Altering patterns of sensorimotor network in patients with different pathological diagnoses and glioma‐related epilepsy under the latest glioma classification of the central nervous system

Abstract Aims We aimed to clarify the relationship between alterations in functional networks and glioma‐related epilepsy (GRE) in patients with different molecular diagnoses. Methods We enrolled 160 patients with prefrontal gliomas and different histories of GRE. The patients were grouped based on the latest pathological glioma classification and GRE history. Graph theory analysis was applied to reveal alterations in the sensorimotor networks among various subgroups. Binary logistic regression was used to identify risk factors for preoperative GRE onset. Results Decreasing shortest path length was found in patients with GRE, regardless of the chromosome 1p/19q status. Nodes located in the premotor and supplementary motor areas showed decreased nodal betweenness centrality and vulnerability in patients with GRE and chromosome 1p/19q intact. Additionally, the node on the primary motor area showed decreased nodal vulnerability but the node on the sensory‐related thalamus increased in patients with GRE and chromosome 1p/19q co‐deletion. Decreased shortest path length, grade 2, and decreased nodal betweenness centrality of the premotor area were risk factors for GRE. Conclusion Decreased shortest path length was a characteristic alteration in GRE and prefrontal glioma. Alterations in global properties were similar, but nodal properties were different in patients with GRE and different chromosome 1p/19q statuses.


| INTRODUC TI ON
Glioma-related epilepsy (GRE) is an abnormal brain network-like disease. [1][2][3][4] Glioma with a specific genetic background is susceptible to GRE. However, the relationship between brain network alterations and genetic background in patients with GRE onset is not well known.
Relying on resting-state functional MRI (rs-fMRI) and applying topological properties 5,6 to reveal alterations in the conveying ability of brain-function networks is valuable for identifying important nodes related to specific symptoms. 7,8 Previous studies 9-12 have demonstrated that different tumor locations cause alterations in different functional networks near the tumor. For instance, increased functional connectivity (FC) of nodes located in the premotor area and decreased shortest path length of the sensorimotor network are specific alterations in patients with prefrontal glioma and GRE onset. 13 Oligodendroglioma 14,15 and astrocytoma with isocitrate dehydrogenase (IDH) mutation 16,17 were considered to be risk factors for GRE onset. One of the potential mechanisms to induce GRE at the microcosmic level was that the d-2-hydroxyglutarate products of IDH mutations potentially increase neuronal activity by mimicking the activity of glutamate on the NMDA receptor. 18 However, at the microcosmic level, the effect of chromosome 1p/19q co-deletion in causing GRE remains unknown. Simultaneously, the explanation for these phenomena through the sight of alterations of functional network has not been sufficient. At present, adult diffuse glioma was classified based on IDH and chromosome 1p/19q status, and the oligodendroglioma and astrocytoma were both considered IDH mutations according to the latest World Health Organization (WHO) central nervous system (CNS) tumor classification. 19,20 Hence, investigation the different alterations of functional networks induced by astrocytoma and oligodendroglioma will help reveal the role of chromosome 1p/19q co-deletion in GRE. Furthermore, due to different molecular diagnoses corresponding to different strategies in glioma patients, 21 we considered that the therapy for controlling GRE onset might be different.
In this study, a new cohort of patients with prefrontal glioma and different histories of GRE onset compared to those in a previous study was reviewed. 13 We aimed to (1) validate previous conclusions, (2) determine the relevance of different IDH statuses, functional network alterations, and GRE onset, and (3) investigate specific alterations in patients with a different history of GRE onset and different chromosome 1p/19q status.

| MATERIAL S AND ME THODS
The study protocol was approved by the local institutional review board.

| Participants
Between January 2019 and February 2022, 210 patients from Beijing Tiantan Hospital who had primary prefrontal lobe gliomas were reviewed. The included patients were: (1)

| Clinical information collection
Information regarding preoperative seizures, type of GRE onset, and history of antiepileptic drugs were extracted from inpatient records.
Follow-up information regarding epileptic control was obtained via telephone interviews at 1 year postoperatively.

| MRI acquisition
In brief, preoperative (72 h before surgery) and postoperative (within 24 h after surgery) T1-3D images with gadolinium contrast enhancement and T2 fluid-attenuated inversion recovery (FLAIR) were used to acquire tumor and tumor resection images. Additionally, echoplanar imaging was used to acquire rs-fMRI data. The parameters of these MRI sequences were the same as those previously used 9,13 and are shown in the Appendix S1.

| Functional MRI preprocessing
In brief, rs-fMRI preprocessing was supported by graph theoretical network analysis (GRETNA) software (https://www.nitrc.org/proje cts/gretna). 22,23 The preprocessing pipeline is the same as previously described. 9 The preprocessing parameters are provided in the Appendix S1.

| Regions of tumor invasion and extent of tumor resection
The regions of gliomas with IDH mutation and glioblastoma with IDH wildtype were manually drawn based on T2 FLAIR images and the T1-contrast enhanced region, respectively. Tumor volume was calculated using the volumetric method, based on the individual tumor masks. Regarding the extent of tumor resection (EOR), the residual tumor was manually drawn by using the same MRI sequence as was used for tumor volume. The EOR was calculated with the following formula 24 : To show the tumor invasion, the Montreal Neurological Institute standard space was used to normalize all tumor masks using SPM8 (http://www.fil.ion.ucl.ac.uk/spm/softw are/spm8/) software ( Figure 1).

| Template for functional connectivity calculation
The sensorimotor template for FC calculation was extracted from the brain atlas "brainnetome 274" (http://www.brain netome.org/). 25 To avoid bias during normalization of the individual image to echoplanar imaging template 26 caused by tumor invasion, the regions of the extracted sensorimotor template that were invaded by glioma were excluded. Detailed information on the sensorimotor template is provided in the Appendix S1.

| Network construction
The FC was calculated using Pearson's correlation analysis, and all FCs were generated into a matrix. To further investigate the topological properties, all negative FC in the matrices were excluded, and a Z-score transformation was performed for all matrices.

| Graph theoretical measures
Global properties were calculated using "brant" software (http:// www.brain netome.org/), including cluster coefficient, fault tolerance, global efficiency, local efficiency, the shortest path length, transformation, and vulnerability. Nodal properties were calculated using GRETNA software, including nodal betweenness, nodal cluster coefficient, nodal degree centrality, nodal efficiency, nodal local efficiency, and nodal vulnerability (using "brant" software). 27 The number of random networks was 10,000. A well-acceptable sparsity was applied (from 0.15 to 0.40, interval 0.01). 28

| Statistical analyses
All statistical analyses were performed using SPSS software (

| Demographic characteristics
The number of patients in each subgroup according to the principle of classification is shown in Figure 2. Except for EOR (left,

| Alterations of global properties in patients with different molecular diagnoses and history of preoperative epilepsy
After post hoc analysis with Sidak correction (Tables S1 and

| Alterations of nodal properties in patients with different molecular diagnoses and history of preoperative epilepsy
After Sidak correction, several significant alterations of nodal properties were found between the three groups (grp-E, grp-nE, and grp-GnE; Tables S3-S14; Figures 3 and 4)

| Alterations of nodal properties in patients with different chromosome 1p/19q co-deletion status and history of preoperative epilepsy
Regarding grp-AE and grp-AnE, after Student's t-test comparison, several significant alterations of nodal properties were found (Tables S17-S22)

Grp-GnE (mean ± SEM)
p Value   Using EEG to diagnose GRE onset or not

| DISCUSS ION
In this study, we investigated alterations in functional networks in glioma patients with preoperative GRE onset and further investigated specific alterations in patients with different pathologies, based on the latest version of the molecular pathological classification. 19,20 Our findings indicated that decreasing the shortest path length was a common alteration in patients with GRE onset, regardless of molecular pathological diagnosis. However, alterations in nodal properties in BA 6 were specific to patients with astrocytoma.
Moreover, alterations in nodal properties in BA 4 and the sensoryrelated thalamus were specific for patients with oligodendroglioma.
In our study, the ratio of preoperative GRE onset was 39.4%, which is lower than those in previous reports. 14,29 The tumor was far from the eloquent area, which is the main reason for this phenomenon. Pallud et al. 29 found that the risk ratio of GRE onset was 2.89 times higher when the glioma was close to the eloquent area 30 than when it was distant from the eloquent area. In this study, all tumors were located in the prefrontal lobe, which is distant from the sensorimotor cortices. Hence, compared to reports 29,31 of frontal gliomas close to the eloquent area, our ratio of GRE onset is low.
Preoperative GRE onset was related to IDH mutation but was not related to chromosome 1p/19q co-deletion. The mechanism by which IDH mutation induced preoperative GRE onset was the product of d-2-hydroxyglutarate activity in glioma, with IDH mutation potentially increasing neuronal activity by mimicking the activity of F I G U R E 4 Results of nodal properties in each subgroups when the glioma located in the right prefrontal lobe. The grp-E (n = 24), patients with IDH mutation with GRE; grp-nE (n = 22), patients with IDH mutation without GRE; and grp-GnE (n = 16), patients with glioblastoma without GRE group of patients with glioma-related epilepsy.
glutamate on the NMDA receptor. 32 When we analyzed all patients (160 patients in all sub-groups) using their clinical information alone (including molecular histopathology data) ( Decreased shortest path length was an important factor that induced GRE onset in prefrontal glioma. 13 We enrolled a new cohort in which the number of patients was more than twice that in the previous study, 13 and verified that decreasing shortest path length was an independent risk factor for GRE onset. The shortest path length means the lowest cost during the conveying of information in a network. 9 The decreased shortest path length was advantageous for conveying information, but it reduced the threshold of epilepsy onset. 37 Hence, patients with a decreasing shortest path length are susceptible to GRE onset. Due to the small sample size of patients in grp-GE (patients with glioblastoma and GRE onset), in our analysis of network properties, all patients with GRE had IDH mutation ( Table 1). IDH mutation facilitated anaerobic glycolysis, which works against network stability. 38 In other words, it was easier for IDH mutation to induce network alterations than it was for IDH wildtype.
Network alterations increased the probability that the originally stable network shortened the path length. It also explains why patients with IDH mutation were susceptible to GRE onset.
In addition, the contralesional node A6cdl was located on the premotor area that is responsible for integrating and controlling motor function. 39 Nodal betweenness centrality represents the number of shortest pathway through a node, 40 and nodal vulnerability represents the altering ratio of network efficiency if the node is removed. 10 Our results showed that the betweenness Notably, these two nodes were near the tumor. However, compared to grp-AE, in grp-OE, both the node (A4ul) near the tumor and that (Stha) far from the tumor showed alterations in nodal properties.
These two nodes were located in the lesional hemispheric primary motor area (A4ul) and the contralateral hemispheric sensory-related thalamus (Stha). We believe the different alterations between grp-AE and grp-OE might be related to the different histopathological characteristics and mechanisms of network reorganization.
Oligodendroglioma grows more slowly than astrocytoma. 42 Slow tumor growth was an advantage for network reorganization. 43 TA B L E 2 Factors for preoperative GRE onset through binary logistic regression analysis without patients in the grp-GE. Based on the classical theory of network reorganization, the surrounding tissue first participates in network reorganization before the distant area. 44 Hence, our findings support the concept that glioma with different molecular classification would cause different network alterations.

Univariate analysis
Our findings built a bridge between network alterations, specific symptoms, and molecular pathological diagnosis. Additionally, they provided some evidence pertaining to potential targets of intervention, for instance, A4ul on the lesional hemisphere for patients with oligodendroglioma and A6cdl on the contralesional hemisphere for patients with astrocytoma, to be effectively treated with neuro-electro-physiological therapy, such as via transcranial magnetic stimulation 45 and transcranial electrical stimulation. 46 However, a major limitation of the study is that we excluded patients with IDH wildtype grade 2/3 because of a lack of information on epidermal growth factor receptor amplification, chromosome +7/−10, and telomerase reverse transcriptase status. In the future, we will obtain this information and investigate network alterations.

| CON CLUS ION
IDH mutation was meaningful to induce preoperative GRE onset.
Moreover, decreasing shortest path length was a common characteristic alteration of GRE onset for patients with IDH mutation.
Importantly, the GRE onset-related alterations of the nodes in the sensorimotor network were different in patients with different chromosome 1p/19q statuses.

ACK N OWLED G M ENTS
We thank Dr. Meng Lanxi for imaging data acquisition.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no competing financial interests.

DATA AVA I L A B I L I T Y S TAT E M E N T
Anonymized data will be made available on request.