Precision therapy for a new disorder of AMPA receptor recycling due to mutations in ATAD1

Identification of recessive loss-of-function ATAD1 mutations.

To obtain a diagnosis in a large consanguineous Kuwaiti family with multiple neonates affected with hypertonia, seizures, respiratory failure, and death (figure 1, supplemental results), we used clinical exome sequencing and single nucleotide polymorphism (SNP)-based homozygosity analysis. The proband (individual IV-6, figure 1) was symptomatic at birth and presented to our center at age 9 months. He was nonresponsive to tactile, visual, and auditory stimuli with no spontaneous movement and extreme hypertonia. EEG at 9 months showed hypsarrhythmia, a chaotic interictal pattern suggestive of severe brain dysfunction. He did not demonstrate infantile spasms as his extreme hypertonia precluded almost all movement. Brain MRI at 2 months of life was normal, but repeat at 9 months showed marked interval reduction in cerebral volume. Prior investigations (table 1) were unrevealing.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1 Subjects' pedigree demonstrates multiple affected individuals in a highly consanguineous family

The proband is patient IV-6; his cousin who was subsequently evaluated and treated is patient IV-3. The ATAD1 variant, p.E276X, was identified by whole-exome sequencing and tracked with disease in all available family members (+ represents the normal allele and − represents the p.E276X allele). See also figure e-1.

The second patient was the proband's 2-month old male cousin (individual IV-3, figure 1). He also had a history of hypertonia and clinical seizures at birth. On first evaluation at our center at 2.5 months of age, he had little spontaneous movement and was extremely stiff (video 1). EEG at presentation while on phenobarbital demonstrated a background lacking normal mixture of frequencies and poorly formed sharps but no seizures. Brain MRI was unremarkable.

Exome sequencing of patient IV-6 revealed no mutations in known disease-causing genes. However, a homozygous nonsense variant of unknown significance in ATAD1 (c.826 G>T, p.E276X) was reported (figure e-1A). His parents each carried 1 copy of the change. The variant was classified as having a potential relationship to a disease phenotype. There are no previous reports of ATAD1 mutations in humans; however, mice harboring mutations in ATAD1 have frequent seizures, hypertonia, and early death.⁷ Compound heterozygote variants of unknown significance were also found in SZT2, which causes autosomal recessive epileptic encephalopathy type 18; however, the variants did not segregate in his affected sister (IV-5) and he did not fit this phenotype (supplemental results).

SNP analysis confirmed that the family was consanguineous (table e-1). Patient IV-6 was found to have a de novo 1.25 Mb duplication of 17p13.3 (for more details, see supplemental results). Overlapping duplications have been reported with variable phenotypes including mild dysmorphic features, learning disability, hypotonia, autism spectrum disorder, and mild brain abnormalities.¹¹ This duplication was not present in affected individuals IV-3 or IV-5. Furthermore, given the extreme hypertonia and early respiratory failure seen in individual IV-6, we did not feel that this could fully explain his phenotype.

Homozygosity mapping revealed that the patient (IV-6), his deceased affected sister (IV-5), and affected cousin (IV-3) shared only one region of homozygosity (ROH) that was not shared with any unaffected relative (III 1–4 and IV-1, 2, and 7). This 6.5 Mb shared ROH (chr10:83,619,940-90,095,915, hg19 coordinates) contained 28 protein-coding genes, 14 of which are associated with a known disease phenotype (tables e-2 and e-3). ATAD1 is located in this shared ROH, while SZT2, the other candidate gene identified through exome sequencing, is not. Molecular analysis confirmed that all affected family members were homozygous for the p.E276X variant, while unaffected family members were not (figure 1). Multipoint genetic linkage analysis of 8 SNPs as markers in this region demonstrated a logarithm of the odds (LOD) score of 3.7 (table e-4), while the LOD score calculated directly using only the ATAD1 p.E276X variant was 3.5. An LOD score ≥3.3 has been shown to correspond to a genome-wide significance level of <0.05.¹²

The effect of the p.E276X variant on ATAD1 expression was investigated in lymphoblastoid cells derived from the patient and an unaffected control. ATAD1 messenger RNA expression was decreased by 78% in p.E276X cells. This degree of reduction is consistent with nonsense-mediated decay,¹³ as noted by the absence of Thorase protein on Western blotting (figure e-1B). Investigation of ATAD1 variation in control populations revealed that in the Exome Aggregation Consortium (exac.broadinstitute.org/gene/ENSG00000138138) database, only 3 loss-of-function variants were observed in the gene, while the expected number of loss-of-function variants is 14.9. This suggests selection against loss-of-function mutations in ATAD1.

Evaluation of perampanel as a therapeutic approach.

Given that the p.E276X variant tracked with disease, decreased ATAD1 expression, and that ATAD1 knockout mice died of seizures, we concluded that this mutation was likely pathogenic. ATAD1 mutations increase AMPA receptor–mediated excitatory signaling due to impaired receptor recycling.⁷ Considering this mechanism of disease, we hypothesized that perampanel,^14,15 a noncompetitive AMPAR antagonist could have therapeutic benefit. Therefore, we investigated the efficacy of perampanel in ATAD1^−/− mice. Perampanel dosing in mice was based on prior studies⁸ and a dose escalation study (figure e-2).

Perampanel response in mice.

Ex vivo brain MRI of ATAD1^−/− mice (figure 2A) demonstrated a volume reduction in all areas examined as compared to controls. Perampanel treatment did not show significant prevention of volume loss; however, there was a trend toward improvement (figure 2B). Loss of ATAD1 was also associated with increased T2-signal intensity throughout the brain, which is indicative of brain pathology with increased fluid content (figure 2C). Perampanel improved whole-slice signal intensity in ATAD1^−/− animals by 30.3% (95% confidence interval 11.3%–59.5%); further analysis of individual brain regions revealed improvement of signal intensity in the thalamus and striatum (figure 2D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2 Perampanel improves brain MRI differences in ATAD1^−/− mice

(A) Representative ex vivo brain MRIs of wild-type and ATAD1^−/− perampanel-treated and perampanel-untreated mice. (B) Untreated ATAD1^−/− mice (light blue bars) demonstrated volume reduction in all brain areas measured as compared to ATAD1^+/+ mice (black bars). Perampanel therapy (dark blue bars) was associated with a trend toward normalization of brain volumes. (C) Whole-slice T2 MRI signal intensity was significantly increased in ATAD1^−/− mice. This change was reversed with perampanel therapy. (D) T2-signal intensity was increased in ATAD1^−/− mice throughout the brain. Perampanel therapy normalized signal intensity in the striatum and thalamus. For all panels, n = 6. *p = 0.01–0.05, **p = 0.001–0.009, and ***p < 0.001.

On open-field activity testing, ATAD1^−/− mice showed both an abnormal activity pattern as compared to wild-type mice (figure 3, A–H) and decreased spontaneous movements as measured by the distance traveled and the total rest time (figure 3, I and J). Perampanel corrected these behavioral deficits, while phenobarbital, an antiepileptic that targets γ-aminobutyric acid (GABA) receptors rather than AMPA receptors, did not. ATAD1^−/− mice showed no difference in rearing activity or average speed as compared to wild-type littermates (figure e-3). Extensive daily observations revealed that ATAD1^−/− mice would have seizures on the day that they died. Video recording revealed that 5 of 5 vehicle-treated, as compared to zero of 5 perampanel-treated, ATAD1^−/− mice demonstrated seizures (video 2). Finally, perampanel significantly prolonged average survival in ATAD1^−/− mice by 115% (mean of 20 vs 43 days), while phenobarbital did not significantly increase survival (figure 3K). These results demonstrate the efficacy of targeted AMPAR blockade by perampanel in mice with impaired AMPAR endocytosis due to mutations in ATAD1.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3 Perampanel normalizes behavior and prolongs survival in ATAD1^−/− mice

(A–H) Perampanel, but not phenobarbital, normalized open-field assessments in ATAD1^−/− mice as seen by quantification (A–D) and representative pathview images (E–H) from (A, E) vehicle-treated, ATAD1^+/+, (B, F) vehicle-treated, ATAD1^−/−, (C, G) perampanel-treated, ATAD1^−/−, and (D, H) phenobarbital-treated, ATAD1^−/− mice (n = 6). Open circles represent peripheral activity; closed circles represent central activity. (I and J) On behavioral analysis, ATAD1^−/− mice (red diamonds) cover less distance (L) and rest more (M) than their wild-type littermates (black filled-in circles). Perampanel (blue triangles) reverses the movement deficits, while phenobarbital does not (n = 6). Two mice in the ATAD1^−/− perampanel and phenobarbital-treated groups seized during this observation. The seizures were characterized by high-speed involuntary movements; therefore, they are shown in the figure (symbol surrounded by a black circle), but they are not included in the statistical analysis. (K) Survival is prolonged from 20 days in vehicle-treated ATAD1^−/− mice (red line) to 43 days in ATAD1^−/− perampanel-treated mice (blue line). There is no statistical survival difference between ATAD1^−/− mice treated with vehicle (red line) or phenobarbital (green line). For all panels, *p = 0.01–0.05, **p = 0.001–0.009, and ***p < 0.001. See also figures e-2 and e-3.

Perampanel response in humans.

Given the efficacy of perampanel in mice, we next considered a therapeutic trial in patients with ATAD1 mutations. Perampanel was recently approved by the FDA for the treatment of seizures and demonstrated minimal off-target toxicity in human studies¹⁵; we obtained permission for off-label compassionate use for treatment of affected individuals. Patient IV-6 was started on perampanel at 16 months of age; at this time, he had severe brain atrophy on MRI (figure 4A), hypertonia, absent cranial nerve reflexes, no spontaneous respiratory drive, and hypsarrhythmia on EEG. Within 8 days of starting perampanel, his tone improved and he developed a pupillary response to light. He continued to show dose-dependent neurologic improvement until his neurologic status plateaued at a dose of approximately 0.5 mg/kg/d (figures 4C, e-4, e-5). He also had resolution of hypsarrhythmia on EEG (figure 4, D–G). He was weaned off all nonperampanel antiepileptics, which included phenobarbital (dose at start of trial 6 mg/kg/d), topiramate (dose at start of trial 6 mg/kg/d), and valium (dose at start of trial 0.9 mg/kg/d). Despite therapy, he continued to have severe neurologic deficits and worsening hydrocephalus requiring endoscopic third ventriculostomy.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4 Perampanel improves hypertonicity, prevents seizures, and slows neurodegeneration but cannot normalize neurologic function in patients with ATAD1 mutations

(A) Photographs and brain MRIs of the proband, individual IV-6. Initial brain MRI at 2 months showed normal parenchyma, but subsequent studies at 9 and 21 months showed progressive volume loss and hydrocephalus. (B) Photographs and brain MRI of patient IV-3. Brain MRI at 2 months of age was normal. Perampanel was started at 2.5 months of age and follow-up MRI at 9 months of age showed no changes (unlike the progressive neurodegeneration seen in untreated patient IV-6 over the same age range, panel A). (C) Timeline of clinical improvements observed in patient IV-6 and patient IV-4 after the initiation of perampanel therapy. Patient IV-4 also had progression of disease (marked in red) including respiratory failure that was not prevented with perampanel. The bottom panel shows an EEG recorded from subject IV-6 at 15 months of age, prior to therapy with perampanel. (D) Hypsarrhythmia is seen on EEG (gain = 7 μV/mm). (E) The pretreatment recording from panel A is shown at a gain of 30 μV/mm. (F) An example of one of many focal seizures captured on routine EEG prior to perampanel. The lower panel shows the EEG of the same patient 10 weeks after starting perampanel. (G) While on perampanel, the hypsarrhythmia resolved and no focal seizures were noted on follow-up 60-minute routine EEG. This improved EEG pattern was noted on all subsequent follow-up EEGs recorded over a several month period of observation. See also figure e-4.

In light of the clinical improvement seen in the proband, patient IV-3 was started on perampanel at 2.5 months of age. Given his young age, his neurologic function was more intact as compared to his cousin's at the start of therapy. While he was profoundly hypertonic, he had normal cranial nerve reflexes and respiratory drive. After starting perampanel, he had rapid improvement in tone and normalization of his EEG (video 1). Like his cousin, he was weaned from his other seizure medications including phenobarbital (dose at start of trial of 2.5 mg/kg/d). Repeat brain MRI at 9 months of age showed no parenchymal volume loss while on perampanel (figure 4B), a finding that is dramatically different from what was seen in the proband at age 9 months (figure 4A, middle MRIs). Despite these markers of improvement, he developed respiratory failure requiring intubation. However, his limited function was better as compared to his cousin who was untreated at the same age: he maintained all brain stem reflexes, sleep/wake cycles, the ability to withdraw and vocalize to pain, and spontaneous movements of all extremities (figure 4C, and supplemental information).

Functional improvement was also measured using the functional status score, a validated pediatric outcome measurement.¹⁰ This outcome assesses 6 functional domains (mental status, sensory, communication, motor, feeding, and respiratory function), each on a scale from 1 (normal) to 5 (severely impaired), and a composite score is calculated. Fully functioning children receive a score of 6, while maximal impairment is assigned a score of 30. Over the course of therapy, patient IV-6 had a score improvement from 29 to 24 (figure e-4, A and B). Patient IV-4 began the trial with a functional status score of 19 and showed an initial improvement to a low score of 17. However, his overall functional score worsened despite therapy at 3–4 months of age due to his respiratory failure (figure e-4, C and D). At the end of the therapeutic trial, his functional status score plateaued at 23, as compared to his untreated cousin's score of 29 at the same age. This suggests that perampanel may improve functional outcomes but cannot completely prevent progression of disease.