The objective of this study was to explore differences of functional connectivity in patients with MTLE associated to unilateral (right or left) HS as compared to control subjects by means of resting state fMRI.
Patients and controls had similar age, educational level, handedness and hemispheric dominance for language. There was no difference in age of seizure onset, duration of epilepsy, seizure frequency and antiepileptic drugs used between the two patients' groups. They also presented similar degree of hippocampal atrophy (Figure 1 and Table 1).
Several studies have demonstrated that the left cerebral hemisphere is dominant for language for the majority of subjects [25, 26] and damage in this hemisphere, such as HS, impairs more aspects of language than analogous damage in the contralateral hemisphere . The basis of hemispheric language specialization deserves more studies, especially in terms of putative asymmetries in anatomical structures and/or functional patterns. Despite controversies, predictions of language lateralization can be accessed from gray matter probabilistic maps  as well as from white matter by means of diffusion tensor imaging . Indeed, these works found high correlations between the density of gray and white matters and the dominant cerebral hemisphere for language. One may suppose that this asymmetric aspect is extendible to left/right hippocampal systems and try to evaluate this assumption by means of functional connectivity procedures . Thus, we hypothesized that, in normal subjects with left hemisphere dominance for language, the network associated to the left hippocampus might have a higher level of functional connectivity than its analogous network on the right side, following the same pattern presented by previous structural studies [27, 28]. To investigate this hypothesis, we excluded from this study subjects with atypical (bilateral or right-sided) language lateralization.
Taking into account that MTLE may cause cognitive deficits in human brain functions, such as memory and language, the sort of asymmetric pattern of functional connectivity discussed above might also occur in patients with this kind of epilepsy. In addition, decreases of the level of functional connectivity were observed ipsilateral to the damage in the MTL during the interictal period . Although reduction of the levels of functional connectivity is not directly linked to the increase of functional damage, the progression of white and gray matter atrophies tends to be more intense in patients with left MTLE . Indeed, asymmetrical extra-hippocampal gray matter loss related to hippocampal atrophy can encompass the ipsilateral and contralateral hemisphere, particularly the contralateral hippocampus, more pronounced in patients with left MTLE . This asymmetrical pattern was also found in cryptogenic temporal lobe epilepsy, which has a distinct neuronal network, and the damages were, again, more widespread in patients with left-sided seizure focus .
By considering the groups of patients and their performance on neuropsychological tests, the group with left HS had lower IQ and worse performance on verbal memory, general memory and delayed recall than patients with right HS . In fact, the correlations between the loss of gray matter and performance on a variety of neuropsychological domains also demonstrated higher scores for patients with left MTLE while no positive associations were found for patients with right MTLE . These correlations were present at a "global level", and again, asymmetrical, suggesting involvement of a more pervasive network of these functions, which fits to the same pattern of functional connectivity that we found.
Vis-à-vis the convergence of the patterns of functional connectivity to the structural and functional evidences that came out with the previous works, which our study corroborates, we have looked for asymmetries in functional connectivity in three groups of subjects (patients with left MTLE, patients with right MTLE, and controls) as regards the functional network emulated by the left and right hippocampi during resting state in subjects with left hemisphere dominant for language.
The functional connectivity from the left to the right hippocampus and vice-versa was seen only in the control group (Figures 2A-RS and 3A-LS) as small spots of positive correlations contralateral to the seed location. Similar outcomes were not found in patients' group, neither for the right MTLE group (Figures 2B-RS and 3B-LS) nor for the left MTLE group (Figures 2C-RS and 3C-LS). These findings show that the patterns of functional connectivity of MTLE patients are disrupted.
Still concerning to the intragroup comparison, the major cluster and the maximum t-score were found in areas of the MTL ipsilateral to the seed. For the three groups, the patterns of functional connectivity were higher on left hemisphere. This asymmetry fits precisely to those works that related language lateralization prediction from brain structures imaging [27, 28], but here, the basis of the asymmetry is directly related to the hippocampal basal functional network.
Besides the patterns of functional connectivity were higher for the left seed, there was a hierarchy among the three groups. Controls presented higher correlations than the right MTLE group, which exhibited higher levels than the left MTLE group (Figures 2A, 2B and 2C, respectively). Identical order was found for the right seed which is illustrated by Figures 3A, 3B and 3C respectively. Since the hierarchy of these groups concerned to the homogeneity of the seeds, the patients with left MTLE presented more variability than the other two groups for both hippocampi. By the same rule, the right MTLE had more variability than control subject for hippocampi time series. For these patients with unilateral MTLE, these findings suggests that the functions which depend on hippocampi may be damaged, displaced or rearranged due to the pathological condition  or to the compensatory mechanisms .
In general, controls exhibited higher patterns of functional connectivity than patients, but the t-scores and the cluster sizes of these intergroup comparisons had distinct values in accordance of two elements: the used seed and the side of HS.
For the seed ipsilateral to the HS, both patient groups presented important reductions of functional connectivity in several brain regions. The Figures 4B and 5A illustrate these regions for the left and the right MTLE groups respectively. These findings could be explained by the ipsilateral hippocampal atrophy associated to the epileptogenic dysfunction, which may reduce the functional connectivity . This explanation, based on a structural hypothesis, indicates an intrinsic relationship between the anatomical injury and the disrupted basal functional connectivity .
Besides the fact that a decrease of the functional connectivity patterns encompasses several brain regions for patients, the left MTLE group presented more of these regions with reduction than the right MTLE group. Indeed, these findings are illustrated by two aspects. First, the maximum t-scores were 13.41 for controls versus left MTLE group (Figure 4B) and 10.83 for controls versus right MTLE group (Figure 5A). Second, the amount of significant voxels was higher for the comparisons between controls and left MTLE group as shown in Figure 6. These intergroup comparisons make evident the reduction of the functional connectivity patterns in MTLE patients and suggest that left HS is associated to higher decrease of functional connectivity than right HS.
For the seed contralateral to HS, we also found a reduced functional connectivity in the left MTLE group as compared to the right MTLE group, as demonstrated by t-scores (Figure 4) and number of significant voxels shown in Figure 6. These outcomes were unexpected since we evaluated the "healthier" hippocampi in each group. This astonishing result strongly suggests the existence of distinct functional basal brain processes associated to the left or to the right HS, since left MTLE group holds off controls where right MTLE group approaches to then.
In order to confirm the hierarchy of the three groups, which were firstly suggested by intragroup and then reinforced by intergroup comparisons, we searched the brain regions that were reduced in the left MTLE when compared to right MTLE. By using the left seed, we found two small clusters in both hippocampi (Figure 4C). The highest t-score was 7.49. By using the right seed, we found two similar clusters (Figure 5C) but the highest t-score was 6.96. Taken together, these findings pose stronger evidences that, in the context of a similar degree of hippocampal atrophy, left HS is associated to more reduction of the hippocampal functional connectivity than the right HS.
The unilateral HS was correlated with abnormal patterns of connectivity in other parts of the brain, not only ipsilateral to the hippocampal atrophy . Indeed, the patterns of disrupted connectivity diverged in the MTLE patients' groups, being even lower in the left MTLE group. This result could be explained by more structural damage (i.e. more intense hippocampal atrophy) in the left MTLE. However, this was not the case here, since both patients' groups had similar degree of hippocampal atrophy confirmed by MRI volumetric analysis. Another possibility could be differences in the regional distribution of atrophy within hippocampi, with more pronounced damage in specific parts of each hippocampus, affecting different fiber projections to and from the hippocampal system, thus, causing distinct functional disruptions [39, 40]. Secondly, for patients with left hemisphere dominance for language, left MTLE may cause more functional connectivity damage in the contralateral hippocampus than right MTLE does, because a disruption in the non-dominant hemisphere (the right one in this case) paradoxically may accelerate verbal processes , and may not disturb the basal brain organization in a critical way. Thirdly, in control subjects with left hemisphere dominance for language, the influence of the left hippocampus on the right hippocampus could be more significant than the reverse influence during the resting state, probably due to the hierarchy of brain organization. This is supported by the fact that the seed in the left hippocampus had a higher correlation with the right hippocampus than vice versa (Figures 2A and 3A). These findings indicate that the time series of the left hippocampus is more homogeneous than the time series of the right hippocampus for the control group. In any case, these assumptions are based on evidence from functional MRI and require further experiments for confirmation, such as perfusion, tensor imaging, etc. However, these results are robust enough to demonstrate that although no MRI structural damage could be seen in the hippocampus contralateral to the seizure focus, significant functional brain plasticity might have occurred in these patients. This plasticity differed between left and right MTLE, suggesting that we should use distinct ways of addressing the functional organization of the brain in patients with MTLE.
In addition to the above caveats, it should be noted that our modest sample size represents an important limitation although a large number or runs were performed for each subject. The issue of the statistical power might be more relevant for patients, especially for left MTLE group which exhibited higher levels of variability than the other groups. A further reservation should be stated about the filters applied during the pre-processing period. As functional connectivity correlates two time series, there might exist functional connectivity between signals oscillating out of the band pass that we used. These interactions were practically impossible to remove if they were synchronized to the physiological noise and, to prevent any contamination; we took the risk of excluding genuine connectivity by filtering frequencies out of the band pass. In addition to the fact that this option is quite conservative, we applied the same steps for all groups.
Another important limitation is related to the mask used. Although, the automated anatomical labeling (AAL) atlas  was made for MNI coordinates, it was segmented concerning only one subject. Moreover, this segmentation was made on a structural image, different from the EPI used for our functional data. To reduce the bias added by the masking period, we applied the same mask (left and right hippocampi masks) for all subjects.