The Early Auditory Gamma-Band Response Is Heritable and a Putative Endophenotype of Schizophrenia

Introduction

There is growing interest in measuring brain electrical oscillations to study abnormal brain dynamics, synchronization, and connectivity in schizophrenia, particularly the gamma frequency band (30- to 80-Hz range, centered around 40 Hz).^1,2 Stimulus-evoked gamma-band responses (GBRs) have been associated with widespread sensory and cognitive processes including perceptual and associative learning,³ object representation,⁴ and selective attention.⁵ The early phase-locked (evoked, exogenous) GBR occurring within 100 milliseconds after stimulus onset has been hypothesized to reflect the synchronization of neural assemblies involved in perceptual processing of sensory input within a local area and across different regions of the brain in order for coherent sensory registration and integration of stimulus events.^6–8

There is supporting evidence from electroencephalogram (EEG) studies suggesting that the core pathophysiology of schizophrenia is related to disturbed neural synchrony in the gamma frequency band within and across different specialized brain areas leading to impaired perceptual experiences and cognitive dysfunction.^2,9–13 Early GBR can be (1) automatically generated in response to direct, repetitive stimulation (ie, the gamma driving response) or (2) “cognitively evoked” when performing higher order cognitive or perceptual tasks (ie., the GBR to tones or visual stimuli). Reduced auditory gamma-driving responses during auditory steady-state evoked potential paradigms have been reported in schizophrenia,^14–17 unaffected family members of schizophrenia patients,¹⁶ and first hospitalized psychosis patients,^18,19 suggesting that neural synchrony or functional connectivity abnormalities might be a biological marker of this disorder at early stage of perceptual processing. Reductions in evoked GBRs have been reported for tasks involving higher order cognitive tasks using visual or auditory oddball paradigms. These tasks typically involve complex cognitive functions such as attention and working memory. However, studies of using higher order cognitive tasks in patients with schizophrenia are fewer, and results are inconsistent. For example, in the visual modality, Spencer et al^20,21 found that schizophrenic patients had abnormal early evoked phase-locking GBRs compared with healthy individuals in a Gestalt perception task. In the auditory modality, Roach and Mathalon²² found reduced early auditory gamma-band response (EAGBR) phase locking to standard tones in patients with schizophrenia using an oddball task. Three other studies using similar but not identical paradigms, however, failed to find such EAGBR impairments.^21,23,24

In addition to the equivocal findings on EAGBR reductions in schizophrenia, little is known about the genetic (heritability) and environmental influences on the EAGBR or its potential as a biological marker (endophenotype) of schizophrenia. Endophenotypes are heritable, disease-associated neurophysiologic, cognitive, or neurobiological traits that are believed to be in the etiological pathway (ie, intermediate) between risk genotype and the clinical syndrome more proximally related to the genetic substrate than is the higher order construct of a “disorder.”^25–28 The use of endophenotypes has been proposed as a strategy to accelerate gene identification^25–28 and characterization²⁹ for psychiatric disorders.

In this study, we sought to examine whether EAGBR measures (both phase-locking factor [PLF] and power) to standard stimuli during an auditory oddball task were in fact reduced in schizophrenia and might serve as putative endophenotypes of schizophrenia. Specifically, we wanted to determine whether EAGBR measures were (1) impaired in patients with schizophrenia and their unaffected monozygotic (MZ) co-twin members, (2) heritable traits, and (3) genetically associated with schizophrenia. The current study used twin design and employed sophisticated structural equation modeling analyses to optimally examine and quantify the heritability of the EAGBR measures and the genetic and environmental overlap between EAGBR measures and schizophrenia.^30–32 Comparing the resemblance (covariance) of MZ twin pairs for a trait with that of DZ twin pairs provides an indication of the extent to which genetic and environmental variation contributes to phenotypic variation of that trait.³³ When multiple traits (such as schizophrenia diagnosis and GBRs) are collected in the same MZ and DZ twin pairs, it is possible not only to estimate the heritability of GBRs but also to examine whether an overlapping set of genes accounts for variation in both phenotypes (eg, schizophrenia and impaired GBRs), and if so, to what extent these shared genes explain the covariation between the phenotypes.^32,34 This is, to our knowledge, the first twin study in schizophrenia of EAGBR.

Materials and Methods

The study was approved by the UK Multi-centre Research Ethics Committee. Probands were ascertained from UK national psychiatric services and the Maudsley Twin Study of Schizophrenia. Control twins were recruited from the Institute of Psychiatry Volunteer Twin Register and through advertisements. Written informed consent was obtained from all participants.

Results

Discussion

To our knowledge, this is the first study to assess EAGBR-evoked power and PLF in twins with schizophrenia. We found that EAGBR measures of power and trial-to-trial phase synchrony to standard stimuli in an auditory oddball task were in fact reduced in schizophrenia. We further demonstrated, using twin design and model fitting analyses, that both power and phase synchrony measures were heritable traits and that impairments in each of these phenotypes were significantly associated with schizophrenia. (Data from Fz revealed equivalent results; details are available upon request).

For a trait to be an appropriate endophenotype, it should be heritable, associated with disease, and observed in genetically at-risk but behaviorally unaffected relatives of patients.^25,31,42,43 In this study, significant higher MZ than DZ correlations were observed in both power and PLF measures suggesting genetic contributions. Significant heritability estimates (0.65 for the power and 0.63 for the PLF) of each measure suggested that they satisfy the first endophenotype criteria.

Because an endophenotype is conceptualized as an expression of the genetic liability for a disorder, it should occur in patients with the illness and appear more prevalent in individuals who are at risk for the disorder (ie, unaffected relatives of patients). Analyses of means between groups revealed that patients with a diagnosis of schizophrenia and their unaffected identical co-twins had reduced EAGBR measures as compared with control subjects. However, the fact that a trait “runs in families” is not sufficient evidence to assume that the observed association is genetic because families may share predisposing environments as well as genes. In order to partition the association between an endophenotype and a disorder into genetic and environmental components, we employed a twin study design.³² Modeling fitting analyses of twin data revealed significant associations (ie, phenotypic correlations) between schizophrenia and reduced power and smaller PLF responses. The analyses were able to further decompose and quantify the source of these observed associations into genetic, shared, and unique environmental components. Results suggested that in each phenotype, shared genes rather than shared environment were likely the main contributors to the observed associations with schizophrenia. In addition to genetic overlapping, individual-specific environment factors also contributed significantly to the observed association between schizophrenia and PLF. These results suggested that reduced EAGBR power and PLF measures had fulfilled the second and third endophenotypes criteria as well and therefore could be considered as putative endophenotypes for schizophrenia.

The 2 metrics of the gamma oscillation, evoked power and PLF, are different measures that are independent but highly correlated. Evoked power is derived from the individual's averaged response and captures the magnitude of the oscillations time locked to task events across trials. Oscillations that are not strongly time locked (out of phase) with respect to stimulus onset across trials are canceled out during averaging. Evoked power reflects a true event-related oscillation (ERO) whose morphology is dependent on the overall amplitude of the response trial to trial and the temporal jitter of the response from trial to trial. In contrast to evoked power, PLF is based on activity on each trial. It measures the variance of EEG phase across single trials independently of amplitude. Rather than measuring differences in phase angle between sites, it measures differences in phase angle between single trials at a specific site. In this regard, it is a measure of temporal stability of a specific evoked response and is sensitive to oscillatory behavior that is averaged out in power measures. The PLF provides a measure of temporal consistency in neural synchrony. It is possible that a large trial-to-trial signal with a low PLF may result in a smaller ERO than a small trial-to-trial signal with a high PLF. Both measures have given complementary information on the functional component of cognitive processing, although the precise cognitive and neurophysiological concomitants of each measure are unknown at present.

Attention has been shown to modulate early GBRs.^44,45 A study in healthy individuals has shown that increased task difficulty and mental effort was associated with greater evoked GBR amplitude.⁴⁶ It is not clear the degree to which attentional modulation of the EAGBR on this active target detection task contributes to group differences. The present data are unable to distinguish between a purely sensory deficit vs a top-down attentional modulation deficit. We are currently examining this issue using a different experimental paradigm.

The reduced EAGBR power and PLF in patients with schizophrenia found in the present study was in accord with Roach and Mathalon²² but not Galliant et al²³ and Spencer et al.²¹ It is not entirely clear why 2 other studies failed to find such impairments. We hypothesize that methodological difference may account for these discrepant results. Specifically, our paradigm used a total of 400 stimuli (320 standards) as compared with a total of 180 stimuli with 150 standards in Spencer et al²¹ and 230 stimuli with 175 standards in Gallinat et al²³ who also used click pairs as standard stimuli rather than single tones. Because of the number of standards used, our paradigm is likely to produce greater signal-to-noise ratios, which leads to better power to distinguish individuals with deficits and those without.

In our sample, we observed no significant correlations between any of the clinical parameters (including medication, age of onset, duration of illness, and SAPS and SANS) and the EAGBR measures. Moreover, medications effects are unlikely to account for the observed EAGBR deficits in twins with schizophrenia as the well co-twins from the discordant twin pairs who were free from psychotropic medications had reduced EAGBR as well.

In addition to genetic associations between EAGBR deficits and schizophrenia, we observed a significant individual-specific environmental overlapping effect between schizophrenia and PLF, raising the possibility that individual specific environment factors such as illness progression also play important roles in EAGBR deficit. Correlations of PLF with clinical parameters (medication, age of onset, duration of illness, SAPS, and SANS) were nonsignificant, however. It remains to be determined which aspects of individual-specific environments are contributory. One possibility is obstetric complications at birth. Studies have found that some obstetric complications seem to be more common in the affected co-twins from MZ pairs discordant for schizophrenia.⁴⁷ Unfortunately, such data are unavailable in the present study. A future twin study is needed to resolve this issue.

The cellular mechanisms of EAGBR are believed to involve networks of γ-aminobutyric acid (GABA)ergic, glutamate, and acetylcholine neurotransmitter systems. The fast synaptic inhibitory GABA neurons appear to be critically involved in generating gamma-band oscillations.⁴⁸ In patients with schizophrenia, evidence of reduced numbers of GABA neurons and reduced synaptic connectivity between GABA neurons and postsynaptic cells have been observed. These cellular deficits, coupled with deficits in glutamate receptor–mediated excitation of interneurons,⁴⁹ may contribute to the observed gamma oscillation dysfunction in patients with schizophrenia and their relatives and are consistent with current theory suggesting that the mechanism underlying abnormal gamma activity is due to lack of inhibitory neuronal pathways.^2,50,51 That is, dysfunction of gamma oscillation may reflect a state of neuron hyperarousal (hyperexcitability) because of reduced inhibitory input from GABAergic interneurons to efficiently suppress spontaneous neuronal activity that is triggered by inputs unrelated to the stimulus being studied or to block unrestrained local circuit processing spread of activation. As a result, synchronized neuronal activity is reduced and this is reflected in reduced scalp gamma oscillations. Future study combining molecular genetics and basic neuroscience may provide important insight into the neurobiological mechanism of the glutamate receptor–mediated excitation and the GABA neuron–mediated inhibition feedback loop underlying the EAGBR deficits in schizophrenia and facilitate the development of therapeutic intervention.^49,50

Using click trains stimuli, deficits in evoked gamma oscillatory activity have been found in patients with schizophrenia,^14,15,19 bipolar disorder,⁵¹ early-onset psychosis,^18,19 and autism,⁵³ as well as in the unaffected family members of schizophrenia probands.¹⁶ Similar deficits are documented in parents of children with autism⁵⁴ using a transient stimulus, suggesting that abnormal gamma driving responses that are automatically generated in response to direct, repetitive stimulation may also be a putative endophenotype for bipolar disorder and autism. However, family studies of gamma oscillations employing an auditory oddball task in patients with bipolar disorder or autism are few. Whether EAGBR deficits are also putative endophenotypes for these disorders will need to be clarified in the future.

The sample size of the present study was relatively small. In particular, we had limited statistical power in separating genetic from environmental correlations across all 3 sets of models. For example, shared genetic factors only emerged as statistically significant when schizophrenia heritability was assumed to be as high as 90% (model 1) for evoked power. Nonetheless, the greater MZ-to-DZ cross-trait cross-twin correlations and the best fitted submodels of each phenotype suggested that shared genes rather than shared environment are more likely the main contributors to the observed phenotypic associations. Abnormal lateralization of gamma power²³ has been reported in patients with schizophrenia compared with control subjects. Unfortunately, the low-resolution electrode array use in this study and small oscillation signals observed at temporal-parietal regions have limited the power of our study for examining hemisphere differences between groups and for performing modeling analyses.

In summary, evidence of reduced EAGBR power and PLF measures in twins with schizophrenia and their unaffected co-twin members, significant heritability of each phenotype, and substantial genetic overlapping between schizophrenia and these measures support EAGBR power and PLF measures as putative endophenotypes for schizophrenia.

Funding

Harvard Medical School (Kaplen Fellowship to M.-H.H.); National Alliance for Research on Schizophrenia and Depression (Sidney R. Baer, Jr. Foundation Young Investigator Award to M.-H.H.); Adam Corneel Young Investigator Award (to M.-H.H.); National Institutes of Mental Health (T32 MH016259-29A1 Fellowship to M.-H.H., MH RO1-58704 to D.F.S.); Wellcome Trust (Research Training Fellowship 064971 to M.P.).