Altered effective brain connectivity at early response of antipsychotics in first-episode schizophrenia with auditory hallucinations

Accepted Manuscript Altered effective brain connectivity at early response of antipsychotics in firstepisode schizophrenia with auditory hallucinations Leilei Zheng, Weibo liu, Wei He, ShaohuaYu, Guodong Zhong PII: DOI: Reference:

S1388-2457(17)30058-5 http://dx.doi.org/10.1016/j.clinph.2017.02.004 CLINPH 2008060

To appear in:

Clinical Neurophysiology

Received Date: Revised Date: Accepted Date:

17 September 2016 2 February 2017 3 February 2017

Please cite this article as: Zheng, L., liu, W., He, W., ShaohuaYu, Zhong, G., Altered effective brain connectivity at early response of antipsychotics in first-episode schizophrenia with auditory hallucinations, Clinical Neurophysiology (2017), doi: http://dx.doi.org/10.1016/j.clinph.2017.02.004

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Effective connection and hallucination

1

Altered effective brain connectivity at early response of antipsychotics in first-episode schizophrenia with auditory hallucinations Leilei Zheng1, Weibo liu1*,Wei He2,3, ShaohuaYu 1, Guodong Zhong4

1

Department of Psychiatry, Second Affiliated Hospital, School of Medicine, Zhejiang

University, Hangzhou, China 2

Department of Cognitive Science, Macquaire University, New South Wales, Australia

3

Australian Research Council Centre of Excellence in Cognition and Its Disorders, Macquaire

University, New South Wales, Australia 4

Electroencephalogram laboratory, Second Affiliated Hospital, School of Medicine, Zhejiang

University, Hangzhou, China

*

Correspondence to:

Dr. Weibo Liu, MD. Department of Psychiatry, Second Affiliated Hospital, School of Medicine, Zhejiang University 88 Jiefang Road, Hangzhou, Zhejiang 310009, China Tel.: 86-571-87783773 E-mail: [email protected]

Effective connection and hallucination

2

Highlights 1. We investigated effective connectivity of cortical networks in schizophrenia with auditory hallucinations using dynamic causal modeling of resting-state EEG data. 2. Medicated schizophrenia showed improvement of fronto-temporal connectivity. 3. This study provided first evidence for early drug response-related alterations in effective connectivity of cortical networks in schizophrenia.

Abstract Objective: This study aimed to examine alterations of cortical connectivity in first-episode schizophrenia (FES) with auditory hallucinations at early response of antipsychotics. Methods: This was a non-experimental control of medication study. We measured cortical activity from 20 medicated FESs, 19 non-medicated FESs and 22 healthy controls using electroencephalogram during an eye-open resting state. Source-reconstruction analysis was performed to determine the brain regions of significant group difference. A dynamic causal modeling (DCM) analysis was used to estimate effective connectivity between sources. Result: Both groups of FESs expressed increased activity in the right middle frontal

Effective connection and hallucination

3

gyrus (RMFG) and left/right superior temporal gyrus (L/RSTG) relative to controls (p < 0.05), and the non-medicated FESs presented even higher activity than medicated ones did (p< 0.05). The effective connectivity from RMFG to LSTG was weaker in non-medicated FESs relative to medicated FESs (p< 0.01), while the medicated FESs had no difference with healthy controls in RMFG to L/RSTG connections. The Bayesian model selection analysis found modulatory lateralization in non-medicated FESs. Conclusion: The FES patients showed fronto-temporal hyperactivity and dysconnectivity. The effective connections accompanied with modulation were improved when hallucination diminished at early response of routine medication. Significance: This study provided first evidence for early drug response-related alterations in effective brain connectivity.

Keywords: schizophrenia, hallucination, electroencephalogram, dynamic casual modelling, effective connection.

Effective connection and hallucination

4

1. Introduction The core impairments in schizophrenia include positive symptoms, negative symptoms and cognitive deficits. The most common symptom, auditory hallucination, appears in 70-75% of affected individuals (Nayaniand David, 1996). According to patients with schizophrenia, the voices can be inside (i.e. pseudo-form) or outside their heads. Disordered brain connectivity has long been suggested to be a central pathophysiological feature of schizophrenia (Andreasen et al., 1998; Friston, 1999). Recently, a number of neuroimaging and neurophysiological studies have made reference to hypoactivity (Ehrlich et al., 2012) or hyperactivity (Hoffman et al., 2011) in the frontal and temporal areas, and their altered connectivity per se (Quintana et al., 2003; Kam et al., 2013) during auditory hallucinations. It is not clear, however, whether dysconnectivity is the influencing factor for the progress of schizophrenia or whether it is an outcome of the illness or its treatment (Konrad and Winterer, 2008). A recent reviewer of studies in first-episode schizophrenia (FES) on structure neuroimaging and electrophysiology demonstrate that there is evidence for altered connectivity in the early stage of this disorder (Begre and Koenig, 2008), i.e. Crossley et al (2009) has demonstrated increased superior temporal and middle frontal gyrus connectivity in schizophrenia patients under working memory task. As these studies have largely been performed in chronic schizophrenia patients, the question is whether the alterations in brain functional integration between regions are secondary to psychotic illness or its treatment. In the present study, the patients were all at their

Effective connection and hallucination

5

first present of psychosis with typical auditory hallucinations. Our interest points were the alterations of neural network at onset of disease and early response of 2-week treatment, namely, the mechanism difference between drug-naïve and the drug early onset state, which, to our knowledge, has seldom been reported. Following the seminal proposition that communication within neuronal networks is mediated by synchronous neural oscillations (Engel et al., 2001; Fries, 2005; Singer, 1999), there has also been growing interest in the oscillatory mechanisms underlying these connectivity disturbances as revealed by electroencephalogram and magnetoencephalography. Previous electrophysiological studies have demonstrated that Dynamic Causal Modelling (DCM) can provide valid information on mechanisms which represent potential key dimensions of psychiatric disease, e.g., excitation-inhibition balance, synaptic plasticity by N-methyl D-aspartate receptors, or its regulation by neuromodulatory transmitters such as dopamine or acetylcholine (Moran et al., 2011; Schmidt et al., 2013). For example, there is evidence that link gamma and alpha-band to neurotransmitter systems involving parvalbumin-positive GABAergic interneurons and glutamatergic pyramidal cells (Bartos et al., 2007; Sohal et al., 2009; Lozano-Soldevilla et al., 2014). The relationship between beta-band and dopaminergic system (Cagnan et al., 2015) has also been reported. DCM can explain measured data by a network model, which consists of a several dynamically interacting sources. This network model is inverted using a Bayesian approach, and one can make inferences about connections between sources (Friston et

Effective connection and hallucination

6

al., 2003; Stephan et al., 2007). The DCM for cross spectral densities used for steady state responses has been previously reported (Penny et al., 2009; Friston et al., 2012; Friston et al., 2014). The canonical microcircuit (CMC)-type neural mass model is a refinement of the Jansen and Rit convolution models that explicitly accommodates the neuronal sources of forward and backward connections in cortical hierarchies (Bastos et al., 2012). The second-generation antipsychotic medication with fewer side effects relative to first-generation antipsychotics is associated with rapid improvement of positive psychotic symptoms in FES patients (Sanger et al., 1999; Lieberman et al., 2003; Schooler et al., 2005). Atypical antipsychotics produce high temporal cortex D2/D3 occupancy, which is involved in antipsychotic efficacy (Stone et al., 2009). In clinical practice, the early response of atypical antipsychotics was observed within 2 weeks in schizophrenia patients (Leucht et al., 2005; Levine andLeucht, 2012). A recent research reports a short-term effect of antipsychotic treatment can increase regional synchronous neural activity in frontal and temporal areas (Lui et al., 2010). But the relevant mechanism of brain effective connection change at early response stage of antipsychotics, i.e. after 2-week medication in FES patients, is still unclear. The first aim of the present study was to examine the brain activity difference between drug-naïve FES and the FES with the onset response of medication. We used electroencephalography to get time-frequency oscillations across all electrodes, and gain the source location by source reconstruction method. We then used DCM

Effective connection and hallucination

7

(Fristonet al., 2003; Moran et al., 2009) to estimate the effective connections that these regions exerted on each other. Effective connection is distinct from functional connection, which reflects non-directional correlations between activation in different regions with each other. Our principal hypotheses were: (1) based on the results of source analysis, the regional activities would be different between medicated and non-medicated FES patients; (2) the connectivity exerted by altered regions would be different between medicated and non-medicated FES patients. Evidence would be given for brain network alteration when routine medication responded.

2. Methods 2.1 Participants The study was approved by the Ethics Committee of Second Affiliated Hospital of Zhejiang University School of Medicine, and all participants gave their written informed consent to participate. Patients at their first presentation of psychosis were recruited from psychiatry department of Second Affiliated Hospital, Zhejiang University School of Medicine. A total of 45 patients were enrolled, but for the sake of their cooperation, 39 of them completed the whole experiment and got analyzable data, which were brought into the later analysis. They were diagnosed as having paranoid schizophrenia with auditory hallucinations in the acute phase according to the ICD-10 criteria (WHO, 1992) after a semi-structured clinical interview. The patients had experienced their first acute episode with the disease durations from 1.50 to 4.50 months (mean 2.78 ± 1.01) at the

Effective connection and hallucination

8

time of EEG recording and met DSM-5 (American Psychiatric Association, 2013) criteria for paranoid schizophrenia when subsequently reassessed 6 months after first presentation. Nineteen patients were drug naïve. The other twenty had been treated with atypical antipsychotics for about 2 weeks at chlorpromazine equivalent mean dosage of 446.23 mg/day. Retrospectively, before treatment, the medicated FES had no difference in disease duration and symptom severity compared to non-medicated FES (Table 1). We only recruited the hallucinating patients, which made the comparison focused, and the medication was non-experimental controlled. The severity of clinical symptoms in the medicated and non-medicated groups before medication and on the day of EEG recording was assessed with the Positive and Negative Syndrome Scale (PANSS) (Kay et al., 1987) by an experienced psychiatrist trained in its use. Twenty-two healthy volunteers from medical staff and the local community were recruited. They received a semi-structured clinical interview to exclude any current or lifetime evidence of psychiatric disorder. The socio-demographic data of the three groups were matched (Table 1), and the subjects who had a history of neurological disorder or met the DSM-5 criteria for a substance abuse disorder, were excluded. Before EEG recording, all subjects accepted a 3.0T magnetic resonance imaging (MRI) scan (Signa HDx, GE Medical System, Milwaukee, WI, USA). The Routine axial T1-weighted fluid attenuation inversion recovery (TR=500ms, TE=15ms) MR images were obtained to ensure all subjects

Effective connection and hallucination

9

were free from any brain structural abnormalities. All participants were right-handed. 2.2 EEG Recording Participants were asked to sit in an armchair in a semi-darkened, electrical-shielded, and quiet room, with eyes focusing on a white cross to keep vigilance on a black background of a computer screen 1 m in front of them. EEG recordings were performed at 9-10 a.m. with 32 Ag-AgCl scalp electrodes placed according to the International 10-20 system (Electro-Cap International, Inc., Eaton, Ohio, USA). The linked mastoids of two sides were attached as the reference electrode. The electrodes placed at the outer and supra-orbitally to the right eye were the bipolar recordings of electro-ocular activity (EOG). EEG was continuously recorded with a sampling rate of 1,024 Hz. The impedances were kept below 5 kΩ. The recording lasted for 5 minutes. 2.3 Data Analysis 2.3.1 Pre-processing Offline preprocessing was performed using SPM12 (http://www.fil.ion.ucl.ac.uk/spm), running in Matlab R2013b (Mathworks Inc. Natick, MA, USA). EEG signals were band-pass filtered offline (0.1–50 Hz). A sweep in which the EEG exceeded ±70 µV was excluded from analysis. EEG trials containing electrooculogram, electrocardiogram, or electromyogram were eliminated using independent component analysis with the extended Infomax algorithm implemented

Effective connection and hallucination

10

in the FieldTrip toolbox (fieldtrip.fcdonder.nl). Spectral estimates were derived in successive temporal windows of 2 sec duration, according to the frequency of five bands: delta (0.5 - 3.5 Hz), theta (3.5 - 7.5 Hz), alpha (7.5 - 12.5 Hz), beta (12.5 - 30 Hz) and gamma (30 - 45 Hz). Forty artifact free 2-s epochs were averaged for further source-reconstruction and dynamic casual modeling analysis. 2.3.2 Source-reconstruction The source locations of brain electrical activity were using 3D source reconstruction toolbox of SPM 12. The EEG imaging pipeline was constructed with four consecutive steps, which were Source space modeling, Coregistration, Forward computation and Inverse reconstruction. In terms of empirical priors, the algorithm transformed the EEG data and automatically selected multiple cortical sources with compact spatial support (Friston et al., 2008). The specific 3-D models were constructed in the SPM software followed the default approach. Based on an empirical Bayesian formalism (Friston et al., 2008; Mattout et al., 2006; Phillips et al., 2005), the source space was constructed using a canonical cortical mesh, defined in a standard stereotactic space. This mesh was warped, in a nonlinear fashion, to match the subjects' anatomy (Mattout et al., 2006). 512 dipoles were generated and constrained to be normal to a cortical surface meshin MNI space. Based on a spatial normalization transformation (Friston et al., 2008), we generated each source of the mesh in subject space, which was directly corresponding with a location in MNI space.

Effective connection and hallucination

11

Followed these steps, the 3D images of each subject were generated for further group comparison. 2.3.3 Dynamic casual modelling We use DCM (Friston et al., 2003) as implemented in the SPM12 software, and the EEG data were elaborated as conventional spatial forward models consisting of a few sources (Penny et al., 2009). A DCM of steady-state responses in electrophysiological data was summarized in terms of their cross-spectral density (Moran et al., 2009). The cross-spectral density is a description of the dependencies among the observed outputs and comprises a network of neuronal sources, expressing the steady-state dynamics. In equation (1) (Moran et al., 2009), gij (ω) represented cross-spectral densities, at radial frequencies ω, between channels i and j. The k meant k-th input
  represented the Fourier transforms. Γk(ω) was the cross-transfer function times the spectral density of that innovation, g    . 

Γ    =  Κ      
    

g   = ∑ Γ   g  

(1)

In the present study, the particular source model used here was the CMC model (Bastos et al., 2012), which explained effects of sources in terms of Endogenous connectivity and it spontaneous modulation at resting state (Marreiros et al., 2008). The fit frequency used here was 3.4-45 Hz. The results of the CMC model analysis described a DCM analysis that focused on explaining activity on right middle frontal gyrus (RMFG), left superior temporal gyrus (LSTG), and right superior temporal

Effective connection and hallucination

12

gyrus (RSTG) in their connectivity in terms of source analysis and group comparison as described in the result section. The connectivity parameters were then derived by the estimation of underlying neural activity (Friston et al., 2003; Moran et al., 2009). The choice of subject-specific coordinates was guided by group differences (Fig. 1). Based on statistical differences (details are included in the next section), we determined the three regions as the key region, and modeled each area (each node of the model, Fig. 2a) with a single equivalent current dipole with a prior variance of 12mm as RMFG [40 38 0], LSTG [-52 -40 20] and RSTG [56 -44 15] that were in the centre of a radius of the group maxima difference. The bidirectional endogenous connection between all regions (RMFG, LSTG, and RSTG) was specified as the basic model. This basic three-region model was then systematically derived fifteen alternative model variants. Specifically, we were interested in the role of RMFG in FES patients and how connections between RMFG, LSTG, and RSTG were modulated. Following the “top-down” processingand the supervisor role of frontal lobe (van der Meeret al., 2013), we define the RMFG self-connection as the model 1. The models 2-4 were the bidirectional connection between RMFG, LSTG, and RSTG respectively. Fig. 2b shows the three-region models 1-4, the addition 11 models yielded by combining these models (M) (M5 = 1+2, M6 = 1+3, M7 = 1+4, M8 = 2+3, M9 = 2+4, M10 = 3+4, M11 = 1+2+3, M12 = 1+2+4, M13 = 1+3+4, M14 = 2+3+4, M15 = 1+2+3+4).

Effective connection and hallucination

13

The random effects Bayesian model selection (BMS) was performed after CMC modeling to identify the best model based on the evidence for each model of each subjects (Stephan et al., 2009). Random effects BMS is a hierarchical form of model selection, which allows different models to be assigned to each subject and then evaluates the relative probability of different models over subjects. The exceedance probability k is the probability that a given model k is more likely than any other model, here K represented all tested models and   |; !  was the conditional

model probability (Equation 2) (Stephan et al., 2009). The calculation of the inference follows Gaussian assumption. The details are described by Rigoux et al. (2014) elsewhere. The calculations were conducted through DCM-BMS toolbox. In SPM12, the protected exceedance probability is the significantly corrected inference in model selection.

∀#$ %1 … Κ|# ≠
):


= + >  -; !.

(2)

We optimized our three-region model (details are included in the next section) by considering each group separately. Using random effects BMS (Stemphan et al., 2009), we gained two different models with the distinct protected exceedance probability, and we reported the results for each group separately. Bayesian parameter averaging was then used to estimate the effective connection or modulatory parameters. The averaging weighted all the 15 models of each subject, and these data were prepared for group comparison.

Effective connection and hallucination

14

2.4 Statistical analysis We perform a standard voxel-wise statistical analysis to identify source activations in each participant independently. The general linear model was used to calculate the parameters for all brain voxels, and contrasts were calculated for each group. We then entered the subject-specific contrast images into a second level ANOVA to gain the main effect of brain activity among medicated, non-medicated and healthy control groups. When main effect was significant, a post-hoc t test was performed between each two groups. Statistical inferences were made at a whole brain corrected cluster level (p < 0.05 with a standard voxel-level threshold of p< 0.05). The connection and modulatory parameters were contrasted using one-way ANOVA among three groups and a post-hoc t test with Bonfferoni correction was performed when group effect was significant. A p value < 0.05 was considered to be significance.

3. Results 3.1 Clinical Symptoms The medicated FES and non-medicated FES matched for disease duration, and there was no significant difference in PANSS total, negative and general scores between the two groups at the time of EEG recording. With regard to the positive and hallucinatory symptoms, the medicated FES had both lower scores than the non-medicated ones (Table 1). 3.2 Regional Activity

Effective connection and hallucination

15

The significant main effects among three groups were only found in RMFG, LSTG and RSTG on theta, alpha, beta and gamma frequency band. There was no significant result founded on delta band. Post-hoc analysis revealed higher brain activities of medicated FES than those of healthy controls in RMFG (theta and gamma band), LSTG (beta band) and RSTG (alpha, and gamma band). Also, there were higher brain activities in non-medicated FES than those in healthy controls of RMFG (beta and gamma band), LSTG (alpha, and gamma band) and RSTG (alpha, beta and gamma band). When comparing the two FES groups, the higher activities were found in RMFG (alpha and beta band), LSTG (alpha and beta band) in non-medicated FES than those in medicated ones (Fig. 1 and Table 2). Based on statistical differences, we determined these three regions as the key region, and modeled each node with a single equivalent current dipole with a prior variance of 12 mm as RMFG [40 38 0], LSTG [-52 -40 20] and RSTG [56 -44 15] that were in the centre of a radius of the group maxima difference (Fig. 2a). 3.3 Effective Connection Following inversion of 15 alternative DCMs in each subject (see methods section and Fig. 2b), random effects BMS (Stephan et al., 2009) showed that Model 8 distinct surpassed all other models, with the protected exceedance probability of 0.76 (Fig. 2c) in 85% (17/20) medicated FES patients and 0.80 (Fig. 2e) in 90% (20/22) healthy controls respectively. While the Model 5 gain the winning model in 95% (18/19) non-medicated FES subjects, with the protected exceedance probability of 0.70 (Fig.

Effective connection and hallucination

16

2d). The optimal Model 5 contained bidirectional endogenous connections from RMFG to LSTG with two spontaneous modulations; self modulation in the right frontal area and modulation of the forward connection between RMFG and LSTG. The optimal Model 8 contained bidirectional endogenous connections from the RMFG to bilateral STG with the modulations between RMFG to L/RSTG. A statistical group analysis was then performed on both endogenous and modulatory parameters. ANOVA revealed a significant group effect in the connection of RMFG to LSTG [F(2, 58) = 6.49, p = 0.003] and RMFG to RSTG [F(2, 58) = 3.72, p = 0.03] indicating an overall group different in network architecture. Post-hoc analysis of endogenous connectivity parameters showed significantly decreased RMFG to LSTG connectivity in non-medicated patients relative to medicated patients (p = 0.003) and healthy controls (p = 0.01). Also, the RMFG to RSTG connectivity was significantly decreased in non-medicated patients relative to healthy controls (p = 0.01), but not relative to medicated patients (Fig. 3 and Table 3).

4. Discussion In the present study, the main clinical differences between medicated and non-medicated FES were the positive symptoms, i.e. the hallucinations (Table 1). The result indicated the rapid improvement of hallucinating symptoms in FES might be the feature of early response of medication (Leucht et al., 2005; Levine and Leucht, 2012). Both groups of FES patients expressed significantly increased activation in the

Effective connection and hallucination

17

right middle frontal and bilateral temporal gyrus across 3.5-45 Hz frequency band relative to the healthy controls, and the non-medicated FES had even higher activity in these regions than medicated patients did. The electroencephalogram studies of resting states in schizophrenia have repeatedly reported both reduced and increased engagement of neuronal resources within the frontal and superior temporal cortex (Ehrlich et al., 2012; Hoffman et al., 2011). One possible explanation for this inconsistence is that the difference of disease duration of schizophrenia, namely, patients at chronic or acute phase may express different cortical activities following the progress of brain lesion. In our study, the increasing spontaneous activity in FES, especially in the drug-naïve ones, may be interpreted as reflecting the dopamine dysfunction and support the hypothesis of excitation-inhibition unbalance (Lisman, 2012). The possibility that this increased activity activation was related to the modulating of dopamine in both excitatory glutamate transmission and local inhibition mediated by GABA, namely, dopamine modulates the excitation-inhibition balance in frontal lobe (O'Donnell, 2011). The second-generation antipsychotic agents that regulate the dopamine presynaptic transmission may modulate the excitability of neural circuit (Lyon et al., 2011). As previously reported, the early response of atypical antipsychotics was observed within 2 weeks in schizophrenia patients (Leucht et al., 2005; Levine and Leucht, 2012), and our result verified the modulative phenomena of early onset effect in acute FES patients (Sambataro et al., 2010). To investigate the effective fronto-temporal connectivity in medicated and

Effective connection and hallucination

18

non-medicated FES patients, we constructed a series of three-region DCMs according to the source analysis. Considering the supervisory function of RMFG (van der Meer et al., 2013), we investigated the possibility that increased temporal activity could represent a disinhibitory mechanism in FES, with the RMFG influencing activity in temporal areas and thereby also influencing their connectivity. As revealed by the DCM analysis, the functional coupling between right middle frontal lobe and bilateral superior temporal cortex was altered in non-medicated FES relative to controls. These results are consistent with the idea that the abnormal interactions within a distributed network of regions can best characterize the neural correlates of schizophrenia, which may relate to the abnormal N-methyl-D-aspartic acid-dependent synaptic plasticity (Friston and Frith., 1995; Gold and Weinberger, 1995; Stephan et al., 2006). Interestingly, the medicated FES presented increasing connection from RMFG to LSTG relative to the non-medicated patients, and this inference had no difference compared to the healthy controls. In the present study, all the FES patients were at their acute phase with typical auditory hallucinations. The medicated patients were on the early effect onset of antipsychotics, and had lower PANSS positive and hallucinatory scores than the non-medicated ones. We inferred that the increasing of RMFG-L/RSTG effective connection might help on improvement of positive especially hallucinatory symptoms. Using BMS, we systematically compared 15 competing models of RMFG, LSTG and RSTG integration during resting state. Across all the subjects of each group, the

Effective connection and hallucination

19

winning model of non-medicated FES was model 5, while the medicated FES and controls had the outer performed model of model 8. The difference between each model was the modulatory connection. Model 5 was the modulation of RMFG self-connection plus RMFG-LSTG connection, while model 8 was the modulation of RMFG to bilateral STG connection. The result indicated the modulatory lateralization was significant in non-medicated FES patients. When reference to the regional connection and activity of drug-naïve FES, they presented lower coupling connection from RMFG to bilateral STG and the hyperactivity in these three regions. The results inferred that the altered neurotransmitter induced unbalanced excitation-inhibition activity, which presented the weak fronto-temporal connection and lateralized modulation in drug-naïve hallucinating brain. These altered connection and modulation could be corrected in 2-week antipsychotic medication, which evidenced by drug early-responded FES patients. One should bear in mind the limitation of our study: (1) the sample size may be small for the pair-wise analyses, which made the number of significant results limited, further research should recruit more subjects, especially the drug-naïve FES; (2) using a 32-channel EEG system for source reconstruction might have some insufficient in accurate precise of localization, more channels should be applied in the future study. Meanwhile, our study provided a resting-state brain network difference between 2-week medicated and non-medicated FES patients, which gave the neural evidence on early response of medication. This was a non-experimental control of medication

Effective connection and hallucination

20

study and the further research should be conduct on cohort design and bring other psychiatric disorders into the comparison. In summary, using electroencephalogram and dynamic casual modeling, we found evidence of fronto-temporal hyperactivity and dysconnectivity in FES patients with auditory hallucinations. The effective connections accompanied with modulation were improved when hallucination diminishedeven at early response of routine medication.

Effective connection and hallucination

21

Conflict of interest The authors declare no conflict of interest.

Acknowledgements This work was supported by the Natural Science Foundation of Zhejiang Province in China (Contract grant number: LY16H090009).

Effective connection and hallucination

22

References American Psychiatric Association. The Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5). 2013. Andreasen NC, Paradiso S, O’Leary DS. “Cognitive dysmetria” as an integrative theory of schizophrenia: a dysfunction in cortical-subcortical-cerebellar circuitry? Schizophr Bull 1998; 24: 203-18. Bartos M, Vida I, Jonas P. Synaptic mechanisms of synchronized gamma oscillations in inhibitory interneuron networks. Nat Rev Neurosci 2007; 8: 45-56. Bastos AM, Usrey WM, Adams RA, Mangun GR, Fries P, Friston KJ. Canonical microcircuits for predictive coding. Neuron 2012; 76: 695-711. Begre S, Koenig T. Cerebral disconnectivity: an early event in schizophrenia. Neuroscientist 2008;14: 19-45. Cagnan H, Duff EP, Brown P. The relative phases of basal ganglia activities dynamically shape effective connectivity in Parkinson's disease. Brain 2015; 138: 1667-78. Crossley NA, mechelli A, Fusar-poli P, Broome MR, Matthiasson P, Johns LC, Bramon E, Vamaggia L, Williams SC, McGuie PK. Superior temporal lobe dysfunction and frontotemporal dysconnectivity in subjects at risk of psychosis and in first-episode psychosis. Hum Brain Mapp 2009; 30: 4129-37. Ehrlich S, Brauns S, Yendiki A, Ho BC, Calhoun V, Schulz SC, Gollub RL, Sponheim SR. Associations of cortical thickness and cognition in patients with

Effective connection and hallucination

23

schizophrenia and healthy controls. Schizophr Bull 2012; 38: 1050-62. Engel AK, Fries P, Singer W. Dynamic predictions: oscillations and synchrony in top-down processing. Nat Rev Neurosci 2001; 2: 704-16. Fries P. A mechanism for cognitive dynamics: neuronal communication through neuronal coherence. Trends Cogn Sci 2005; 9: 474-80. Rigoux L, Stephan KE, Friston KJ, Daunizeau J. Bayesian model selection for group studies - revisited. Neuroimage 2014; 84: 971-85. Friston KJ, Bastos A, Litvak V, Stephan KE, Fries P, Moran R. DCM for complex valued data: Cross-spectra, coherence and phase-delays. Neuroimage 2012; 59: 439-55. Friston KJ, Bastos AM, Oswal A, van Wijk B, Richter C, Litvak V. Granger causality revisited. Neuroimage 2014; 101: 796-808. Friston KJ, Frith CD. Schizophrenia: a disconnection syndrome? Clin Neurosci 1995; 3: 89-97. Friston KJ, Harrison L, Daunizeau J, Kiebel S, Phillips C, Trujillo-Barreto N, Henson R, Flandin G, Mattout J. Multiple sparse priors for the M/EEG inverse problem. Neuroimage 2008; 39: 1104-20. Friston KJ, Harrison L, Penny WD. Dynamic Causal Modelling. Neuroimage 2003; 19: 1273-1302. Friston KJ. Schizophrenia and the disconnection hypothesis. Acta Psychiatr Scand Suppl 1999; 395: 68-79.

Effective connection and hallucination

24

Gold J, Weinberger D. Cognitive deficits and the neurobiology of schizophrenia. Curr Opin Neurobiol 1995; 5: 225-30. Hoffman RE, Pittman B, Constable RT, Bhagwagar Z, Hampson M. Time course of regional brain activity accompanying auditory verbal hallucinations in schizophrenia. Br J Psychiatr 2011; 198: 277-83. Kam JW, Bolbecker AR, O’Donnell BF, Hetrick WP, Brenner CA. Resting state EEG power and coherence abnormalities in bipolar disorder and schizophrenia. J Psychiatr Res 2013; 47: 1893-1901. Kay SR, Fiszbein A, Opler LA. The Positive and Negative Syndrome Scale (PANSS) for schizophrenia. Schizophr Bull 1987; 13: 261-76. Konrad A, Winterer G. Disturbed structural connectivity in schizophrenia primary factor in pathology or epiphenomenon? Schizophr Bull 2008; 34: 72-92. Leucht S, Busch R, Hamann J, Kissling W, Kane JM. Early-onset hypothesis of antipsychotic drug action: a hypothesis tested, confirmed and extended. Biol Psychiatry 2005; 57: 1543-49. Levine SZ, Leucht S. Early symptom response to antipsychotic medication as a marker of subsequent symptom change: An eighteen-month follow-up study of recent episode schizophrenia. Schizophr Res 2012; 141: 168-72. Lieberman JA, Tollefson G, Tohen M, Gree AI, Gue RE, Kahn R, McEvoy J, Perkins D, Sharma T, Zipursky R, Wei H, Hamer RM, HGDH Study Group. Comparative efficacy and safety of atypical and conventional antipsychotic drugs in

Effective connection and hallucination

25

first-episode psychosis: a randomized, double-blind trial of olanzapine versus haloperidol. Am J Psychiatry 2003; 160: 1396-1404. Lisman J. Excitation, inhibition, local oscillations, or large-scale loops: what causes the symptoms of schizophrenia? Curr Opin Neurobiol 2012; 22: 537-44. Lozano-Soldevilla D, terHuurne N, Jensen O. GABAergic modulation of visual gamma and alpha oscillations and its consequences for working memory performance. Curr Biol 2014; 24: 2878-87. Lui S, Li T, Deng W, Jiang L, Wu Q, Tang H. Yue Q, Huang X, Chan RC, Collier DA, Meda SA, Pearlson G, Mechelli A, Sweeney JA, Gong Q.Short-term effects of antipsychotic treatment on cerebral function in drug-naïve first-episode schizophrenia revealed by "resting state" functional magnetic resonance imaging. Arch Gen Psychiatry 2010; 67: 783-92. Lyon GJ, Abi-Dargham A, Moore H, Lieberman JA, Javitch JA, Sulzer D. Presynaptic regulation of dopamine transmission in schizophrenia. Schizophr Bull 2011;37: 108-17. Marreiros AC, Daunizeau J, Kiebel SJ, Friston KJ. Population dynamics: Variance and the sigmoid activation function. Neuroimage 2008; 42: 147-57. Mattout J, Phillips C, Penny WD, Rugg M, Friston KJ. Meg source localization under multiple constraints: an extended Bayesian framework. Neuroimage 2006; 30: 753-67. Moran RJ, Rosalyn J, Symmonds M, Stephan KE, Friston KJ, Dolan RJ. An in vivo

Effective connection and hallucination

26

assay of synaptic function mediating human cognition. Curr Biol 2011; 21: 1320-25. Moran RJ, Stephan KE, Seidenbecher T, Pape HC, Dolan RJ, Friston KJ. Dynamic causal models of steady-state responses. Neuroimage 2009; 44: 796–811. Nayani TH, David AS. The auditory hallucination: a phenomenological survey. Psychol Med 1996; 26: 177-89. O'Donnell P. Adolescent onset of cortical disinhibition in schizophrenia: insights from animal models. Schizophr Bull 2011; 37: 484-92. Penny WD, Litvak V, Fuentemilla L, Duzel E, Friston KJ. Dynamic Causal Models for phase coupling. J Neurosci Meth 2009; 183: 19-30. Phillips C, Mattout J, Rugg MD, Maquet P, Friston KJ. An empirical Bayesian solution to the source reconstruction problem in EEG. Neuroimage 2005; 24: 997-1011. Quintana J, Wong T, Ortiz-Portillo E, Kovalik E, Davidson T, Marder SR, Mazziotta JC. Prefrontal-posterior parietal networks in schizophrenia: primary dysfunctions and secondary compensations. Biol Psychiatry 2003; 53: 12-24. Sambataro F, Blasi G, Fazio L, Caforio G, Taurisano P, Romano R, Di Giorgio A, Gelao B, Lo Bianco L, Papazacharias A, Popolizio T, Nardini M, Bertolino A. Treatment with olanzapine is associated with modulation of the default mode network in patients with Schizophrenia. Neuropsychopharmacol 2010; 35: 904-12.

Effective connection and hallucination

27

Sanger TM, Lieberman JA, Tohen M, Grundy S, Beasley C, Tollefson GD. Olanzapine versus haloperidol treatment in first-episode psychosis. Am J Psychiat 1999; 156:79-87. Schmidt A, Diaconescu AO, Kometer M, Friston KJ, Stephan KE, Vollenweider FX. Modeling ketamine effects on synaptic plasticity during the mismatch negativity. Cereb Cortex 2013; 23: 2394-2406. Schooler N, Rabinowitz J, Davidson M, Emsley R, Harvey PD, Kopala L, McGorry PD, Van Hove I, Eerdekens M, Swyzen W, De Smedt G, Early Psychosis Global Working Group. Risperidone and haloperidol in first-episode psychosis: a long-term randomized trial. Am J Psychiat 2005; 162: 947-53. Singer W. Neuronal synchrony: a versatile code for the definition of relations? Neuron 1999; 24: 49-65, 111-25. Sohal VS, Zhang F, Yizhar O, Deisseroth K. Parvalbumin neurons and gamma rhythms enhance cortical circuit performance. Nature 2009; 459: 698-702. Stephan KE, Harrison LM, Kiebel SJ, David O, Penny W, Friston KJ. Dynamic causal models of neural system dynamics: current state and future extensions. J Biosci 2007; 32: 129-44. Stephan K, Baldeweg T, Friston K. Synaptic plasticity and disconnection in schizophrenia. Biol Psychiat 2006; 59: 929-39. Stephan KE, Penny WD, Daunizeau J, Moran RJ, Friston KJ. Bayesian model selection for group studies. Neuroimage 2009; 46: 1004-17.

Effective connection and hallucination

28

Stone JM, Davis JM, Leucht S, Pilowsky LS. Cortical dopamine D2/D3 receptors are a common site of action for antipsychotic drugs- an original patient data meta-analysis of the SPECT and PET in vivo receptor imaging literature. Schizophr Bull 2009; 35: 789-97. van der Meer L, de Vos AE, Stiekema AP, Pijnenborg GH, Nolen WA, David AS, Aleman A. Insight in schizophrenia: involvement of self-reflection networks? Schizophr Bull 2013; 39: 1288-95. World Health Organization. The ICD-10 Classification of Mental and Behavioural Disorders: Clinical Descriptions and Diagnostic Guidelines. Geneva, WHO.1992.

Effective connection and hallucination

29

Figure Legends Fig. 1. Statistical parametric maps of 3 groups. The column panels represented the five frequency bands (delta, theta, alpha, beta and gamma). The row panels represented the comparison among three groups including main effects and post-hoc analysis. The first row represented the main effects of ANOVA among medicated/non-medicated FES and healthy controls; the second row represented the differences between medicated FES and healthy controls; the third row represented the differences between non-medicated FES and healthy controls; the bottom row presented the differences between non-medicated FES and medicated FES. The detailed information was presented in Table 2. The figures were generated using xjView (http://www.alivelearn.net/xjview). FES = first episode schizophrenia.

Fig. 2. (a) Prior locations for the nodes in the models. Sources of activity were modeled as equivalent dipoles. Their prior mean locations in mm are superimposed in an MRI of a standard brain in MNI space. (b) Competing three-region DCMs of effective connectivity constructed with bidirectional connections to/from RMFG, LSTG, RSTG. Model 1 to 4 specify different location for modulatory of Endogenous inputs, an additional 11 models were constructed by combinations of these models (M) (M5 = 1+2, M6 = 1+3, M7 = 1+4, M8 = 2+3, M9 = 2+4, M10 = 3+4, M11 = 1+2+3, M12 = 1+2+4, M13 = 1+3+4, M14 = 2+3+4, M15 = 1+2+3+4). (c) Protected

Effective connection and hallucination

30

exceedance probabilities for three-region models 1-15 in medicated FES. (d) Protected exceedance probabilities for three-region models 1-15 in non-medicated FES. (e) Protected exceedance probabilities for three-region models 1-15 in controls. X-axis = model number, Y-axis = Protected exceedance probability, the diagrams of the winning models were presented beside each bar graph.

Fig. 3. The endogenous connectivity differences among three groups. Significant main effects presented in the backward connections from RMFG to bilateral STG. * non-medicated FES vs. healthy controls (p < 0.05); # non-medicated FES vs. medicated FES (p < 0.05). More information regarding the parameter estimates of the three-region models was detailed in Table 3. FES = first episode schizophrenia; RMFG = right middle frontal gyrus; LSTG = left superior temporal gyrus; RSTG = right superior temporal gyrus.

Effective connection and hallucination

31

Table 1.Mean and standard deviations for demographic characteristics of three group participants and symptom ratingsin schizophrenia patients

Age

medicated FES

non-medicated FES

Heath controls (n

(n = 20)

(n = 19)

= 22)

21.75 (4.70)

21.21(6.72)

22.91 (6.91)

Statistic value

F (2,58) = 0.41, p =

0.67

Gender (male/female)

12/8

7/12

7/15

χ2 = 3.78, df = 2, p

= 0.15

Education-self (years)

15.55 (2.67)

15.79 (2.04)

15.41 (1.53)

F (2,58) = 0.17, p =

0.85

Education-father (years)

13.15 (2.08)

13.74 (2.38)

14.77 (2.20)

F (2,58) = 2.89, p =

0.07

Education-mother

12.40 (2.44)

13.74 (2.67)

14.10 (2.10)

(years)

Disease duration

F (2,58) = 2.60, p =

0.08

3.10 (1.03)

2.58 (0.96)

NA

(months)

F (1,37) = 2.65, p =

0.12

PANSS total (before /

64.50 (7.72) /

after medication)

59.70 (9.40)

64.26 (8.90)

NA

F (1,37) = 0.10, p =

0.93 /

F (1,37) = 1.67, p =

0.20

PANSS positive (before

17.60 (3.53) /

17.68 (2.65)

NA

F (1,37) = 0.07, p =

Effective connection and hallucination

/

32

0.93 /

14.40 (3.97)

after medication)

F (1,37) = 9.15, p =

0.005

PANSS negative (before

10.25 (2.34) /

/

10.45 (3.19)

9.11 (1.70)

NA

F (1,37) = 3.04, p =

0.09 /

after medication)

F (1,37) = 2.66, p =

0.11

PANSS general (before /

36.35 (4.90) /

after medication)

35.40 (5.11)

37.52 (7.54)

NA

F (1,37) = 0.34, p =

0.57 /

F (1,37) = 1.07, p

=0.31

PANSS hallucination

4.45 (0.60) /

(before /

3.50 (0.61)

4.58 (0.51)

NA

F (1,37) = 0.87, p

=0.35

after medication)

F (1,37) = 36.08,

p<0.001

Medication (n)

CPZ equivalents

Olanzapine (7)

NA

NA

NA

Quetiapine (5)

NA

NA

NA

Risperidone (5)

NA

NA

NA

Paliperidone (3)

NA

NA

NA

446.23 (68.36)

NA

NA

NA

Notes: FES = first episode schizophrenia. PANSS = positive and negative syndrome

Effective connection and hallucination

scale. CPZ = chlorpromazine. NA = not applicable

33

Effective connection and hallucination

Table 2. Coordinates of remarkable differences in three sources among/between groups on four bands (only significant results were presented) x

y

z

LSTG

-60

-40

12

RSTG

57

-43

14

Statistic value

theta band three group comparison

F(2,58) = 7.35; p = 0.01

medicated FES vs control RMFG

t = 2.45; p = 0.01 48

28

5

alpha band three group comparison

F(2,58) = 7.35; p = 0.01

LSTG

-59

-45

15

RSTG

56

-42

16

medicated FES vs control RSTG

t = 2.45; p = 0.01 51

-44

16

non-medicated FES vs control LSTG

t = 2.15; p = 0.02 -52

-39

12

RMFG

45

20

5

LSTG

-52

-43

15

non-medicated FES vs medicated FES

t = 2.45; p = 0.01

beta band three group comparison

F (2,58) = 7.35; p = 0.01

RMFG

42

38

0

LSTG

-58

-50

15

RSTG

56

-42

16

medicated FES vs control LSTG

t = 2.15; p = 0.02 59

-50

18

non-medicated FES vs control

t = 2.15; p = 0.02

RMFG

40

36

-2

RSTG

53

-44

12

non-medicated FES vs medicated FES

t = 2.57; p = 0.01

RMFG

42

38

0

LSTG

-56

-42

10

60

-42

8

gamma band 3-group comparison LSTG

F(2, 58) = 7.35; p = 0.01

medicated FES vs control

t = 2.45; p = 0.01

RMFG

42

40

2

RSTG

53

-44

16

non-medicated FES vs control

t = 2.45; p = 0.01

RMFG

38

36

-2

LSTG

-53

-44

12

34

Effective connection and hallucination

RSTG 55 -44 12 Notes: FES, first episode schizophrenia; RMFG, right middle frontal gyrus; LSTG, left superior temporal gyrus; RSTG, right superior temporal gyrus.

35

Effective connection and hallucination

Table 3. DCM modulatory (bilinear terms) and endogenous parameters for all connections in medicated/non-medicated and healthy participants. medicated

non-medicated

Heathy

F

FES

FES

controls

(2,

n = 20

n = 19

n = 22

58)

Modulatory

Mean

Mean (S.E.M)

Mean

parameters

(S.E.M)

RMFG self

0.04

connection

(0.02)

RMFG →

0.04

LSTG

(0.03)

RMFG →

0.03

RSTG

(0.02)

LSTG →

-0.10

RSTG

(0.04)

p

(S.E.M)

0.05 (0.04)

0.03

0.09

0.91

0.91

0.41

0.05

0.96

0.40

0.67

0.06

0.94

0.24

0.78

(0.03)

0.06 (0.09)

0.05

(0.04)

0.04 (0.02)

0.05

(0.03)

-0.07 (0.03)

-0.09

(0.08)

Endogenous

parametera

LSTG →

0.02

RMFG

(0.03)

RSTG →

0.03

RMFG

(0.02)

0.03 (0.01)

0.05

(0.03)

-0.01 (0.02)

0.04

(0.03)

Post-hoc analysis

36

Effective connection and hallucination

RMFG →

0.21

LSTG

(0.11)

-0.05 (0.09)

0.10

6.49

0.003

37

medicated FES vs Heathy controls:

n.s.

(0.03)

non-medicated FES vs Heathy

controls: t = -2.67; p = 0.01

non-medicated FES vs medicated

FES: t = -3.14; p = 0.003

RMFG →

-0.02

RSTG

(0.02)

-0.03 (0.04)

0.14

3.72

0.03

medicated FES vs Heathy controls:

n.s.

(0.05)

non-medicated FES vs Heathy

controls: t = -2.71; p = 0.01

non-medicated FES vs medicated

FES: n.s.

LSTG →

0.04

RSTG

(0.02)

RSTG →

-0.04

LSTG

(0.02)

0.10 (0.09)

0.04

1.05

0.36

0.59

0.56

(0.04)

-0.08 (0.05)

-0.04

(0.03)

Notes: Bayesian parameter averaging weighted all the 15 models of each subject. FES, first episode schizophrenia; n.s., no significance; RMFG, right middle frontal gyrus; LSTG, left superior temporal gyrus; RSTG, right superior temporal gyrus.

Effective connection and hallucination

38

Effective connection and hallucination

39

Effective connection and hallucination

40