Natural history of epilepsy in FOXG1 Syndrome

Abstract

Background:

FOXG1 syndrome is rare neurodevelopmental disorder with microcephaly, brain malformations, epilepsy, and cognitive and motor disabilities as major features. Knowledge of the clinical features is primarily from case series and a foundation sponsored registry. We expand insight into epilepsy in FOXG1 syndrome by examining longitudinal data from 94 individuals from a multi-site natural history study and local cohorts.

Methods:

Clinical information on severity, seizure type, and seizure features was collected from 68 individuals enrolled in the Rett Syndrome and Related Disorders Natural History Study and extracted via retrospective chart review from 15 individuals seen at the Children’s Hospital of Philadelphia Rett Syndrome Center of Excellence and 11 individuals from Children’s Hospital of Colorado. Genotype-phenotype and other correlations were assessed using non- and semiparametric analyses.

Results:

78.7% of participants had seizures, beginning at a median age of 1.0 years. Individuals were followed for a median of 4.9 years after first seizure onset. Taken independently, over 70% of seizures were partial or tonic-clonic, occurred less than weekly, and lasted less than 5 minutes. 1/3 of seizures resolved in a median time and age of 1.1 and 3.3 years. Age had a weak non-linear association with average seizure frequency, and, for those with seizures, smaller head circumference correlated with increased disease severity.

Conclusions:

We further characterize disease severity and epilepsy in FOXG1 syndrome and demonstrate that smaller head circumferences are associated with more severe disease and age is weakly non-linearly correlated with average seizure frequency. This information will improve clinical care and aid therapeutic development.

1.0. Introduction:

FOXG1 syndrome is a rare, autosomal dominant, neurodevelopmental disorder with an estimated incidence of at most 1:30,000 (Brockmann and Staudt., 2024; Lopez-Rivera et al., 2020). Heterozygous pathogenic variants in the forkhead box protein G1 (FOXG1) gene on chromosome 14 (14q12) cause the disease and typically occur de novo though rare cases of parental mosaicism have been reported (Ariani et al., 2008; Diebold et al., 2014). FOXG1 encodes for an evolutionarily conserved, pleiotropic transcription factor essential for regulating telencephalon development (Wong et al., 2019). Deletions or intragenic mutations leading to FOXG1 presumed loss-of-function manifest clinically with severe global developmental delays, absent or minimal speech, inability to ambulate, visual impairments, sleep disorders, early-onset dyskinesia and hyperkinesia, stereotypies, postnatal microcephaly, cerebral malformations – including corpus callosum hypoplasia or agenesis – and variable epilepsy phenotypes (Wong et al., 2019; Kortum et al., 2011; Mitter et al., 2018; Cellini et al., 2016; Brimble et al., 2023). FOXG1 duplications produce a different syndrome: these individuals are usually normocephalic and without brain abnormalities, have more variable developmental delay and better speech and ambulation (Wong et al., 2019; Brunetti-Pierri 2011). Epilepsy in the FOXG1 duplications usually presents as infantile spasms, which resolve with therapy (Seltzer et al., 2014). Historically, FOXG1 syndrome was known as the congenital variant of Rett syndrome (RTT); however, it is now clear that FOXG1 is its own neurodevelopmental syndrome as FOXG1 individuals often do not regress and are typically more severe across domains like language and ambulation (Wong et al., 2019; Kortum et al., 2011; Cutri-French et al., 2020).

Epilepsy is a major feature of FOXG1 syndrome and has been reported in 61% to 87% of individuals (Brimble et al., 2023; Seltzer et al., 2014). Seizure types vary and include infantile spasms, partial, complex partial, generalized tonic, generalized tonic-clonic, atonic, and myoclonic seizures (Kortum et al., 2011; Mitter et al., 2018; Brimble et al., 2023; Seltzer et al., 2014; Vegas et al., 2018). About half of individuals reportedly experience refractory epilepsy with multiple seizure types (Vegas et al., 2018). Age of epilepsy diagnosis is significantly earlier, around 3–7 months, for duplication individuals than for those with intragenic mutation or deletions, who typically begin seizing around age 2 years (Mitter et al., 2018; Seltzer et al., 2014; Vegas et al., 2018). Truncating mutations and deletions have shown a more severe clinical phenotype than those with missense variants (Kortum et al., 2011; Mitter et al., 2018; Brimble et al., 2023). Missense mutations in the FOXG1 forkhead conserved site 1 are associated with a milder phenotype than nonsense or frameshift mutations in the rest of the forkhead or the N-terminal domains (Mitter et al., 2018). For epilepsy features, besides infantile spasms’ predominance in duplication individuals and the difference in age at diagnosed epilepsy onset, genotype-phenotype correlations have not been demonstrated. Two hotspot mutations, c.256dupC and c.460dupG, demonstrate variable epilepsy phenotypes, leading to the suggestion that factors beyond genetic mutation alone could have larger roles in seizure presentation (Wong et al., 2019).

Since seizures are a core feature of FOXG1, developing a deeper understanding of the epilepsy phenotype will improve clinicians’ and researchers’ ability to care for and develop treatments for this condition. To generate a thorough understanding of FOXG1-related epilepsy, we analyzed data from 94 participants with confirmed pathogenic FOXG1 variants. Sixty-eight were part of the Natural History Study (NHS) of Rett syndrome and Related Disorders. Two additional cohorts came from retrospective chart review of 15 individuals from Children’s Hospital of Philadelphia (CHOP) Rett Syndrome Center of Excellence and 11 individuals from Children’s Hospital of Colorado (CHCO). The NHS cohort represents the largest prospectively followed FOXG1 population. This paper presents detailed information about clinical features, genotype-phenotype, and other correlations with regards to severity and epilepsy in FOXG1 syndrome.

2.0. Methods:

2.1. Data Sources:

Sixty-eight participants were recruited from 15 clinical sites that are part of the multicenter NIH-funded Natural History Study of Rett Syndrome and Related Disorders (clinicaltrials.gov NCT02738281). The study protocol received prior approval by the appropriate Institutional Review Boards of Baylor College of Medicine, CHOP, Cleveland Clinic, Gillette Children’s Hospital, and Vanderbilt University. The remaining sites (Boston Children’s Hospital, Cincinnati Children’s Hospital, Greenwood Genetic Center, Oakland Children’s Hospital, Rush Medical Center, UCSD, University of Colorado-Denver, and Washington University) relied on the single-IRB agreement provided by the University of Alabama at Birmingham. Informed consent was obtained for each subject. Assent to participate was waived as these individuals were incapable of providing such. Data were extracted from participant visits between November 2006 and July 2021. A neurologist or geneticist confirmed each participant’s diagnosis of FOXG1 syndrome either due to intragenic mutations and deletions or duplications. Data extraction included date of birth (DOB), sex, enrollment date, FOXG1 mutation, seizure semiology and features such as frequency and duration, date of seizure onset and resolution (if resolved), maximum Clinical Severity Scale (CSS) score and seizure frequency, maximum Motor Behavioral Assessment (MBA) score, and most recent head circumference, which was converted to a Z-score. 13 participants had data on if they had a seizure but did not have seizure semiology and features recorded. Data drawn from retrospective chart review of an additional 15 individuals from CHOP and 11 from CHCO not otherwise in the NHS study were included. The CHOP IRB (IRB 21–018453) and CHCO IRB (COMIRB 13–2020) granted an exemption for this review, and informed consent was waived. Data were extracted from participant visits between February 2000 for CHOP and October 2016 for CHCO to November 2024 for both. A neurologist or geneticist confirmed each participant’s FOXG1 syndrome diagnosis. Data extraction included the same items as the NHS participants except for CSS and MBA.

2.2. Data Definitions and Categorization:

NHS data came from the Rett Consortium database, hosted by the Rare Diseases Clinical Research Network (RDCRN) through the Data Management and Coordinating Center at the University of South Florida. CHOP and CHCO data came from EPIC by chart review.

The CSS was previously published by Schanen et al. (Schanen et al., 2004) and has been applied in both clinical studies (Neul et al., 2008; Bebbington et al., 2008) and an RTT clinical trial (Glaze et al., 2017; Glaze et al., 2019). Most investigators had previous training and experience with this scale as part of industry-sponsored trials (Glaze et al., 2017; Glaze et al., 2019). The MBA is a clinician-reported measure first developed by Fitzgerald et al. (Fitzgerald et al., 1990) to assess RTT and has been evaluated as a potential clinical endpoint for RTT (Raspa et al., 2020). All investigators were trained on the CSS and MBA at site initiation.

Participants were categorized into one of 6 groups by variant type: frameshift (n=44, in-frame deletion (n=2), missense (n=22), nonsense (n=12), deletion (n=12), or duplication (n=2). To increase sample sizes, these groups were combined into 2 higher-level groups: loss of genetic material, which included nonsense, in-frame deletion, and deletion individuals, and change in genetic material, which included frameshift and missense individuals. Duplication individuals were excluded for correlational analyses.

Investigators combined their clinical assessments and parent reports to map seizure semiology onto a prespecified list of seizure types, which included absence, atonic, infantile spasm, myoclonic, partial, tonic, tonic-clonic, or not a seizure. Electroencephalograms were not consistently available or used to determine seizure type. For seizures not matching the prespecified set of types, they were reclassified as closely as possible as follows: generalized seizures to tonic-clonic, complex partial seizures as partial, and epileptic spasms as infantile spasms. To increase sample sizes, these seizure types were combined into 2 higher-level groups: generalized seizures, which included absence, atonic, infantile spasms, myoclonic, tonic, and tonic-clonic seizures, and focal, which included partial seizures.

For each seizure, age at seizure onset was calculated by finding the time between DOB and the date at onset. If only the date at onset’s month and year was available, the day was set to 15. If only the year was available, the date was set to July 1^st unless that was before the DOB, in which case the date was set to December 31^st. For each participant, censor age was calculated by finding the time between DOB and the most recent date of data update.

For each seizure, age at seizure resolution (if the seizure had resolved) was calculated by finding the time between DOB and the date of resolution. If only the date of resolution’s month and year was available, the day was set to 15 unless that would make the resolution date after that participant’s censor date, in which case censor date was set as the date of resolution. If only year was available, the date was set to July 1^st unless that would make the resolution date after the censor date, in which case the censor date was set as the date of resolution.

Seizures were labeled “resolved” if they had a resolution date and “ongoing” if they did not. We calculate the “Number of seizures” as the sum of distinct seizure types for each participant. Some distinct seizures were of the same type by clinical assessment or bucketing. Individuals could also have had more than 1 seizure type (e.g. absence and tonic-clonic would count as 2), discontinuous spells (e.g. absence seizures from age 1–3 years and then absence seizures again from age 6–8 years would count as 2), or a combination of the above.

Average and maximum seizure frequency values were documented using a scale from absent (i.e. not reported between two yearly visits but not judged to have stopped) to less than monthly, less than weekly to monthly, weekly, more than weekly, daily, and multiple times a day. The CSS scale did not include the last category. Average and maximum seizure duration values were reported using a 5-point scale from less than 5 seconds to 5–30 seconds, greater than 30 to 60 seconds, 1–5 minutes, greater than 5 to 10 minutes, and greater than 10 minutes.

To prevent floor effects of percentiles, head circumference (HC) measurements were converted to Z-scores using the LMS method (Cole et al., 1998) according to:

$$Z=\frac{\left({\left(\frac{A}{M}\right)}^L\right)-1}{L\text{*}S}$$

where A is the HC (in cm), and L, M, and S are data extracted from the normal HC curves from Rollins et al. (Rollins et al., 2010). Cutri-French et al. demonstrated the accuracy of this method for converting raw HC measurements to Z-scores for FOXG1 (Cutri-French et al., 2020).

Data indicated as “missing” in the results were not available.

2.3. Statistical Analysis:

Median and range were used to describe continuous variables and frequency and percentage for categorical variables. Statistical analysis was done on first seizure data for individuals with seizures. To assess group differences, the Fischer’s exact test was used for categorical variables, Kruskal-Wallis test for comparisons between categorical and ordinal or continuous variables, Spearman’s rank correlation for comparisons between ordinal and ordinal or continuous data, and the Pearson’s product-moment correlation for continuous versus continuous data comparisons. For analysis by mutation or seizure type, the high-level binning was tested first. Unless results were significant at that level, analysis on the lower level is considered exploratory only. For comparisons, a strict Bonferroni correction was used with an adjusted p-value for significance of 0.0004386 (two tailed alpha = 0.05, therefore, the corrected p-value = 0.025/number of comparisons; 57 comparisons total). For analysis of the CSS dataset, this study used Spearman’s rank correlation and a generalized additive model (GAM) to assess the relationship between age and seizure frequency. For comparisons, a strict Bonferroni correction was used with an adjusted p-value for significance of 0.0125 (two tailed alpha = 0.05, therefore, the corrected p-value = 0.025/number of comparisons; 2 comparisons total). The paper reports raw p-values only. All statistical analyses were performed using RStudio Version 2024.04.0+735.

3.0. Results:

This study includes 94 confirmed FOXG1 individuals (Table 1). 78.7% (n = 74) had at least 1 seizure by data censor, and 21.3% (n = 20) were seizure free (Table 1). 54.3% of individuals were female (n = 51), and 45.7% were male (n = 43) (Table 1). Median study enrollment age was 5.6 years (Table 1). Frameshift mutations were most common (46.8%), followed by missense (23.4%), nonsense (12.8%), deletions (12.8%), in-frame deletions (2.1%), and duplications (2.1%) (Table 1).

Table 1.

Demographic and Clinical Summary of FOXG1 Syndrome Study Population. Length of seizure follow-up defined as time between date of first seizure onset and date of data censor.

	Seizure (N=74)	Seizure Free (N=20)	Total (N=94)
Sex:
Female	40 (54.1%)	11 (55.0%)	51 (54.3%)
Male	34 (45.9%)	9 (45.0%)	43 (45.7%)
Age at Study Onset (years):
Median [Min, Max]	5.81 [0.449, 34.9]	3.39 [0.337, 11.0]	5.59 [0.337, 34.9]
Missing	1 (1.4%)	0 (0%)	1 (1.1%)
Age at Seizure Onset (years):
Median [Min, Max]	1.00 [0.0329, 14.2]	NA	1.00 [0.0329, 14.2]
Missing	10 (13.5%)	20 (100%)	30 (31.9%)
Length of Seizure Follow-Up (years):
Median [Min, Max]	4.93 [0.318, 22.7]	NA	4.93 [0.318, 22.7]
Missing	11 (14.9%)	20 (100%)	31 (33.0%)
Mutation Type:
Deletion	11 (14.9%)	1 (5.0%)	12 (12.8%)
Duplication	2 (2.7%)	0 (0%)	2 (2.1%)
Frameshift	37 (50.0%)	7 (35.0%)	44 (46.8%)
In-frame deletion	1 (1.4%)	1 (5.0%)	2 (2.1%)
Missense	13 (17.6%)	9 (45.0%)	22 (23.4%)
Nonsense	10 (13.5%)	2 (10.0%)	12 (12.8%)
Number of Diagnosed Seizures:
0	0 (0%)	20 (100%)	20 (21.7%)
1	26 (36.1%)	0 (0%)	26 (28.3%)
2	24 (33.3%)	0 (0%)	24 (26.1%)
3	10 (13.9%)	0 (0%)	10 (10.9%)
4	6 (8.3%)	0 (0%)	6 (6.5%)
5	5 (6.9%)	0 (0%)	5 (5.4%)
7	1 (1.4%)	0 (0%)	1 (1.1%)
Missing	2 (2.7%)	0 (0%)	2 (2.1%)
Clinician Severity Score:
Median [Min, Max]	30.0 [10.0, 42.0]	23.0 [0, 35.0]	28.5 [0, 42.0]
Missing	23 (31.1%)	7 (35.0%)	30 (31.9%)
Motor Behavioral Assessment:
Median [Min, Max]	51.5 [14.0, 71.0]	38.0 [0, 55.0]	50.0 [0, 71.0]
Missing	22 (29.7%)	7 (35.0%)	29 (30.9%)
Head Circumference Z-Score:
Median [Min, Max]	−3.76 [−6.35, 1.05]	−2.79 [−7.30, 0.0340]	−3.62 [−7.30, 1.05]
Missing	6 (8.1%)	0 (0%)	6 (6.4%)

3.1. Clinical Severity:

Median CSS score was 28.5 (30.0 seizure group, 23.0 seizure-free group), and median MBA score was 50.0 (51.5 seizure group, 38.0 seizure-free group), but these values were not available for the CHOP or CHCO cohorts as well as 4 and 3 NHS individuals respectively (Table 1). Median HC Z-score was −3.62 (−3.76 seizure group, −2.79 seizure-free group) (Table 1). For the seizure group, median CSS score ranged from 19.5 for missense (n=10) to 34.0 for individuals with deletions (n=7) or in-frame deletions (n=1); median MBA ranged from 37.0 for missense (n=11) to 58.0 for deletions (n = 7); and median HC Z-score ranged from −4.15 for frameshifts (n=33) to −1.71for missense (n=13) (Table 2). HC Z-score had a statistically significant negative linear relationship with both CSS (p = 9.545 * 10⁻⁶, correlation −0.5863668, 95% CI −0.7447636 to −0.3654228) and MBA (p = 2.443 * 10⁻⁴, correlation −0.496603, 95% CI −0.6808394 to −0.2532635) (Supplementary Table 1). Smaller head circumferences are associated with more severe disease for individuals with seizures with intragenic mutations or deletions. Differences in CSS and MBA were not significant by mutation or seizure type at either level, age at seizure onset, or sex (p > 0.0004386, Supplementary Table 1). CSS and MBA by lower-level mutation type had p = 0.007 and 0.091 respectively, which did not meet the significance threshold (Supplementary Table 1).

Table 2.

Demographic and Clinical Summary of FOXG1 Syndrome Study Population with Seizures by Mutation Type. Time to seizure resolution defined as time between date of seizure onset and seizure resolution (if resolved). Data for second and third seizures available in Supplementary Table 2.

		Mutation Type
	Deletion (N=11)	Mutation Duplication (N=2)	Type Frameshift (N=37)	Inframe deletion (N=1)	Missense (N=13)	Nonsense (N=10)	Total (N=74 )
Age at Study Onset (years):
Median [Min, Max]	4.74 [1.26, 34.9]	8.87 [4.42, 13.3]	6.00 [0.449, 19.2]	13.9 [13.9, 13.9]	6.13 [0.559, 20.0]	3.29 [0.794, 12.4]	5.81 [0.449, 34.9]
Missing	1 (9.1%)	0 (0%)	0 (0%)	0 (0%)	0 (0%)	0 (0%)	1 (1.4%)
Number of Diagnosed Seizures:
1	4 (36.4%)	1 (50.0%)	13 (36.1%)	1 (100%)	4 (33.3%)	3 (30.0%)	26 (36.1%)
2	4 (36.4%)	1 (50.0%)	12 (33.3%)	0 (0%)	4 (33.3%)	3 (30.0%)	24 (33.3%)
3	0 (0%)	0 (0%)	5 (13.9%)	0 (0%)	2 (16.7%)	3 (30.0%)	10 (13.9%)
4	1 (9.1%)	0 (0%)	3 (8.3%)	0 (0%)	1 (8.3%)	1 (10.0%)	6 (8.3%)
5	2 (18.2%)	0 (0%)	2 (5.6%)	0 (0%)	1 (8.3%)	0 (0%)	5 (6.9%)
7	0 (0%)	0 (0%)	1 (2.8%)	0 (0%)	0 (0%)	0 (0%)	1 (1.4%)
Missing	0 (0%)	0 (0%)	1 (2.7%)	0 (0%)	1 (7.7%)	0 (0%)	2 (2.7%)
Clinician Severity Score:
Median [Min, Max]	34.0 [28.0, 42.0]	NA	30.0 [22.0, 40.0]	34.0 [34.0, 34.0]	19.5 [10.0, 33.0]	30.0 [22.0, 39.0]	30.0 [10.0, 42.0]
Missing	4 (36.4%)	2 (100%)	13 (35.1%)	0 (0%)	3 (23.1%)	1 (10.0%)	23 (31.1%)
Motor Behavioral Assessment:
Median [Min, Max]	58.0 [47.0, 71.0]	NA	52.0 [32.0, 71.0]	56.0 [56.0, 56.0]	37.0 [14.0, 65.0]	50.0 [41.0, 64.0]	51.5 [14.0, 71.0]
Missing	4 (36.4%)	2 (100%)	13 (35.1%)	0 (0%)	2 (15.4%)	1 (10.0%)	22 (29.7%)
Head Circumferenc e Z-Score:
Median [Min, Max]	−4.14 [−6.32, −1.47]	NA	−4.15 [−6.35, −0.956]	−3.44 [−3.44, −3.44]	−1.71 [−5.00, 1.05]	−3.64 [−5.32, −2.62]	−3.76 [−6.35, 1.05]
Missing	0 (0%)	2 (100%)	4 (10.8%)	0 (0%)	0 (0%)	0 (0%)	6 (8.1%)
First Seizure: Diagnosis
Absence	1 (9.1%)	0 (0%)	1 (2.7%)	0 (0%)	0 (0%)	0 (0%)	2 (2.7%)
Infantile spasms	1 (9.1%)	1 (50.0%)	7 (18.9%)	0 (0%)	2 (15.4%)	0 (0%)	11 (14.9%)
Partial	5 (45.5%)	0 (0%)	12 (32.4%)	1 (100%)	4 (30.8%)	6 (60.0%)	28 (37.8%)
Tonic-clonic	4 (36.4%)	1 (50.0%)	11 (29.7%)	0 (0%)	5 (38.5%)	3 (30.0%)	24 (32.4%)
Myoclonic	0 (0%)	0 (0%)	2 (5.4%)	0 (0%)	0 (0%)	0 (0%)	2 (2.7%)
Tonic	0 (0%)	0 (0%)	3 (8.1%)	0 (0%)	1 (7.7%)	1 (10.0%)	5 (6.8%)
Missing	0 (0%)	0 (0%)	1 (2.7%)	0 (0%)	1 (7.7%)	0 (0%)	2 (2.7%)
First Seizure: Age at Onset (years):
Median [Min, Max]	1.12 [0.496, 14.2]	0.456 [0.329, 0.583]	0.999 [0.0329, 11.7]	1.06 [1.06, 1.06]	0.757 [0.241, 9.00]	0.956 [0.120, 4.40]	1.00 [0.0329, 14.2]
Missing	0 (0%)	0 (0%)	5 (13.5%)	0 (0%)	3 (23.1%)	2 (20.0%)	10 (13.5%)
First Seizure: Resolution
Ongoing	8 (72.7%)	0 (0%)	27 (73.0%)	1 (100%)	5 (38.5%)	7 (70.0%)	48 (64.9%)
Resolved	3 (27.3%)	2 (100%)	9 (24.3%)	0 (0%)	7 (53.8%)	3 (30.0%)	24 (32.4%)
Missing	0 (0%)	0 (0%)	1 (2.7%)	0 (0%)	1 (7.7%)	0 (0%)	2 (2.7%)
First Seizure: Age of Resolution (years):
Median [Min, Max]	1.82 [1.00, 36.4]	3.02 [1.62, 4.42]	7.05 [0.706, 9.67]	NA	1.59 [0.246, 20.0]	3.33 [2.42, 7.42]	3.07 [0.246, 36.4]
Missing	8 (72.7%)	0 (0%)	28 (75.7%)	1 (100%)	6 (46.2%)	7 (70.0%)	50 (67.6%)
First Seizure: Time to Resolution (years):
Median [Min, Max]	0.561 [0, 22.2]	2.56 [1.29, 3.83]	3.40 [0, 7.66]	NA	0.0329 [0, 11.0]	2.30 [2.17, 5.92]	2.05 [0, 22.2]
Missing	8 (72.7%)	0 (0%)	29 (78.4%)	1 (100%)	6 (46.2%)	7 (70.0%)	51 (68.9%)

3.2. Seizures:

Across all seizures, the most common were partial seizures (39.1%), followed by tonic-clonic (35.8%), infantile spasms (9.9%), absence (7.3%), tonic (4.0%), myoclonic (2.6%), and atonic (1.3%) seizures (Table 3). Partial and tonic-clonic seizures were the most common at each age and present across the largest range of ages (0–23 and 0–36 years, respectively) (Figure 1A). Partial (37.8%) and tonic-clonic (32.4%) seizures were the most common first seizures as well (Table 2).

Table 3.

Demographic and Clinical Summary of FOXG1 Syndrome Study Population by Seizure Type. Time to seizure resolution defined as time between date of seizure onset and seizure resolution (if resolved). Table excludes unclassified seizures (n = 2).

		Seizure Type
	Absence (N=11)	Atonic (N=2)	Infantile spasms (N=15)	Myoclonic (N=4)	Partial (N=59)	Tonic (N=6)	Tonic-clonic (N=54)	Total (N=151)
Sex:
Female	9 (81.8%)	1 (50.0%)	4 (26.7%)	3 (75.0%)	30 (50.8%)	3 (50.0%)	28 (51.9%)	78 (51.7%)
Male	2 (18.2%)	1 (50.0%)	11 (73.3%)	1 (25.0%)	29 (49.2%)	3 (50.0%)	26 (48.1%)	73 (48.3%)
Mutation Type:
Deletion	4 (36.4%)	0 (0%)	1 (6.7%)	0 (0%)	8 (13.6%)	1 (16.7%)	9 (16.7%)	23 (15.2%)
Duplication	0 (0%)	0 (0%)	1 (6.7%)	0 (0%)	0 (0%)	0 (0%)	2 (3.7%)	3 (2.0%)
Frameshift	5 (45.5%)	1 (50.0%)	11 (73.3%)	3 (75.0%)	29 (49.2%)	3 (50.0%)	26 (48.1%)	78 (51.7%)
In-frame deletion	0 (0%)	0 (0%)	0 (0%)	0 (0%)	1 (1.7%)	0 (0%)	0 (0%)	1 (0.7%)
Missense	1 (9.1%)	0 (0%)	2 (13.3%)	1 (25.0%)	10 (16.9%)	1 (16.7%)	11 (20.4%)	26 (17.2%)
Nonsense	1 (9.1%)	1 (50.0%)	0 (0%)	0 (0%)	11 (18.6%)	1 (16.7%)	6 (11.1%)	20 (13.2%)
Age at Seizure Onset (years):
Median [Min, Max]	1.73 [0.564, 7.00]	7.69 [6.37, 9.00]	0.832 [0.329, 3.92]	0.652 [0.353, 0.999]	1.54 [0.0329, 20.7]	1.40 [0.246, 2.38]	1.81 [0.0329, 18.0]	1.48 [0.0329, 20.7]
Missing	1 (9.1%)	0 (0%)	2 (13.3%)	0 (0%)	6 (10.2%)	0 (0%)	4 (7.4%)	13 (8.6%)
Seizure Resolution:
Ongoing	7 (63.6%)	0 (0%)	9 (60.0%)	4 (100%)	40 (67.8%)	3 (50.0%)	37 (68.5%)	100 (66.2%)
Resolved	4 (36.4%)	2 (100%)	6 (40.0%)	0 (0%)	19 (32.2%)	3 (50.0%)	17 (31.5%)	51 (33.8%)
Age of Seizure Resolution (years)
Median [Min, Max]	4.68 [1.00, 9.67]	13.7 [11.4, 16.0]	1.27 [0.591, 1.82]	NA	5.60 [0.241, 20.0]	7.37 [0.246, 9.67]	3.16 [0.706, 36.4]	3.33 [0.241, 36.4]
Missing	7 (63.6%)	0 (0%)	9 (60.0%)	4 (100%)	40 (67.8%)	3 (50.0%)	37 (68.5%)	100 (66.2%)
Time to Seizure Resolution (years)
Median [Min, Max]	3.28 [0, 6.20]	6.00 [5.00, 7.00]	0.167 [0, 1.29]	NA	2.02 [0, 11.0]	5.56 [0, 7.29]	0.627 [0, 22.2]	1.14 [0, 22.2]
Missing	7 (63.6%)	0 (0%)	10 (66.7%)	4 (100%)	41 (69.5%)	3 (50.0%)	38 (70.4%)	103 (68.2%)

Figure 1.

Recorded seizure features roughly correspond with expected differences between seizure types (Figure 1B). For example, 100.0% of tonic and 79.6% of tonic-clonic but 0.0% of atonic seizures were recorded as having arm or leg stiffness; additionally, 44.4% of tonic-clonic but 0.0% of tonic seizures were recorded as having arms and legs moving (Figure 1B). Differences in seizure type at either level were not significant by mutation type at either level or sex (p > .05, Supplementary Table 1).

The median age at first seizure onset was 1.0 years, and the median length of follow-up after seizure initiation was 4.9years (Table 1). Median age at first seizure onset ranged from 0.5 years for duplication (n = 2) to 1.1 years for deletion individuals (n = 11) (Table 2). Median age at second seizure onset was 1.8 years and ranged from 0.8 years for duplication (n = 1) to 2.2 for missense individuals (n = 6) (Supplementary Table 2). For all seizures, median age at onset ranged from 0.7 years for myoclonic (n = 4) to 7.7 years for atonic seizures (n = 2) (Table 3). Differences in age at first seizure onset were not significant by mutation or seizure type at either level, sex, or HC Z-score (p > .05, Supplementary Table 1).

Overall, 66.2% of seizures were ongoing and 33.8% were resolved at time of data censoring (Table 3). 66.7% of first seizures were ongoing and 33.3% were resolved at time of data censoring (Table 2). Overall, median resolution time was 1.1 years at a median age of 3.3 years (Table 3). For first seizures, median resolution time was 2.1 years at a median age of 3.1 years (Table 2). 50.0% overall and 63.8% of those with seizures had more than 1 distinct seizure, with the most being an individual who had 7 (Table 1). The most distinct seizures experienced simultaneously was 5 (Figure 2A). The most common combinations of simultaneous seizures were partial and partial, tonic-clonic and tonic-clonic, and partial and tonic-clonic, and Figure 2B shows the individual seizure journeys for each participant (Figure 2B). Differences in first seizure resolution rate and seizure count were not significant by mutation or seizure type at either level or by sex, age at onset, average or maximum duration or frequency (p > .0004386, Supplementary Table 1).

Figure 2.

The most common average frequencies for all seizures were less than monthly (38.8%), less than weekly to monthly (19.7%) or weekly (12.2%), followed by absent (11.6%), daily (8.2%), multiple times a day (5.4%), or more than weekly (4.1%) (Figure 3A). Across ages, most seizures had an average frequency of weekly at most (Figure 3B). The most common maximum frequencies were multiple times a day (26.5%), followed by less than monthly (22.7%), weekly (20.5%), daily (17.4%), and less than weekly to monthly (12.9%) (Figure 3C). The most common average durations for all seizures were 1–5 minutes (31.0%), followed by 5–30 seconds (26.9%), greater than 30 to 60 seconds (17.2%), less than 5 seconds (10.3%), greater than 10 minutes (7.6%), and greater than 5 to 10 minutes (6.9%) (Figure 4A). Across ages, over 75% of seizures had an average duration of 1–5 minutes or less (Figure 4B). The most frequent maximum duration was 1–5 minutes (29.1%), followed by greater than 10 minutes (19.1%), greater than 30 to 60 seconds (16.3%), greater than 5 to 10 minutes (15.6%), 5–30 seconds (13.5%), and less than 5 seconds (6.4%) (Fig 4C). Differences in average or maximum first seizure frequency or duration were not significant by mutation or seizure type at either level, age at onset, sex, or HC Z-score (p > 0.0004386, Supplementary Table 1). For the CSS data, which recorded average seizure frequency at every visit, age was not linearly associated with average seizure frequency (p > 0.0125); however, a GAM with a Gaussian distribution and identity link function showed a significant non-linear relationship between age and seizure frequency (edf = 3.62, F = 3.17, p = 0.0123), with age explaining 9.7% of the deviance in average seizure frequency (adjusted R² = 0.077) (Supplementary Table 1). Figure 5 plots this trend on the scatter plot of age vs seizure frequency (Figure 5).

Figure 3.

Figure 4.

Figure 5.

4.0. Discussion:

FOXG1 is a rare neurodevelopmental disorder that was only recognized as having a distinct etiology from RTT in 2008 (Ariani et al., 2008). This study presents the most comprehensive information yet on the epilepsy phenotype and clinical course in FOXG1. While we did not identify significant genotype-phenotype or other correlations for seizure type or features in FOXG1 syndrome except for a weak non-linear association between age and average seizure frequency, these descriptive results will aid providers in counseling individuals and their caregivers on prognosis related to epilepsy in this disorder. This improved understanding of the natural history of FOXG1 will also support clinical development.

4.1. Clinical Severity:

This study corroborates findings from the literature. We report postnatal microcephaly for individuals with intragenic mutations or deletions (median HC Z-score −3.62, Table 1). In the seizure population, lower HC Z-scores were significantly associated with higher CSS and MBA scores and not associated with age at seizure onset, results which recapitulate those found by Cutri-French et al. with some of the same FOXG1 individuals from the NHS (Cutri-French et al., 2020). The CSS result is slightly confounded by the fact that head growth is one of the 13 equally weighted subdomains of the scale, but none of the 37 subdomains of the MBA relate to head size measurements. Neither level of mutation type was associated with Z-score, MBA, or CSS. Previous reports have suggested that individuals with missense mutations are generally less affected than those with nonsense mutations (Kortum et al., 2011; Mitter et al., 2018; Brimble et al., 2023). While this study did not recapitulate this finding using strict statistical significance, we did find that missense individuals with seizures had the lowest median CSS and MBA scores (19.5 and 37.0 respectively vs. 30.0 and 51.5 overall) and the median Z-score closest to 0 of any mutation group (−1.71 vs. −3.76 overall) (Table 2), all of which suggest a milder phenotype. The lack of a statistically significant genotype-phenotype correlation may have been due to the large variability in the 10 and 11 missense individuals with CSS and MBA scores available compared to the variability of the overall population (Table 2). Given the relationship between HC Z-score and severity, a larger sample of missense individuals would likely replicate results showing missense individuals are significantly milder.

4.2. Seizures:

This study finds that 78.7% of FOXG1 individuals have at least one distinct seizure, within the previously reported range of 61–87% (Table 1) (Brimble et al., 2023; Seltzer et al., 2014). We also present a similarly broad set of seizure types to those in the literature (Table 3) (Kortum et al., 2011; Mitter et al., 2018; Brimble et al., 2023; Seltzer et al., 2014; Vegas et al., 2018). The 1.0-year median age at first seizure onset is earlier than the published mean of 2–3 years (Table 2) (Mitter et al., 2018; Seltzer et al., 2014; Vegas et al., 2018). We report median because our data is skewed, but the mean value of 1.9 years is consistent. In accordance with Vegas et al., half of all individuals had more than 1 distinct seizure (Table 1) (Vegas et al., 2018). Only 1 of 3 seizures for duplication individuals were infantile spasms, but since there were only 2 duplication individuals, this finding does not refute the result that duplication individuals mainly have infantile spasms (Table 3, Supplementary Table 2) (Seltzer et al., 2014). We also could not confirm that duplication individuals had significantly earlier seizure onset than those with intragenic mutations or deletions, but the duplication individuals did have a median age at onset of 0.5 years, which is within the 3–7-month range described previously (Table 2) (Mitter et al., 2018; Seltzer et al., 2014; Vegas et al., 2018). Our data largely matches key findings from earlier research.

This study adds new depth to the epilepsy phenotype in FOXG1 syndrome. Considering each characteristic independently, over 70% of seizures are partial or tonic-clonic, occur on a weekly basis or less, and last less than 5 minutes on average – over 50% last less than 1 minute (Table 2, Figure 3A, 4A). Only 1/3 of seizures resolve within a median 4–5-years of follow-up, and of those that resolve, the median time to resolution is 1–2 years at a median age of 3–4 years old (Table 1, 2, 3). Over half of individuals have more than 1 distinct seizure (Table 1). Overlapping seizures of the same type are due to clinical assessment of distinctness or recategorization (see section 2.2), but this did not affect statistical analysis since we only used first seizure data (Figure 2B). Sex and mutation type are not associated with when seizures first start, their type, their frequency, their duration, or their resolution status and count; onset, type, frequency, and duration are not correlated with downstream characteristics either (Supplementary Table 1). The CSS data recorded average seizure frequency across visits enabling assessment of the relationship between age and average seizure frequency rather than just age at seizure onset and first seizure frequency. A linear relationship was insignificant, but a non-linear GAM was narrowly significant with p = 0.0123 versus a threshold for significance of p < 0.0125 (Supplementary Table 1). However, age only explains 9.7% of the variance in average seizure frequency, so age is not a powerful predictor on its own. There is also no obvious pathophysiological reason why the relationship would be negative from age 0–7, positive from 7–21, and then negative again from 21 on (Figure 9). Rather, the modeled relationship may instead be driven by decreased data density beyond about age 13 (Figure 9). More and more evenly spread data is needed to better characterize any potential relationship. Overall, across datasets, there is no clear association between age and age at seizure onset and average seizure frequency.

4.3. Limitations:

This study has limitations. The CSS and MBA measures of severity were designed for RTT not FOXG1, but, considering the historical categorization of FOXG1 as a RTT variant and that CSS has been shown to distinguish developmental encephalopathies including FOXG1 and RTT, use of these scales is appropriate (Cutri-French et al., 2020). Further work is needed to validate and adapt these scales for interventional trials. Another limitation is potential bias and inaccuracy in seizure type and feature classification. This study had clinicians use their own assessments and caregiver reports to categorize seizure type and code for features. While this medical interpretation should reduce some bias and inaccuracy from caregiver recall, it also potentially introduces new clinician or site-level biases. Across features, while certain seizure types like tonic and atonic are clearly distinct, partial and tonic-clonic overlap (Figure 1B). We did not find significant associations with seizure type, and it is possible that the seizure categories used do not reflect the true epilepsy phenotype in FOXG1. Electroencephalograms would both help more objectively define seizure type and determine if there are unique seizure semiologies for FOXG1 that should be used as categories instead of the typical International League Against Epilepsy terminology. Another limitation was missing data in key variables. Day, month, and year of seizure onset were all missing for approximately 10% of seizures (10/74 first seizures and 15/153 all seizures) (Table 1, 3). Many other seizure dates had missing months or days requiring assigning dates by the rules described in the methods section. CSS and MBA scores were missing for 30 and 29 of 94 study participants, including all 26 CHOP and CHCO individuals (Table 1). Six participants were missing HC Z-scores (Table 1). This study also did not investigate the impacts of anti-epileptic drugs, which are used prominently in FOXG1 without an understanding of their effect on seizures – except that infantile spasms in duplication individuals are often responsive to treatment (Brimble et al., 2023; Seltzer et al., 2014). Additional work in this area will be important for developing treatment algorithms.

5.0. Conclusions:

This study analyzed overall severity and epilepsy features in a large cohort of FOXG1 syndrome individuals. For individuals with intragenic mutations or deletions and seizures, we demonstrate that smaller head circumferences are associated with more severe disease. We corroborated several findings about epilepsy in FOXG1 and add new detail about seizure type, frequency, duration, resolution, count, and other features. This information will aid clinicians in providing more accurate prognoses to individuals and inform clinical trial development.

Highlights:

• 78.7% of participants had seizures starting at a median age of 1.0 years.

• 1/3 of seizures resolved and did so in a median time of 1.1 years at 3.3 years old.

• Over 70% of seizures were partial or tonic-clonic in nature.

• For those with seizures, smaller heads were associated with more severe disease.