Frequency‐specific changes in the default mode network in patients with cingulate gyrus epilepsy

Abstract To identify abnormal functional connectivity of the default mode network in cingulate gyrus epilepsy, which may yield new information about the default mode network and suggest a new cingulate gyrus epilepsy biomarker. Fifteen patients with cingulate gyrus epilepsy (mean age = 21 years) and 15 healthy controls (mean age = 24 years) were studied in the resting state using magnetoencephalography. Twelve brain areas of interest in the default mode network were extracted and investigated with multifrequency signals that included alpha (α, 8–13 Hz), beta (β, 14–30 Hz), and gamma (γ, 31–80 Hz) band oscillations. Patients with cingulate gyrus epilepsy had significantly greater connectivity in all three frequency bands (α, β, γ). A frequency‐specific elevation of functional connectivity was found in patients compared to controls. The greater functional connectivity in the γ band was significantly more prominent than that of the α and β bands. Patients with cingulate gyrus epilepsy and controls differed significantly in functional connectivity between the left angular gyrus and left posterior cingulate cortex in the α, β, and γ bands. The results of the node degree analysis were similar to those of the functional connectivity analysis. Our findings reveal for the first time that brain activity in the γ band may play a key role in the default mode network in cingulate gyrus epilepsy. Altered functional connectivity of the left angular gyrus and left posterior cingulate cortex may be a new biomarker for cingulate gyrus epilepsy.

The International League Against Epilepsy (ILAE; Commission on Classification and Terminology of the International League Against Epilepsy., 1989) classified cingulate gyrus epilepsy as a type of frontal lobe epilepsy in the proposed classification of epilepsy and epileptic syndromes in 1989. ILAE described cingulate gyrus epilepsy as consisting of complex partial seizures with complex motor gestural automatisms at onset, autonomic signs, and changes in mood and affect. However, cingulate gyrus epilepsy is disputed because it may have diverse clinical features or features that may overlap with other frontal lobe epilepsy syndromes. Due to its anatomical location, recording epileptic discharges directly from the cingulate gyrus using scalp electrodes is difficult. Therefore, locating the epileptogenic zone in the cingulate gyrus is difficult; thus, studies of clinical symptoms and electrophysiology in patients with cingulate gyrus epilepsy are rarely conducted or reported.
The default mode network (DMN) was first proposed by Raichle and colleagues and describes the areas of the brain that are more active during a state of alert, awake but not actively goal-directed behavior (Raichle et al., 2001). Many anatomically separated cortical regions compose the DMN, which contains the posterior cingulate cortex (PCC), precuneus (PCUN), inferior parietal cortex (IPC) (mainly angular gyrus; AG), medial temporal (MT) lobes, medial prefrontal cortex (MPFC), and anterior cingulate cortex (ACC; Greicius, Krasnow, Reiss, & Menon, 2003;Greicius & Menon, 2004;Raichle et al., 2001). The PCC is suggested as the only node in the DMN that directly interacts with almost all other nodes according to Fransson and Marrelec (2008). Various brain regions that comprise the DMN are involved in the integration of autobiographic, self-monitoring, and social cognitive functions (Spreng, Mar, & Kim, 2009). However, the precise functions collectively served by the DMN are still largely unknown. Previous evidence suggested that the activity of the DMN was bound to be related to alpha (8-13 Hz) and beta (14-30 Hz) oscillations (Knyazev, Slobodskoj-Plusnin, Bocharov, & Pylkova, 2011;Mantini, Perrucci, Del Gratta, Romani, & Corbetta, 2007).
Studies have found altered functional connectivity (FC) within the DMN in several types of epilepsy, including idiopathic generalized epilepsy (Gotman et al., 2005;Kay et al., 2013), temporal lobe epilepsy (Haneef, Lenartowicz, Yeh, Engel Jr, & Stern, 2014;Liao et al., 2011;Robinson et al., 2017), absence epilepsy (Luo et al., 2011), benign epilepsy with centrotemporal spikes (Li et al., 2019), and infantile spasms (Wang, Li, Wang, Chen, & Huang, 2017 Before recording data, three coils were fixed to the left and right preauricular points and nasion of each subject. A head localization procedure was implemented before and after each acquisition to locate the patient's head relative to the coordinate system fixed to the MEG system. Subjects were put in a supine position with their eyes slightly closed and requested to remain still. Head movement was limited to 5 mm during each recording. If a patient's head moved more than 5 mm, the data were excluded and a new dataset was recorded for the subject. All MEG data were recorded at a sampling rate of 1,000 Hz. In addition, empty room MEG recordings were routinely performed to monitor system and environmental noise.

| MRI scan
Three-dimensional MRI was acquired using a 3-T scanner (Siemens Magneton Vision; Siemens, Munich/Erlangen, Germany). Three fiduciary marks were laid on identical locations to the positions of the three coils used in the MEG recordings with the aid of digital photographs to achieve an accurate co-registration of the two data sets. Subsequently, all anatomic landmarks were made visible in the MRI scans.

| Preprocessing and source reconstruction
The MEG data were preprocessed with the Elekta Maxfilter using a band-stop filter to remove power-line interference (50 Hz Previous evidence has suggested that the activity of the DMN is related to alpha (8-13 Hz) and beta (14-30 Hz) oscillations (Knyazev et al., 2011;Mantini et al., 2007). In this study, we chose alpha (8-13 Hz), beta (14-30 Hz), and gamma (31-80 Hz) band oscillations to reconstruct sources with resting-state MEG data.
To study the neuromagnetic network at source level, we localized significant neuromagnetic activities using accumulated source imaging (Babiloni et al., 2005;Xiang et al., 2014 that was defined as the volumetric summation of source activity over a period of time. Accumulated source imaging was based on the following equation: In Equation (1), Asi delegates accumulated source strength at location r; s represents the time slice; t stands for the time point of MEG data; n indicates total time points of MEG data, and Q indicates the source activity at source r and at time point t. We defined that s ≥ 1 and s ≤ n/2.
We applied two-step beamforming to compute the source activity (Barnes, Hillebrand, Fawcett, & Singh, 2004;Sekihara, Nagarajan, Poeppel, Marantz, & Miyashita, 2001;Xiang, Tenney, et al., 2015). The first step calculated lead fields for each source. The second step produced matrices with MEG data. In the third step, we chose sensors for partial sensor coverage for each voxel with lead field , defined as voxel-based partial sensors. Next, we computed the covariance of voxel-based partial sensors. Then, we computed two sets of magnetic source images through a vector beamformer .
The next step estimated a coherent source and source orientation using the covariance matrix-vector beamformer. When the source direction was worked out, the last step generated the source activity with the scalar beamformer . The detailed mathematical algorithms were described in previous reports (Xiang et al., 2014;Xiang, Tenney, et al., 2015). In our study, the whole brain was scanned at a 6-mm resolution (around 17,160 sources). When the distance between two voxels was shorter than 10 mm, they were regarded as one source.

| Dynamic functional connectivity estimation
Based on previous research (Dai, Zhang, Dickens, & He, 2012;Xiang et al., 2014;, functional networks were analyzed at the source level. In this study, whole-brain source neural networks were estimated by analyzing the correlation of all voxels signals through the aforementioned algorithms. In particular, the correlation of signals from a two source pair was statistically analyzed using the computing correlation coefficient. The correlation factors were based on the following equation: where R(X a ,X b ) represents the correlation of a source pair in two locations ("a" and "b"). The X a and X b represent signals in each of the two sources, which were paired for calculating connection. C (X a , X b ) indicates the mean of the signals in the two sources. SX a and SX b indicate the standard deviation of the signals from the two sources. Moreover, every possible connection was analyzed for each dual-source pair at the source level to reduce possible variance.
The same data analyses were also applied to the MEG data obtained from the 15 healthy subjects. MEG Processor software (Cincinnati, OH) was used to perform the aforementioned calculation.

| Degree of node estimation
In this study, we also estimated the node degree using graph theory.
In graph theory, the degree si of a node i was the number of edges linking to the node and was defined as: where w ij represented the edge that connected node i and node j. The node degree, si, can be used to determine the extent to which the node was central in the graph. We think that investigating both the connection strength and the node degree may reveal different information about functional integrations of the brain network. The above calculation was performed using MEG Processor software (Cincinnati, OH).

| Regions of interest for the DMN
Different DMN coordinates were suggested in previous resting-state fMRI studies (Grady, Luk, Craik, & Bialystok, 2015;Sasai et al., 2014). Intraoperative electrocorticography (ECoG) was used to identify epileptogenic areas and guide epilepsy surgery in all patients.
Three patients (20%) were almost seizure-free. Gender, age, epilepsy duration, and MRI (positive or negative) were not significantly associated with a seizure-free outcome. Pathological abnormalities were found in all patients; 14 patients (93%) had focal cortical dysplasia (FCD; 7 ILAE Type I, 5 Type II, and 2 ILAE Type III); one had malformation of cortical development. The existence of FCD in the surgical specimen was not associated with a seizure-free outcome. The detailed clinical characteristics of the patients are summarized in Table 3.

| FC analysis of the DMN connectivity between patients with cingulate gyrus epilepsy and controls
In the FC analysis, we found that the FC within the DMN was significantly greater than 0 in the control group and patient group. The mean DMN FC strengths of the two groups and the difference of DMN FC strength between the two groups are shown in Figure 1. As shown in Figure 1, similar FC strength was shown in homologous bilateral brain regions. The higher the frequency was, the greater and more common were differences of FC between the two groups.
The between-group differences in the DMN, which was defined by the strength of FC with seeds in the DMN, were analyzed by an independent sample t-test. subjects in only two pairs (LPCC-RITG and LAG-RITG) in the alpha band. However, significantly (p < .05) higher FC strength was found for all three frequency bands (alpha (8-13 Hz), beta (14-30 Hz), and gamma (31-80 Hz) band), as shown in Figure 2. In the alpha band, only two FC strengths (LAG-LPCC and RAG-RPCC) were significantly higher in the cingulate gyrus epilepsy group. In the beta band, significantly higher FC in the patient group was observed in more locations than in the alpha band (26 locations had significantly higher FC). In the gamma band, the patients had significantly higher FC more frequently than controls (42 FC strengths were significantly higher). The details of the significantly higher FCs of the patients are shown in Figure 2. The number of significantly higher FC strengths was equal in bilateral hemispheres. Notably, as shown in Figure 2, the effect of cingulate gyrus epilepsy upon DMN connectivity on the left side was more obvious than that on the right side.

| Node degree analysis between patients with cingulate gyrus epilepsy and controls
Not surprisingly, in the node degree analysis, the results were similar to those of the FC analysis. We found that the difference in node F I G U R E 1 Matrices of the mean functional connectivity (FC) strength of the default mode network (DMN) regions of interest for α, β, and γ bands. Matrices shown separately for the control group (Normal) and cingulate gyrus epilepsy group (Patient), and the difference in DMN FC strength between the two groups degree between the control group and cingulate gyrus epilepsy group was most obvious when the correlation factor was >0.3. With increasing MEG frequency, the difference in node degree between the two groups became significantly greater.
Compared with healthy control subjects, no nodes had a significantly lower node degree in patients with cingulate gyrus epilepsy.
However, in all frequency bands, nodes with a significantly higher node degree were found (p < .05), as shown in Figure 3. In the alpha band, only two nodes (LAG and RAG) had a significantly higher node degree in the cingulate gyrus epilepsy group. Other nodes had a lower degree, but there was no statistically significant difference (p < .05). In the beta band, all selected ROIs had a significantly higher node degree. Likewise, in the gamma band, all selected ROIs had a significantly higher node degree, but the differences were greater than in F I G U R E 2 (a-c) Significantly different functional connectivity (FC) within the default mode network (DMN) in a pseudoanatomical organization between the control group and cingulate gyrus epilepsy group. (d-f) Regional connectivity pairs with a significant between-group difference the beta band. As shown in Figure 3, the bilateral AG showed the most significantly higher node degree in all selected bands.

| γ band neural activity may play a key role in DMN connectivity
Surprisingly, we found a special phenomenon in this study. The significantly higher FC between the cingulate gyrus epilepsy and control groups became more marked with greater MEG frequency (shown in Figure 2). In the alpha band, only two FCs were significantly higher in the patient group. This increased to 26 in the beta band. The gammaband had 42 significantly higher FCs in the patient group. Moreover, for the same FC, the difference between study groups was greater with higher frequency. Similar to the FC, with higher frequency, the difference in node degree between the two groups was increasingly more significant (shown in Figure 3). In the alpha band, only two nodes showed a significantly greater node degree in the cingulate gyrus epilepsy group. In the beta band, all 12 selected ROIs had a significantly higher node degree. In the gamma band, all selected ROIs also had a significantly higher node degree, but the difference was greater than for the beta band. Therefore, the most significant F I G U R E 3 Bar graph comparing the node degree for each region of interest (ROI) within the default mode network (DMN) of the two study groups. The symbols *, **, ***, and **** denote a significant difference of p < .05 corrected, p < .01, p < .001, and p < .0001, respectively, between the cingulate gyrus epilepsy group (Patient) and control group (Normal) differences between cingulate gyrus epilepsy and controls were observed in the γ band. Thus, we consider that γ band neural activity may be a key contributor to MEG-based DMN connectivity.

| Spatially specific FC between the LAG and LPCC in the seed-based maps
In the present study, we found that one FC pair (LPCC-LAG) had the most significant differences between the cingulate gyrus epilepsy and control groups in all three frequency bands. In the alpha band, only two significantly higher FCs (LAG-LPCC and RAG-RPCC) were found.
The LAG-LPCC FC was significantly higher than the RAG-RPCC FC. In the beta and gamma bands, the significantly higher FCs were greater than in the alpha band. However, the most remarkable difference was located in the FC between the LAG and LPCC. Moreover, the between-group difference was increasingly significant with greater frequency: the γ band had the most significant difference (shown in Figure 4). Consequently, we considered that the strength of the FC between the LAG and LPCC was significantly associated with cingulate gyrus epilepsy. It may be useful as a biomarker for cingulate gyrus epilepsy.  we found only two lower FCs (LPCC-RITG and LAG-RITG) in the alpha band in patients with cingulate gyrus epilepsy compared to controls. However, significantly higher FC was found in all selected frequency bands. Different types of epilepsy have distinct effects on the DMN. Chronic epilepsy generally impairs cognition, but it might also contribute to functional reorganization and plasticity (Elger, Helmstaedter, & Kurthen, 2004). A previous study suggested that the functional integration of the DMN could be comparatively weak during childhood, and then become gradually stronger during development (Fair et al., 2008). Therefore, early epileptic seizures in patients with epilepsy may disturb the development of the DMN and change the structure of the DMN. Likewise, chronic cingulate gyrus epilepsy impairs intrinsic brain activity of the DMN and causes functional reorganization and plasticity of the DMN. Further research is needed to determine the clinical relevance of this abnormal FC.

| Activity of the DMN is likely related to γ band neural activity
In previous studies, research on FC in the DMN has uncovered a wide spectrum of oscillatory components, among which, the alpha and beta frequency bands are the most frequently reported (Brookes et al., 2011;de Pasquale et al., 2010de Pasquale et al., , 2012Foster et al., 2016;Hipp, Hawellek, Corbetta, Siegel, & Engel, 2012;Jerbi et al., 2010;Mantini et al., 2007;Marzetti et al., 2013). One study suggested a positive correlation between fluctuations of the blood-oxygen-level-dependent (BOLD) signals in the DMN and alpha power fluctuations (Mantini et al., 2007). Recent MEG evidence suggested that the DMN was a hub of inter-network cortical interactions in the resting state, especially in the alpha and beta frequency bands (de Pasquale et al., 2012).
However, in our study, we found that the most significant differences between cingulate gyrus epilepsy and controls in terms of FC in the DMN were in the γ band. As the cingulate gyrus is one of the core nodes in the DMN, and the PCC is considered as the only node in the DMN that directly interacts with almost all other nodes, we suggest that FC in the DMN was likely related to γ band oscillations (31-80 Hz).
4.5 | MEG is useful for investigating intrinsic brain activity MEG, a well-established technology with better temporal resolution than functional MRI, has shown the ability to contribute to investigations of intrinsic brain activity. In our study, we investigated the DMN in patients with cingulate gyrus epilepsy at the source level, whereas many previous studies have analyzed the DMN at the sensor or electrode level. The entire brain was analyzed in this study, which may make the findings more objective than those using a few contacts.

| CONCLUSION
Our findings identify specific components of the DMN that are affected in cingulate gyrus epilepsy. The results of this study suggest that γ band neural activity may be a key contributor to MEG-based DMN connectivity. Furthermore, the high FC between the LAG and left PCC may be useful as a biomarker for cingulate gyrus epilepsy.