Background: Although epilepsy is considered a grey-matter disorder, changes in the underlying brain connectivity have important implications in seizure generation and propagation. Abnormalities in the temporal and extratemporal white matter of patients with temporal-lobe epilepsy (TLE) and mesial temporal sclerosis (MTS) have previously been identified. Patients with TLE but without MTS often show a different course of the disorder and worse surgical outcome than patients with MTS. The purpose of this study was to determine if said white-matter abnormalities are related to the presence of MTS or if they are also present in non-lesional TLE.
Methods: Seventeen patients with TLE and MTS (TLE+uMTS), 13 patients with non-lesional TLE (nl-TLE) and 25 controls were included in the study. Diffusion tensor imaging (DTI) was used to assess tract integrity of the fornix, cingulum, external capsules and the corpus callosum.
Results: The white-matter abnormalities seen in the fornix appear to be exclusive to patients with MTS. Although the cingulum showed an abnormally high overall diffusivity in both TLE groups, its anisotropy was decreased only in the TLE+uMTS group in a pattern similar to the fornix. The frontal and temporal components of the corpus callosum, as well as the external capsules, demonstrated reduced anisotropy in TLE regardless of MTS.
Conclusions: While some white-matter bundles are affected equally in both forms of TLE, abnormalities of the bundles directly related to the mesial temporal structures (ie, the fornix and cingulum) appear to be unique to TLE+uMTS.
Statistics from Altmetric.com
As the electrical impulses that produce seizures originate in neurons, epilepsy is typically considered a grey-matter disease. Although seizures are transmitted along conventional pathways, white-matter tracts have been considered to have a passive role in epileptogenesis restricted to seizure propagation. Nonetheless, volumetric MRI has demonstrated extensive white-matter abnormalities in patients with temporal-lobe epilepsy (TLE).1 2 Correlations between cerebral white-matter volume and disease duration suggest that seizure-related degeneration could play a role in progressive cognitive decline observed in patients with epilepsy.1 The demonstration that transection of the fornix can create an epileptogenic state in an animal model provides indirect evidence that white-matter pathology may be a factor in initiating epileptogenesis.3 Additionally, heterotopic neurons have been found in temporal-lobe white matter of patients with TLE.4 Thus, white-matter pathology has been implicated in both epileptogenesis as well as seizure-related comorbidity.
While volumetric MRI is a non-specific measure of white-matter pathology, diffusion tensor imaging (DTI) is sensitive to microscopic tissue characteristics.5 This is particularly evident in white matter, which shows a high degree of architectural coherence and is organised into tightly packed bundles of axons with similar orientations, whose membranes and myelin sheaths result in anisotropic water diffusion (ie, not equal in all directions).6 Water diffusion abnormalities of white-matter fascicles have been reported in patients with TLE,7–9 which did not normalise upon seizure freedom following surgery.10 The diffusion changes seen in the fornix, cingulum, corpus callosum and external capsules suggested structural abnormalities of the white matter, such as reduced axonal density, myelin abnormalities, or both.6 11 12 These studies have however been restricted to TLE with mesial temporal sclerosis (MTS) or have not distinguished between patients with and without MTS.
The purpose of this study was to perform a direct comparison of the diffusion properties of temporal and extratemporal white-matter structures between TLE patients with unilateral MTS (TLE+uMTS) and non-lesional (nl-TLE).
SUBJECTS AND METHODS
Settings and participants
Approval of the research protocol was obtained from the University of Alberta Health Research Ethics Board, and informed consent was obtained from all participants.
Thirty patients with medically intractable TLE and 25 healthy controls were studied.
TLE with unilateral MTS (TLE+uMTS, n = 17, 39 (SD 11) years old (range 20–59 years); five men, 12 women)
All patients underwent presurgical evaluation. EEG video-telemetry demonstrated unilateral temporal lobe ictal onset. T2 relaxometry demonstrated hippocampal T2 values above two standard deviations from the mean of the controls (supplemental fig 1).13 In two patients, the contralateral hippocampus was also found to have abnormal T2, but in both cases there was clear asymmetry. No other lesions were identified on clinical imaging. Eleven patients included in this group have been previously reported.7 8 10
Non-lesional TLE (nl-TLE, n = 13, age 41 (SD 12) (range 17–62 years); five men, eight women)
All patients had primary ictal semiology of complex partial seizures and temporal lobe epileptic EEG abnormalities. Patients with MRI evidence of MTS (hippocampal T2 greater than 2 SD of control values) or extratemporal structural lesions were excluded. Eleven of 13 subjects had ictal and interictal EEG recordings (seven with prolonged scalp EEG-video recordings and four with both scalp and intracranial electrodes). For the two subjects who did not have ictal recordings, both had unequivocal temporal interictal epileptic EEG abnormalities, and both experienced typical complex partial seizures as their primary seizure pattern. EEG lateralisation was right temporal for two patients, left temporal for six and bitemporal in the remaining five.
Healthy controls (n = 25, age 33 (SD 10) (range 18–58 years); 17 men, eight women)
None of the subjects suffered from any neurological or psychiatric condition. All subjects lacked any visible anomalies on clinical MRI.
The mean age was not significantly different between groups (ANOVA, p = 0.1). Disease duration was similar between the two TLE groups, although the TLE+uMTS group had a slightly longer disease duration (27 (SD 13) years) than the nl-TLE group (18 (13) years) (Student t test, p = 0.053).
Imaging and data analysis
Cerebrospinal fluid-suppressed diffusion tensor images were acquired in 9:30 min using a 1.5 T Siemens Sonata MRI scanner (Siemens Medical Systems, Erlangen, Germany). The sequence consisted of 26 contiguous 2 mm thick axial slices with an in-plane resolution of 2×2 mm2 (interpolated to 1×1×2 mm3). Diffusion-sensitised images were acquired in six directions, with a b value of 1000 s/mm2. Full details of the DTI protocol have been previously provided.7 14 T2 images for the quantification of hippocampal T2 were obtained using a modified CPMG sequence with 32 echoes (TR = 4.43 s; TE1 = 9.1 ms, echo spacing = 9.1 ms), producing 10 coronal 3 mm thick slices with a 3 mm interslice gap in 8:13 min (voxel size 1.2×1.2×3 mm3 interpolated to 0.6×0.6×3 mm3). High-resolution T1-weighted images were acquired using a three-dimensional magnetisation-prepared rapid gradient echo imaging sequence (3D_MPRAGE) with voxel dimensions of 1×1×1 mm3.
Diffusion tensor tractography
Six white-matter structures were studied, namely the fornix, cingulum and external capsule, as well as three portions of the corpus callosum: the frontal (genu), occipital and temporal (tapetum) components of the splenium (fig 1). These six structures were selected for analysis based on previously reported abnormalities in TLE subjects.7–9 15 The fornix, cingulum and the occipital and temporal callosal fibres were depicted with tractography (fibre assignment by continuous tracking algorithm),16 17 whereas the genu of the corpus callosum and the external capsules were analysed using manually placed regions of interest (ROI) on a single axial slice (parallel to the AC-PC line) at the level of the central portion of the thalamus. The tractography algorithm was started in all white-matter voxels in the brain, and the tracts were virtually dissected using a priori anatomical knowledge of their trajectories.18 Placement of tract-selection regions for the depiction of the fornix and cingulum has been reported previously.7 14 Fibres passing through the splenium of the corpus callosum were selected using a mid-sagittal slice, and the temporal and occipital tracts were separated using a second large tract-selection region drawn on the corresponding lobe. The FA threshold for tractography was set as 0.3 (start and stop criteria) for all tracts. Specific portions of the tractography-defined structures were analysed to yield summary diffusion parameters. The crus of the fornix and the temporal portion of the cingulum were analysed between the axial levels of the mammillary bodies and the fusion of the crura of the fornices; the tapetum and occipital components of the corpus callosum were analysed between the lateral ventricles (fig 1). The micro-structural integrity of the tracts was assessed by evaluation of the average quantitative diffusion parameters measured from the voxels within the structures (namely fractional anisotropy (FA), mean diffusivity (MD) and diffusivities parallel and perpendicular to the tracts (λ|| and λ⊥, respectively)). Average diffusion parameters were calculated for each white-matter structure in every subject.
There was no significant left–right asymmetry of the diffusion anisotropy of the fornix or external capsules in the control group (p = 0.11 and p = 0.9, respectively). Also, in the controls, the cingulum showed a slight right>left asymmetry of FA (mean difference = 0.011, p = 0.035). In patients with TLE+uMTS, the fornix and cingulum showed larger abnormalities in the structures ipsilateral to MTS, as compared with the contralateral side, although these differences were small and not present in all patients (fig 2). The issue of laterality and asymmetry in patients with nl-TLE is complicated due to the presence of patients with independent bilateral seizure onset, and the terms “ipsilateral” and “contralateral” are difficult to apply. However, fig 3 shows the apparently low dependence of seizure lateralisation on the anisotropy measures of the paired structures. Due to minimal asymmetry of these paired structures, our previous report of bilateral white-matter abnormalities in patients with unilateral MTS7 8 and to simplify subsequent between-group comparisons, the individual parameters of the paired bundles are reported as (left+right)/2.
Quantitative hippocampal T2 relaxometry
Quantification of hippocampal T2 was performed by fitting a monoexponential curve to the multiecho coronal images and obtaining the mean exponent (ie, T2) within a manually drawn region of the hippocampus encompassing three slices.7
Quantitative hippocampal volumetry
Hippocampi were segmented automatically using the method described in Collins et al.19 The ANIMAL (automated non-linear image matching and labelling) procedure was used to estimate the non-linear transformation required to align all voxels from a subject’s MRI with those in a target MRI model. Hippocampal labels, defined on the target MRI, were mapped through the inverse of the recovered transformation onto the subject’s MRI, thus segmenting the hippocampus. Hippocampal volumes were computed by counting voxels in each hippocampal label set and multiplying the sum by the voxel size in mm3. The high-resolution anatomical images required for this analysis were available in 20/25 controls, 8/13 nl-TLE patients and 15/17 TLE+uMTS patients.
Multivariate analyses of variance (MANOVA) were performed for each white-matter structure. If the p value of the test (Wilk lambda) was <0.05, univariate analyses of variance (ANOVA) were conducted for each diffusion parameter and, when necessary, between-group comparisons were performed using Tukey’s post-hoc test. The DTI parameters of the white-matter bundles were linearly correlated with disease duration and age of seizure onset, using the Pearson correlation coefficient (r). Additionally, we performed partial correlations controlling for age (when analysing the effect of disease duration) or disease duration (for the analysis of age at seizure onset).
Hippocampal T2 and volumetry (supplemental fig 1)
Control subjects showed hippocampal volumes of 3991 (613) mm3 and T2 of 114 (3) ms. Eight of 17 TLE+uMTS patients with hippocampal T2>2SD of controls also demonstrated abnormal hippocampal volumes (<2SD of controls). The hippocampal T2 of patients with TLE+uMTS was 136 (9) and 117 (5) ms (ipsilateral and contralateral, respectively). The ipsilateral hippocampal volume in the same group was 2908 (683) mm3, while the contralateral hippocampi had volumes of 4039 (749) mm3. In the nl-TLE group, the hippocampal T2 was 111 (4) and 112 (6) ms (ipsilateral and contralateral, respectively). The hippocampal volume in nl-TLE patients was 4027 (413) (ipsilateral) and 3886 (627) mm3 (contralateral). Patients with nl-TLE and independent bilateral seizure foci had hippocampal volumes and T2 values that were similar between hemispheres and to the rest of the nl-TLE patients. Of the 17 subjects with increased hippocampal T2, 14 have undergone surgery, with all 14 demonstrating histopathological evidence of MTS. Of the seven patients with abnormal hippocampal T2 and normal volumes, surgical pathology reports were available in six patients, with all six demonstrating MTS. Based on this observation in combination with a previous report of increased sensitivity of quantitative T2 over volumetry13 we chose to categorise our patients as TLE+MTS (n = 17) or nl-TLE (n = 13) based on T2. While histopathology is not available for all subjects, all TLE+uMTS who have had surgery demonstrated histopathological features of MTS, while MTS was absent in all nl-TLE patients who have undergone surgery (four of 13).
DTI of the fornix (figs 4, 5A,B, 6; table 1)
The TLE+uMTS group showed significant diffusion abnormalities when compared with the control and the nl-TLE groups. FA was reduced, and λ⊥ increased in the TLE+uMTS group. These findings were also present when fornices ipsilateral and contralateral to MTS were analysed separately (data not shown). The diffusion parameters of the nl-TLE group were not significantly different from the control group but differed from the TLE+uMTS group. The TLE+uMTS group did not show any significant correlation between disease duration or age at seizure onset with any of the diffusion parameters. The nl-TLE group showed significant correlations between disease duration and FA (r = −0.68, p = 0.01) and λ⊥ (r = 0.74, p = 0.004) (fig 6). When controlling for age, the correlation between disease duration and FA in the nl-TLE group did not reach statistical significance (r = −0.53, p = 0.077), but the correlation with λ⊥ remained significant (r = 0.63, p = 0.028). No other diffusion parameter correlated with disease duration. FA in the nl-TLE group also correlated with age at seizure onset, but only when controlling for disease duration (r = −0.59, p = 0.045).
The TLE+uMTS group showed reduced FA and increased MD and λ⊥ as compared with the control group. Similarly to the fornix, these findings were present both ipsilateral and contralateral to MTS (not shown). Patients with nl-TLE showed increased mean, parallel and perpendicular diffusivities, but normal FA when compared with controls. Furthermore, the increase in MD and λ⊥ was not as large in the nl-TLE group as in the TLE+uMTS group. The TLE+uMTS showed a lower diffusion anisotropy (p = 0.001) and higher mean (p = 0.03) and perpendicular (p = 0.001) diffusivities than the nl-TLE group. The disease duration did not correlate with any of the diffusion parameters in any of the TLE groups. In the nl-TLE group, the age at seizure onset correlated with MD (r = 0.38, p = 0.046) and λ⊥ (r = 0.58, p = 0.049) only when controlling for disease duration.
Both TLE groups showed abnormalities, characterised by reduced FA and increased MD and λ⊥, with no significant differences between these two groups. In the TLE+uMTS group, both hemispheres showed similar abnormalities (not shown). The nl-TLE group showed a significant correlation between λ|| and disease duration, when controlling for age (r = 0.60, p = 0.04).
Both TLE groups showed reduced FA and increased λ⊥, but only the TLE+uMTS group showed increased MD when compared with controls. There were no significant differences between the two TLE groups in any of the diffusion parameters. There were no significant bivariate or partial correlations.
Only the TLE+uMTS group showed a significant, albeit small, reduction in FA and increase in MD and λ⊥ when compared with the controls. The two epilepsy groups had similar diffusion parameters. In the TLE+uMTS group, FA correlated with age at seizure onset, when controlling for disease duration (r = −0.60, p = 0.01).
Both TLE groups showed reductions in FA and elevations of λ⊥. TLE+uMTS patients appeared to have the most marked changes, although the comparison of both TLE groups did not reach statistical significance. Only the TLE+uMTS group had increased MD as compared with the controls. The nl-TLE group showed a significant bivariate correlation between disease duration and λ⊥ (r = 0.61, p = 0.03), but it did not reach statistical significance when controlling for age (r = 0.53, p = 0.07).
In this study, we have shown white-matter abnormalities with distinct spatial distribution and diffusion characteristics for TLE with and without MTS. In general, TLE patients with MTS have a higher degree and more extensive white-matter abnormalities, in particular of the primary limbic pathways.
The previously identified diffusion abnormalities of the fornix7 appear to be exclusive to patients with MTS. The reported reduction in FA and increase in λ⊥ are interpreted to reflect myelin abnormalities, reduced axonal density or both.6 11 12 With the fornix being the principal output pathway from the hippocampus, it is not surprising to find ipsilateral abnormalities in the presence of MTS.20 21 It is notable, however, that such diffusion abnormalities are bilateral7 and that the contralateral structures do not normalise upon seizure freedom following epilepsy surgery,10 suggesting that they are not only due to downstream axonal or myelin degeneration secondary to hippocampal neuronal death. DTI and tractography cannot differentiate hippocampal afferents from efferents, and so the true nature of these white-matter abnormalities cannot be directly elucidated.
Patients with non-lesional TLE showed diffusion anisotropy of the fornix that was not statistically different from controls. TLE with and without MTS did, however, share abnormalities of the genu and tapetum of the corpus callosum and the external capsules. A previous study also found diffusion abnormalities of the external capules, genu and splenium of the corpus callosum, in patients with TLE.9 The underlying reason for the abnormalities seen in the genu in TLE is not clear, as it holds few temporal lobe fibres. The findings in the tapetum, and their presence in TLE with and without MTS, are more straightforward to interpret, as these fibres directly interconnect the two temporal lobes. Although the temporal and occipital callosal fibres were virtually dissected, it is possible that some temporal axons partially join the path of the occipital axons and are thus incorrectly classified. This could explain why the TLE+uMTS group also showed abnormalities of the occipital callosal segment. Alternatively, these callosal changes could reflect diffuse white-matter pathology in TLE patients.
The external capsule contains long association fibres that are part of the inferior fronto-occipital, uncinate and superior longitudinal fascicles.18 The uncinate fasciculus, in particular, interconnects the frontal and temporal lobes and has recently been reported to have abnormally low diffusion anisotropy in patients with TLE and right MTS, and is believed to be implicated in the spread of seizure activity.22
Although reductions of FA of the fornix with age (range 18–88 years) have been demonstrated,23 we did not find any significant correlation between FA of the fornix and age in our control group, in which 88% of the subjects were less than 50 years old (r = 0.18, p = 0.39). The age range in our study, as well as the use of CSF suppression,14 could account for such a discrepancy. The only significant correlation we found between diffusion anisotropy and age was located in the genu of the corpus callosum (r = −0.47, p = 0.02, in line with previous reports24 25). Furthermore, considering age as a covariate in a multivariate analysis of covariance (MANCOVA) showed that it had a significant effect only in the fornix and genu (Wilk’s lambda, p = 0.047 and p = 0.01, respectively). Re-analysis of the fornix using MANCOVA showed similar differences between groups as our original analysis. The use of age as a covariate in the reanalysis of the genu showed that the differences seen in FA and λ⊥ of the genu in the nl-TLE group were not significant with respect to the control group (FA: p = 0.22; λ⊥: p = 0.15).
Ipsilateral grey-matter volume reductions of the entorhinal cortex have been reported in patients with nl-TLE,26 and reductions of grey-matter concentration of the extended limbic system were present in patients with clearly defined unilateral MTS.27 White-matter volume reductions have also been reported in patients with TLE ipsilateral to the suspected seizure focus,28–30 but also bilaterally.2 31 In a voxel-based DTI study of grey and white matter, reduced FA and increased MD were found within and outside the temporal lobe of patients with TLE+uMTS although, contrary to our reports, the contralateral hemisphere only showed abnormalities of mean diffusivity and not FA.32 Similarly, magnetic resonance spectroscopy has shown metabolic abnormalities that are not exclusive to the ipsilateral temporal lobe.33 34 These reports and our current findings demonstrate that TLE presents with brain abnormalities that extend beyond the seizure focus.
The issue of progressive structural changes in response to ongoing seizures in humans remains controversial.35 Of interest, a correlation between white matter asymmetry and duration of TLE has been observed using quantitative volumetric MRI.2 This finding suggests that white-matter changes can occur in response to recurrent seizures in humans. Our report of a correlation between disease duration and DTI parameters of the fornix of patients with nl-TLE (fig 6) suggests that the micro-structural characteristics of this tract are progressively affected by seizures. In patients with TLE+uMTS, however, this correlation was not found. This lack of correlation could be explained by the fact that several patients with TLE+uMTS have abnormal diffusion parameters even with short disease duration. Alternatively, different patho-physiological mechanisms could be responsible for the differences between the two TLE groups.36 Correlations with clinical time-related measures are difficult to evaluate, as confounding factors must be accounted for, such as age-related changes, different susceptibility states to an initial insult at different ages, episodes of status epilepticus, etc. Thus, it is still unknown if the identified white-matter abnormalities predate the development of TLE or if they are a secondary effect of ongoing seizures. It is also plausible that disease duration might be in part responsible for the differences found between the two TLE groups. The two patient groups had disease durations that were slightly different, although this did not reach statistical significance (p = 0.053). We investigated this using a MANCOVA and found that the disease-duration effect was only borderline significant for the fornix (Wilk lambda, p = 0.048), but not for any of the other structures. Given the cross-sectional design of our study and the relatively small sample size, it would be inappropriate to derive definitive conclusions from our correlative analyses. Nonetheless, these data suggest a plausible interaction between disease duration and abnormalities of the fornix.
While epilepsy is certainly a disorder of the grey matter, the present study shows evidence of abnormality in the integrity of the connecting white matter. Coupled with the findings reported by other groups of diffuse white and grey-matter abnormalities, there is considerable evidence to suggest that these abnormalities form part of a dysfunctional network in patients with TLE and that along with different mesial temporal pathology, TLE patients with and without MTS also have distinctly different extratemporal white-matter involvement. The clinical relevance, as well as the cause or effect relationship with seizures, of the white-matter diffusion abnormalities remains to be elucidated.
Operating support by the Canadian Institutes of Health Research (DWG, CB), the Savoy Foundation and the University of Alberta Hospital Foundation (DWG). Salary support by the Alberta Heritage Foundation for Medical Research (CB) and Promep (LC). MRI infrastructure from the Canada Foundation for Innovation, Alberta Science and Research Authority, Alberta Heritage Foundation for Medical Research and the University of Alberta Hospital Foundation. Fibre-tracking software kindly provided by H Jiang and S Mori (NIH grant P41 RR15241-01).
Competing interests: None.
Ethics approval: Ethics approval was provided by the University of Alberta Health Research Ethics Board.
Patient consent: Obtained.
▸ An additional figure is published online only at http://jnnp.bmj.com/content/vol80/issue3
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.