|Home | About | Journals | Submit | Contact Us | Français|
Memory impairment is one of the most prominent cognitive deficits in temporal lobe epilepsy (TLE). The overall goal of this study was to explore the contribution of cortical and hippocampal (subfield) damage to impairment of auditory immediate recall (AIMrecall), auditory delayed recall (ADMrecall), and auditory delayed recognition (ADMrecog) of the Wechsler Memory Scale III (WMS-III) in TLE with (TLE–MTS) and without hippocampal sclerosis (TLE-no). It was hypothesized that volume loss in different subfields determines memory impairment in TLE–MTS and temporal neocortical thinning in TLE-no.
T1 whole brain and T2-weighted hippocampal magnetic resonance imaging and WMS-III were acquired in 22 controls, 18 TLE–MTS, and 25 TLE-no. Hippocampal subfields were determined on the T2 image. Free surfer was used to obtain cortical thickness averages of temporal, frontal, and parietal cortical regions of interest (ROI). MANOVA and stepwise regression analysis were used to identify hippocampal subfields and cortical ROI significantly contributing to AIMrecall, ADMrecall, and ADMrecog.
In TLE–MTS, AIMrecall was associated with cornu ammonis 3 (CA3) and dentate (CA3&DG) and pars opercularis, ADMrecall with CA1 and pars triangularis, and ADMrecog with CA1. In TLE-no, AIMrecall was associated with CA3&DG and fusiform gyrus (FUSI), and ADMrecall and ADMrecog were associated with FUSI.
The study provided the evidence for different structural correlates of the verbal memory impairment in TLE–MTS and TLE-no. In TLE–MTS, the memory impairment was mainly associated by subfield-specific hippocampal and inferior frontal cortical damage. In TLE-no, the impairment was associated by mesial–temporal cortical and to a lesser degree hippocampal damage.
Mesial temporal lobe structures, particularly, the hippocampus, play a major role in declarative memory [Eichenbaum, 2004]. Nonetheless, they are just one element in the widespread networks of cortical and subcortical brain structures supporting declarative memory functions. Different subtypes of declarative memory use different brain structures or networks although there is also some overlap. For example, posterior–inferior parietal, middle, and mesial temporal; dorsomedial, ventromedial, and inferior–lateral prefrontal; and posterior cingulate have been implicated in semantic memory, while hippocampus, mesial temporal, anterior and inferior–lateral prefrontal, precuneus (PRE), lateral parietal, lingual regions, and anterior cingulate have been implicated in episodic memory [Binder et al., 2009; Burianova and Grady, 2007; Desganges et al., 1998; Moscovitch et al., 2005]. Within a network, different brain structures and sometimes even different parts of the same brain structure are specialized in different aspects of the process. An example for specialization between different brain structures is the finding of functional imaging studies, showing activity in the left prefrontal cortex during encoding of verbal information and in the right prefrontal cortex during retrieval [Cabeza and Nyberg, 2000; Desgranges et al., 1998]. The hippocampus on the other hand provides an example for specialization within a brain structure. It consists of several histologically distinct subfields: subiculum, cornu ammonis sectors (CA) 1–3, and dentate gyrus (DG) [Duvernoy, 2005]. These histological differences also translate into functional differences. Animal studies suggest that DG and CA3 are primarily responsible for encoding and early retrieval while CA1 is responsible for delayed retrieval and novelty detection, that is, functions associated with recognition [Acsady and Kali, 2007; Hunsaker and Kesner, 2008; Nakazawa et al., 2004; O’Reilly et al., 2001; Rolls and Kesner, 2006; Wan et al., 1999]. Considering the complexity of memory processes and the widespread networks of specialized cortical and subcortical structures supporting them, any interruption within a network regardless of the localization can be expected to impair the network performance and thus also the function supported by it. The severity of the deficit though will vary depending on the function of the affected structure and the ability of the remaining network to compensate for its dysfunction.
Medial temporal lobe epilepsy (TLE) is characterized by seizures originating in mesial temporal lobe structures, and, consequently, declarative memory deficits are one of the most common and most prominent cognitive deficits in TLE [Jokeit and Schacher, 2004]. Based on imaging and histopathologicial findings, two types of medial TLE are distinguished: TLE with mesial–temporal lobe sclerosis (TLE–MTS, about 60–70%) is characterized by an atrophied hippocampus with MR signal abnormalities and severe neuron loss in the histological examination (MTS), and TLE with a normal appearing hippocampus on the magnetic resonance imaging (MRI; TLE-no, about 30–40%) and no or mild neuron loss in the histological examination. Impairment of declarative memory is found in both subtypes although the deficits tend to be more severe in TLE–MTS [Alessio et al., 2004; Dulay et al., 2004]. The pronounced ipsilateral hippocampal atrophy of TLE–MTS provides an obvious structural correlate for the memory impairment in this TLE subgroup. Furthermore, given the functional specialization of different subfields, different patterns of subfield atrophy [Mueller et al., 2009a) might be associated with different aspects of memory impairment. However, recent studies [Berhardt et al., 2008, Lin et al., 2007] have demonstrated widespread temporal and extratemporal cortical thinning in TLE–MTS, which also affects memory relevant cortical regions. It is therefore possible that damage to these cortical regions contributes to the memory impairment caused by the hippocampal damage. Widespread temporal lateral and extratemporal cortical thinning have also been demonstrated in TLE-no [Mueller et al., 2009b]. This suggests cortical thinning in memory relevant regions as the structural correlate for the memory impairment in this subgroup. However, it cannot be excluded that subtle hippocampal damage, for example, mild DG damage also contributes to the memory impairment in TLE-no. The overall goal of this study was therefore to explore how cortical and hippocampal damage contribute to verbal memory impairment as measured by the auditory immediate recall, delayed recall, and auditory delayed recognition scores of the Wechsler Memory Scale III (WMS-III). In particular, the following hypotheses were tested. (1) Memory impairment in TLE–MTS is mostly caused by hippocampal damage. Specifically, it will be tested if impaired immediate memory performance reflecting a deficient encoding process is associated with volume loss in CA3 and dentate and impaired recognition performance with volume loss in CA1. Cortical thinning, particularly in mesial temporal and prefrontal regions, that is, regions that are involved in episodic memory [Desgranges et al., 1998; Nyberg, 1998] and are known to be affected in TLE–MTS aggravates the memory impairment caused by the hippocampal damage. (2) Memory impairment in TLE-no is associated with cortical thinning in mesial temporal [fusiform and parahippocampal gyrus (PARA)] and prefrontal lateral regions. Subtle hippocampal damage in TLE-no contributes to the impairment of verbal memory caused by the damage to cortical structures.
The committees of human research at the University of California, San Francisco (UCSF), California Pacific Medical Center, San Francisco, and VA Medical Center, San Francisco approved the study, and written informed consent was obtained from each subject according to the Declaration of Helsinki. Forty-three patients with drug-resistant TLE undergoing evaluation for epilepsy surgery who agreed to undergo a dedicated imaging research protocol at 4 Tesla were recruited between mid 2006 and end of 2009 from the Pacific Epilepsy Program, California Pacific Medical Center and the Northern California Comprehensive Epilepsy Center, UCSF (Table I). Eighteen patients (mean age 41.2 ± 11.9; mean years of education: 14.3 ± 2.0; left TLE/right TLE: 7/11, females/males 12/6) had evidence for mesial temporal lobe sclerosis on their 1.5-T MR images (TLE–MTS) by visual assessment. The lateralization of the MTS was concordant with the side of seizure onset in 16 and discordant in 2. Twenty-five patients (mean age 39.9 ± 10.1; mean years of education: 15.3 ± 1.9; left TLE/right TLE: 13/8, bilateral 4; females/ males: 10/15) had normal appearing hippocampi on their 1.5-T MR exams (TLE-no). Twenty-two TLE-no had completely normal 1.5 T MR reads. Of the remaining TLE-no, one had a small cystic lesion in the left inferior temporal pole, another had an ectopic gray matter mass adjacent to the right amygdala, and one had a small left frontal venous angioma. The identification of the epileptogenic focus was based on seizure semiology and prolonged ictal and interictal video/EEG/telemetry recordings in all patients. All patients had been seizure free for at least 24 h before the MRI. TLE–MTS were significantly younger at the onset of epilepsy (TLE–MTS 11.4 ± 9.5 years. TLE-no: 25.6 ± 12.6 years, P = 0.0003) and had a significantly longer duration of epilepsy (TLE–MTS: 29.8 ± 12.7 years, TLE-no: 14.4 ± 13.6 years, P = 0.0005) than TLE-no. The control population consisted of 22 healthy volunteers (mean age 34.9 ± 11.4; mean years of education: 15.1 ± 2.2; females/ males: 17/5). All subjects underwent an extensive neuropsychological test battery to assess their cognitive function. TLE–MTS performed significantly worse compared to controls but not to TLE-no regarding full-scale intelligence and performance intelligence (Wechsler Adult Intelligence Scale: Full IQ: Controls: 112.3 ± 14.7; TLE–MTS: 98.6 ± 14.0; TLE-no: 104.1 ± 15.2, P = 0.016. Performance IQ: Controls: 112.5 ± 15.1; TLE–MTS: 96.8 ± 10.5; TLE-no: 105.4 ± 15.1). The three groups were not different regarding verbal intelligence performance (Wechsler Adult Intelligence Scale Verbal IQ: Controls: 110.4 ± 13.9; TLE–MTS: 100.6 ± 16.3; TLE-no 102.4 ± 14.4, P = 0.08). From this battery, the following standard scores of the WMS-III were selected for the purpose of this study: auditory immediate memory tested by recall (AIMrecall), auditory delayed memory tested by recall (ADMrecall), and auditory recognition delayed memory tested by recognition (ADMrecog). AIMrecall can be regarded as a measure of rapid encoding and early consolidation and ADMrecall as a measure of retrieval performance while ADMrecog provides a measure of memory performance by recognition.
All imaging was performed on a Bruker MedSpec 4 T system. The following sequences were acquired. (1) For the measurement of hippocampal subfields, a high-resolution T2-weighted fast spin echo sequence (TR/TE: 3,990/ 21 ms, echo train length 15, 18.6-ms echo spacing, 160° flip angle, 100% oversampling in ky direction, 0.4 × 0.4 mm in plane resolution, 2-mm slice thickness, 24 interleaved slices without gap, acquisition time 5:30 min, and angulated perpendicular to the long axis of the hippocampal formation). (2) For the measurement of total hippocampal volume, a volumetric T1-weighted gradient echo MRI (MPRAGE) (TR/TE/TI = 2,300/3/950 ms, 7° flip angle, 1.0 × 1.0 × 1.0 mm3 resolution, and acquisition time, 5:17 min) and 3. For the determination of the intracranial volume (ICV), a T2-weighted turbospin echo sequence (TR/ TE 8,390/70 ms, 150° flip angle, 0.9 × 0.9 × 3-mm nominal resolution, 54 slices, and acquisition time 3:06 min).
The method used for subfield marking including assessment of measurement reliability and its limitations has been described in detail previously [Mueller et al., 2007, 2009a]. To briefly summarize it, the marking scheme depends on anatomical landmarks, particularly on a hypointense line representing myelinated fibers in the stratum moleculare/lacunosum [Eriksson et al. 2008], which can be reliably visualized on these high resolution images. The distance between this hypointense line and the outer boundary of the hippocampus provides an estimate of the cortical thickness of the hippocampus at this point. Additional external and internal hippocampal landmarks are used to further subdivide the hippocampus into subiculum, CA1, CA1-2 transition zone (CA1-2 transition), CA3, and DG. The latter two are lumped together (CA3&DG), because there are no macroscopic landmarks to separate them (cf. Fig. 1a). CA1-2 transition is in the dorsal medial region of the hippocampus and consists mostly of CA2. However, due to the landmarks used for labeling it, its volume is influenced by the thickness of the dorsal CA1. To reflect this fact, the sector is called CA1-2 transition rather than CA2. The volumes from the left and right hemisphere were combined, that is, added to provide a single measure from each subfield to be used in the analysis. ICV was determined from the T2-weighted image, which was skull-stripped using the BET program (FMRIB Image Analysis Group, Oxford University, www.fmrib.ox.ac.uk/fsl). To correct for volume differences due to different head sizes, all volumes were normalized to the ICV using the following formula: normalized volume = raw volume × 1,000 ccm/ICV ccm.
All T1 images were segmented using the Expectation Maximization algorithm [van Leemput et al., 1999a,b]. The bias field maps and tissue maps obtained from this process were used for bias correction and skull stripping of the T1 image. FreeSurfer (version 3.05, https://surfer.nmr.mgh.harvard.edu) was used for cortical surface reconstruction and estimation of the mean cortical thickness of 34 anatomical regions in each hemisphere [Dale et al., 1999; Desikan et al., 2006]. The following cortical regions commonly associated with episodic verbal memory [Desgranges et al., 1998] of each hemisphere were selected as cortical regions of interest (ROI) for this project: PARA, fusiform gyrus (FUSI), pars orbitalis, pars opercularis (POP), pars triangularis (PT), caudal middle frontal and rostral middle frontal, superior frontal gyrus, PRE, lobulus parietalis superior, lobulus parietalis inferior, and supramarginal gyrus (cf. Fig. 1b). The thickness measures from the left and right hemisphere were combined, that is, added to provide a single measure from each region. This approach was chosen (1) to account for the findings of functional imaging studies in healthy controls [Desgranges et al., 1998], which did not show a clear preference for the left or right side in verbal tasks but rather provided evidence that both sides are involved although differently; (2) to account for potential compensatory inter-hemispheric shifts of the verbal memory function, which has been described in TLE [Bettus et al., 2009; Campo et al., 2009; Wagner et al., 2008; Richardson et al., 2003].
MANCOVA tests with subfield volumes, cortical ROIs, and memory scores as dependent variables were used to identify nuisance variables explaining a significant amount of the variation of these parameters. In addition to group (TLE–MTS, TLE-no, and control), age, gender, and years of education (log transformed) were modeled as nuisance variables, and thus their contribution to the variability of each of the dependent variables (subfield, cortical ROI, and verbal memory) was tested. Multiple linear regression analysis with group and the co-variates identified in the previous steps followed by Tukey’s post hoc tests were used to test for group differences of normalized hippocampal subfield volumes, cortical thickness and of AIMrecall, ADMrecall, and ADMrecog.
The population was divided into two groups: (a) TLE– MTS and controls and (b) TLE-no and controls to test for associations between subfield volumes or thickness measures and memory scores in each of the TLE subgroups. TLEs were combined with controls after ensuring that there was an overlap between the two groups by correlation plots to allow for the assessment of the full range of structural– functional associations. In a first step, partial correlation analyses (corrected for log of years of education) were used to identify subfields and cortical ROIs, which were significantly correlated with the memory scores. However, subfield volumes and thickness measures of neighboring ROI were correlated with each other (subfield volumes r: range, 0.02–0.76; cortical ROI r: range, 0.08–0.9). Therefore, stepwise linear regression analyses (mixed, probability to enter/ leave P = 0.05) with the memory score as dependent, and those subfield volumes, which had been identified as being significantly correlated with this score in the previous analysis as independent variables, were performed in the next step. Years of education (log transformed) were forced into the model as independent variable, so that only subfields explaining a significant amount of the variation of the memory performance in addition to that explained by years of education could be included in the stepwise regression model. Group was not modeled, because it was intended to test for structural–cognitive correlations over the whole range of memory impairment and subfield volumes represented in the two groups. The contribution of the cortical ROI to memory performance was tested with the same approach. Finally, to determine if cortical ROI or the hippocampal subfields were stronger associated with memory performance, a third stepwise regression was performed with the cortical ROIs and hippocampal subfields selected by the previous analysis as dependent variables. Given the a priori hypotheses outlined in the introduction, corrections for multiple comparisons during the initial partial correlation analyses were not performed for correlations between memory scores and hippocampal subfields, mesial temporal, and frontal ROIs. For parietal ROIs, for which no a priori hypotheses existed, the threshold for significance was set at P < 0.004 (Bonferroni correction based on 12 cortical ROI). All statistical analyses were done in JMP 8 (SAS Institute).
The MANCOVA for subfields showed a significant effect for group [Roy’s Max Root 1.24, F(5, 55) = 13.66, P < 0.0001] but not for age, gender, or years of education. The MANCOVA for cortical ROIs showed a trend for group [Roy’s Max Root 0.57, F(12, 49) = 1.76, P = 0.08], there were no significant effects for gender, age, or years of education. The MANCOVA for memory scores finally showed a significant effect for group [Roy’s Max Root: 0.58, F(3, 58) = 11.34; P = <0.0001] and years of education [0.33, F(3,57) = 6.25; P = 0.001] but not for gender or age. AIMrecall and ADMrecall were significantly lower in TLE–MTS and TLE-no compared to controls but not different between the two TLE groups (cf. Fig. 2a); ADMrecog was worse in TLE–MTS compared to controls; TLE-no was not different from the other two groups. Left and right TLE did not differ from each other regarding verbal memory performance (cf. Fig. 2b). Figures 3 and and44 provide an overview of the subfield volumes and thickness measures in the three groups.
Please refer Figure 4 showing the corresponding scatter plots. TLE–MTS and controls: Table II displays the results of the partial correlation analyses. The results of the stepwise regression analysis were as follows: on the hippocampal level, AIMrecall (β = 0.23, P = 0.0012) was associated with CA3&DG. ADMrecall (β = 0.19, P = 0.0003) and ADMrecog (β = 0.11, P = 0.0261) were both associated with CA1. On the cortical level, AIMrecall and ADMrecall were both associated with inferior lateral prefrontal cortical thickness. AIMrecall was associated with POP thickness (β = 32.17, P = 0.0140) and ADMrecall with PT thickness (β = 33.71, P = 0.004). When CA3&DG and POP were entered together in the model, both were significantly associated with AIMrecall [CA3&DG: β = 0.20, P = 0.0037; R2 adjusted (education, CA3&DG) = 0.35, POP: β = 24.32; P = 0.0416, R2 adjusted (education, CA3&DG, POP) = 0.40]. Similarly, when CA1 and PT were both entered into the model for ADMrecall, both fulfilled the selection criteria [CA1: β = 0.17, P = 0.0006, R2 adjusted (education, CA1) = 0.26) and PT: β = 27.35, P = 0.0074, R2 adjusted (education, CA1, PT) = 0.34].
TLE-no and controls: Table III displays the results of the partial correlation analysis. The results of the stepwise regression analysis were as follows: on the hippocampal level, AIMrecall was associated with CA3&DG (β = 0.15, P = 0.0431) and on the cortical level, with FUSI thickness (cf Table III). When CA3&DG and FUSI were both entered into the model, both contributed significantly to AIMrecall variability and thus were selected by the model [FUSI: β = 17.54, P = 0.0164, R2 adjusted (education, FUSI) = 0.30; CA3&DG: β = 0.15, P = 0.0290, R2 adjusted (education, FUSI, CA3&DG) = 0.36]. FUSI was also the cortical region explaining most of the variability of ADMrecall (β = 24.2, P = 0.0012) and was the only cortical region, which was significantly associated with ADMrecog (cf. Table III).
There were two major findings in this study. (1) As predicted by our a priori hypothesis, hippocampal volume loss was the strongest determinant of verbal memory deficits in TLE–MTS. CA3&DG was associated with impaired early recall (AIMrecall) and CA1 with impaired delayed recall (ADMrecall) and recognition (ADMrecog), suggesting a functional specialization by different hippocampal subfields. Cortical thinning of inferior lateral prefrontal cortical regions contributed to AIMrecall and ADMrecall impairment. (2) Verbal memory deficits in TLE-no were associated with mesial temporal cortical (FUSI) and to a lesser degree also with hippocampal damage. The associations on the hippocampal level were in accordance with the subfield specialization demonstrated in TLE–MTS. Prefrontal cortical damage contributed less to the memory impairment in this group than in TLE–MTS. These findings suggest that verbal episodic memory impairment has different structural correlates in TLE–MTS and TLE-no.
In accordance with our a priori hypothesis regarding TLE–MTS, we found a strong association between hippocampal subfield volume loss and AIMrecall impairment. The major finding in this regard was that AIMrecall was associated with volume loss in CA3&DG but not with volume loss in CA1 even though the volume loss in the two subfields was of equal severity. This supports the notion that CA3&DG plays a crucial role in rapid encoding, early consolidation, and retrieval in TLE–MTS and thus provides evidence for a functional specialization, which is in accordance with the subfield specialization shown for the normal hippocampus in animal studies [Acsady and Kali, 2007; Florian and Roullet, 2004; Gloor, 1997; Hunsaker and Kesner, 2008; Kesner, 2007; Lee et al., 2005; Nakashiba et al., 2008; Nakazawa et al., 2004; Rajji et al., 2006].
Structural–functional correlations in TLE–MTS, however, were not restricted to the hippocampus. AIMrecall was also associated with inferior lateral prefrontal (POP) thickness which was thinner in TLE–MTS than in controls although not significantly so. When CA3&DG and POP were entered together in the stepwise regression model predicting AIMrecall, both were selected. This finding supports the close interaction between mesial temporal and prefrontal regions during memory processes which has been described by other studies [Andersen et al., 2010; Burianova and Grady, 2007; Desgranges et al., 1998; Grady et al., 2005]. Although the exact contribution of the prefrontal cortex to the memory performance is not fully understood, it is generally assumed that the prefrontal cortex monitors retrieval processes and suppresses potential intrusions [Cruse and Wilding, 2009; Vallesi and Shallice, 2006]. Therefore, the significant association between hippocampal and inferior lateral prefrontal regions with AIMrecall shown in this study could indicate that CA3&DG damage results in a less reliable retrieval on the hippocampal level, which requires an enhanced monitoring by prefrontal regions if the early verbal retrieval performance is to be sustained. However, in TLE–MTS with hippocampal and inferior lateral prefrontal damage this compensation mechanism fails and consequently AIMrecall performance suffers.
Subfield volumes in TLE-no were within the control range. Nonetheless, TLE-no showed a strong association between AIMrecall and CA3&DG volume. The finding of such an association even without obvious structural damage further supports the hypothesis that CA3 and DG play an important role in encoding and early consolidation. In accordance with our a priori hypothesis of a contribution of medial temporal cortical damage to memory impairment in TLE-no, AIMrecall was also strongly associated with FUSI, which was thinner in TLE-no than in controls. When FUSI and CA3&DG were entered together in the stepwise regression model, both explained a significant amount of the variation of AIMrecall. This suggests that AIMrecall impairment in TLE-no is related to the dysfunction of a larger hippocampal–neocortical temporal region. In contrast to TLE–MTS, there was no evidence that prefrontal thinning contributed to the AIMrecall impairment in this group.
Our a priori hypotheses that memory impairment is mostly determined by hippocampal damage in TLE–MTS and by neocortical damage in TLE-no were also confirmed for ADMrecall and ADMrecog, which were both associated with CA1 in TLE–MTS and with FUSI in TLE-no. The fact that on the hippocampal level significant correlations with ADMrecall and ADMrecog were limited to CA1 in TLE–MTS provides further evidence for a functional specialization of the different hippocampal subfields and is in accordance with animal and computational studies, which also show a special role for CA1 in late retrieval/recognition [Lai et al., 2005; Leutgeb et al., 2004; Takahashi and Sakurai, 2009]. CA1 has extensive projections into the mesial–inferior temporal lobe [Ichinohoe and Rockland, 2005], a region that includes FUSI, that is, the only neocortical region showing significant associations with ADMrecog in TLE-no. Interestingly, the Freesurfer label FUSI encompasses the perirhinal cortex in its anterior section. The perirhinal cortex is the cortical region, which has consistently been implicated in recognition performance although there is still a debate as to the exact contributions of the hippocampus and perirhinal cortex to recognition [Aggleton and Brown, 2006; Squire et al., 2007; Kumaran et al., 2007].
In TLE–MTS, ADMrecall impairment was associated with inferior lateral prefrontal cortical thinning (PT). When CA1 and PT were entered together into the stepwise regression analysis, both contributed significantly to ADMrecall thus indicating that inferior–lateral prefrontal damage might contribute to ADMrecall impairment similarly as it does to AIMrecall impairment. In TLE-no prefrontal cortical thinning correlated with ADMrecall impairment in the initial partial correlation analyses, but these associations became nonsignificant in the stepwise regression after FUSI was selected by the model. This finding could indicate that in TLE-no prefrontal damage also contributes to ADMrecall impairment but to a lesser degree than it does in TLE–MTS.
This study has several limitations. (1) The number of patients in the study is small, and the findings need to be validated in larger patient population. This will allow to test if the memory—structural associations found in the combined groups, that is, controls/TLE–MTS and controls/TLE-no are also found when each TLE group is analyzed separately. In the current study, the patient groups were too small for these associations to reach significance. Furthermore, a larger population will also allow to analyze left TLE and right TLE groups separately and to assess the influence of side of seizure onset and language lateralization on memory performance. (2) Only verbal memory impairment was assessed in this study. One of the main objectives of this study was to replicate the findings of animal studies showing an association of CA3 and DG with encoding and early retrieval and of CA1 with recognition. AIMrecall and ADMrecog were the only scores in the whole clinical neuropsychological test battery, which allowed to test this association in this population, because they measure not only early retrieval (AIMrecall) but also recognition (ADMrecog) of the same information. (3) Although we used a stepwise regression analysis and thus a relatively conservative statistical approach, it is important to keep in mind that a significant association between structure and function does not necessarily also imply a causal relationship. (4) In contrast to functional neuroimaging studies, which detect not only the networks involved in a specific cognitive function but also potential compensatory mechanisms, structural studies only detect regions whose damage is associated with impairment of a particular function, that is, the failure to find a structural–functional correlation does not allow for the conclusion that a brain region is not involved in a particular cognitive process. (5) There is evidence from functional studies that the anterior and posterior hippocampus have different specializations [Giovanello et al., 2009]. In this study, subfield measurements were only obtained from the anterior third of the hippocampal body and so we were not able to study the anterior–posterior effects of MTS on memory function. (6) Three TLE-no subjects had extrahippocampal lesions. Although these lesions were not included in any of the ROIs assessed in this study, we can not exclude that they had an influence on the memory impairment in these patients. When these subjects were excluded from analysis the findings did not change, indicating that the structural–functional correlations in the TLE-no group were not driven by these three subjects.
In conclusion, this study provided evidence for a different structural correlate of the verbal memory impairment in TLE–MTS and TLE-no. In TLE–MTS, the impairment was mostly associated with neuron loss in the hippocampus but was exaggerated by prefrontal thinning while it was associated with damage to hippocampal–mesial–temporal cortical regions in TLE-no. The memory impairment however was of similar severity in both groups. This indicates that it is the interruption of the network supporting a specific function and less the anatomical localization of the interruption within the network, which determines the memory impairment. The finding of a different structural correlate for memory impairment in the two TLE groups also further supports the notion that TLE-no is not just a mild form of TLE–MTS but a different entity of TLE.
Contract grant sponsor: National Institutes of Health; Contract grant number: RO1-NS31966.
Additional supporting information may be found in the online version of this article.