Access to this full-text is provided by Wiley.
Content available from CNS Neuroscience & Therapeutics
This content is subject to copyright. Terms and conditions apply.
CNS Neurosci Ther. 2023;29:1109–1119.
|
1109wileyonlinelibrary.com/journal/cns
Received: 21 July 202 2
|
Revised: 7 December 2022
|
Accepted: 26 December 2022
DOI: 10.1111/cns.14087
ORIGINAL ARTICLE
Shared functional network abnormality in patients with
temporal lobe epilepsy and their siblings
Kangrun Wang1,2,3,4 | Fangfang Xie5 | Chaorong Liu1 | Ge Wang1 | Min Zhang1 |
Jialinzi He1 | Langzi Tan1 | Haiyun Tang5 | Fenghua Chen2 | Bo Xiao1,3 |
Yanmin Song6 | Lili Long1,3,4
This is an op en access arti cle under the ter ms of the Creative Commons Attribution L icense, which pe rmits use, dis tribu tion and reprod uction in any med ium,
provide d the original wor k is properly cited.
© 2023 The Authors . CNS Neuroscience & Therapeutics published by John Wiley & Sons Ltd.
The fir st two author s contributed e qually to this wor k.
1Depar tment of Neurology, Xiangya
Hospit al, Central South Universit y,
Changsha, China
2Depar tment of Neurosurger y, Xiangy a
Hospit al, Central South Universit y,
Changsha, China
3Clinical Resea rch Center for Epil eptic
disease of Hunan Province, X iangya
Hospit al, Central South Universit y,
Changsha, China
4Nationa l Clinic al Research Center for
Geriat ric Disorders , Xiang ya Hospital,
Centra l South U niversity, Changsha, China
5Depar tment of Radiol ogy, Xiangya
Hospit al, Central South Universit y,
Changsha, China
6Depar tment of Emergency, Xiang ya
Hospit al, Central South Universit y,
Changsha, China
Correspondence
Lili Long and Yanmin So ng, Xiangya
Hospit al, Central South Universit y, 87
Xiang ya Road, C hangsha, Hunan 41000 0,
China.
Email: longlili1982@126.com and
csuxysym@csu.edu.cn
Funding information
Key Research and Development Progr am
of Hunan Province of China, Grant/Award
Number : 2022S K20 42; Nation al Natur al
Science Foundation of Chin a, Grant/
Award Number: 82171454; The N ational
Multidisciplinary Cooperative Diagnosis
and Treatment Capacity Project for Major
Diseases of Xiangya Hos pital , Central
South Un iversi ty, Grant/Award Number:
z027001
Abstract
Aim: Temporal lobe epilepsy is a neurological network disease in which genetics
played a greater role than previously appreciated. This study aimed to explore shared
functional network abnormalities in patients with sporadic temporal lobe epilepsy
and their unaffected siblings.
Methods: Fifty- eight patients with sporadic temporal lobe epilepsy, 13 unaffected
siblings, and 30 healthy controls participated in this cross- sectional study. We exam-
ined the task- based whole- brain functional network topology and the effective func-
tional connectivity between networks identified by group- independent component
analysis.
Results: We observed increased global efficiency, decreased clustering coefficiency,
and decreased small- worldness in patients and siblings (p < 0.05, false discovery rate-
corrected). The effective network connectivity from the ventral attention network
to the limbic system was impaired (p < 0.001, false discovery rate- corrected). These
features had higher prevalence in unaffected siblings than in normal population and
was not correlated with disease burden. In addition, topological abnormalities had a
high intraclass correlation between patients and their siblings.
Conclusion: Patients with temporal lobe epilepsy and their unaffected siblings showed
shared topological functional disturbance and the effective functional network con-
nectivity impairment. These abnormalities may contribute to the pathogenesis that
pr omotes the suscepti bili t y of se izur es and la ngu a ge de clin e in tem poral lobe ep ilep s y.
KEYWORDS
endophenotype, func tional magnetic resonance imaging, graph theory, independent
component analysis, temporal lobe epilepsy
1110
|
WA NG e t al.
1 | INTRODUC TION
Temporal lobe epilepsy (TLE) is the most common drug- resistant
focal epilepsy syndrome in adults,1 pathologically characterized by
hippocampal sclerosis (HS) in two- thirds of the cases.2 Previous
MRI studies proposed the idea that TLE was a network disorder.
Specifically, TLE was related to atrophy of neocortex,3– 5 cerebellar,6
hippocampus5 and thalamus,7,8 and derangements of structural7, 9
and functional connectivity8,10 across the whole brain. It remains
unresolved whether the network disruptions are consequences or
predisposing risk factors for continued seizures.
Endophenotypes are heritable, measurable characteristics that
link genotype to phenotype.5 ,11 Studying endophenotypes help
deconstruct complex clinical symptoms and increase the analytical
power of genetic mapping for polygenetic disorders such as sporadic
TLE. Since endophenotypes have a higher prevalence in unaffected
family members of patients than in the normal population, study-
ing unaffected siblings of patients is a popular approach to explore
endophenotypes.5,9,12,13 In our previous study, we observed shared
hippocampal distortion in patients with TLE and their unaffected
siblings.5 Other studies also identified white matter and thalamic at-
rophy,14 temporal morphological alteration,15,16 and decreased mean
diffusivity in left superior longitudinal fasciculi and corticospinal
tract9 in the affec ted/unaffected sibling pairs of TLE. However, how
do these structural endophenotypes lead to functional malfunction,
including seizure propagation and language decline, is unresolved.17
Functional endophenotypes are yet to be explored to link structural
endophenotypes to clinical symptoms.18 ,19
Graph theor y provided insights into the topological disturbance
in TLE and mathematical tools to quantitatively characterize com-
prehensive cerebral networks. Functional brain network topology
modulated the threshold for seizure propagation,20 surgical out-
comes,6,21,22 and cognitive function23 in patients with TLE. Previous
graph theoretical works on cortical thickness,6,22 white matter
tracts,2 2,24 and functional MRI21,23,25,26 advanced the knowledge
of TLE as a network disorder, showing a less small- world topology
in TLE. While graph theory unfolds the functional net work at the
whole- brain level, group- independent component analysis (group-
ICA) separates the whole brain into functional segregated net-
works.27, 28 Then, effective functional network connectivity (eFNC)
could probe into directed connectivity from one network to another
network af ter discounting the effect by other networks. Abnormal
connections between limbic system, default mode network (DMN),
and attention networks contributed to the comorbidity of cognitive
decline.10,29
While resting- state functional MRI focuses primarily on the
spontaneous activities in rest, task- based functional MRI (tb- fMRI)
provides direct information for language impairment.30 Notabl y, ver-
bal fluency tasks mobilize a network roughly overlapping with the
hippocampal epileptogenic network, providing an ideal tool to study
the underlying malfunctional network and explore potential func-
tional endophenotypes in TLE.8,31
With a well- tested Chinese character verbal fluency task,31 ,32
this study aimed to explore the shared task- related functional net-
work disorders in patients with TLE and their unaffected siblings.
We applied graph theoretical approaches to investigate the topol-
ogy of the whole- brain functional network. We analy zed the effec-
tive connectivity between networks identified by group- ICA. At
last, we used three approaches to test their potential as functional
endophenotypes.
2 | MATERIAL AND METHODS
2.1 | Subjects
From December 2018 to January 2021, we sequentially re-
cruited 58 patients with sporadic TLE from the Department of
Neurology, Xiangya Hospital. Patients visited our institution due
to recent seizures. We identified 30 patients with HS (TLE- HS, left
TLE (LTLE):right TLE (RTLE) = 13:17) and 28 patients without HS
(TLE- NHS, LTLE:RTLE = 12:16). The clinical semiology, electroen-
cephalography, and MRI evidence were evaluated by two experi-
enced epileptologists (BX and LL) to diagnose and lateralize TLE.33
Thirteen5,12 unaffected siblings of patients participated in this
study. Thirty age, sex, and educational level matched healthy con-
trols (HC) were recruited from the same social background. For all
subjects, the exclusive criteria included (1) brain lesions other than
HS; (2) history of neurological or psychiatric disease except for TLE;
(3) under 16 or over 65 years of age; (4) unable to comprehend our
language paradigm. All subjects are right- handed native Chinese
speakers.
All subjects routinely underwent T2WI, T2WI fluid- attenuated
inversion recovery, and 3DT1 sequences, which were visually
assessed by experienced neuroimagers to diagnose HS or other
potential structural abnormalities. The hippocampi of subjects
were segmented by Hipposeg, an automatic online hippocampal
segm ent ation algorithm developed for TLE, to calculate the hippo-
campal volume (http://nifty web.cs.ucl.ac.uk/progr am.php?p=HIP-
POSEG),34 which was then corrected for total intracranial volume.
The 95% confidence inter val (CI) of HC was calculated as the ref-
erence volume. Hippocampal sclerosis was defined as (1) visually
decreased hippocampal volume, asymmetrical hippocampus, or
loss of internal structure on T1- weighted MRI; (2) increased signal
in hippocampi on T2- weighted MRI; and (3) smaller than the ref-
erence hippocampus volume. Criteria one and two were carried
out by two neuroimager specialized in TLE (FF and HY). According
to a well- established protocol,35 disagreement between the neu-
roimagers and automatic segmentation was reconciled by a blind
rater (LL) who has rich experience in both visual assessment and
automatic segmentation.5
Written informed consent was obtained from all participants. The
study was approved by the ethics committee of Xiangya Hospital of
Central South University.
|
1111
WANG et al.
2.2 | Neuropsychological tests
Participants routinely underwent neuropsychological tests,
including the Self- Rating Anxiety Scale (SAS),36 Self- Rating
Depression Scale (SDS),37 Boston Naming Test (BN), Montreal
Cognitive Assessment (MoCA),38 and verbal fluency Pinyin test
(VFP).32 Befo re t he t b- fMRI sca n, subjec ts were given a verbal flu-
ency Char ac ter tes t (VFC), 32 same as our lan guage task, to confirm
that they understand the procedure and estimate their perfor-
mance during the scan.
2.3 | MR data acquisition
With a Siemens MAGNETOM Prisma 3.0T MR scanner and stand-
ard head coils at the MRI center of Xiangya Hospital, MRI images
were collected, including the whole- brain structural images ob-
tained using a magnetization- prepared rapid acquisition with
gradient echo sequence (field of view 233 mm, repetition time
[TR] = 2.11 s, echo time [TE] = 3.18 ms, flip angle = 9°, 320 × 320
matrix), and the whole- brain blood oxygenation level- dependent
(B OLD ) sig nal s prov ide d by a grad ien t echo pl ana r T2- we igh ted se -
quence (field of view 225 mm, TR = 1 s, TE = 37 ms, flip angle = 52°,
90 × 90 matrix).
A previously described and well- tested Chinese character ver-
bal fluency task31,32 was carried out during the fMRI scan. Subjects
needed to silently come up with words that initiate with the single
Chinese character exhibited on the screen.
2.4 | fMRI image preprocessing
Imaging data were preprocessed with the Statistical Parametric
Mapping 12 software (https://www.fil.ion.ucl.ac.uk/spm/). The pre-
processing pipeline includes realignment, coregistration, segmen-
tation, normalization to the Montreal Neurologic al Institute (MNI)
space, and spatial smoothing (6 mm).31
2.5 | Whole- brain functional connectivity and
graph theory analysis
We used toolbox CONN v.20.b39 (http://www.nitrc.org/proje cts/
conn) to conduct two different types of functional connectivity
analysis with our tb- fMRI data. Task- based functional connectivit y
(t b- FC ) anal yze d the fun c t i ona l conne c t i ons du r ing ta s k s. By re g ress -
ing out task- specific signals, the task residual- functional connectiv-
ity (tr- FC) approach extracted resting- state functional connectivity
from tb- fMRI data.31
The image data of ever y subject were parcellated into 272 re-
gions of interest (ROI), including 210 cortical region, 36 subcor tical
regions,40 and 26 cer eb el la r regions.41 Outlying sca ns were detected
by the func tional outlier detection tool embedded in the CONN.
Head motion, outlying scans, the effect of modules, and BOLD sig-
nals inside white matter and cerebral spinal fluid were defined as
confounders and were regressed out.39 BOLD signals were filtered
with a band- pass filter [0.009– 0.08] Hz to remove task- evoked sig-
nals in the tr- FC analysis.31 A wider [0.009– 0.10] Hz band- pass filter
was applied for tb- FC analysis.42
A generalized linear model (GLM) with bivariate- correlation39
was used for the individual- level analysis. The Pearson correlation
coefficients were calculated for every pair of ROIs to generate a
272 × 272 weighted matrix for each subject. Weighted matrices were
processed into binarized undirected matrices with connection den-
sity ranging from 5% to 40% , in steps of 1%.8 For each connection
density, we explored four global metrics: (1) global efficiency (GE);
(2) average shortest path leng th (APL); (3) clustering coefficiency
(CC); (4) small- worldness (σ) = (CC/CCrand)/(APL/APLrand). Both GE
and APL reflect the integration level of the network, and the CC
measures the segregation level of the network. The property of
healthy brains accords with a small- world organization, character-
ized by relatively higher CC and lower APL than that of a random
network. CCrand and APLrand are the CC and APL of equivalent ran-
dom networks. The theoretical values are:
To investigate the between- group differences across different
connection densities, we computed the area under curve (AUC),
which provided an overall scalar value integrating differences under
all densities.
2.6 | Group- ICA and post hoc eFNC
The toolbox CONN v.20.b was used to conduct group- ICA and post
hoc eFNC.27, 39 According to previous studies, the whole brain was
divided into 20 independent components (ICs) to identify the DMN
and task- related networks.27, 2 8, 4 3 IC s were la bel ed by th e dic e coef fi-
cient of sp atial overl ap and visual ass essment.28,4 3 Then, group- level
ICs were reconstructed into individual- level ICs.27 The time courses
in individual- level ICs were extracted for eFNC analysis. The de-
noising and filtering steps were the same as tb- FC described above.
GLM with semipartial- correlation was used to construct individual-
level eFNC matrices.39 Effective FNC was compared bet ween three
groups by analysis of covariance (ANCOVA) and post hoc pairwise
comparisons, with age, sex, educational level, and MoCA controlled
as covariates of no interest. Connec tivit y was considered significant
under a threshold of false discovery rate (FDR)- corrected, p < 0.05.
2.7 | Statistical analysis
We used the IBM SPSS Statistics 23 (https://www.ibm.com/produ
c t s / s p s s - s t a t i s t i c s ) for statistical analysis. Kolmogorov– Smirnov
CC
rand =
average degrees of nodes
number of nodes
; APLrand =
ln(number of nodes)
ln(average degrees of nodes).
1112
|
WA NG e t al.
test was used to assess data distribution. Data without a normal
distribution or homogeneity of variance would be compared be-
tween groups with nonparametric approach. When comparing
qualitative variables, we used Kruskal– Wallis H- test, ANCOVA,
Quade nonparametric ANCOVA and post hoc pairwise compari-
sons, or Mann– Whitney U test when applicable. Considering the
difference in sample size, we additionally reported the effect
sizes.44 To compare categorical variables between groups, Chi-
square test or Fisher exact test was applied. A ge, sex, educational
level, and MoC A were controlled as covariates of no interest
when comparing neuropsychological scores and graph theoretical
metrics.
FDR procedures were applied for the multiple comparisons. A
p < 0.05 was considered statistically significant.
2.8 | Sensitivity analysis
2.8.1 | Endophenotypes verification
We applied three widely used methods to verify the endopheno-
types: receiver operating characteristic (ROC) curve analysis,12 ,17
correlation analysis,13 and intraclass correlation coefficient (ICC)
analysis.13,14
We employed the functional net work parameters in the logis-
tic regression and ROC cur ve analysis to discriminate siblings from
healthy controls. High discriminative accuracy indicates that siblings
have higher prevalence of functional network abnormalities than the
normal population.
We investigated the relation between endophenotype can-
didates and language test scores. To assess the impact of disease
severity and psychiatric comorbidity, we additionally evaluated the
correlation of endophenotype candidates with SAS, SDS, age of
onset (AOO), disease duration, seizure frequency, and the number of
antiseizure medications (ASMs) in patients.
To assess the heritability of potential endophenotypes, we cal-
culated the ICC of graph theory metrics and eFNC between patient-
sibling pairs.13,14 One correspon ding patient pr esented bilatera l TLE,
one failed the fMRI scan due to claustrophobia, and one had poor
imaging quality. Hence, there were 10 patient- sibling pairs in our
cohort.
2.8.2 | Subgroup analysis
Patients were divided into two groups according to HS and NHS.
There were seven unaffected siblings of patients with TLE- HS
(Sib- HS), and six unaffected siblings of patients with TLE- NHS (Sib-
NHS). We explored the difference between HC, two patient groups,
and two sibling groups.
Besides, differences within LTLE and RTLE subgroups were ex-
plored to address the laterality effect.
3 | RESULTS
3.1 | Demographic and clinical data
No significant group difference of age, sex, or educational level was
noted. But HC performed better in MoCA than patients (p = 0.008,
FDR- corrected; F = 9.55 , η2 = 0.089) and siblings (p = 0.03, FDR-
corrected; F = 14.4 4, η2 = 0.128). Patients had higher anxiety level
than HC (p = 0.0 02, FDR- cor re cted; F = 8.76 , η2 = 0.0 85) and sib li ng s
(p = 0. 03, FDR- co rre cte d; F = 7. 24 , η2 = 0.071) , and hig h er de pre ssion
level than HC (p = 0.006, FDR- corrected; F = 11 .17, η2 = 0.106). The
HC group outperformed patients in VFP (p = 0.002, FDR- corrected;
F = 15.62, η2 = 0.142) and BN (p = 0.02, FDR- corrected; F = 8.18,
η2 = 0.080). Details of demographic and clinical data were listed in
Table 1.
3.2 | Task residual- functional
connectivity topology
On the overall scale (Figure 1), patients presented significantly
higher GE (p = 0.01, FDR- corrected; F = 9. 26; η2 = 0.090), lower
CC (p < 0.0 01, FDR- corrected; F = 15.09; η2 = 0.138), and lower σ
(p = 0.0 06 , FDR- cor re c te d; F = 10.26; η2 = 0.0 98). Th ou gh th e dif fer-
ence was not statistically significant, siblings showed the same but
milder abnormality as patients.
The group difference of functional connectivity topology under
specific densities was presented in Figure S1.
3.3 | Task- based functional connectivity topology
On the overall scale (Figure 2), patient s had higher GE (p = 0.0 05,
FDR- corrected; F = 10.85; η2 = 0.103), lower CC (p < 0.001, FDR-
corrected; F = 14.30; η2 = 0.132), and lower σ (p = 0.002, FDR-
corrected; F = 12.64; η2 = 0.119) in comparison with HC. Compared
to HC, siblings also had higher GE (p = 0.0 4, FDR- corrected;
F = 5.23; η2 = 0.053) and lower CC (p = 0.05, uncorrected; F = 4.33;
η2 = 0.044).
The group difference of functional connectivity topology under
specific densities was presented in Figure S2.
3.4 | Group- ICA and post hoc eFNC
The group- ICA successfully separated the DMN, limbic system,
ven tr al at te nt io n net wo rk (VA N), an d other task- r el ated net works.
The detailed labeling and presenting of all 20 ICs are listed in
Figure S3.
The effective connectivity from the IC 13 (VAN) to the IC 17
(limbic system) was significantly different across the three groups
(p < 0.05, FDR- corrected; F = 14. 01; η2 = 0.230; Figure 3A,B). Post
|
1113
WANG et al.
hoc pairwise comparison revealed reduced eFNC in the siblings
(p < 0.001, FDR- corrected; F = 16.62; η2 = 0.150) and patients
(p < 0.001, FDR- corrected; F = 23.31; η2 = 0.199) compared to HC
(Figure 3C).
3.5 | Sensitivity analysis
3.5.1 | Endophenotypes verification
Both graph theoretical metrics (AUC, 0.767; 95% CI, 0.595– 0.938)
and the eFNC from the VAN to the limbic system (AUC, 0.772; 95%
CI, 0.627– 0.917) reached high accuracy when discriminating siblings
from HC. The combination of two methods had an AUC = 0.813,
95% CI = 0.672– 0.954 (Figure 4).
Graph theory metrics or eFNC were not correlated with disease
burden (AOO, disease duration, and seizure frequency), or number
of ASMs (p > 0.05, uncorrec ted). Stronger connectivity between the
VAN and the limbic system was correlated with higher VFP scores
(p = 0.02, uncorrected; r = 0.32).
GE and CC in tr- FC and tb- FC, and σ in tb- FC were highly cor-
related bet ween patients with TLE and their unaffected siblings
(Table 2).
3.5.2 | HS and NHS subgroups
Patients with HS and their siblings had relatively worse language
performance and more severe topological disturbance than patients
without HS and their siblings. The eFNC from the VAN to the limbic
TAB LE 1 Demographic and clinical data.
HC Sibling TLE statistic η2 or Φ
p
value
N30 13 58 – – –
Age, years, median (IQR) 26.0 (18.0) 31.0 (7.0) 29.0 (10.0) 0.87a0.012 0.65
Sex, Male/Female 14/16 3/10 29/29 3.13b0 .176 0.21
Education, median (IQR) 12.0 (7.0) 9.0 (5.0) 12.0 (7.0) 1.95a0.001 0.38
MoCA , median (IQR) 29.0 (3.0) 25.0 (5.0) 26.0 (5.0) 11.0 5a0.092 0.009*
VFC, median (IQR) 24.5 (16.0) 18.0 (12.0) 18.0 (9.0) 3.31c0.066 0.04
VFP, mean (SD) 34. 2 (17.6) 23.9 (9.6) 24.4 (13.1) 7.82 d0.143 0.004*
BN, median (IQR) 27.0 (5.0) 24.0 (7.0) 25.0 (5.0) 4.09c0.080 0.04*
SAS, median (IQR) 38.0 (8.0) 40.0 (10.0) 45.5 (14.0) 8.82c0.158 0.003*
SDS, mean (SD) 39.4 (10.3) 42.3 (10.5) 4 8.4 (9.8) 6.04d0.114 0.009*
duration, years, median (IQR) 10.0 (16.0)
AOO, years, median (IQR) 17.0 (10.0)
Laterality, Left/Right 25/33
Febrile convulsion histor y 3 (5.2%)
SGS history 46 (79.3%)
Number of A SMs
136
221
3 1
Seizure frequency
Every year 15
Every month 21
Every week 12
Every day 10
Abbreviations: AOO, age of onset; ASMs, antiseizure medications; BN, Boston Naming Test; HC, healt hy controls; IQR, interquar tile range; MoCA ,
Montreal Cognitive Assessment; SAS, Self- Rating Anxiety Sc ale; SD, standard deviation; SDS, Self- Rating Depression Scale; SGS, secondary
generalized seizures; TLE , patients with temporal lobe epilepsy; VFC, verbal fluency charac ter test; VFP, verbal fluency Pinyin test.
aH value of Kruskal– Wallis H- test.
bχ2 value of Chi- squared test .
cF value of Quade nonparametric analysis of covariates.
dF value of analysis of covariates.
*p values are FDR corrected across nine comparisons. The uncorrected p = 0.00 4, <0.001, 0.02, <0.001, and 0.004, in order.
1114
|
WA NG e t al.
system was disrupted at a similar level in patients with and without
HS (see details in Material S1, Table S1, Figure S 4 – S 6 ).
3.5.3 | LTLE and RTLE subgroups
The results roughly replicated that in the whole group analyses.
Meanwhile, LTLE- NHS also presented milder but impaired tr- FC
topological features compared to HC (see details in Material S1,
Figure S 7 – S 9 ).
3.5.4 | SAS and SDS as covariates of no interest
Since the limbic system is related to psychiatric comorbidity, we ex-
amined the group difference of eFNC after controlling for SAS and
SDS scores. HC presented stronger functional connectivity com-
pared to patients ( p < 0.001, FDR- corrected; F = 14 .6 3; η2 = 0.186)
and siblings (p < 0.001, FDR- corrected; F = 16.15; η2 = 0.201). In
addition, the eFNC strength was not correlated with SAS or SDS in
the patients (p > 0.05, uncorrected).
3.5.5 | Effect of intellectual disability
Since patients and siblings presented lower MoCA scores compared
to HC , we calculated the partial correlation bet ween MoCA scores
and functional statistics. The MoCA score was not correlated with
functional net work connectivity and graph statistics (p > 0.8).
4 | DISCUSSION
With a Chinese character version of verbal fluency task, we ob-
served shared func tional network abnormalities in patients with
TLE and their unaffected siblings. The topological disturbance and
network- level impairment, especially the eFNC from the VAN to the
limbic system, could be considered functional endophenotypes that
FIGURE 1 Topological parameters
during the rest. The AUC comparison of
three groups for APL , CC, GE, and small-
worldness throughout all densities. APL,
average shortest path leng th; AUC, area
under cur ve; CC, clustering coefficiency;
GE, global efficiency; HC, healthy
controls; TLE, temporal lobe epilepsy.
|
1115
WANG et al.
might contribute to the susceptibilit y of seizures and cognitive de-
cline in TLE.
Verbal fluency decline is a common deficit in TLE. A Chinese
character verbal fluency task mobilizes orthographic, tonal, pho-
nological, and semantic processing procedures.32 The Chinese
character verbal fluency task triggered widespread activation and
deactivation ef fect s, including the deactivation of the DMN and the
activation of task- related networks, ensuring an ideal tool for evalu-
ating language decline and net work disruption.31
By exploring the graph theory metrics, we observed a less small-
world and more random topology in patients with TLE and their
unaffected siblings. A small- world network balances information
local processing and global transferring, with economical energy
cost.45 This property was impaired, indicating reduced brain effi-
ciency in patients and their unaffected siblings.2 3 – 2 6 We also found
shared lower CC in patient s and their unaffected siblings, repre-
senting a reduced local information processing ability.20,23,26 In TLE,
disconnection within local net works such as the lef t temporal lobe
language network46 and DMN47 impeded local information process-
ing and resulted in language decline. While the ictal network bears a
more regular topology,48,49 the interictal functional network would
be a more random network.20 Our study supp orted this observation
as patients and their unaffected siblings showed lower CC and higher
GE. Neuronal synchronization is an important mechanism underlies
seizure generation and propagation.50 – 52 A random functional net-
work has lower threshold for synchronization.49,5 3 In the ma the mat ic
hippocampal model established by Netoff et al., with the increasing
of random connections, the network had lower threshold for syn-
chronization, and the neural activit y shifted to seizure- like activi-
ties.54 These indicated a predisposed susceptibility of seizures in
patients with TLE and their unaffected siblings.
In addition, we explored the eFNC bet ween networks identi-
fied by group- ICA . In patients and their unaffected siblings, we ob-
served reduced connectivity from the VAN to limbic system. The
FIGURE 2 Topological parameters
during the task. The AUC comparison of
three groups for APL , CC, GE, and small-
worldness throughout all densities. APL,
average shortest path leng th; AUC, area
under cur ve; CC, clustering coefficiency;
GE, global efficiency; HC, healthy
controls; TLE, temporal lobe epilepsy.
1116
|
WA NG e t al.
VAN ser ved to reg ul at e on e' s at tenti on when une xpe cted st imuli ap-
pear ed during a t ask . 55 Al o n g wi t h hi p p o c a m pi, the limb ic system wa s
at the center of seizure generation. Previous studies had observed
macrostructural56,57 and microstructural58 damage of the lim bic sys-
tem in TLE. Girardi- Schappo et. la. observed slower information flow
in and out the limbic system in TLE, which was partially controlled
by hippocampal volume.10 Another independent study also reported
disrupted hierarchical organization originated from the limbic sys-
tem . The severit y of org an iz at io n alte ration was correlated with cog-
nitive performance.29 We observed a positive correlation between
the language test performance and the connectivit y strength. Our
findings suggested that the disconnection between the VAN and the
limbic system during tasks could had contributed to the language
impairment. This indicated that, at least in some cases, the cognitive
decline in patients with TLE and their unaffected siblings59 was ge-
netically predisposed.
The genetic back ground of seizure and cognitive decline in
TLE is a topic of ongoing debate. While structural abnormalities
con nect ed direc tl y to the genot ype t hat influen ce cor te x de velop-
ment, functional disturbances lie closer to seizure generation and
spreading, representing an essential step from genot ype to phe-
notype.12 ,17 A recent resting- state electroencephalography- fMRI
study noted impaired ability to suppress sensorimotor activities
FIGURE 3 The results of eFNC analysis. (A) ANCOVA matrix of eFNC analysis. the eFNC from the IC 13 to the IC 17 (red squared) was
different across three groups (p < 0.05, FDR- corrected); (B) brain regions of IC 13 and IC 17; (C) group comparison of the eFNC from the IC
13 to the IC 17. ANCOVA, analysis of covariates; eFNC, effective functional network connectivity; HC, healthy controls; IC, independent
component; TLE, temporal lobe epilepsy.
FIGURE 4 Siblings discriminating using functional abnormalities.
eFNC, ef fective functional network connectivity; IC, independent
component.
TAB LE 2 Intraclass correlation coefficients.
Parameter ICC
95% CI
Lower
95% CI
Upper
p
value
A P L _ t r - F C 0.45 −0.21 0.83 0.09
CC_tr- FC 0.77 0.32 0 .94 0.009*
G E _ t r - F C 0.66 0.09 0.90 0.03*
σ_ t r - F C 0.13 − 0. 51 0.68 0.36
A P L _ t b - F C −0.56 −0. 87 0.06 0.96
CC_tb- FC 0.59 −0.02 0.88 0.03
G E _ t b - F C 0.77 0.31 0.94 0.009*
σ_ t b - F C 0.87 0.57 0.97 0.003*
eFNC 0.26 −0.40 0.75 0.22
Abbreviations: APL, average shortest path length; CC, clustering
coefficiency; CI, confident interval; eFNC , effective functional
network connectivity; GE, global efficiency; ICC, intraclass correlation
coefficient; tb- FC, task- based func tional connectivit y; tr- FC, t ask
residual- functional connectivity; σ, small- worldness.
*p values were FDR- corrected across all parameters. The uncor rected
p = 0.003, 0.02, 0.003, <0.001, for CC_tr- FC, GE_tr- FC, GE_tb- FC, and
σ_tb- FC separately.
|
1117
WANG et al.
in patients with TLE and their unaffecte d relatives.18 We reported
t wo no vel fu n c tional fe a t u r e s sh a r e d by p at i e n t s wi th T L E an d th eir
unaffected siblings. These features may not be the consequence
of seizures but were likely to be predisposed traits determined by
shared genetic and (or) environmental factors. To explore their po-
tential as functional endophenotypes, we conducted three analy-
ses. The ROC cur ve a nalyses demonstrated a higher prevalence of
functional net work impairment in unaf fected siblings of patients
than in the normal population. ICC analyses showed high consis-
tency between patients and their siblings, fur ther suggesting the
genetic background of functional network abnormalities. Also,
topological parameters and eFNC were not related to disease bur-
den. These results suggested that the graph theoretical metrics
and eFNC conformed to the definitions of endophenotypes,60 and
were connected to TLE at the population level. Our findings indi-
cated that the disruption of functional connectivity in TLE were
prior to and might promote the onset of se izure s an d cognitive im-
pairment under certain environmental factors. Unaffected siblings
did not develop epilepsy despite bearing a genetic susceptibility,
potentially indicating that additional environmental factors are
needed to cause seizures. Alternatively, milder topological abnor-
malities in siblings may not be sufficient to induce epilepsy. We
also provided candidates for functional endophenotypes, which
can be used as quantitative traits in future genetic searching, im-
proving the success and power of such studies.
We conducted sensitivity analyses in subgroups to explore the
effect of HS and lesion lateralit y. Patients with HS and their siblings
presented more severe topological disturbance compared to patients
without HS and their siblings . This obs er vat ion not only further sup-
ported the heritability of network topology but also suggested that
HS was related to whole- brain network topology. A possible expla-
nation is that HS is a protruding phenotype of genetic fac tors that
affect whole- brain development.14 Alternatively, genetic factors
may only lead to the maldevelopment of the hippocampus,5 and the
whole- brain functional alteration is a downstream event of HS. Did
not differ between patients with and without HS, the eFNC from the
VAN to the limbic system may be a universal feature independent
to hippocampal pathologies in TLE. Meanwhile, patients with LTLE
and patients with RTLE had similar functional network patterns, in-
dicating that lesion laterality was not a confounder. Also, patients
had relatively lower MoCA score compared to HC. Hence, we de-
fined MoCA scores as a covariate of no interest in group comparison
and correlation analysis. Also, MoCA scores was not correlated with
function statistics.
There are limitations. First, our cohort of siblings was rela-
tively small. Hence, we collected clinical details and applied well-
established and robust methods. Even though the power of analyses
for unaffected siblings was lowered by the small sample size, we still
observed significant differences between siblings and healthy con-
trols, suggesting our results are reliable. Fur ther sensitivity analyses
also verified that the functional network abnormalities are promising
endophenotypes. Still, the results may not be representative, and
a larger cohort of unaf fected siblings would benefit future studies.
Second, some patients had HS despite normal hippocampal volume
and MRI signal.61 Though we designed a thorough procedure to
distinguish HS, missed diagnosis in MRI- negative patients shall not
be ignored. However, our cohort would represent patients with ap-
parent HS, who might have more profound genetic backgrounds for
HS and TLE. Third, potential admission rate bias should be admitted
since our study was based on subjects from Xiangya Hospital. Before
generalizing our results, they should be tested in external data.
In conclusion, we observed shared whole- brain functional net-
work topological disruption and impaired eFNC from the VAN to
the limbic system in patients with TLE and their unaffected siblings.
The functional disturbance could be considered endophenotypes
that precede and even promote seizure susceptibility and cognitive
decline in TLE. These imaging traits could be applied as predictors
for seizure onset and language decline in future prospective studies.
Our findings would also benefit future genetic searching aimed for
common variants in TLE.
AUTHOR CONTRIBUTIONS
Kangrun Wang designed and conceptualized study, played a role in
the acquisition of data, analyzed the data and functional connec-
tivity, and drafted the manuscript. Fangfang Xie designed and con-
ceptualized study, analyzed the data, major role in the acquisition of
data, and drafted the manuscript. Charong Liu, Ge Wang, and Min
Zhang analyzed the data and played a role in the acquisition of data.
Jialinzi He and Langzi Tan set the parameters of the task and played
a role in the acquisition of data. Haiyun Tang played a role in the
acquisition of data. Fenghua Chen and Bo Xiao designed and con-
ceptualized study. Yanmin Song supervised the study, played a role
in the acquisition of data, and revised the final manuscript. Lili Long
designed and conceptualized the study, supervised the study, and
revised the final manuscript.
ACKNOWLEDGMENTS
This study was supported by The National Natural Science
Foundation of China (82171454), The Key Research and
Development Program of Hunan Province (2022SK2042), and The
National Multidisciplinary Cooperative Diagnosis and Treatment
Capacity Project for Major Diseases of Xiang ya Hospital, Central
South University (z027001).
CONFLICT OF INTEREST
The authors declare that they have no conflict of interest.
DATA AVAIL AB I LI T Y STATE MEN T
The data that support the findings of this study are available from
the corresponding author upon reasonable request.
INFORMED CONSENT
Written informed consent was obtained from all participants.
ORCID
Kangrun Wang https://orcid.org/0000-0002-8145-3321
1118
|
WA NG e t al.
Fangfang Xie https://orcid.org/0000-0003-2575-0425
Fenghua Chen https://orcid.org/0000-0002-4462-0880
Bo Xiao https://orcid.org/0000-0001-5204-1902
Yanmin Song https://orcid.org/0000-0003-3829-0683
Lili Long https://orcid.org/0000-0001-5078-8770
REFERENCES
1. A sadi- Pooya AA, Stewart GR, Abrams DJ, Sharan A. Prevalence
and incidence of drug- re sistant mesial temporal lobe epilepsy in
the United States. World Neurosurg. 2017;99:662- 666.
2. Wieser HG. ILAE commission repor t. Mesial temporal lobe epilepsy
with hippocampal sclerosis. Epilepsia. 200 4;45(6) :695- 714.
3. Galovic M, de Tisi J, McEvoy AW, et al. Resec tive surger y pre-
vents progressive cortical thinning in temporal lobe epilepsy. Brain.
2020;143(11):3262- 3272.
4. Bernhardt BC, Worsley KJ, Kim H, Evans AC, Bernasconi A,
Bernasconi N. Longitudinal and cross- sectional analysis of at-
rophy in pharmacoresistant temporal lobe epilepsy. Neurology.
2009;72(20):1747- 1754.
5. Long L, Galovic M. Shared hippoc ampal abnormalities in spo-
radic temporal lobe epilepsy patients and their siblings. Epilepsia.
2020;61(4):735- 746.
6. Bernhardt BC, Chen Z, He Y, Evans AC, Bernasconi N. Graph-
theoretical analysis reveals disrupted small- world organization of
cortical thickness correlation networks in temporal lobe epilepsy.
Cereb Cortex. 2011;21(9):2147- 2157.
7. Kim CH, Koo BB, Chung CK, Lee JM, Kim JS, Lee SK. Thalamic
changes in temporal lobe epilepsy with and without hippo-
campal sclerosis: a diffusion tensor imaging study. Epilepsy Res.
2 0 1 0 ; 9 0 ( 1 – 2 ) : 2 1 - 2 7 .
8. Caciagli L, Allen LA, He X. Thalamus and focal to bilateral seizures:
a multiscale cognitive imaging study. Neurology. 2020;95(17):e2427
- e2441.
9. Whelan CD, Alhusaini S, O'Hanlon E, et al. White matter alterations
in patients with MRI- negative temporal lobe epilepsy and their as-
ymptomatic siblings. Epilepsia. 2015;56(10):1551- 1561.
10. Girardi- Schappo M, Fadaie F, Lee HM, et al. Altered communica-
tion dynamics reflect cognitive deficits in temporal lobe epilepsy.
Epilepsia. 2021;62(4):1022- 1033.
11. Leboyer M, Bellivier F, Nosten- Bertrand M, Jouvent R, Pauls
D, Mallet J. Psychiatric genetic s: search for phenotypes. Tre nds
Neurosci. 1998;21(3):102- 105.
12. Caciagli L , Wandschneider B, Centeno M. Motor hyperactivation
during cognitive tasks: an endophenotype of juvenile myoclonic
epilepsy. Epilepsia. 2020;61(7):1438- 1452.
13. Stier C, Elshahabi A, Li Hegner Y, et al. Heritability of magnetoen-
cephalography phenotypes among patients with genetic general-
ized epilepsy and their siblings. Neurology. 2021;97(2):e166- e177.
14. Scanlon C, Ronan L, Doherty CP, et al. MRI- based brain st ruc-
ture volumes in tem poral lo be epilepsy patients and their unaf-
fected siblings: a preliminary study. J Neuroimaging. 20 13 ;2 3(1):
6 4 - 7 0 .
15. Yaakub SN, Barker GJ, Carr SJ, Abela E, Koutroumanidis M.
Abnormal temporal lobe morphology in asymptomatic relatives of
patient s with hippocampal sclerosis: a replic ation study. Epilepsia.
2019;60(1):e1- e5.
16. Alhusaini S, Whelan CD, Doherty CP, Delanty N, Fitzsimons M,
Cavalleri GL . Temporal cor tex morphology in mesial temporal lobe
epilepsy patients and their asymptomatic siblings. Cereb Cortex.
2016;26(3):1234- 1241.
17. Caciagli L , Wandschneider B, Xiao F, et al. Abnor mal hippocampal
struc ture and func tion in juvenile myoclonic epilepsy and unaf-
fected siblings. Brain. 2019;142(9):2670- 2687.
18. Yaakub SN, Tangwiriyasakul C. Herit ability of alpha and sensorim-
otor network changes in temporal lobe epilepsy. Ann Clin Transl
Neurol. 2020;7(5):667- 676.
19. Whittaker JR, Foley SF, Ackling E, Murphy K , Caseras X. The
functional connectivity between the nucleus Accumbens and the
ventromedial prefrontal cortex as an endophenotype for bipolar
disorder. Biol Psychiatry. 2018;84(11):803- 809.
20. Reijneveld JC, Ponten SC , Berendse HW, Stam CJ. The application
of graph theoretical analysis to complex networ ks in the brain. Clin
Neurophysiol. 20 07;118 (11):2317- 2331 .
21. Doucet G E, Rider R, Taylor N, et al. Presurgery resting- state local
graph- theor y measures predict neurocognitive outcomes after brain
surger y in temporal lobe epilepsy. Epilepsia. 2015;56(4):517- 526.
22. Gleichgerrcht E, Keller SS, Drane DL, Munsell BC, Davis KA,
Kaestner E. Temporal lobe epilepsy surgical outcomes can be in-
ferred based on structural connectome hubs: a machine learning
stud y. Ann Neurol. 2020;88(5):970- 983.
23. Vlooswijk MC, Vaessen MJ, Jansen JF, et al. Loss of network ef-
ficiency associated with cognitive decline in chronic epilepsy.
Neurology. 2011;77(10):938- 944.
24. Bernhardt BC , Fadaie F, Liu M, et al. Temporal lobe epilepsy: hippo-
campal pathology modulates connectome topology and controlla-
bility. Neurology. 2019;92(19):e2209- e2220.
25. Liang X, Pang X, Liu J, Zhao J, Yu L , Zheng J. Comparison of topo-
logical proper ties of functional brain networks with graph theor y
in temporal lobe epilepsy with different duration of disease. Ann
Transl Med. 2020;8(22):1503.
26. Liao W, Zhang Z, Pan Z, et al. Altered func tional connectiv-
ity and small- world in mesial temporal lobe epilepsy. PloS One
2010;5(1):e8525.
27. Calhoun VD, Adali T, Pearlson GD, Pekar JJ. A method for making
group inferences from functional MRI data using independent com-
ponent analysis. Hum Brain Mapp. 2001;14(3):140- 151.
28. Gürsel DA , Reinholz L, Bremer B, et al. Frontoparietal and salience
network alterations in obsessive– compulsive disorder: insights
from independent component and sliding time window analyses. J
Psychiatry Neurosci. 2020;45(3):214- 221.
29. Fadaie F, Lee HM, Caldairou B, et al. Aty pical functional con-
nectome hierarchy impacts cognition in temporal lobe epilepsy.
Epilepsia. 2021;62(11):2589- 2603.
30. Nenning KH, Fösleitner O, Schwartz E, et al. The impact of hip-
pocampal impairment on task- positive and task- negative lan-
guage networks in temporal lobe epilepsy. Clin Neurophysiol.
2021;132(2):404- 411.
31. Wang K, Xie F, Liu C, et al. Abnor mal functional connectivity pro-
files predic t drug responsiveness in patients with temporal lobe
epilepsy. Epilepsia. 2022;63(2):463- 473.
32. Ci H, van Graan A , Gonzálvez G, Thompson P, Hill A, Duncan
JS. Mandarin functional MRI language paradigms. Brain Behav.
2016;6(10):e00525.
33. Fisher RS, Acevedo C, Arzimanoglou A, et al. ILAE official report: a
practical clinical definition of epilepsy. Epilepsia. 2014;55(4):475- 482.
34. Winston GP, Cardoso MJ, Williams EJ, et al. Automated hippocam-
pal segmentation in patients with epilepsy: available free online.
Epilepsia. 2013 ;54(12):2166- 2173.
35. Cook MJ, Fish DR, Shorvon SD, Straughan K, Stevens JM.
Hippocampal volumetric and morphometric studies in frontal and
temporal lobe epilepsy. Brain. 1992;115(Pt 4):1001- 1015.
36. Zung WW. A rating instrument for anxiety disorders. Psychosomatics.
1971;12(6):371- 379.
37. Zung W W. A self- rating depression scale. Arch Gen Psychiatry.
1 9 6 5 ; 1 2 : 6 3 - 7 0 .
38. Nasreddine ZS, Phillips NA, Bédirian V, et al. The Montreal cogni-
tive assessment, MoCA: a brief screening tool for mild cognitive
impairment. J Am Geriatr Soc. 2 005;53(4):695- 699.
|
1119
WANG et al.
39. Whit field- Gabrieli S, Nieto- Castanon A. Conn: a functional con-
nectivity toolbox for correlated and anticorrelated brain networks.
Brain Connect. 2012;2(3):125- 141.
40. Fan L, Li H, Zhuo J, et al. The human brainnetome atlas: a new
brain atlas based on connectional architecture. Cereb Cortex.
2016;26(8):3508- 3526.
41. Tzourio- Mazoyer N, Landeau B, Papathanassiou D, et al. Automated
anatomical labeling of activations in SPM using a macroscopic
anatomical parcellation of the MNI MRI single- subject brain.
Neuroimage. 2002;15(1):273- 289.
42. Amiri S, Mehvari- Habibabadi J, Mohammadi- Mobarakeh N,
et al. Gr aph theory application with functional connectivity to
distinguish left from right temporal lobe epilepsy. Epilepsy Res.
2020;167:106449.
43. Smith SM, Fox PT, Miller KL, et al. Correspondence of the brain's
functional architecture during activation and rest. Proc Natl Acad
Sci U S A. 20 09;106(31):1304 0- 13045.
44. Fritz CO, Morris PE, Richler JJ. Effect size estimates: current use,
calculations, and interpretation. J Exp Psychol Ge n. 2012;141(1):2- 18.
45. Watts DJ, Strogatz SH. Collective dynamics of ‘small- world’ net-
works. Nature. 1998;393 (6684):440 - 442.
46. Trimmel K, van Graan AL, Caciagli L, et al. Left temporal lobe
language network connectivity in temporal lobe epilepsy. Brain.
2018;141(8) :240 6- 2418.
47. Zanão TA, Lopes TM, de Campos BM, Yasuda CL, Cendes F. Patterns
of default mode network in tempor al lobe epilepsy with and with-
out hippocampal sclerosis. Epilepsy Behav. 2019;121:106523 .
48. Otte WM, Dijkhuizen RM, van Meer MP, et al. Characterization of
functional and structural integrity in experimental focal epilepsy:
reduced network efficienc y coincides with white matter changes.
PloS One. 2012;7(7):e39078.
49. Ponten SC , Bartolomei F, Stam CJ. Small- world networks and epi-
lepsy: graph theoretical analysis of intracerebrally recorded mesial
temporal lobe seizures. Clin Neurophysiol. 2007;118(4):918- 927.
50. Aupy J, Wendling F, Taylor K , Bulacio J, Gonzalez- Martinez J,
Chauvel P. Cortico- striatal synchronization in human focal seizures.
Brain. 2019;142(5):1282- 1295.
51. Ren X, Brodovskaya A, Hudson JL, Kapur J. Connectivity and neuro-
nal synchrony during seizures. J Neurosci. 2021;41(36):7623- 7635.
52. Avoli M. Mechanisms of epileptiform synchronization in cortical
neuronal networks. Curr Med Chem. 2014;21(6):653- 662.
53. Chavez M, Hwang DU, A mann A, Boccaletti S. Synchronizing
weighted complex networks. Chaos. 2006;16(1):015106.
54. Netof f TI, Clewley R , A rno S, Keck T, White JA . Epilepsy in small-
world net works. J Neurosci. 200 4;24(37):8075- 80 83.
55. Farrant K, Uddin LQ. Asymmetric development of dorsal and ven-
tral attention networks in the human brain. Dev Cogn Neurosci.
2015;12:165- 174.
56. Düzel E, Schiltz K, Solbach T, et al. Hippocampal atrophy in tempo-
ral lobe epilepsy is correlated with limbic systems atrophy. J Neurol.
2006;253(3):294- 300.
57. Besson P, Dinkelacker V, Valabregue R , et al. Structural connectiv-
ity differences in left and right temporal lobe epilepsy. Neuroimage.
2014;15(100 ):135- 14 4.
58. Bernhardt BC, Fadaie F, Vos de Wael R, et al. Preferential sus-
ceptibility of limbic cortices to microstructural damage in tempo-
ral lobe epilepsy: a quantit ative T1 mapping study. Neuroimage.
2018;15(182): 294- 3 03.
59. Tan L, Chen Y, Wu W, Liu C, Fu Y. Impaired cognitive abilities in
siblings of patients with temporal lobe epilepsy. Neuropsychiatr Dis
Tre at . 2020;16:3071- 3079.
60. Gottesman II, Gould TD. The endophenotype concept in psy-
chiatry: etymology and strategic intentions. Am J Psychiatry.
2003;160(4):636- 645.
61. Mo J, Liu Z. Automated detection of hippocampal sclero-
sis using clinically empirical and radiomics features. Epilepsia.
2019;60(12):2519- 2529.
SUPPORTING INFORMATION
Additional supporting information can be found online in the
Suppor ting Information section at the end of this article.
How to cite this article: Wang K, Xie F, Liu C, et al. Shared
functional net work abnormality in patients with temporal
lobe epilepsy and their siblings. CNS Neurosci Ther.
2023;29:1109-1119. doi:10.1111/cn s.140 87
Available via license: CC BY 4.0
Content may be subject to copyright.
