Introduction

Cortical gamma band (∼40 Hz) neural oscillations play a pivotal role in integrating sensory information across distributed cortical areas [1]. The involvement of gamma-band oscillations in multi-modal cognitive activities has been suggested in spatiotemporal integration of perception [1], speech [2], associative learning [3], visual attention [4], and feature binding [5]. Abnormal gamma band oscillation has been hypothesized to be related to cognitive deficits in these areas, many of which are known to be significantly impaired in schizophrenia. The auditory steady-state response (ASSR) entrains neural oscillations in the brain to a specific frequency of auditory stimuli and has been used to assess the integrity of cortical oscillatory activity [6,7,8]. Reduced gamma band ASSR has been consistently observed in patients with schizophrenia [6, 9,10,11] and replicated in EEG [10, 12,13,14,15,16,17,18,19,20,21,22,23,24] and MEG studies [25,26,27,28,29,30], observed not just in chronic schizophrenia, but also in first-episode psychosis [14, 22], the ultra-high risk for psychosis [22], and non-ill first degree relatives of the patients [10]. Although there are also studies showing null or even reversed findings in patients [11, 31,32,33,34], it is one of the more robust electrophysiological biomarkers in schizophrenia. Correcting gamma band deficits have been argued to be important for develo** more effective treatment for cognitive deficits in schizophrenia [23, 35], which may, in part, rely on a better understanding of the underlying brain circuitry.

The ASSR generators are believed to be at the primary auditory cortex or the superior temporal plane [8, 36,37,38,39] based on MEG source localization studies [27, 38,39,40,41,42,43,44]. EEG dipole modeling supported similar sources in the bilateral auditory cortices [43]. In a functional magnetic resonance imaging (fMRI) study, Heschl’s gyrus, along with the medial geniculate body and inferior colliculus, were found to be associated with 40-Hz amplitude-modulated tones [45]. Positron emission tomography (PET) studies also suggest that activation of bilateral auditory cortices may be associated with 40-Hz ASSR [46]. The 40 Hz ASSR deficits in schizophrenia have been localized to the superior temporal plane [20] and primary auditory cortex [27, 28] with dipole models or by associations to the reduced cortical volume of the superior temporal gyrus [34], although it remains unknown whether dysregulated auditory cortex activity fully explains 40 Hz ASSR deficits in schizophrenia.

The 40 Hz ASSR is acquired using continuous stimulations over a long duration (typically around 300 to 500 ms), which would likely engage the extensive brain networks known to be associated with the primary auditory cortex [47,48,49,50]. We propose that a deficit in the underlying functional network of the primary auditory cortex may lead to the inability to sustain synchronization of the gamma band, contributing to the impaired 40 Hz ASSR in schizophrenia. To examine the functional connections between brain regions that may account for the gamma band ASSR deficit in schizophrenia, resting-state functional connectivity (rsFC) was obtained using the left and right primary auditory cortex as the initial seeds to identify rsFC that are associated with 40 Hz ASSR in patients with schizophrenia and healthy individuals. We then tested whether the gamma band ASSR deficit in schizophrenia was mediated by those functional connections with a mediation analysis. Revealing the neural underpinnings of this gamma band ASSR deficit could shed light on the mechanism of auditory-related gamma synchronization abnormalities in schizophrenia.

Materials and methods

Participants

The study included 66 patients with schizophrenia (n = 62) or schizoaffective disorder (n = 4) (referred together as schizophrenia for brevity) and 85 healthy individuals (Table 1). Patients were recruited from the Maryland Psychiatric Research Center and neighboring mental health clinics in the Baltimore area. Controls were recruited from local media advertisements. The Structured Clinical Interview for DSM-IV was used to confirm the diagnoses in patients and the absence of current DSM-IV Axis I diagnoses in healthy controls. Exclusion criteria were major medical and neurological illnesses, head injury, and substance dependence or substance abuse (except nicotine). Eight schizophrenia patients were not on antipsychotic medications, 51 were atypical, and 12 were on typical, including five on both atypical and typical antipsychotics (Table 1). No patients took benzodiazepines at the time of scanning. All subjects gave their written informed consent approved by the local Institutional Review Board. The ASSR EEG data and resting-state functional MRI data were collected in separate sessions. There was no significant group difference in intervals between the two sessions, although it was longer in the patient group (median intervals for schizophrenia and healthy individuals were 10 and 4 weeks, respectively, 95% CI (−25.4 – 0.3), p > 0.05). The current sample of patients and controls (66 and 85, respectively) was a subset of the previously reported sample (128 and 108, respectively) on 40 Hz ASSR [10] who had also completed fMRI; the fMRI-ASSR data are not previously reported.

Table 1 Demographic and clinical characteristics.
Full size table

ASSR paradigm

The details of the ASSR paradigm have been reported elsewhere [10, 11]. Briefly, trains of click sounds at 72 dB and of 1 ms duration were delivered via headphones at 40 Hz. Each train consisted of 15 clicks that last for 375 ms. There were 75 stimulus trains (trials) with 750 ms intervals between the end of a train and the beginning of the next. The total durations were 1 min and 25 s.

Electroencephalography (EEG) recording was performed in a sound-attenuated chamber using a 64-channel Quick-Cap with sintered silver/silver-chloride (Ag/AgCl) electrodes and a Neuroscan SynAmp2 amplifier (Compumedics, Charlotte, NC). The EEG data were recorded at a sampling rate of 1000 Hz with a 0.1–200 Hz bandpass filter. Impedance was kept below 5 kΩ. Linked mastoid electrodes served as the reference. The EEG data were re-referenced to average reference, high-pass filtered at 0.8 Hz, and detrended during offline analysis. Ocular artifacts were removed using the time-shift-PCA algorithm [51], with ocular channels as references. Participants were instructed to relax, remain alert, and keep eyes open during the recording.

Normalized ASSR power

Rather than using individual channels (e.g., CZ or FZ), we adapted the denoising source separation (DSS) algorithm to maximize ASSR response reliability, where individual EEG channels are spatially combined [10, 52,53,54]. DSS is specifically designed for use with data from multi-trial evoked responses or narrowband signals and works by enhancing stimulus-driven activity over stimulus-unrelated activity, with its components ordered according to their reliability [52,53,54]. Raw 40 Hz ASSR power was obtained at the 40 Hz frequency and background power was calculated by averaging spectral power over 1 Hz width frequency bands (on either side of the 40 Hz frequency, after leaving a guard band of 0.5 Hz on either side). Normalized 40 Hz ASSR power was then calculated as the ratio of raw ASSR power and respective background power as in our previous studies [10, 53]. This normalization with respect to background power remarkably reduces subject-to-subject variability of frequency response profiles [40,41,42,43,44]. Source localization of ASSRs in healthy individuals identified a wide range of sources both within and outside of the primary auditory cortex [43, 76, 77]. In patients with schizophrenia, interhemispheric phase locking for the primary auditory cortices was reduced in comparison to healthy controls [78]. This is consistent with the current finding that the strength of rsFC between bilateral STG (i.e., left STG—right STG) is positively linked to the power of 40 Hz ASSR (Table 2).

However, the study also suggested that the left medial prefrontal cortex and left precentral gyrus are also important brain areas for gamma band ASSR in schizophrenia. The role of their functional connection with STG in gamma band ASSR was first identified with STG as the seed, and then, confirmed by using MPFC or PrG as the seeds. The involvement of frontotemporal connection in 40 Hz ASSR illustrated here was in line with previous source localization evidence [43, 46, 47]. For example, a PET study found that 40-Hz ASSR activated not only primary auditory cortices but also the middle frontal gyrus;[46], and a dipole modeling to EEG data study found that sources for 40 Hz ASSR include the left frontal lobe [43]. The prefrontal and primary auditory cortices are anatomically connected [47]. Chen and others showed that besides STG, there was less activity in frontal regions during the auditory tasks in schizophrenia patients compared to healthy controls [79, 80]. Reduced gamma-band response in schizophrenia patients has been linked to impaired frontal network processing [81]. Dysfunction of the medial prefrontal cortex, which is the anterior midline node of the default mode network, in schizophrenia has been well demonstrated. Pomarol-Clotet and others used three different whole-brain voxel-based imaging techniques and identified the medial prefrontal cortex as a prominent site of abnormality in schizophrenia [82]. MPFC has also been suggested in a comparison of auditory-evoked gamma-band responses between patients with schizophrenia and healthy control subjects using an auditory reaction task [83]. They found reduced gamma band responses in schizophrenia, which was due to reduced activity in the auditory cortex and the medial frontal gyrus region. This is consistent with our findings addressing the essential role of STG and MPFC, although we used a different passive auditory task. Our data suggest that the 40 Hz ASSR deficits may be associated with reduced functional connectivity between STG and frontal areas in schizophrenia patients. i.e., the reduced left STG—left MPFC rsFC in schizophrenia, although the neurophysiological mechanism of this association remains unclear, in part because of the still limited understanding in the neurophysiology of rsFC.

Besides left STG—left MPFC rsFC, the mediation results suggest that left MPFC—left PoG and left PrG— right PoG rsFC also mediated the deficit of 40 Hz ASSR in schizophrenia. The findings with precentral and postcentral gyri may seem unusual, but precentral and postcentral gyri lesions are associated with Wernicke’s type of aphasia [84] and the interhemispheric connections of precentral and postcentral gyri were also associated with positive symptoms of schizophrenia [90,91], which limits its application in identifying at-risk individuals or the effectiveness of treatment in young first-episode patients. Moreover, deficits of low-frequency auditory responses have also been robustly observed in schizophrenia and the low-frequency activities could sometimes provide better separation between schizophrenia patients and healthy controls [32, 92,93,94,95,96]. The neural network origins of those low-frequency abnormalities in schizophrenia should be explored in future studies.

Other limitations of the study include that the potential effects of antipsychotic medications on current findings were unknown, although adding CPZ as an extra covariate did not significantly affect the results. The findings of similar correlation coefficients between 40 Hz ASSR and rsFC in healthy controls also suggest that these correlations are unlikely to be mainly driven by medication effects. Still, another limitation is that we used seed-based functional connectivity methods, with a limited number of seeds (i.e., left/right STG, left MPFC, and left PrG), to explore the rsFC underlying gamma band ASSR in schizophrenia. There might be other networks which do not functionally link to those seeds and also modulate 40 Hz ASSR (e.g., networks with auditory brainstem and thalamic nuclei as seeds). Conversely, it is not clear whether the rsFC identified by 40 Hz ASSR is specifically linked to 40 Hz ASSR or it is more generally associated with auditory encoding neural processes in frequencies lower than 40 Hz, as simultaneous resting fMRI and EEG recordings usually showed that, across all EEG bands, rsFC correlations with EEG are the highest at the lower frequencies [97]. Our goal here is more limited by focusing on the gamma band, which limits the specificity conclusion, although the near complete mediation of the 40 Hz effect on diagnosis by rsFC was quite surprising and encouraged future research to examine the neurophysiological interpretations for such associations. Other limiting factor is that we did not have handedness information [98, 99] on many participants so the analysis was not performed but laterality is relevant here as right hemispheric laterality for 40 Hz ASSR has been reported [74, 75].

In summary, the study explored the neural circuits underlying gamma band ASSR deficits in schizophrenia by examining the associations between ASSR and resting-state functional connectivity. We found an auditory-parietal-prefrontal network that potentially explains most of the 40 Hz ASSR deficit in schizophrenia. These findings shed light on further understanding of the mechanism of neural oscillatory deficit in schizophrenia.