Contralesional functional network reorganization of the insular cortex in diffuse low-grade glioma patients

Diffuse low-grade gliomas (DLGGs) growing on the insular lobe induce contralesional hemispheric insular lobe compensation of damaged functioning by increasing cortical volumes. However, it remains unclear how functional networks are altered in patients with insular lobe DLGGs during functional compensation. Thirty-five patients with insular DLGGs were classified into the left (insL, n = 16) and right groups (insR, n = 19), and 33 healthy subjects were included in the control group. Resting state functional magnetic resonance imaging was used to generate functional connectivity (FC), and network topological properties were evaluated using graph theoretical analysis based on FC matrices. Network-based statistics were applied to compare differences in the FC matrices. A false discovery rate was applied to correct the topological properties. There was no difference in the FC of edges between the control and insL groups; however, the nodal shortest path length of the right insular lobe was significantly increased in the insL group compared to the control group. Additionally, FC was increased in the functional edges originating from the left insular lobe in the insR group compared to the control group. Moreover, there were no differences in topological properties between the insR and control groups. The contralesional insular lobe is crucial for network alterations. The detailed patterns of network alterations were different depending on the affected hemisphere. The observed network alterations might be associated with functional network reorganization and functional compensation.

www.nature.com/scientificreports/ Furthermore, removal of the entire insular lobe in the lesioned hemisphere does not always cause serious neurologic or neuropsychological impairments 7 . However, this does not mean that the insular lobe is insignificant. The reason for the lack of serious neurological deficits is thought to be related to an increase in contralesional homotopic compensation that frequently increases functional connectivity (FC) in the contralesional hemisphere accompanied with decreases in FC within the ipsilateral damaged network 2,19 . A previous study indicated that diffuse insular low-grade gliomas (DLGGs) cause contralateral hemispheric compensation by increasing the volume of the contralateral insular cortex 11 . This increase in cortical volume provides a basis for network reorganization 20 . Hence, the increase in gray matter volume in the contralesional insular lobe of patients with DLGGs implies that the contralesional insular lobe might join in functional compensation through network reorganization. To our knowledge, no study to date has reported how contralesional insular lobe network reorganization leads to the compensation of damaged functions in DLGG patients. Graph theoretical analysis is a reliable technique that analyzes topological functional properties using resting state functional MRI (rs-fMRI) in order to identify alterations in functional networks 21,22 . Identifying differences between patients and healthy subjects may help elucidate the altered characteristics of functional networks in various diseases. In this study, we investigated changes in the functional connectivity (FC) and topological properties of the insular lobe in patients with DLGG.

Results
Demographic characteristics. In total, 35 right-handed patients with a insular lobe glioma (male, n = 14; age, 42.0 ± 10.8 years) were recruited for this study. All patients were subdivided into the left (insL) and right (insR) groups based on the tumor-affected hemisphere (insL, n = 16). The motor function and muscle strength of all patients was normal. Additionally, 33 healthy participants who were matched for sex, age, and education level were recruited and included in the control group (male, n = 18; all right-handed; age, 38.2 ± 1.5 years). The demographic characteristics of enrolled subjects are shown in Table 1.
There were no differences in age (p = 0.48), sex (p = 0.12), and educational level (p = 0.52), MMSE scores (p = 0.52) between the patient and control groups. Moreover, There were no differences in histopathology (p = 0.72) and tumor volume (p = 0.96) between the insL and insR groups. Additionally, there were no differences in IDH mutation (p = 0.71) and chromosome 1p/19q co-deletion (p = 0.72) between the patient groups. Table 1. Demographic and clinical characteristics of patient groups. * Values are presented as means ± standard deviations, unless indicated otherwise. Using Chi-square test to compare difference of sex distribution among three groups. Using one-way ANOVA test to compare differences of age and education level among three groups. Using one-way ANOVA test with nonparametric test to compare the difference of MMSE scores among three groups. Using Student's t test to compare difference of tumor volume between insL group and insR group. Using Chi-square test to compare differences of histopathology, IDH mutation, and chromosome 1p/19q co-deletion between the insL and insR groups. www.nature.com/scientificreports/ Templates of the healthy hemisphere and whole brain. After excluding tumor-invaded regions and the cerebellum, we extracted templates of the healthy hemisphere and whole brain from the "brainnetome atlas" (including 274 sub-regions). The healthy hemisphere and whole brain templates included 118 and 235 subregions, respectively (Supplemental Table 1 and 2).
Functional connectivity in the healthy hemisphere template. After NBS comparison, the FC was increased in 33 edges originating from the left insular lobe in the insR group as compared with that in the control group (threshold of p-value = 0.0001, Fig. 1A,B; detailed functional edges are shown in Supplemental Table 3). In addition, there was no difference in the FC of functional edges originating from the right insular lobe between the control and the insL groups (threshold of p-value = 0.0001).
Functional connectivity in whole brain template. After NBS comparison, the FC was increased in 53 edges originating from the left insular lobe in the insR group as compared with that in the control group after NBS correction (threshold of p-value = 0.0001). Among the 53 edges, 20 were connected to the lesioned hemisphere (Fig. 1C,D; detailed functional edges are shown in Supplemental Table 3). Additionally, between the control and the insL groups, no difference was found in the FC of functional edges that originated from the right insular lobe (threshold of p-value = 0.0001).
Topological properties in the healthy hemisphere template. We observed multiple alterations in the topological properties between the control group and the insL (Fig. S2) and insR (Fig. S3) groups (after www.nature.com/scientificreports/ FDR correction, detailed information in Table 2). For example, when the tumor grew on the left hemisphere, the global efficiency (p = 0.020) was significantly decreased in the insL group as compared with that in the control group. Additionally, the local efficiency (p = 0.003) and shortest path length (p = 0.019) were significantly increased in the insL group as compared with those in the control group. For the right insular lobe node, nodal efficiency was significantly increased (p = 0.002) and nodal shortest path length was decreased (p = 0.001) as compared with those in the healthy group (p-value threshold after FDR correction was 0.002).
When the tumor grew on the right hemisphere, the local efficiency was significantly increased in the insR group as compared with that in the control group (p < 0.001). However, there were no significant differences in global properties and nodal properties between the insR and healthy groups.
Topological properties in whole brain template. We observed multiple alterations in the topological properties between the control group and the insL (Fig. 2) and insR ( Fig. 3) groups (after FDR correction, detailed information in Table 3).  www.nature.com/scientificreports/ When the tumor grew on the left hemisphere, global efficiency was significantly decreased (p = 0.003) in the insL group as compared with that in the control group. Additionally, local efficiency (p = 0.006) and shortest path length (p = 0.004) were significantly increased in the insL group as compared with those in the control group. For the right insular lobe node, nodal efficiency (p = 0.001) was significantly decreased (p-value threshold was 0.002) and nodal shortest path length (p = 0.001) was significantly increased (p-value threshold was 0.002) in the insL group as compared with those in the control group.
When the tumor grew on the right hemisphere, local efficiency (p < 0.001) and shortest path length (p = 0.001) were significantly increased and global efficiency was significantly decreased (p = 0.001) in the insR group as compared with those in the control group. Additionally, no significant alterations in nodal properties between the insR and control groups were detected after FDR correction (p-value threshold was 0.002).

Discussion
In this study, we investigated the alterations of functional networks in patients with insular DLGGs. Our findings indicate that insular gliomas in different hemispheres can induce different alterations of functional networks, and these alterations may be related to functional compensation 10 .  www.nature.com/scientificreports/ Rs-fMRI and graph theoretical analysis are widely accepted as reliable methods for investigating alterations in functional networks 23 . Additionally, previous studies have suggested that the alterations in functional networks may be associated with neuroplasticity in DLGG patients. For instance, van Dokkum et al. reported that the right inferior parietal lobe compensated for impaired language function in DLGG patients through the reorganization of attention networks 24 .
The current study investigated the alterations of functional networks in patients with insular gliomas by graph theoretical analysis. The results showed that DLGGs induced functional networks alterations in the whole brain, especially for the mirror insular lobe in the healthy hemisphere. Global efficiency was significantly decreased, and the shortest path length was significantly increased in both the insL and insR groups as compared with those in the control group. These findings are consistent with the findings from previous studies [25][26][27][28] , and indicates that the pathways of information conveying change as DLGGs grow.
The alterations of networks that occurred on each side of the insular lobe were different. In the insL group, the nodal shortest path length of the right insular lobe was significantly increased as compared with that in the control group. This finding implies that the pathway of information conveying was prolonged because the edges that originated from the left insular lobe were damaged 21 . Therefore, it can be inferred that when the DLGG originated in the left insular lobe, the right insular lobe replaced the left to convey information through the edges that connected the right insular lobe to some nodes (such as nodes from dorsal prefrontal lobe and parietal lobe) on the lesion hemisphere. Conversely, in the insR group, the FC of some edges originating from the left insular lobe was significantly increased in the insR group as compared with that in the control group, but no significant alterations of the nodal shortest path length and nodal efficiency were found. Most of these functional edges connected the insular lobe to the bilateral precentral gyrus, which is responsible for motor functions 29 ; postcentral gyrus, which is responsible for sensory function; inferior parietal lobe, which is responsible for cognitive functions (e.g., calculation, reading, and visual spatial cognition) 30 ; and superior and middle temporal gyrus in the left hemisphere, which are responsible for language functions 31 . These findings indicate that when the DLGG grew in the right insular lobe, the networks that were related to left insular lobe altered via strengthening the original functional edges of the whole brain.
Regarding our findings, two potential explanations existed. The one is that contralesional insula participated in functional network reorganizations. Since, the glioma is commonly thought to disrupt functional networks and decrease FC 25,32,33 . Hence, the increasing FC implied that network reorganization occurred similar to what other studies have shown 3,10,19,20 . No significant sensorimotor and abnormal mental status were found in our patients. Thus, our finding might indicate that the damage in these functions might be compensated for through the different network reorganizations. However, the other viewpoint supported that these increasing FC were intrinsic alterations when brain lesions occurred 33 . These atypical increasing FC only reflected a functional adaption in a short-term but not indicated functional compensation 34 . Even some abnormal increasing FCs were related to worse cognitive performance in patients with glioma 35 . Moreover, some studies found that these increasing FC did not ultimately aid in cognitive function [36][37][38] . Hence, our findings only verified that the contralesional insula participated in brain network alterations. Whether these alterations meant functional network reorganization or functional compensation was controversial.
Furthermore, we propose that these different alterations in insula-related networks are associated with whether DLGGs grew in the dominant hemisphere or not. As is well known, language and other advanced cognitive functions are mainly mediated by the dominant hemisphere [39][40][41][42] . Importantly, all enrolled subjects in the current study were right-handed. Accordingly, for those patients, DLGGs involving the left insular lobe were more likely to affect cognitive functions than those involving the right insular lobe [43][44][45][46][47][48] . Consequently, lesions in the left insular lobe may require greater degrees of alterations of the contralateral insular lobe by building novel functional edges instead of merely strengthening existing ones.
The main limitation of this study was that no effective cognitive tests (such as the Montreal Cognitive Assessment, West Aphasia Battery) were used to evaluate alterations of cognition in glioma patients. Hence, we lacked strong evidence to verify functional plasticity occurring in contralesional insula lobe. In the future, more comprehensive cognitive tests will be applied to validate our findings in future studies. Moreover, histopathology and molecular subtyping may affect the alterations of functional networks in patients with DLGGs. Fortunately, there were no differences in the distributions of histopathology and molecular subtyping of the enrolled patients. However, the small sample size is still a limitation of the study and future large-scale studies are needed to verify our results.

Methods
The study was approved by the institutional review board of Beijing Tiantan Hospital. Written informed consent was obtained from all study participants. All methods used were carried out in accordance with relevant guidelines and regulations.
Participants. This study retrospectively reviewed 44 patients who were diagnosed with DLGG in the insular lobe at Beijing Tiantan Hospital between January 2017 and January 2018. The inclusion criteria were as follows: (a) older than 18 years; (b) gliomas growing on the insular lobe; and (c) no history of biopsy, radiotherapy, or chemotherapy. The exclusion criteria were as follows: (a) gliomas involving the bilateral hemispheres, (b) head motion greater than 1° in rotation or 1 mm in translation, and (c) tumor directly invaded other brain lobes (except insula lobe).

MRI acquisition.
All imaging data were acquired with a MAGNETOM Prisma 3 T MR scanner (Siemens, Erlangen, Germany). Parameters of the T1-magnetization prepared rapid acquisition gradient echo were as fol- Regions of tumor invasion. The extent of DLGG invasion (shown in Fig. S1) was independently and manually drawn by two neuro-radiologists according to the enhanced regions of the T2/FLAIR images. If the images drawn varied by more than 5%, a third neuro-radiologist with 20 years of experience made the final decision regarding the region location.

Regions of interest selection.
Regions of interest (ROIs) were extracted from the "brainnetome atlas" (http://www.brain netom e.org/) 51 to build the FC matrices. The ROI in the healthy hemispheric insular lobe was combined in order to investigate how functional networks compensated. Additionally, two templates (only including the healthy hemisphere and whole brain) were extracted in order to explore how intra-hemispheric and inter-hemispheric functional networks were reorganized.
Network construction. Pearson correlation coefficients were applied to construct the weighted FC matrix by comparing regional mean time series for all possible pairs of nodes. Consequently, two different sizes of FC matrices were constructed based on the healthy hemisphere template and whole brain template.

Graph theoretical analysis.
To identify the characteristics of topological properties, both negative and positive connections were retained. The weighted FC matrices were transformed into binary FC matrices with absolute values and were processed by GRETNA 21,52 . Both global and nodal topological properties, including the global efficiency, shortest length, local efficiency, nodal global efficiency, nodal local efficiency, and nodal shortest path length, were calculated. All matrices were binarized and absolutized to further analyze the topological properties. Moreover, the graph measurements were normalized using randomized networks (random network number: 10000) during calculation of the topological properties.
Statistical analysis. Statistical analyses were performed using GraphPad Prism 7 software. Clinical characteristics were compared between the patient and control groups using Student's t-test, Mann-Whitney U test, one-way ANOVA test, or chi-square test according to the type of data. Additionally, we applied a series of sparsity thresholds (from 0.15 to 0.50, interval 0.01) to explore group differences in network topological properties, and each property was evaluated according to the corresponding matrix. Subsequently, some matrices were generated according to the levels of sparsity. From these generated matrices, the value of each property was calculated. Then, a corresponding curve was made. The area under the curve was applied to compare differences between the patient and control groups. The differences in FC matrices were assessed using a two-sample t-test between the patient and control groups. Network-based statistics (NBS) were applied to compare differences in FC matrices. The original threshold was equal to 0.001, and the time of permutation was 10,000. The false discovery rate (FDR) correction was applied with an original threshold (p = 0.05) to evaluate differences in nodal topological properties between patient and control groups. Clinical information (age, sex, and education) was used to regress out in statistical comparison.
Ethical approval. This study was approved by the IRB of Beijing Tiantan Hospital.
Informed consent. All participants wrote the informed consent of this study.

Conclusions
Insular DLGGs induced alterations in brain functional networks, and the contralesional insular lobe is a crucial role. Interestingly, when the DLGG grew in the left insular lobe, the right insular lobe replaced the left insular lobe to convey information through the edges that connected the right insular lobe to the same nodes on the www.nature.com/scientificreports/ Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creat iveco mmons .org/licen ses/by/4.0/.