Understanding cognitive functioning in glioma patients: The relevance of IDH‐mutation status and functional connectivity

Abstract Introduction Cognitive deficits occur frequently in diffuse glioma patients, but are limitedly understood. An important marker for survival in these patients is isocitrate dehydrogenase (IDH) mutation (IDH‐mut). Patients with IDH‐mut glioma have a better prognosis but more often suffer from epilepsy than patients with IDH‐wildtype (IDH‐wt) glioma, who are generally older and more often have cognitive deficits. We investigated whether global brain functional connectivity differs between patients with IDH‐mut and IDH‐wt glioma, and whether this measure reflects variations in cognitive functioning in these subpopulations beyond the associated differences in age and presence of epilepsy. Methods We recorded magnetoencephalography and tested cognitive functioning in 54 diffuse glioma patients (31 IDH‐mut, 23 IDH‐wt). Global functional connectivity between 78 atlas regions spanning the entire cortex was calculated in two frequency bands (theta and alpha). Group differences in global functional connectivity were tested, as was their association with cognitive functioning, controlling for age, education, and presence of epilepsy. Results Patients with IDH‐wt glioma had lower functional connectivity in the alpha band than patients with IDH‐mut glioma (p = 0.040, corrected for age and presence of epilepsy). Lower alpha band functional connectivity was associated with poorer cognitive performance (p < 0.034), corrected for age, education, and presence of epilepsy. Conclusion Global functional connectivity is lower in patients with IDH‐wt diffuse glioma compared to patients with IDH‐mut diffuse glioma. Moreover, having lower functional alpha connectivity relates to poorer cognitive performance in patients with diffuse glioma, regardless of age, education, and presence of epilepsy.


| INTRODUC TI ON
Diffuse gliomas are characterized by poor survival and cognitive deficits (Taphoorn & Klein, 2004). An important marker for survival in patient with diffuse glioma is isocitrate dehydrogenase (IDH) mutation (IDH-mut), which is associated with better prognosis but higher epilepsy prevalence (Chen et al., 2017;Yan et al., 2009). Patients with wildtype glioma (IDH-wt) are generally older and more often have cognitive deficits (Wefel, Noll, Rao, & Cahill, 2016), even though they less often have epileptic seizures that are generally linked to poorer cognitive functioning (Chen et al., 2017;Klein, Engelberts, et al., 2003). Although the relevance of IDH status for survival is relatively straightforward (Yan et al., 2009), the higher occurrence of cognitive deficits at diagnosis in patients with IDH-wt glioma as compared to patients with IDH-mut glioma is incompletely understood. This is in part because of the mentioned group differences in age and prevalence of epilepsy. IDH-wt-related cognitive deficits have been linked to deterioration in cortical thickness networks (Kesler, Noll, Cahill, Rao, & Wefel, 2017). However, cortical thickness is highly age-dependent (Thambisetty et al., 2010), indicating that age differences between IDH-mut and IDH-wt patients may obfuscate how much variance in this brain correlate of cognition is explained by the mutation itself.
Regardless of IDH-mutation status, cognitive deficits in glioma patients are associated with altered brain functional connectivity.
We performed MEG and cognitive testing in a cohort of de novo glioma patients and investigated whether theta and alpha global brain functional connectivity differed according to IDH-mutation status and whether this measure reflects cognitive functioning in these subpopulations beyond differences in age and presence of epilepsy. We expected lower functional connectivity to relate to poorer cognitive functioning (when correcting for age and presence of epilepsy), and thus functional connectivity to be lower in IDH-wt patients as compared to IDH-mut patients.

| Patients
Patients visiting the VUmc CCA Brain Tumor Center Amsterdam between 2010 and 2017 with suspected diffuse glioma were eligible to participate. Part of this patient cohort has been reported on before (Carbo et al., 2017;van Dellen et al., 2012van Dellen et al., , 2014Derks et al., 2018). Inclusion criteria were (a) age over 17 years and (b) ability to participate in neuropsychological testing. After testing and MEG recording, all patients were diagnosed with World Health Organization (WHO) grade II, III, or IV diffuse glioma according to the WHO 2007 classification (Louis, Ohgaki, Wiestler, & Cavenee, 2007). Patients with previous craniotomies or neurological/psychiatric comorbidities were not able to participate in this study. Information on the presence of epilepsy, use of anti-epileptic drugs, use of dexamethasone, and Karnofsky performance status (KPS) was collected (Karnofsky, Abelmann, & Craver, 1948). Level of education was gathered based on a commonly used Dutch scale for highest obtained educational degree, which ranges from level 1 (not completed primary education) to level 7 (academic degree) (Verhage, 1964). Tumors were manually drawn on 3D anatomical magnetic resonance imaging (MRI) images, slice by slice (LD), using both contrast-enhanced T1-weighted and FLAIR images. Tumor volume was assessed by calculating the volume of the voxels containing tumor. The ethical review board of the VU University Medical Center approved this study, and all patients gave written informed consent before participation.

| IDH-mutation status
IDH-mutation status was assessed with immunohistochemistry on formalin-fixed paraffin-embedded tissue according to IDH-mutation diagnostic routine, identifying 90% of all IDH-mutant diffuse cognitive performance (p < 0.034), corrected for age, education, and presence of epilepsy.

Conclusion:
Global functional connectivity is lower in patients with IDH-wt diffuse glioma compared to patients with IDH-mut diffuse glioma. Moreover, having lower functional alpha connectivity relates to poorer cognitive performance in patients with diffuse glioma, regardless of age, education, and presence of epilepsy.

| Cognitive functioning
Cognitive functioning was extensively measured preoperatively Klein et al., 2002;Klein, Postma, et al., 2003;Taphoorn & Klein, 2004). As measures of cognitive performance, we included the sum score of the five trials, and the delayed recall score of the Rey Auditory Verbal Learning Test (RAVLT) (Rey, 1958) (Stroop, 1935) (attentional functioning), and by the number of words generated during the Categorical Word Fluency test (Luteijn & van der Ploeg, 1983) (executive functioning).

| Magnetoencephalography
Participants underwent MEG recording before neurosurgical intervention, and/or start of any radio-or chemotherapy (as described previously (Carbo et al., 2017;van Dellen et al., 2013van Dellen et al., , 2014). In brief, an eyes-closed resting state recording of 5 min in a magnetically shielded room (Vacuum Schmelze GmbH, Hanua, Germany) with a 306 channel MEG system (Elekta Neuromag Oy, Helsinki, Finland) was acquired. Data were sampled at 1,250 Hz, and a high-pass filter (0.1 Hz) and anti-aliasing filter (410 Hz) were  (Taulu & Hari, 2009;Taulu & Simola, 2006), and then visually inspected for quality. For coregistration of MEG with participants' MRI, the outline of the scalp and four or five head localization coils were digitized using a 3D digitizer (3Space Fastrak, Polhemus, Colchester, VT, USA) and matched to the MRI scalp surface. The coregistered MRI was then spatially normalized to a template MRI, and, using the Automated Anatomical Labeling (AAL) atlas (Tzourio-Mazoyer et al., 2002), the centroid voxels (Hillebrand et al., 2016) in the 78 cortical regions (Gong et al., 2009) were selected for further analyses after inverse transformation to the patient's coregistered MRI. A scalar beamformer implementation (Elekta Neuromag Oy, version 2.1.28) was used to reconstruct broadband (0.5-48 Hz) time series of neuronal activity for these centroids (Hillebrand, Barnes, Bosboom, Berendse, & Stam, 2012). For each patient, 60 consecutive epochs of 3.27 s (4,096 samples) were used to extract theta and alpha time series. These were obtained by digitally filtering the selected epochs using a fast Fourier transform, after which all bins outside the passbands were made zero, and an inverse Fourier transform was performed. Theta and alpha powers were calculated, relative to the power in the broadband signal. The 78 × 78 functional connectivity matrix was then averaged over epochs to yield a single measure of global functional connectivity per subject and per frequency band. A Box-Cox transformation (lambda = −5) (Box & Cox, 1964) was performed on global functional connectivity measures to ensure normality of the data; the transformed data were used in all statistical analyses. Group differences in cognitive functioning were established with (seven) linear regression models for each cognitive test described above, corrected for age, level of education, and presence of epilepsy. A log transformation was performed on the raw test scores of the Concept Shifting Test, the Memory Comparison Test, and of the Stroop Color Word Test to adhere to the assumption of normally distributed standardized residuals in linear regression analyses. The transformed data were used in all following statistical tests.

| Statistical analyses
Group differences in theta and alpha functional connectivity (dependent variable) between IDH subgroups (independent variable) were computed with linear regression models to account for confounding variables (age and presence of epilepsy). In case of significant results, group differences in frequency-specific relative power were investigated with linear regression as well, to ascertain that connectivity differences were not driven by differences in relative power.
In case of significant differences in functional connectivity between the subgroups, associations between functional connectivity and cognitive functioning were assessed. Linear regression models were computed for each cognitive test, with (log transformed) test score as the dependent variable and functional connectivity, age, presence of epilepsy, and level of education as independent variables. In addition, tumor volume was also added to these models, since tumor volume is possibly associated with cognitive functioning, particularly in patients with IDH-wt glioma (Kesler et al., 2017).
Post hoc analyses included testing for the possible confounding effect of dexamethasone use, which may affect cognition (Kostaras, Cusano, Kline, Roa, & Easaw, 2014). First, differences in cognitive performance according to dexamethasone use were tested with seven linear regression models corrected for age, presence of epilepsy, and education. Next, a Student's t test was performed to test for differences in alpha functional connectivity according to dexamethasone use. Furthermore, histological WHO grade may confound functional connectivity changes beyond IDH-mutation status. Therefore, we tested for differences in alpha functional connectivity between grade II/III and grade IV IDH-wt patients with linear regression controlling for age and presence of epilepsy, as this subgroup had enough variation in WHO grade to do statistical testing.
All linear regression analyses met the criteria of normally distributed standardized residuals and homoscedasticity by visual inspection. A p-value lower than 0.05 was considered significant.

| Functional connectivity differs according to IDH status
Patients with IDH-wt glioma had significantly lower alpha functional connectivity compared to patients with IDH-mut glioma while controlling for age and presence of epilepsy (B = 138.209, CI = 6.575-269.842, p = 0.040, Figure 1). We did not control for tumor volume in this analysis because tumor volume was not significantly different between groups. These findings were specific to functional connectivity, as relative alpha power did not differ between groups

| Functional connectivity explains cognitive variance across groups
In the entire cohort of patients, there were significant associations between alpha functional connectivity and cognitive test scores (after false F I G U R E 1 Violin plots showing alpha functional connectivity (before normalization) for IDH-mut and IDH-wt patients, the crosses within the plots indicate mean per group

| D ISCUSS I ON
Patients with IDH-wt glioma have lower functional connectivity in the alpha band compared to patients with IDH-mut glioma, even when controlling for age and presence of epilepsy. Moreover, lower functional connectivity is associated with poorer cognitive performance in the entire cohort.
Previous studies in glioma patients have amply shown whole brain alterations in functional connectivity compared to healthy controls (Derks et al., 2014), which are specifically relevant for cognition in theta van Dellen et al., 2012;Douw et al., 2010) and alpha (Bosma et al., 2009;Carbo et al., 2017;van Dellen et al., 2013) frequencies. The association between alpha band functional connectivity and cognition has been evidenced longitudinally as well (Carbo et al., 2017), showing in- These previous studies mainly reported on connectivity differences in glioma patients diagnosed according to the 2007 WHO classification, based on histopathology only (Derks et al., 2014;Louis et al., 2007). As the neuro-oncology field has moved toward incorporating molecular markers like IDH-mutation F I G U R E 2 Associations between alpha functional connectivity and cognitive performance. Alpha functional connectivity (after normalization) is plotted on the x-axis. The y-axis reflects cognitive performance: RAVLT recall score ( Wefel and colleagues were the first to describe worse cognitive performance in patients with IDH-wt glioma compared to patients with IDH-mut glioma (Wefel et al., 2016). The same group investigated cortical thickness covariation as a neural correlate of this difference, showing that thickness covariation patterns indeed differ between subgroups and associate with cognition (Kesler et al., 2017).
This provides a first insight into possible mechanisms underlying IDH-mutation-related cognitive differences and also suggests that tumor growth rate might contribute to cognitive problems (Klein, 2016;Wefel et al., 2016). However, the observed group differences in covarying cortical thickness were not corrected for age or presence of epilepsy. Since the cortex thins over time, regardless of the presence of glioma, as cognition also deteriorates, the reported results may have partly been due to normal aging (Thambisetty et al., 2010). Our results show alpha functional connectivity group differences irrespective of age and presence of epilepsy. Associations between alpha functional connectivity and cognitive functioning remained significant after controlling for these confounders. Our findings suggest that lower functional connectivity has a particular contribution to cognitive deterioration in IDH-wt patients.
The fact that functional connectivity differed between IDH subgroups may reflect the impact of variable tumor growth rate on global connectivity (Klein, 2016;Wefel et al., 2016), while molecular mechanisms related to functional connectivity may also be at play. The overexpression of D-2-hydroxyglutarate in IDH-mut glioma is of particular interest, as a preclinical study reported that this protein activates NMDA receptors specifically, thereby mimicking glutamatergic neuronal activation (Chen et al., 2017). The related increase in neuronal spiking could be the mechanism responsible for the higher incidence of epilepsy in patients with IDH-mut glioma compared to patients with IDH-wt glioma (Chen et al., 2017). Speculatively, the mimicry of glutamatergic neuronal signalling by D-2-hydroxyglutarate might also underlie alpha functional connectivity differences found in the current study, as glutamate plays an important role in the synchronization of neuronal oscillations (Angulo, Kozlov, Charpak, & Audinat, 2004;Fellin et al., 2004).
A limitation of this study is that we only tested gliomas for the R132 variant of IDH1, detecting approximately 90% of the IDH-mutated gliomas (Ichimura et al., 2015). Hence, a few patients classified as having IDH-wt glioma might actually have IDH-mut glioma, meaning that we may have underestimated the group difference in global functional connectivity.

| CON CLUS ION
Alpha functional connectivity is lower in patients with IDH-wt glioma as compared to patients with IDH-mut glioma, regardless of age and presence of epilepsy. Moreover, alpha functional connectivity is positively associated with cognitive performance, irrespective of TA B L E 3 Linear regression models for alpha functional connectivity and cognition Note. Regression models were corrected for age, education and presence of epilepsy. B-values and 95% confidence intervals are displayed for alpha functional connectivity. CI, 95% confidence interval. *p < 0.05, **p < 0.01, a significant after false discovery rate correction.
IDH status. These findings contribute to our understanding of cognitive functioning in patients with diffuse glioma in general, and cognitive deficits in patients with IDH-wt glioma specifically.

ACK N OWLED G M ENTS
The authors would like to thank Nico Akemann, Ndedi Sijsma, Karin Plugge, Marieke Alting Siberg, Marlous van den Hoek and Peter-Jan Ris for the MEG acquisitions.

CO N FLI C T O F I NTE R E S T
None declared.