Hippocampal malformation as a cause of familial febrile convulsions and subsequent hippocampal sclerosis

Twenty-five to forty percent of patients with drug-resistant temporal lobe epilepsy (TLE) had febrile convulsions during their childhood.^1-5 This subpopulation of TLE patients almost invariably has hippocampal sclerosis (HS), and the HS in these patients shows a more diffuse and severe pattern than the HS of TLE patients without febrile convulsions.^1,2,6,7 Surgical treatment of TLE patients with antecedent febrile convulsions is a favorable option; more than 90% of these patients remain seizure-free following medial temporal lobe resection.¹ An antecedent febrile convulsion is the preoperative variable with the strongest predictive value for good postoperative outcome.¹ These clinical and pathologic characteristics suggest that TLE, unilateral HS, and febrile convulsions are causally connected. Whether HS is a cause or consequence of febrile convulsions remains unclear.^3,4,6,8 To our knowledge, there are only a few case reports that clearly support the assumption that HS is the consequence of a seizure during early childhood or adulthood.^9-12 Nevertheless, there is no good explanation for how systemic stimuli such as fever and more or less generalized seizures induce such a focal lesion as a lateralized HS. One often proposed hypothesis is that a pre-existing alteration of the medial temporal lobe on one side facilitates febrile convulsions and acts as a crystallization point for the development of HS.^2,6,13 This hypothesis is supported by preliminary data from the University of Sheffield study of complicated early childhood convulsions.¹⁴ The findings of that study led to the suggestion that an asymmetry of hippocampal volumes (HVs) detected shortly after a prolonged febrile convulsion is more likely to be a pre-existing anomaly than a consequence of the febrile convulsion.

Except for sporadic cases, a predisposition to febrile convulsions is an inherited trait with a genetic basis. Major gene effects with genetic heterogeneity might be responsible for febrile convulsions in some families.^15-18 The genetic basis of familial febrile convulsions and the hypothesis of a pre-existing alteration of the medial temporal lobe raise the possibility that, in familial febrile convulsions with subsequent TLE, a structural alteration is the primary phenotypic manifestation that contributes to the generation of febrile convulsions and the development of HS. This alteration does not lead to febrile convulsion in each affected subject, but may appear in first-degree relatives who have not had any seizures. Thus, we proposed that alterations of medial temporal lobe structures in clinically unaffected subjects as well as in subjects with febrile convulsions would support the hypothesis that these structural alterations are the cause, not the consequence, of febrile convulsions.

Visual and quantitative MRI investigations have proven useful in detecting structural abnormalities of the hippocampus.^{2,5,7,14,19-34} In these studies, asymmetric HVs were associated with dysgenic or sclerotic hippocampal structures. Although HS leads to increased signal intensities in T2-weighted or, with higher sensitivity, fluid-attenuated inversion recovery images (FLAIR),^20,28,33,34 dysgenic hippocampi show normal signal intensities.^23,24,30,31 In our study, we used this approach of visual and quantitative MRI analysis to address the aforementioned hypothesis.

Materials and methods. The families. Twenty-three members of two presumably unrelated families of Caucasian origin were investigated (figure 1). Thirteen had febrile convulsions during their childhood. Their mean age was 32 years (range, 10 to 63 years); six were female. The neurologic and psychiatric status of all investigated subjects was normal. All subjects were right-handed except for subject A-III-8, who was left-handed. Medical histories of the first generation of each family and of the second generation of Family A were obtained by interview only. Clinical data of the remaining subjects was obtained by interviews of the participants and their parents and from medical records.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Pedigrees of Families A and B. TLE = temporal lobe epilepsy; FC = febrile convulsions.

Seizure history data are summarized in table 1. All but one first febrile convulsion occurred before the age of 13 months. Whether Subject A-II-2 had lateralized convulsions is unknown; lateralized convulsions were denied in all other subjects with febrile convulsions. Subject A-IV-6 had a single prolonged seizure during a respiratory infection with fever when he was 7 years old. All other subjects had more than one febrile convulsion. These convulsions were generally brief; only three subjects had prolonged convulsions requiring emergency admission to a hospital. The mean age of cessation of febrile convulsions was 4 years. Of the five subjects with more than 25 febrile convulsions or prolonged convulsions(>2 hours), two developed chronic afebrile convulsions and the other three showed a regression or delay in language and motor development. Subject A-III-19 developed afebrile seizures at age 6, a few days (<1 week) after a fever episode with several febrile convulsions. In addition to unprovoked seizures, he had convulsions during each fever episode until age 10. Subject B-II-7 developed afebrile seizures at age 9, 5 years after his last febrile convulsion. Both subjects with afebrile seizures have complex-partial seizures with epigastric auras followed by loss of consciousness, behavioral arrest, early oroalimentary automatisms, and slow postictal reorientation. Patient A-III-19 also has secondarily generalized seizures. The clinical diagnosis of TLE was made in both subjects based on this seizure semiology. Subjects A-III-8 and A-IV-6 had an immediate language regression following a prolonged convulsion. They both improved over more than 5 years, but still have some impairment in word finding and vocabulary. Subject B-II-5 had delayed development after a febrile convulsion. He learned to walk without assistance at age 5 and to speak in complete sentences at age 6.

None of the investigated subjects reported any incidents during pregnancy or birth. There was no evidence of encephalitis, meningitis, or significant head trauma. Subject A-III-20 had one occasional grand mal seizure without any focal signs following alcohol excess at age 22. Subjects A-III-7 and A-III-14 have migraine without aura that began at age 13 and age 17, respectively. Subject A-I-2 was not investigated in this study because he was killed during the Second World War. He was reported to have had febrile convulsions during early childhood and epileptic seizures during adulthood, but no further details were available. A-III-1 died within 24 hours of birth due to unknown cause. A-III-6 died in a drowning accident at age 3.

Normal controls. Twenty-three age- and sex-matched controls(mean age, 32 years; range, 14 to 70) were investigated. Their neurologic and psychiatric status was normal. None of their first-degree relatives have a neurologic or psychiatric disorder. Twenty-one controls were right-handed and two were left-handed. Each participant, study subjects as well as controls, gave informed, written consent for the study.

Magnetic resonance imaging. Data acquisition. All MRIs were obtained using a 1.5-T Siemens Magnetom Vision imager with a circular polarizing head coil (Siemens Medizintechnik, Erlangen, Germany). T2-weighted turbo-spin-echo sequences were obtained in the sagittal plane(repetition time [TR] = 5400 millisecond; echo time [TE] = 99 millisecond; flip angle = 90°; section thickness = 6 mm; intersection gap = 0.6 mm; field of view = 25 cm). Further T2-weighted turbo-spin-echo sequences were acquired in adjusted coronal and axial planes (5400/99/90/3/0.0/25). These adjusted planes were perpendicular and parallel to the longitudinal axis of the hippocampal body. Coronal FLAIR images were also adjusted to this plane(9999/105/180/3/0.3/25; inversion time [TI]= 2400 millisecond). T1-weighted spin-echo sequences were obtained in a conventional axial plane (462/12/90/6/0.6/25). The matrix size was 256× 256 in every sequence. Hard copy images of all scans were obtained for visual assessment. The adjusted coronal T2 images were transferred to a workstation for quantitative assessment (Advantage Windows 2.0 Analysis Workstation, General Electric, Milwaukee, WI). Each imaging data set and hard copy were coded by a four-digit code for masking purposes. Data of study subjects and controls were randomly intermixed in all further analyses.

Visual assessment. To assess hippocampal signal, internal pattern, and outer shape, hard copies of the T2 and FLAIR images were independently assessed by two raters using a detailed checklist. The raters were masked to other subject data. In T2 and FLAIR images, the hippocampal signal was compared with the signal of the contralateral hippocampus and the neocortex. In addition to diffuse signal changes, raters paid special attention to circumscribed signal increases. To estimate the integrity of the internal hippocampal structure, the MRI pattern of rolled-up neuronal layers was compared between both hippocampi. If this pattern was not identifiable in at least three contiguous MRI slices, architectural disturbances were present. To detect an abnormal outer shape of the hippocampus, the ratio of height and width in each coronal image was visually compared with the ratio of the contralateral hippocampus.

Quantitative assessment. The coronal T2 images were analyzed using Advantage Windows 2.0 analysis software. Right and left HV and their right/left ratio (RHV) were determined. For these measurements, the coronal areas of all hippocampal slices were assessed. The volume was calculated as the sum of the computed areas multiplied by the thickness of the slices (3 mm). To obtain anatomic consensus, hippocampal boundaries were initially determined in each subject by both raters without tracing according to the criteria of Jack et al.^26,35 Afterwards, each rater independently outlined the boundaries using a mouse-controlled cursor. The coronal section on which the crus of the fornix could be seen in full profile was used to define the posterior boundary. In the body and tail regions, the boundaries were defined by CSF and the gray-white matter junction. Anteriorly, the entire hippocampal head was included. If no distinction between the head and the overlying amygdala was possible, a straight horizontal line drawn between the midportion of the gyrus ambiens medially and the most superomedial portion of the temporal horn laterally was defined as the superior hippocampal boundary. The raters were masked to other subject data. The correlation coefficient between the corresponding values of each rater was calculated to measure the interrater reproducibility. Mean results of both raters were used for further analysis. For individual identification of subjects with hippocampal asymmetry, the right/left ratios were compared with the mean ratio of normal controls using three times the standard deviation (SD) as the critical limit.

Results. Accuracy and normal values of MRI findings. In normal controls, the mean HV was 2932 ± 377 mm³ for the left and 2945 ± 376 mm³ for the right hippocampus. Therefore, the mean RHV was 1.00 ± 0.01 (figure 2). The interrater reproducibility coefficients were 0.9675 for left HV and 0.9570 for right HV. Initially, three disagreements regarding the visual findings occurred between raters. The presence of a blurred internal pattern of the left hippocampus in one subject with febrile convulsions (A-IV-6) and in one control subject and a right hippocampal signal increase detected in the FLAIR images of one control subject were initially discrepant between raters. These discrepancies were solved by mutual re-evaluation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Left hippocampal volumes as a function of right hippocampal volumes. The oblique line represents symmetric measures. TLE = subjects with febrile convulsions and subsequent temporal lobe epilepsy; FC = subjects with febrile convulsions; no FC = subjects without febrile convulsions.

Patients with temporal lobe epilepsy. Both subjects who subsequently developed TLE showed significant asymmetries of HVs (RHV = 1.534 and 2.339) (figure 2;table 2). In Subject B-II-7, visual inspection of the left hippocampus confirmed the characteristics of severe left hippocampal sclerosis with no identifiably internal pattern and increased T2 and FLAIR signal intensities on nearly all coronal slices. The other subject with TLE(A-III-19) had more circumscribed left HS, with T2 and FLAIR signal increase most prominent in the lateral portion of the hippocampal body.

Subjects with febrile convulsions. All subjects who had febrile convulsions but did not develop epilepsy also had significantly asymmetric hippocampal volumes (figure 2^,table 2): the left hippocampus was consistently smaller than the right. This asymmetry of hippocampal volumes had a range of RHV from 1.099 to 1.233 (mean RHV, 1.168). In Family A, four of the smaller hippocampi had a blurred internal pattern when compared with the other side(table 2). Three of these hippocampi with a blurred internal pattern additionally showed a flat body. Three other members of Family A with febrile convulsions had only an abnormal flat body of the smaller left hippocampus. This flatness was most obvious at the lateral portion of the hippocampal body. Members of Family B showed left hippocampi with smaller heads. This produced a shortening of the affected hippocampi that ranged between 1 and 3 coronal slices (3 to 9 mm in length). In all subjects with febrile convulsions, the hippocampal signal intensities (T2, FLAIR) did not show any asymmetry or difference between them and the signal of the neocortex. Additionally, the MRI of Subject A-IV-7 showed bilateral subcortical heterotopias in the parietal and occipital lobes(figure 3).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3. MRI shows additional subcortical heterotopias (Subject A-IV-7). The left side of the brain is on the right side in the figure.

Unaffected relatives. Four of the unaffected subjects of Family A (A-III-10, A-III-14, A-III-16, A-III-20) had symmetric hippocampal volumes(mean RHV, 1.00; range, 0.996 to 1.009) and symmetric hippocampal pattern (figure 2, table 2). The six remaining unaffected relatives of both families (A-III-4, A-III-15, A-III-17, A-III-18, A-III-21, B-II-9) had significant asymmetric hippocampal volumes (mean RHV, 1.16; range, 1.081 to 1.314) and, similar to the subjects affected by febrile convulsions, asymmetries of internal pattern and outer shape (figures 2 and 4;table 2). The left hippocampus was always smaller than the right. No hippocampal signal abnormalities were detected. Using this data, it was possible to distinguish two groups of clinically unaffected subjects: those with abnormal left hippocampi and those with symmetric hippocampi. Additionally, there was no reliable difference between the hippocampal MRI findings of subjects with febrile convulsions alone and those with abnormal left hippocampi who did not have febrile convulsions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4. MRI shows a smaller left hippocampus with blurred internal pattern in a subject who has never had a seizure(A-III-17). The left side of the brain is on the right side in the figure.

Discussion. In this study, we tested the hypothesis that a structural alteration of the hippocampus is the primary phenotypic manifestation in familial febrile convulsions with subsequent TLE. We found that the left hippocampi of all subjects with febrile convulsions who did not develop TLE and six of their clinically unaffected relatives had significantly smaller volumes when compared with the right side. Visual inspection revealed normal signal intensities in the smaller left hippocampi but two types of abnormal structural pattern. The members of Family A had blurred internal structures, flat hippocampal bodies, or both; the members of Family B had smaller hippocampal heads. MRI did not reveal any reliable differences between subjects who had febrile convulsions alone and those who had asymmetric hippocampal volumes without febrile convulsions. The two subjects with TLE had left HS with considerably different MRI patterns than the abnormal left hippocampi in subjects who did not develop TLE. The patterns of HS, however, showed a correlation with the findings in their relatives. Subject A-III-19 exhibited a circumscribed left HS in the lateral portion of the hippocampal body, which is the same area predominantly affected by a volume deficit in his relatives. Subject B-II-7 had a HS of the entire left hippocampus; his relatives had smaller hippocampal heads.

The high interrater reproducibility of volumetric measurements indicates that this quantification was accurate. Additionally, absolute HVs of normal controls correspond with previous findings, especially with findings in T2-weighted images.^{19,21-24,30,36-38} In these studies, there was no agreement whether HVs in healthy subjects are symmetric or slightly asymmetric toward the left or right side. Jack et al.³⁵ explained these small discrepancies of relative HVs as due to different boundary criteria and study populations regarding race, age, and gender, which mostly affect the results in studies with few control subjects. Our controls were matched for age and sex with the investigated family members. We found, with high interrater reproducibility and a small standard deviation, symmetric HVs in normal controls as previously found in a larger series that also used T2-weighted, adjusted coronal images and a similar protocol of analysis.²² However, the measured asymmetries in affected subjects would have been reliably asymmetric even in comparison with less symmetric volumes of normal controls.

The low rate of initial disagreement of visually assessed findings indicates that the presence of hippocampal signal asymmetries and blurred internal pattern was clear in the acquired images. As revealed in surgical studies with histopathologic validation, combined visual assessment of optimized T2 and FLAIR images, as used here, reaches a sensitivity of more than 95% for the detection of a signal increase caused by HS.^20,28,33,34 By combining volumetric and visual MRI assessments, we were able to differentiate the MRI findings of hippocampal sclerosis, hippocampal malformation, and normal medial temporal lobe structures.

The observation that the hippocampal malformation found here is inherited and causally related to febrile convulsions is supported, because the hippocampal asymmetries were present in every subject with febrile convulsions and the affected side and abnormal hippocampal pattern were uniform within each family. The presence of this malformation in six clinically unaffected subjects and the lack of a difference between the findings in clinically unaffected subjects and subjects with febrile convulsions suggests that this hippocampal malformation is a necessary condition and not a consequence of familial febrile convulsions. This data supports our hypothesis that a pre-existing hippocampal malformation facilitates familial febrile convulsions. In this study, we were not able to determine whether the HS in TLE patients is the result of repeated febrile convulsions during early childhood. However, the following findings are consistent with a causal relation between a hippocampal malformation and HS:(1) the HS also occurred on the left side only; (2) the patterns of HS had a structural correlation with the findings in the relatives; and (3) the TLE subjects had more risk factors for subsequent epilepsy (>25 febrile convulsions, >1 febrile convulsion within 24 hours) than their relatives.

All hippocampal malformations and HS that we found were on the left side. Left temporal resections, as a treatment for drug-resistant TLE, are slightly more frequent than right-sided resections in most surgical series, although they are associated with a higher risk of postoperative memory decline.^39-42 Taylor and Ounsted⁴³ explained this predominance of left hemispheric TLE as due to differences of cerebral maturation with different vulnerable phases and gender interactions. To our knowledge, no study has addressed this question specifically, and although the side of surgery is rarely reported differentially for TLE patients with and without febrile convulsions, several reports indicate that left-sided TLE is much more frequent in patients with antecedent febrile convulsions. Cendes et al.² reported a series of TLE patients with febrile convulsions and severe HS, 73% of whom showed left HS. Marsh et al.⁴⁴ reported a series of consecutively recruited men with drug-resistant TLE. In this series, 57% of the patients with left-sided TLE had antecedent febrile convulsions, as compared with 14% of the patients with right-sided TLE. Although HS in TLE patients with antecedent febrile convulsions is generally more severe than in TLE patients without febrile convulsions, Trenerry et al.⁵ found this highly significant correlation between hippocampal volume loss and antecedent febrile convulsions only in left TLE patients. Finally, Wallace⁴⁵ described a predominance of right-sided symptoms among children with lateralized febrile convulsions. Given these reports, our findings suggest an explanation for the left hemispheric predominance of TLE: an inherited malformation that occurs perhaps exclusively in the left hemisphere and facilitates febrile convulsions and subsequent HS with TLE.

With MRI, we were unable to determine the underlying histopathology, but the characteristics of our MRI findings are compatible with impaired migration of hippocampal neurons.^30,46,47 Migrational disturbances lead to higher seizure susceptibility, and there are several hereditary clinical syndromes with impaired neuronal migration.^48-52 Additionally, there is an animal model with inherited disturbed migration of hippocampal neurons.⁵³ In this model, rats suffer from spontaneous limbic-like seizures and, following repeated seizures, their hippocampi develop characteristics of HS. Furthermore, disturbance of hippocampal neuronal migration in immature rats lowers the threshold to hyperthermia-induced behavioral seizures and neuronal damage within the hippocampus.⁵⁴ The development of HS in these studies contrasts with the findings of several studies of Sperber et al. (for a review, see ^{reference 55}). They found no HS in primarily healthy rats after repeatedly induced seizures during early life. Thus, a pre-existing malformation such as a circumscribed neuronal migration disorder within the hippocampus may be necessary for the development of HS following seizures during early life. Further support for the hypothesis that the hippocampal malformation found in our subjects is a neuronal migration disturbance comes from clinical findings. In patients with focal epilepsy, HS frequently coexists with neocortical migration disorders, but rarely with other congenital lesions.^56-60 Additionally, epileptic patients with the dual pathology of HS and neocortical migration disorders often have a history of febrile convulsions and a family history of epilepsy as well as disturbed neocortical migration.^59,61 These clinical and experimental findings regarding the association among cortical migration disorders, seizure susceptibility, and HS development lend plausibility to our speculation that the hippocampal malformation is a neuronal migration disturbance; the coexistence of additional subcortical heterotopias in Subject A-IV-6 fits with this hypothesis.

Although most familial febrile convulsions probably have a polygenic mode of inheritance, the investigated families with a close to dominant mode of inheritance may be useful for mapping febrile convulsion susceptibility genes. If seizure history were used as the only basis for a diagnosis, six pedigree members would have not been classified as affected in our two families. If the hippocampal malformation predisposes to febrile convulsions, it is "closer" to the gene action than the febrile convulsions. Therefore, families such as those we have studied should be suitable for a successful linkage study. So far, previous attempts to map febrile convulsion genes have preliminarily assigned two loci to different regions of the genome(chromosomes 8q13-21¹⁸ and 19p¹⁵), pointing toward genetic heterogeneity. These studies, however, did not report MRI results.

In conclusion, our results suggest that this is an inherited hippocampal malformation that facilitates febrile convulsions and is not the consequence of febrile convulsions. Furthermore, this malformation may act as a crystallization point for the development of HS in TLE patients with antecedent febrile convulsions. The phenotypical distribution observed in our two families suggests a close to dominant mode of inheritance. This proposal of an inherited hippocampal malformation associated with febrile convulsions and subsequent HS and TLE needs further confirmation by histopathologic and genetic studies. Our findings reinforce the usefulness of MRI in identifying subjects with this phenotypic character for genetic analysis.

Acknowledgments

We thank L. Marsh, M. Reuber, and H. Tiemeier for detailed and thoughtful comments on earlier versions of this article. We thank the patients, their relatives, and the control subjects, without whose help and cooperation this study would not have been possible.