a hyperglutamate hypothesis of autism [⁶⁸], which suggests in part that higher glutamate levels in autism may be due to reductions in GAD. Alteration in the balance of cortical excitation and inhibition is the key prediction made by Rubenstein and Merzenich in their theoretical model of autism [⁶⁹]. We propose that the γ-band phase-locking deficit seen in people with ASD and their first-degree relatives is a potential non-invasive biomarker reflective of this change in the excitation/inhibition balance.

The observed correlation between PLF and the AQ communication subscale may be consistent with observations that γ-band activity is sensitive to perception of speech sounds and lexicality [^70-72]. Frontal γ-band EEG power is strongly correlated with expressive and receptive language skill in 24 to 36-month-old children [⁷³]. The observed inverse relationship between SRS scores and tGBR and ASSR γ-band evoked power suggests that there may be a relationship between γ-band dysfunction and important traits in ASD such as social skill, although it seems unlikely that passive auditory γ-band power is directly related in any causal manner to social reciprocity. More likely, the auditory findings we describe are related to those found in other areas of the cerebral cortex that are more directly related to social cognitive function. For example, previous studies have shown γ-band changes in ASD in the context of face perception and eye-gaze processing [³⁷,⁷⁴].

It is worth emphasizing that the pASD sample in this study did not exhibit a strong presentation of ASD/BAP traits based on the results of the AQ and SRS data. Group differences were found only for the attention to local detail subtest of the AQ. It is possible that this was due to the use of singleton families in the study, as previous research has indicated that the BAP is expressed more strongly in multiplex autism families [⁶,⁷⁵]. Nonetheless, it is important that we found differences in a biological marker in the singleton families because simple biomarkers may be more sensitive than behavioral phenotypes to risk for ASD. A future study might usefully compare single- and multiple-incidence family members to assess whether the multiple-incidence families or those who express the BAP strongly exhibit stronger γ-band findings than the single-incidence families. Furthermore, because the number of fathers in the pASD sample was low relative to mothers, it is possible that increasing the number of men in future studies would reveal differences based on parent gender, as some BAP studies have suggested that it is more strongly expressed in fathers than mothers [⁷⁶]

The findings of reduced phase-locking and evoked γ-band power in first-degree relatives of people with ASD are consistent with a heritable neural synchrony endophenotype. In the current study, ASSR γ-band measures were significantly different between groups, whereas tGBR measures were not, suggesting that the ASSR may be a more robust measure in first-degree relatives. Although the findings in this study are consistent with heritability, heritability itself was not directly measured in this study and, strictly speaking, the group differences observed in this study should be considered evidence for familiality. Family studies, particularly twin designs, would be necessary to provide direct measures of heritability. A caveat to these findings is that reduced phase-locking and evoked power in the γ-band range is not specific to ASD. Similar, if not identical, deficits occur in people with schizophrenia and their first-degree relatives [^77-79]. This limits the utility of the finding as a diagnostic biomarker. However, γ-band measures markers should be useful in genetics studies and as biomarkers of drug response in future clinical trials. Future research should focus on establishment of a normal range for each γ-band measure in healthy subjects, definition of useful cut-offs for abnormal values, and the relationship between γ-band markers and measures of cortical GABA and glutamate concentrations in people with ASD and in animal models of ASD.

AM: amplitude modulation; AQ: Autism-Spectrum Quotient; ASSR: auditory steady-state response; BAP: broad autism phenotype (BAP); GABA: Gamma-amino butyric acid; GAD: glutamate decarboxylase; HC: healthy control; MEG: magnetoencephalography; pASD: parents of children with autism spectrum disorder; PLF: phase-locking factor; SSP: source space projection; SRS: social responsiveness scale; tGBR: transient γ-band response; WASI: Wechsler Abbreviated Scale of Intelligence.

The authors declare that they have no competing interests.

DR designed the study, carried out the final statistical analyses and drafted the manuscript. KY, LW and AW assisted with study coordination, administration and scoring of behavioural measures. PT, EK and KM provided key input on MEG data analysis, and wrote most of the customized routines used in the analysis of data. SH provided input on behavioral measure selection and scoring, and was involved in the clinical diagnostic evaluation of probands. All authors read and approved the final manuscript.

This work was supported by grants from the Autism Speaks Foundation and the National Institutes of Health (R01 MH082820) to DR. The funding bodies had no role in the design, collection analysis and interpretation of the data, in the writing of the manuscript or decision to submit the manuscript for publication.