Altered language processing in autosomal dominant partial epilepsy with auditory features

Abstract

Background

Autosomal dominant partial epilepsy with auditory features (ADPEAF) is an idiopathic focal epilepsy syndrome with auditory symptoms or receptive aphasia as major ictal manifestations, frequently associated with mutations in the leucine-rich, glioma inactivated 1 (LGI1) gene. Although affected subjects do not have structural abnormalities detected on routine MRI, a lateral temporal malformation was identified through high resolution MRI in one family. We at-tempted to replicate this finding and to assess auditory and language processing in ADPEAF using fMRI and magnetoencephalography (MEG).

Methods

We studied 17 subjects (10 affected mutation carriers, 3 unaffected carriers, 4 noncarriers) in 7 ADPEAF families, each of which had a different LGI1 mutation. Subjects underwent high-resolution structural MRI, fMRI with an auditory description decision task (ADDT) and a tone discrimination task, and MEG. A control group comprising 26 volunteers was also included.

Results

We found no evidence of structural abnormalities in any of the 17 subjects. On fMRI with ADDT, subjects with epilepsy had significantly less activation than controls. On MEG with auditory stimuli, peak 2 auditory evoked field latency was significantly delayed in affected individuals compared to controls.

Conclusions

These findings do not support the previous report of a lateral temporal malformation in autosomal dominant partial epilepsy with auditory features (ADPEAF). However, our fMRI and magnetoencephalography data suggest that individuals with ADPEAF have functional impairment in language processing.

Autosomal dominant partial epilepsy with auditory features (ADPEAF) is an idiopathic focal epilepsy syndrome with auditory symptoms or receptive aphasia as major ictal manifestations.^1–6 The most common auditory symptoms are simple unformed sounds such as humming, buzzing, or ringing; less common forms are distortions (e.g., volume changes) or complex sounds (e.g., specific songs or voices). Ictal receptive aphasia consisting of a sudden inability to understand language in the absence of general confusion has occurred prominently in some families.^7–10 The prominent auditory symptoms and aphasia strongly suggest localization of the epileptogenic zone to the lateral temporal lobe. Accordingly, the syndrome is also called autosomal dominant lateral temporal epilepsy.³

Mutations in the leucine-rich, glioma inactivated 1 (LGI1) gene have been found in approximately half of families with this syndrome,^4,5,11–14 but not in other forms of familial temporal lobe epilepsy.^14,15 More than 20 disease-causing mutations have been reported,¹⁶ almost all of which have been unique to an individual family. Germline mutations in LGI1 are seldom found in individuals with symptoms consistent with ADPEAF who do not have a family history.^17,18 Two de novo mutations have been identified, among about 77 isolated cases screened (2.6%).^19,20

The mechanism by which LGI1 mutations cause epilepsy remains poorly understood. Initially the gene was hypothesized to influence risk through a mechanism related to CNS development. ¹¹ Mutations found in ADPEAF families lead to defects in secretion of the protein product, suggesting that this is an important part of the pathogenic mechanism.^21,22 Other recent findings have suggested either a potassium channel²³ or glutaminergic^24,25 mechanism.

In families with LGI1 mutations, affected individuals do not have mesial temporal sclerosis or other structural brain abnormalities detected on routine MRI. However, in one family with a mutation, a lateral temporal malformation was reported involving enlargement of the temporal lobes and protrusion of brain parenchyma with an encephalocele-like appearance.²⁶ Another recent study identified a region of increased fractional anisotropy in the left temporal lobe through diffusion tensor imaging of patients with LGI1 mutations. ²⁷ These results support the idea that LGI1 might influence risk through a neurodevelopmental mechanism leading to subtle malformations or cortical disorganization.

The findings of lateral temporal abnormalities and the characteristic symptoms of the disorder suggest potential deficits in auditory and language processing. Abnormal phonologic processing was demonstrated using a fused dichotic listening task in four persons in a Sardinian family. ²⁸ Asymmetry of long-latency auditory evoked potentials (with reduced left N1-P2 amplitudes) was found in a Norwegian family with prominent aphasic seizures.¹⁰ In this study, we investigated auditory and language processing in ADPEAF further, using fMRI and magnetoencephalography (MEG).

METHODS

We invited previous research participants in genetic studies of ADPEAF to travel to the NIH Clinical Epilepsy Center for additional testing. Although we studied families with and without mutations in LGI1, here we describe the results only for families with mutations. Seventeen subjects in seven families with mutations were tested, of whom 10 were mutation carriers with epilepsy, 3 were clinically unaffected mutation carriers, and 4 were unaffected noncarriers (table 1). Each of the families had a different mutation in LGI1. The methods for clinical data collection and the clinical features of epilepsy in the families have been described in detail previously.^2,5,6,11 A control group comprising 26 healthy volunteers recruited by advertisement was also included (9 women; mean age 36.2 ± 1.9 SEM). Prior to enrollment, controls were screened by the NIH Clinical Epilepsy Section with routine physical and neurologic examinations. Each control participant was also given an EEG and a neuropsychological test battery with special emphasis on language measures.

Table 1.

Clinical characteristics of subjects in ADPEAF families

Subject ID (family-individual)	Status	Age at seizureonset, y	Current AEDs	EEG
I–1	LRE	28	PHT	Normal
I–2	IGE	10	TPM	Normal (occasional small sharp spikes)
I–3	Unaffected			Normal
I–4	LRE	6	PHT	Normal
I–5	Unaffected			Normal
I–6	Unaffected			Normal
II–1	LRE	10	CBZ	Normal
II–2	LRE	12	PB, TPM, LTG	Normal
III–1	LRE	20	CBZ	Mild diffuse slow activity
IV–1	LRE	10	CBZ	Normal
V–1	Both IGE and LRE	11	LEV, PGB	4–5 Hz diffuse spikes with frontal predominance
V–2	Both IGE and LRE	13	PHT	4–5 Hz diffuse spikes with frontal predominance
V–3	Unaffected			Normal
VI:1	LRE	12	None	Normal
VI–2	Unaffected			Normal
VII–1	Unaffected			Normal
VII–2	Unaffected			Normal

All affected subjects and three unaffected subjects carried a mutation in LGI1.

ADPEAF = autosomal dominant partial epilepsy with auditory features; AED = antiepileptic drug; LRE = localization-related epilepsy; PHT = phenytoin; IGE = idiopathic generalized epilepsy; TPM = topiramate; CBZ = carbamazepine; PB = phenobarbital; LTG = lamotrigine; PGB = pregabalin.

The study was approved by the National Institute of Neurological Disorders and Stroke and Columbia University Medical Center Institutional Review Boards, and all participants gave informed consent.

Structural and functional MRI

All subjects in ADPEAF families and controls underwent high-resolution structural MRI on a Philips 3-tesla scanner, including three-dimensional magnetization prepared rapid gradient echo (MP-RAGE), susceptibility-weighted, fluid-attenuated inversion recovery (FLAIR), T1- and T2-weighted sequences. In addition to standard clinical interpretation, MRI scans were reviewed by two neuroradiologists (E.H.B. and J.A.B.). For visual evaluation of the temporal lobes, symmetric coronal images were generated by first realigning the three-dimensional MP-RAGE sequences with the Oxford Centre for Functional MRI of the Brain’s linear image registration tool (FLIRT)²⁹ (so that the midsagittal plane was vertical), and then resectioning these data in the coronal plane. Volumetric measurement of the temporal lobes was performed using previously reported methods.³⁰

Eight affected family members, six unaffected family members (two mutation carriers, four noncarriers), and 20 unrelated controls (13 women; mean age 27.4 ± 2.1 SEM) had fMRI using a GE Signa 3T scanner, with a gradient echoplanar sequence. fMRI was not obtained in subjects I-5, III-1, or IV-1 (table 1). Two of the eight affected subjects and one unaffected mutation carrier were left-handed; the remaining family members and controls were right-handed. Each subject had three-dimensional fast spoiled gradient echo recalled (SPGR) anatomic images obtained at the same time as fMRI. Monitored activation paradigms included an auditory description decision task (ADDT), during which subjects heard true and false word definitions (e.g., “A small red fruit is a banana”) and pressed a button when the statement was true.³¹ The control condition for this task was reverse speech; subjects pressed a button on hearing an interspersed beep following a reverse speech cue. Both the control and experimental conditions were 30 seconds in duration and alternated for 5 minutes; each contained 70% true responses and 30% foils. We also performed a musical tone discrimination task in which tones were played in ascending, descending, or random order. The participant was instructed to press the button when the tones were played in order and not to press the button when tones were not in order. The control condition for this task was reverse speech during which beeps were played and subjects were instructed to press a button when they heard a beep. An additional control condition consisted of the visual presentation of blocks. Subjects were instructed to press the button when a white square was presented. Images were realigned, normalized, smoothed to 8 mm, and analyzed with SPM2 (Wellcome Trust Centre for Neuroimaging) using a random effects analysis. Results were examined initially at p < 0.05 corrected for multiple comparisons, and subsequently at p < 0.001 uncorrected.

Magnetoencephalography

All family members and five unrelated controls underwent a resting supine MEG-EEG recording of spontaneous cerebral activity lasting a minimum of 1 hour using a 275-channel CTF (Coquitlam, BC, Canada) whole-head system and 10–20 electrode EEG configuration. In addition, in a subsequent segment of the same recording session, 1 kHz pure tones with a duration of 50 msec and with 5 msec rise and fall times were presented binaurally for 10 minutes at 75 dB sound-pressure level (SPL). The stimuli were delivered via pneumatic tubes, and interstimulus intervals were randomized between 1 and 2 seconds. Spontaneous MEG-EEG data were analyzed visually by an experienced MEG-EEG expert. Equivalent current dipole (ECD) method was used for source localization of identified interictal epileptiform discharges (IIED). A VSM-CTF Toolbox was utilized for analyzing latencies, amplitudes, and dipole fits of at least 100 averaged auditory evoked fields (AEFs) for each participant. High pass filter was set at 1 Hz and low pass filter at 70 Hz. The butterfly displays were analyzed assuming that peak 2 is the most prominent (N100m). If a subject had more than two peaks preceding peak 2, an adjacent peak was considered peak 1, and additional peaks were not included in this analysis. The same analogy was used for more than two peaks following peak 2. The t test for independent samples was used for statistical analysis.

All findings were analyzed blinded to clinical and genetic data.

RESULTS

Clinical and neuropsychological findings

Subjects in ADPEAF families ranged from 20 to 81 years in age (average 48.3 ± 4.6 SEM) and from 13 to 21 years in education; 11 were men and 6 women. The three groups of family members did not differ by age or sex (table 2), but did differ by education, with more years of education in the four noncarriers than in the other two groups (p = 0.046 by analysis of variance).

Table 2.

Demographic information and neuropsychological test results in subjects from ADPEAF families^*

	Affected carriers (n = 10)	Unaffected carriers (n = 3)	Unaffected noncarriers (n = 4)
Sex,%women	40	33	25
Age, y	44.9 ± 6.4	53.0 ± 16.8	53.2 ± 4.4
Education, y†	15.7 ± 0.8	13.7 ± 0.3	18.3 ± 1.1
Verbal IQ	108.0 ± 3.1	119.7 ± 7.2	123.2 ± 5.2
Performance IQ	108.2 ± 4.0	106.7 ± 10.3	125.5 ± 6.9

^*Mean ± SEM for each measure except sex.

^†p < 0.05 For comparison of three groups by analysis of variance.

ADPEAF = autosomal dominant partial epilepsy with auditory features.

All 17 family members had IQ scores within the normal range (verbal IQ range 90–135, performance IQ 85–140). The four noncarriers had higher IQ scores than either the affected subjects or unaffected carriers. Verbal IQ was lower in affected subjects than in both groups of unaffected subjects combined (p = 0.021) but performance IQ did not differ significantly between affected and unaffected subjects.

No subject had an abnormality on routine physical or neurologic examination. Interictal EEG was normal in 12 subjects. One affected family member had mild diffuse slow activity, and two brothers (V-1 and V-2, table 1) had diffuse 4–5 Hz spikes with bifrontal predominance. As described previously,⁵ these two brothers had both primary generalized and focal seizures (table 1).

Structural MRI

We found no evidence of mesial temporal sclerosis (increased signal intensity on T2 or FLAIR sequences), malformations of cortical development, or abnormal gyral patterns in any of the 17 subjects. No vascular anomalies were identified on susceptibility-weighted images.

On visual examination, the architecture of the temporal lobes was within normal variation for all subjects; none had the appearance of an encephalocele such as that described previously.²⁶ Ten AD-PEAF family members (all aged 45 or older) had nonspecific focal or confluent T2-hyperintense signal abnormalities in their white matter that were more numerous than expected for their age. These abnormalities were not associated with disease or mutation status (affected mutation carriers 6/10, unaffected carriers 1/3, noncarriers 3/4). Five of the 10 had clinical evidence of hypertension, which can produce abnormalities in this pattern. Two of these hypertensive subjects had diffuse cerebral atrophy and confluent white matter signal abnormalities. Temporal lobe volumes did not differ significantly between the affected subjects (left 75,818 ± 11,744 SD, right 79,451 ± 11,128 SD) and controls (left 75,018 ± 9,506 SD, right 77,383 ± 8,561 SD).

fMRI

There were too few subjects in the ADPEAF families to create group maps at p < 0.05 corrected for multiple comparisons; hence in initial analyses we created group maps for the controls and visually compared individual ADPEAF family members to the control group maps. On the tone discrimination task, the controls showed bilateral frontotemporal activation, and the activation patterns were similar to those in controls in all three subgroups of family members. On the ADDT, the controls showed predominantly left temporal and inferior frontal activation (figure 1A). On visual inspection, the activation patterns differed from those in controls in seven of the eight tested subjects with epilepsy. Left temporal activation was clearly reduced in four subjects and marginally reduced in a fifth (example in figure 1C). Three subjects, including one of those with reduced left activation, had bilateral activation rather than the left temporal activation pattern observed in controls (example in figure 1B). In the remaining affected subject and all asymptomatic carriers and noncarriers, the activation pattern was similar to that in controls, and none of the controls had an activation pattern similar to that illustrated in figure 1, B or C. None of the affected subjects with reduced or bilateral activation was left handed.

Figure 1. fMRI scan results for auditory description decision task.

(A) Controls group map (p < 0.05 family-wise error, corrected) showing predominant left frontotemporal activation. (B) Affected subject V-2 (p < 0.05 family-wise error, corrected) showing bilateral (but not reduced) activation with right predominance, reversing the normal pattern. (C) Affected subject I-4 (p= 0.001 family-wise error, corrected) showing reduced activation.

We also carried out statistical parametric analysis to compare the affected subjects to the controls, using maps at p < 0.001 without correction because of the small number of affected individuals. We compared the eight affected subjects to eight controls matched for sex and age (within 10 years) (figure 2). (Using all controls would have unfairly weighted the results against affected subjects.) Affected subjects had significantly lower activation than controls in left middle temporal, post central, parahippocampal, and inferior occipital gyri.

Figure 2. Group maps in SPM2 for eight patients and eight age-and sex-matched controls (p < 0.001, uncorrected).

There is significantly less activation for patients than controls in several left hemisphere regions.

It is unlikely that antiepileptic drugs (AEDs) explained the results, because the affected individuals with reduced or bilateral activation varied in their medication regimens. Among these seven subjects, two were taking topiramate (one in combination with phenobarbital), four were taking other medications, and one was on no medications. The patient taking both topiramate and phenobarbital had a normal activation pattern with only marginally reduced activation.

MEG

On MEG with auditory stimuli, peak 2 AEF latency was significantly delayed in affected individuals compared to controls (mean msec 110.0 ± 2.3 SEM vs 102 ± 2.6 SEM; p < 0.05) (table 3). Unaffected family members, both carriers and noncarriers, were similar to controls in peak 2 (table 3). Peak 4 was only identified in a minority of subjects. Four participants had a small peak preceding peak 1, and two had a small peak following peak 4 that was not taken into account. The spatio-temporal dipole analysis of AEFs evoked by binaural stimulation showed similar AEF source location in all groups. There was no effect of age (F-ratio 0.242; r² = 0.013) or sex (mean peak 2 men: 105 ± 8 SEM; women: 106 ± 8 SEM) on latencies. Information on IQ was not available for the controls; however, verbal IQ was unrelated to peak latency among the ADPEAF family members (F-ratio 1.45; r² = 0.108).

Table 3.

Magnetoencephalography auditory evoked field latency (mean ± SEsM)

	Affected carriers (n = 10)	Unaffected carriers (n = 3)	Unaffected noncarriers (n = 4)	Controls (n = 5)
Peak 1	60.9 ± 3.5	65.0 ± 3.3	61.7 ± 4.8	55.3 ± 4.8
Peak 2	110.0 ± 2.3	101.7 ± 4.7	99.5 ± 5.5	102.0 ± 2.6

DISCUSSION

We found no clear evidence of structural abnormalities such as the left temporal lobe malformation reported previously in a family with a mutation in LGI1.²⁶ In the previous report, 10 members of a large Brazilian family were found to have dysgenic features.²⁶ However, these features were not consistently associated with carrying the mutation—they were found in 8/15 mutation carriers with epilepsy, 0/4 carriers without epilepsy, and 2/3 non-carriers. The authors also found reduced hippocampal volumes in four family members (two mutation carriers with epilepsy, one mutation carrier without epilepsy, and one noncarrier). The lack of a consistent association of these features with the mutation suggests that they are unrelated to LGI1, although they could be related to another unidentified genetic variant in the family.

Although we found no evidence for structural abnormalities, we did find evidence for functional deficits in patients with ADPEAF. Our fMRI and MEG data suggest that individuals with ADPEAF have disturbed functional anatomy for language processing. Functional MRI and MEG showed altered activation patterns in affected subjects. The N100m is considered an early response generated from primary auditory cortex.^32,33 Thus our finding of significantly prolonged N100m latency in individuals with ADPEAF suggests that primary auditory cortex and primary association areas within the temporal lobe may be affected by or even involved in the underlying circuitry that generates the ictal auditory phenomenology in this disorder.

These results parallel the study that found a significant reduction in N1-P2 left hemisphere auditory evoked potential amplitudes in patients from an AD-PEAF family with predominantly aphasic seizures and left-sided EEG abnormalities.¹⁰ Dichotic listening, fluency, and lexical abilities were also abnormal in patients with an LGI1 mutation.²⁸ Taken together, these findings suggest functional impairment in auditory and language processing.

We attempted to disentangle the effects of LGI1 mutations from the effects of epilepsy and its treatment by examining unaffected mutation carriers separately from unaffected noncarriers, but the small number of unaffected subjects in each group limited our ability to do this. Since the asymptomatic gene carriers we tested did not show any abnormalities on fMRI or MEG, we cannot prove that the altered activation patterns we observed in affected individuals are directly related to mutation status. We might have missed these effects by chance in the small group of unaffected carriers. Alternatively, functional deficits might occur only in individuals with both an LGI1 mutation and some other unidentified factor that interacts with the mutation to produce disease. Finally, the effects of AEDs and seizures cannot be definitively excluded, although we believe an effect of AEDs is unlikely.

The four noncarriers were generally more educated and had higher IQs than other subjects, possibly reflecting selection bias among highly motivated family members. The same selection bias would be expected in the unaffected carriers, and might have explained the lower verbal IQ in affected subjects than in the two groups of unaffected subjects combined. However, this difference in verbal IQ did not explain our MEG findings, since peak latency was not associated with verbal IQ. It was also unlikely to have explained our fMRI findings because we used relatively simple processing tasks and the verbal IQ of the affected subjects was within the normal range. A previous study showed that left-right activation asymmetry on a language task was not related to IQ in patients with epilepsy.³⁴

Although two patients in our study were taking topiramate, this is unlikely to explain our results. Topiramate has been shown to be associated with language impairment suggestive of frontal dysfunction, particularly word fluency,^35–37 although this has not been found in all studies.³⁸ In fMRI using covert word generation, patients taking topiramate had significantly less activation in the language-mediating regions of the prefrontal cortex.³⁹ Our semantic description task, however, relies heavily on temporal rather than frontal lobe activation. Moreover, in the previously reported topiramate study,³⁹ tone discrimination was impaired as well, which we did not observe.

Three of the subjects tested in this study had idiopathic generalized epilepsy (IGE), as we have described previously.⁵ Our current results confirm these findings in that two of them had generalized EEG discharges on retesting. The third subject with IGE who was included here had only nocturnal seizures, and was previously classified as IGE based on EEG evidence alone. On retesting, the EEG was normal with occasional small sharp spikes (table 1). As we have suggested previously, the occurrence of generalized epilepsies in families with LGI1 mutations could be explained by two genotypes (LGI1 and an unidentified IGE genotype) segregating within the same family, or by an effect of LGI1 mutations on generalized epilepsy.

Finally, we note that our structural MRI results agree with a recent report showing no abnormalities on standard interpretation at 1.5 T in eight patients. ²⁷ On voxel-based morphometry of diffusion tensor imaging data, a cluster of 46 voxels with increased fractional anisotropy was found in the left mid-temporal gyrus. This finding might reflect a very subtle developmental abnormality that could underlie the functional disturbances we report.

GLOSSARY

ADDT

auditory description decision task

ADPEAF

autosomal dominant partial epilepsy with auditory features

AED

antiepileptic drug

AEF

auditory evoked field

ECD

equivalent current dipole

FLAIR

fluid-attenuated inversion recovery

IGE

idiopathic generalized epilepsy

IIED

identified interictal epileptiform discharges

MEG

magnetoencephalography

MP-RAGE

magnetization prepared rapid gradient echo

SPGR

spoiled gradient echo recalled

SPL

sound-pressure level