Exploring the impact of hippocampal sclerosis on white matter tracts and memory in individuals with mesial temporal lobe epilepsy

Abstract Objective To investigate how the presence/side of hippocampal sclerosis (HS) are related to the white matter structure of cingulum bundle (CB), arcuate fasciculus (AF), and inferior longitudinal fasciculus (ILF) in mesial temporal lobe epilepsy (MTLE). Methods We acquired diffusion‐weighted magnetic resonance imaging (MRI) from 86 healthy and 71 individuals with MTLE (22 righ‐HS; right‐HS, 34 left‐HS; left‐HS, and 15 nonlesional MTLE). We utilized two‐tensor tractography and fiber clustering to compare fractional anisotropy (FA) of each side/tract between groups. Additionally, we examined the association between FA and nonverbal (WMS‐R) and verbal (WMS‐R, RAVLT codification) memory performance for MTLE individuals. Results White matter abnormalities depended on the side and presence of HS. The left‐HS demonstrated widespread abnormalities for all tracts, the right‐HS showed lower FA for ipsilateral tracts and the nonlesional MTLE group did not differ from healthy individuals. Results indicate no differences in verbal/nonverbal memory performance between the groups, but trend‐level associations between higher FA of visual memory and the left CB (r = 0.286, P = 0.018), verbal memory (RAVLT) and ‐left CB (r = 0.335, P = 0.005), ‐right CB (r = 0.286, P = 0.016), and ‐left AF (r = 0.287, P = 0.017). Significance Our results highlight that the presence and side of HS are crucial to understand the pathophysiology of MTLE. Specifically, left‐sided HS seems to be related to widespread bilateral white matter abnormalities. Future longitudinal studies should focus on developing diagnostic and treatment strategies dependent on HS's presence/side.


| INTRODUCTION
Individuals with epilepsy may demonstrate persistent and deteriorating memory performance. 1,2 Even though seizures originate in gray matter, epilepsy is a whole-brain network disorder. 3 The structural networks of the brain, which are made up of myelinated axons connecting gray matter areas, are central to the propagation of the electrical signals, which lead to epileptic seizures. 4 Repeated seizures can be associated with microstructural white matter abnormalities. 5 Furthermore, white matter alterations in individuals with epilepsy are related to memory performance, 6 are a marker for pharmacoresistance, 7 are pivotal for surgery planning, 8 and can predict postsurgical cognitive deficits. [9][10][11] Temporal lobe epilepsy (TLE) is the most prevalent form of focal epilepsy in adults, 12 with mesial temporal lobe epilepsy (MTLE) being the most frequent subtype. While most individuals with MTLE have hippocampal sclerosis (HS), 13 which can be diagnosed in vivo by magnetic resonance imaging (MRI), 14,15 up to 20% have a nonlesional or negative MRI. 13 Most current research investigating white matter abnormalities in epilepsy has utilized diffusion-weighted MRI (dMRI). 4 These studies have typically reported the dMRI-derived measure known as fractional anisotropy (FA) as the primary measure of interest. 4,10 FA is often interpreted to serve as a proxy for overall white matter health, with FA reductions generally indicating abnormalities in the structural organization of white matter. 4,10 Indeed, several studies observed FA reductions in, for example, the cingulum bundle (CB), arcuate fasciculus (AF), and inferior longitudinal fasciculus (ILF), individuals with TLE. [16][17][18] Notably, the CB, AF, and ILF are all central for verbal and nonverbal memory performance, a domain often impaired in individuals with MTLE. These findings suggest a critical role of these three white matter bundles in the manifestation of TLE memory impairments, including verbal and nonverbal memory. Although episodic memory is also frequently impaired in individuals with MTLE, the association of the left hemisphere with verbal, and right with nonverbal memory 19 makes it appealing to explore memory relationships with white matter tracts left and right-sided.
Importantly, preliminary studies suggest that the presence of HS is related to the severity and extent of both white matter and cognitive abnormalities. 9,10,[20][21][22][23][24][25] Some studies have indicated that the side of HS is clinically meaningful, and the presence of left-HS is associated with more widespread white matter damage, more memory impairments, and a worse prognosis. 19,26 On the contrary, others described bilaterally FA reductions. 16,27 Little is known about nonlesional MTLE, suggesting the inconsistent results may be due to lateral and mesial mixed etiologies and underpowered studies. Demonstrated evidence indicates that each epilepsy syndrome has a different pattern of white matter alterations 9 and a better characterization may help to understand further the pathophysiology, maybe leading to personalized treatment for each MTLE subgroup. • While individuals with MTLE and right-HS had abnormalities in the right hemisphere only, individuals without HS did not differ from healthy individuals.
• Our findings suggest that it is vital to study the presence and side of HS when trying to understand MTLE and its role in verbal and nonverbal memory performance and when identifying treatment targets.
Therefore, it is currently unclear how the presence and side of HS relate to these findings and if white matter abnormalities in memory-related tracts are associated with cognitive deficits in MTLE, since direct comparison of diffusion measures and memory performance are still scarce. 24,28,29 The present study investigates white matter microstructure in individuals only with MTLE with the principal focus on whether the presence and side of HS yield a specific microstructural impact on white matter tracts associated with memory performance. We collected dMRI and cognitive data from healthy individuals, individuals with either a right or left unilateral HS (right-HS and left-HS group, respectively), and nonlesional MTLE of both lateralities (nonlesional MTLE group). Different from most previous studies, we utilized a machinelearning approach and fiber clustering methods 30 to extract and compare white matter FA of the CB, AF, and ILF between groups. Employing a method distinct from those used in most previous studies could enhance the validation of white matter microstructure changes in MTLE. Furthermore, these tracts are the primary focus considering their role in memory. 6,[31][32][33][34] Based on previous studies, 4,10,23,25 we hypothesized that 1) the left-HS group will demonstrate more widespread white matter abnormalities than the right-HS group, and 2) white matter abnormalities will be less pronounced for the nonlesional MTLE group compared with the HS groups. Finally, we explored how the white matter microstructure of the CB, AF, and ILF is associated with verbal and nonverbal memory performance in individuals with MTLE.

| MATERIALS AND METHODS
The study was approved by the University of Campinas (UNICAMP) Ethical committee, and all participants signed informed consent in accordance with the Declaration of Helsinki. We recruited healthy controls from the UNICAMP community via advertisement and individuals with epilepsy at the outpatient Epilepsy Clinics at UNICAMP, Brazil. This is a retrospective study and the recruitment period occurred from January 2015 to July 2018.
All included subjects were Brazilian Portuguese native speakers between 21 and 70 years. Exclusion criteria for all participants were metal in their body, pregnancy, and any other health conditions that advise against MRI. Additional exclusion criteria for healthy controls were any past or present neurological or neuropsychiatric disorder diagnosed by trained neuropsychologists (TZ and TL) and neurologists (FC and YC). Last, we excluded healthy controls if presenting with any structural brain abnormalities diagnosed by MRI.
All individuals with epilepsy presented with clinical and EEG features consistent with the diagnosis of MTLE according to the International League Against Epilepsy (ILAE) https://www.epile psydi agnos is.org/seizu re/tempo ral-over-view.html 35 ). All individuals underwent serial diagnostic EEGs with at least 25 min of duration each, which were reviewed by a certified neurophysiologist (FC). We did not include individuals with polyspikes, bisynchronous discharges, discharges outside the temporal lobe regions, any neurologic comorbidity, or individuals submitted to brain surgery. As part of Epilepsy Clinics medical care, we employed an MRI epilepsy protocol following the ILAE recommendations for structural MRI in the care of patients with epilepsy on a 3 T scanner (Philips Achieva, Best, the Netherlands). 15 MRI signs of HS were defined by visual analysis of hyperintense T2-weighted and Fluid-attenuated inversion recovery (FLAIR) signal and hippocampal atrophy on high-resolution 1 mm 3D T1-weighted and coronal 3 mm 2D T1-weighed inversion recovery images and confirmed by MRI volumetry as previously described. 36 Individuals were grouped into MTLE with MRI-negative of both lateralities (nonlesional MTLE), MTLE with unilateral HS on the right (right-HS), or left (left-HS). Patients with bilateral HS or other MRI lesions were not included.
Last, we recorded the response to pharmacotherapy of all individuals. Specifically, we inquired about the use of topiramate, as this antiseizure medication can affect cognitive abilities. Regarding the frequency of seizures, we considered seizures with impaired awareness self-report, categorizing individuals who were free or with two or fewer seizures during the previous year as infrequent seizures and those with three seizures or more during the previous year as frequent seizures.
Briefly, diffusion images were axis-aligned, centered, and motion-and eddy-corrected by affine whole-brain registration of each gradient direction to the baseline B 0 using FLIRT (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki). Brain masks (Slicer-software version 4.5, www.slicer.org 37 ) were automatically generated, visually checked, and manually edited to exclude non-brain areas.
We utilized two-tensor Unscented Kalman Filter (UKF) 38 whole-brain tractography (https://github.com/ pnlbw h/ukftr actog raphy), a multi-tensor filter method that uses tractography to drive the local fiber model estimation. Differently from other models that use independent estimation at each voxel, we performed tractography using Gaussian tensors to model the signal in a causal process, starting at a seed point and defining each new position along the fiber, based upon the previous position findings, and tracing the fibers to its termination. Compared with single-tensor model, multi-tensor model is especially relevant for identifying crossing-fibers and more complex diffusion patterns, allowing to find not only the main dominant pathway but also many of the anatomically known existing lateral pathways. The use of two Gaussian tensors enables the fiber estimation directly from the raw signal data without separate preprocessing. The two-tensor fiber model assumes ellipsoidal shape for tensors (preferred λ1 direction, fixed weights). Equally weighted may limit flexibility, but adjust to fit the signal as a fully weighted model. 38 Considering the signal reconstruction is nonlinear and aiming to choose the most consistent direction, UKF was used to estimate the model parameters. 38 Previous studies demonstrated that UKF wholebrain tractography is highly sensitive and consistent. 39 Whole-brain tractography outputs were assessed using the semi-automated quality control tool in the whitematteranalysis software (https://github.com/Slice rDMRI/ white matte ranal ysis). 37 Next, we identified the white matter tracts for each subject using the White Matter Analysis (WMA) package, a robust machine-learning approach with high test-retest reproducibility. 30,39 This method combines a well-established fiber clustering pipeline with a neuroanatomist-curated white matter atlas. 39,40 Each participant's whole-brain tractography was registered to the atlas space. For detailed information, see Zekelman et al. 41 Last, we performed visual and quantitative quality control of the parcellated tracts and extracted fractional anisotropy (FA) values for the CB, AF, and ILF.

| Neuropsychological assessment
Trained neuropsychologists conducted the cognitive evaluation of individuals, which included the Wechsler Memory Scale-Revised (WMS-R) 42 (only Verbal Memory and Visual Memory indexes) and the Rey Auditory Verbal Learning Test (RAVLT) 43 (only the codification subtest).
The tests were selected to evaluate nonverbal memory (Visual Memory WMS-R), and verbal memory (Verbal Memory WMS-R and RAVLT codification) as these are the most consistently affected domains in individuals with MTLE. 44, 45 We did not perform cognitive assessments on healthy control participants. 43

| Statistical analyses
Statistical analyses were conducted using the Statistical Package for Social Sciences, version 26.

| FA analysis
We performed two one-way MANCOVAs to evaluate differences in the white matter tracts between the four groups (nonlesional MTLE, right-HS, left-HS, and healthy controls). We included the group as the independent variable, the FA values of CB, AF, and ILF as dependent variables, and sex and age as covariates. We utilized separate MANCOVAs for the left and right hemispheres to avoid multicollinearity. In the case of a significant group difference in the MANCOVA (P < 0.025, Bonferroni-corrected for two tests), we performed post hoc ANCOVAs for the three tracts. We then conducted pairwise comparisons only in the case of a significant group difference (P < 0.017, Bonferroni-corrected for three tracts). We applied this hierarchical approach of MANCOVAs, ANCOVAs, and post hoc comparisons to limit type I error.

| Neuropsychological analyses
Neuropsychological data were evaluated using three oneway ANCOVAs with the group as the independent variable (nonlesional MTLE, right-HS, and left-HS) and visual memory, verbal memory, and RAVLT codification as the dependent variable, respectively. Age, sex, and years of education were included as covariates, and a P value of <0.017 was considered significant (Bonferroni-corrected for three tests). We conducted post hoc comparisons between the three single groups, only in the case of a significant group difference in the ANCOVA.
Last, we applied Pearson correlations to investigate the association between the FA of each tract and the three neuropsychological tests in individuals with epilepsy. We considered P < 0.003 significant (Bonferroni-corrected for six tracts and three neuropsychological tests).

| Demographic and clinical information
We recruited 99 healthy controls and 92 individuals with MTLE. Thirteen healthy controls and 21 individuals with MTLE failed to meet the diffusion tensor imaging (DTI) quality control standards and were excluded from the analyses. Thus, our final sample consisted of 86 healthy controls and 71 MTLE individuals. The MTLE individuals were divided into nonlesional MTLE, right-HS, and left-HS, considering the presence and side of hippocampal sclerosis (Table 1). There was no significant difference among the groups for the duration of illness or use of topiramate.

| FA analysis
First, we conducted two MANCOVAs (left and right hemisphere) with the group as the independent variable and the FA values of CB, AF, and ILF as dependent variables. We observed a significant group effect for the left (F 9,441 = 3.004, = 0.058, P = 0.002) and right hemisphere (F 9,438 = 5.872, = 0.108, P < 0.001). Next, we conducted post hoc ANCOVAs for the three tracts of the left and right hemispheres separately. After multiple comparison corrections, we observed significant group effects for all tracts (Table 2). Last, we conducted multiple comparisonscorrected post hoc analyses to compare the single groups (Table 3, Figure 1). The tractography of CB, AF, and ILF is represented in Figure 2.
Healthy controls presented higher FA values than (a) individuals with left-HS for all tracts and (b) individuals with right-HS for CB and ILF. The nonlesional MTLE group showed higher FA values than (a) the left-HS group for all left-sided tracts and (b) the right-HS group for the right CB. We did not observe any significant group difference for any tract between (a) healthy controls and nonlesional MTLE individuals, nor (b) between the left-HS and right-HS groups.

| Neuropsychological analyses
There was no difference in the neuropsychological tests for any MTLE groups (Table S1). We, therefore, combined the MTLE groups and performed Pearson correlations to test for an association between the FA of each tract and the three neuropsychological tests. There were no significant correlations at the Bonferroni-corrected threshold for the six tracts and three tests (P < 0.003, Table S2), however, we observed a few correlations that exhibited trend-level significance. Specifically, we saw an association between the T A B L E 1 Clinical and demographic characteristics of healthy controls, nonlesional, right-HS, and left-HS.

| DISCUSSION
This study investigated white matter microstructure in CB, AF, and ILF and the impact of the presence and side of HS in adults with MTLE. The main strengths of our paper are the use of a machine-learning approach with a different procedure for analyzing the white matter tracts, and the correlation of the findings with memory performance that has not been explored in most studies with DTI. We found a graded relationship dependent on the presence and side of HS for white matter abnormalities. As hypothesized, the left-HS group demonstrated more widespread white matter abnormalities than the right-HS group. While the left-HS group showed abnormal white matter for all three tracts in both hemispheres, the right-HS group experienced abnormal white matter for CB and ILF in only the right hemisphere. Our hypothesis was partially confirmed regarding the nonlesional MTLE group. We expected fewer abnormalities than the left-HS and right-HS groups but were surprised to find no abnormalities. The exploratory memory analysis revealed no significant differences in verbal or nonverbal memory performance between the left-HS, right-HS, and nonlesional MTLE groups. In addition, we only observed trend-level significant associations between the white matter microstructure of the CB, AF, and ILF and memory performance.

| White matter abnormalities
Although previous studies have described white matter abnormalities as a core deficit of epilepsy, findings have been inconsistent regarding the extent and severity of white matter damage. Specifically, it is still an area of active debate whether white matter changes in individuals with epilepsy are constrained only to the same hemisphere of the epileptogenic zone (ipsilateral hemisphere) 4,10 or found in both hemispheres. 16,27 Also, as previously mentioned, studies have suggested that the side of HS is relevant for clinical purposes, and the left HS has been associated with a worse prognosis and more impairments in white matter and memory. 19,26 However, previous results are somewhat inconsistent for specific tracts, such as CB, AF, and ILF. Given that a recent meta-analysis has shown that each epilepsy syndrome displays a unique pattern of white matter alterations, but without direct comparison with memory performance, 9  The "+" indicates that the fractional anisotropy value of the group in the first column is larger than the fractional anisotropy value of the group in the second column; the "−" indicates the opposite. The significant values are colored in gray.
Our findings align with our hypothesis that the left-HS group would be the most affected. Individuals with left-HS exhibited widespread bilateral white matter abnormalities in the CB, AF, and ILF compared to healthy individuals. However, the right-HS group displayed lower FA than controls for the ipsilateral (i.e., right-hemispheric) CB and ILF. While we expected that the nonlesional MTLE group would present with fewer FA reductions when compared to the other MTLE groups, we were surprised that the nonlesional group did not exhibit any abnormalities in the white matter tracts we investigated. Contrary to our study, few other authors have proposed that the absence of MRI signs of HS is associated with minor white matter damages instead of our observations of no differences. 11,46 Our findings, however, agree with a smaller study by Concha and colleagues, 18 which also showed that white matter abnormalities were exclusive to MTLE patients with HS compared to those without HS for the CB. They further highlight how crucial it is to separate individuals into pathologically defined subgroups when trying to understand the complex pathophysiology of epilepsy.
Additionally, previous research reported that unilateral MTLE could be related to bilateral alterations of different regions, such as the CB. 16,47 Focke and colleagues, 16 who did not include a nonlesional group, revealed that unilateral MTLE was associated with a different pattern of FA reductions for left-and right-HS. In particular, these authors reported bilateral abnormalities of the AF for left-HS and unilateral abnormalities of the AF for right-HS. Our results partially overlap with these findings as we see bilateral AF abnormalities for the left-HS group but not for the right-HS group. It suggests that, at least in our sample, the AF might be more susceptible to damage associated with the left hippocampus and emphasizes that the right-HS may not affect all ipsilateral tracts with the same severity. Furthermore, our results agree with a few previous TLE studies 25,28 that have reported bilateral FA reductions in the ILF for left-HS, while the right-HS group showed only ipsilateral abnormalities.
Interestingly, in a previous study, 48 we found a similar hierarchical pattern investigating the role of HS in the functional connectivity of the default mode network (DMN). We reported a possible compensatory re-organization of the DMN for the right-HS and nonlesional MTLE groups. In contrast, the left-HS group did not show this pattern. We interpreted these findings to indicate an increasing connectivity disruption hierarchy, with the left-HS group having more disrupted connectivity than the right-HS and MTLE with negative MRI. 48 Future studies should investigate if resting-state functional connectivity is associated with structural white matter microstructure alterations in MTLE. These findings suggest considering the identification of HS as a potentially critical diagnostic measure. Nonlesional, right-, and left-HS seem to be associated with different functional and structural damages, resulting in distinct prognoses and possibly in different treatment strategies.

| Association between white matter microstructure and memory performance
Given that memory deficits are the most common cognitive impairments in individuals with MTLE 49 and verbal and nonverbal memory are considered side-related, we focused our cognitive analyses on verbal and nonverbal memory performance. Additionally, we selected the CB, AF, and ILF as relevant white matter tracts because all have previously been related to verbal and nonverbal memory performance. The CB and AF are linked to verbal 31,50 and visuospatial memory. 51 The AF connects Broca's and Wernicke's areas, and it is related to naming and verbal fluency tasks. 28,50 The ILF is essential for cognitive flexibility, visual (specifically face), and language recognition. 50 Surprisingly, contrary to the aforementioned white matter findings, we observed no gradation differences between the three MTLE groups for verbal and nonverbal memory performance (as measured by the WMS-R and RAVLT). Some factors such as the absence of a control group comparison, low educational level of all three MTLE groups, the long duration of epilepsy (floor effect), uncontrolled onset, and different group sizes may have contributed to the absence of a gradual degree of memory loss. Although we did not investigate verbal and nonverbal memory performance in healthy controls, other studies have shown that WM microstructure is relevant for the adequate communication between gray matter areas associated with memory. 52,53 Interestingly, a longitudinal study reported structural WM improvement 53 induced by working memory training in both young and older healthy adults, suggesting fiber tracts integrity is directly associated with verbal and nonverbal memory.
As we did not observe group differences for any of the cognitive tests, we conducted exploratory correlation analyses for the combined sample of all individuals with MTLE. While no correlations remained significant after correction for multiple comparisons, we found trend-level correlations between the left CB and visual memory, the left and right CB and RAVLT performance, and the left AF and RAVLT performance. Although nonsignificant, better performance was associated with higher FA values for correlational associations, which is in accordance with another study. 51 Thus, further longitudinal studies are needed to follow up on white matter abnormalities' role in memory performance in MTLE.

| Limitations and future directions
One limitation of the study is the unequal sample sizes between the patient groups, specifically the differences between the left-HS and right-HS groups, which could be attributed to the larger sample size of the left-HS group. Also, the nonsignificant result in the nonlesional MTLE group could be caused by its smaller sample size (15 vs. 86), and this should be explored further in larger studies. With a larger sample size, there is a higher likelihood of detecting significant differences between groups and obtaining more robust results. A smaller sample size, on the other hand, may limit the study's ability to identify subtle but meaningful differences.
We could not lateralize most patients in the nonlesional group using EEG, the resulting subgroups were too small for analyses; therefore, we consider left-and right-sided EEG nonlesional MTLE patients as a single group.
Furthermore, multiple testing can increase error type 1 and each of the three tracts was investigated on both right and left sides, we focused on the FA as it is the most used and well-established dMRI measure. Nevertheless, future studies should expand analyses to include mean, radial, and axial diffusivity and additional tracts to better understand the pathophysiology of epilepsy syndromes. Additionally, it will be essential to conduct longitudinal studies to examine the trajectory of structural and functional abnormalities, including memory performance, in individuals with MTLE. These studies may enlighten unanswered questions regarding the progress of the disease (i.e., if the presence of left-HS leads to more widespread white matter damage or alterations were already present from the beginning) 19,26 and help to decide the best treatment options (i.e., the characteristics that indicate best surgery outcome; the best moment for surgery). Future investigations should also include neuropsychological evaluations of healthy controls and covariates for the age of onset of MTLE, type, and dosage of antiseizure medications not available in the present study. Also, memory is a complex higher-order function. The standard memory tests and the implications of these classical cognitive battery performances in real-life memory impairments have been debated in the last decades. 49 As all neuropsychological tests have their own strengths and limitations, it may be important to consider using an additional test battery. More widespread white matter impairment in left-HS may be associated with the severity of the illness and should be further investigated. Additionally, a potential limitation includes the absence of registration of barbiturates, benzodiazepines, and other medications that may affect memory or other cognitive functions.
Another limitation is the significant difference in the self-reported seizures, with the left-HS presenting more impaired awareness seizures than the other MTLE groups. As there is a known association between the frequency of seizures and therapeutic efficacy, future longitudinal studies should focus on investigating if there is a causality between the presence and side of HS in MTLE and the frequency of seizures, or if the frequency of seizures themselves are associated with the white matter impairments. Also, we consider it difficult to accurately register the number of seizures as they are self-reported and individuals with MTLE may not recognize or remember their occurrences.

| CONCLUSIONS
In summary, we observed white matter abnormalities in individuals with epilepsy, dependent on the presence and side of HS. Individuals with left-HS presented widespread bilateral white matter abnormalities. Individuals with right-HS showed abnormalities only in the right hemisphere. The nonlesional MTLE group showed no white matter abnormalities compared with healthy individuals. Importantly, these graded abnormalities align with previous functional findings and highlight that nonlesional, right-HS, and left-HS should be studied when trying to understand the pathophysiology of MTLE. Last, we believe that future longitudinal studies are warranted to examine further the role of these specific structural pathologies in cognitive performance (e.g., memory) and clinical presentation.