CUX2 in the Medical Literature - A Consolidated Review of Findings From 11 Research Reports and One Research Summary

Prepared by the CUX2 Foundation (October 2026)

Important Notice About This Report

This report is intended to give parents and family members of children affected by CUX2 a central place to learn about the CUX2 gene, its variants and the research published on them. It brings together the findings of 11 research reports on CUX2 identified by the CUX2 Foundation, along with one summary of one of those reports (Chatron et al.) written for the public by CURE Epilepsy. It complements the foundation’s parent-led natural history study, ORIGINS. ORIGINS records what families report; this review records what researchers have published. This report has not undergone clinical or scientific peer review.

Nothing in this report or analysis should be used as medical advice, diagnosis or treatment. It does not replace the guidance of a qualified health care provider. Families should consult their child’s neurologist, geneticist or other qualified health care provider about any medical question or decision.

The reports come from research groups in different countries, working at different times. Some reports state that they include patients first described in earlier reports. Patients studied by different groups may or may not be the same people, and different people may have very similar descriptions. For these reasons, this review does not combine patient descriptions from different reports and does not give a combined number of individuals. Where a report states that a patient was described earlier, that statement is reported as the authors made it.

CUX2 at a Glance

• CUX2 is a gene that encodes a transcription factor, a protein that controls other genes. It is active mainly in neurons of the outer layers (layers 2 and 3) of the cerebral cortex.[2,8]
• Variants in CUX2 are associated with developmental and epileptic encephalopathy 67 (DEE67), an autosomal dominant disorder.[3,5]
• Chatron and colleagues describe E590K as a recurrent variant, and Romano and colleagues state that 10 individuals have been reported with it.[3,8] Chatron and colleagues, describing nine patients with E590K, report a spectrum that is predominantly generalized epilepsy, from infantile-onset myoclonic DEE at the severe end to generalized epilepsy with severe static developmental encephalopathy at the milder end.[8]
• Other reported variants include p.Thr945Met in one patient,[5] a frameshift variant in one patient (preprint),[3] a de novo nonsense variant, p.(Arg633*), in one patient from an autism study, cited by Barington and colleagues,[7] and five missense variants in Japanese patients with epilepsy, most with temporal lobe epilepsy.[6]
• Seven of the 11 reports describe patients with CUX2 variants. An eighth, the Epi4K study, does not name CUX2 in the supplied text; later reports cite it as the source of one patient with E590K.[7,8] Nationalities or ethnicities stated in the reports are Japanese, Chinese, Danish and German.[5,6,7] Because some reports include patients first described elsewhere, this review lists each report's patients separately and does not give a combined total.
• Suzuki and colleagues propose increased excitatory signaling, found in mice lacking Cux2, as a possible mechanism for temporal lobe epilepsy.[6] Morcom and colleagues state that DNA damage accumulation could partly explain why CUX2 dysfunction is associated with seizure susceptibility.[1]
• The authors state that functional studies of patient variants and more case reports are still needed.[5,6,8]

Part 1: CUX2 in the Research Literature

1. What CUX2 Is

CUX2 is a gene that encodes a transcription factor of the CUT homeobox family.[8,11] The CUX2 protein contains three DNA-binding regions called CUT domains and one homeodomain, which are important for DNA binding.[3,8] Its closely related gene is CUX1; Cux1 and Cux2 are the vertebrate counterparts of the fruit fly gene Cut.[8]

CUX2 is expressed primarily in neuronal tissue during development and continues to be expressed in postmitotic neurons, where it is a marker of neurons in the upper layers (layers 2 and 3) of the cerebral cortex.[8] Suzuki and colleagues, citing earlier studies, state that CUX2 controls neuronal proliferation, dendrite branching, spine morphology and synapse formation.[6] Mice lacking Cux2 had reduced dendrite length and branching, lower spine density, reduced synaptic function and defects in working memory.[8]

Excitatory neurons in cortical layers 2 and 3 expanded disproportionately during mammalian evolution.[2] Xia and colleagues note that loss of CUX2-expressing neurons has been reported in multiple sclerosis, head trauma in young athletes, Alzheimer's disease and frontotemporal dementia. They add that dysfunction of layer 2 and 3 neurons is consistently linked to temporal lobe epilepsy and schizophrenia, and that these neurons show the most pathway dysregulation in human autism.[2]

In population databases, CUX2 is intolerant to variation, with a probability of loss-of-function intolerance of 1 and a missense Z-score of 3.61.[8] Variants in CUX2 are associated with developmental and epileptic encephalopathy 67 (DEE67), an autosomal dominant disorder.[3,5]

2. What the Condition Looks Like

Figures in this section are given separately for each report and are not combined.

Seizures

In the nine patients with E590K described by Chatron and colleagues, the median age at seizure onset was 6 months, ranging from 2 months to 9 years. Seizures were frequent at onset, with multiple seizures per day, and the most common types at onset were myoclonic seizures or absence seizures with a myoclonic component. Other seizure types often emerged over time, including atypical absences, tonic seizures and generalized tonic-clonic seizures. Six patients had a developmental and epileptic encephalopathy (myoclonic DEE in three, Lennox-Gastaut syndrome in two, West syndrome in one), two had a severe static developmental encephalopathy with genetic generalized epilepsy, and one infant had multifocal epilepsy.[8]

The patient with E590K described by Barington and colleagues had generalized and myoclonic seizures from 12 months, which appeared to be provoked by stress and lack of sleep.[7] The boy with p.Thr945Met had three febrile seizures beginning at age 2.[5] The girl with the frameshift variant had a single generalized seizure at 4 years 7 months.[3] Eight of the nine Suzuki patients had temporal lobe epilepsy.[6]

EEG and brain imaging

In the Chatron report, EEG was abnormal at seizure onset in all patients, most often showing generalized spike-wave or polyspike-wave discharges. Brain MRI was normal in six of nine; the others showed cerebellar atrophy, hippocampal asymmetry and a thin corpus callosum.[8] The boy with p.Thr945Met had discharges during sleep at 2 years 9 months, a normal EEG at 5 years and a normal MRI.[5] The girl with the frameshift variant had multifocal EEG anomalies activated by sleep, and her MRI showed small areas of gray matter heterotopia.[3] In the Suzuki report, all patients with mesial temporal lobe epilepsy had hippocampal sclerosis.[6]

Development, communication and behavior

Chatron and colleagues report that seven of their nine patients had severe intellectual disability and were nonverbal, one had profound impairment, and one was too young to assess. Seven could walk. Two had marked developmental regression that was not related to frequent seizures or epileptiform activity. Two had autistic features; the abstract states six (see Appendix C). The authors state that the cognitive impairment reflects a developmental component, together with an epileptic encephalopathy in those with a DEE.[8]

The Barington patient was diagnosed at age 4 with moderate to severe intellectual disability and infantile autism, and by late adolescence had no language.[7] The boy with p.Thr945Met had a developmental quotient of 56 at 2 years 9 months and 75 at 5.5 years, after three years of home-based rehabilitation.[5] The girl with the frameshift variant had an IQ of 40 at age 8, described as moderate intellectual disability.[3]

Movement and other features

Chatron and colleagues report movement disorders in five of nine patients, often stereotypies such as hand flapping, dyskinesia or athetosis. Inappropriate laughter episodes occurred in three; in two, these had no EEG correlate, and the episodes were not thought to be epileptic in origin.[8] Hypersalivation and constipation were reported in the Barington patient.[7] The girl with the frameshift variant had subtle facial features that the authors state had not been highlighted in other CUX2 patients.[3]

3. What Research Suggests Is Happening in the Brain

Effect on DNA binding

Based on computer analyses, Chatron and colleagues state that their results suggest E590K may alter the stability or specificity of how CUX2 binds DNA, and that functional studies are needed to determine its effect.[8] Zhang and colleagues predict that p.Thr945Met could alter the DNA-binding specificities of CUX2 and state that molecular functional studies will be required.[5]

Increased excitatory signaling

Suzuki and colleagues found that mice lacking Cux2 did not have spontaneous seizures but were more susceptible to seizures induced by the chemical kainate. These mice had more excitatory neurons in the entorhinal cortex and increased excitatory signaling from the entorhinal cortex to the hippocampus. The authors propose this as a possible mechanism for temporal lobe epilepsy.[6]

DNA repair

Barington and colleagues, citing earlier laboratory work, state that the results suggest CUX2, like CUX1, functions as an accessory factor in the repair of oxidative DNA damage, and that it could be speculated that a variant affecting a CUT domain leads to oxidative DNA damage.[7] Morcom and colleagues found that in mice, Cux2 is essential to limit DNA damage and protect layer 2 and 3 neurons against injury, and that adding extra CUX2 and ATF4 reduced DNA breaks in human neurons grown in the laboratory from stem cells.[1] Xia and colleagues found that in developing mice, the protein ATF4 is required for the progenitor cells that give rise to CUX2-expressing neurons to repair DNA damage and survive.[2] Morcom and colleagues state that DNA damage accumulation could partly explain why CUX2 dysfunction is associated with seizure susceptibility.[1]

What these studies did and did not test

In the Suzuki and Morcom studies, Cux2 was inactivated in mice; these mice do not carry patient variants.[1,6] Suzuki and colleagues' kainate and synaptic findings were in mice lacking both copies of Cux2.[6] Xia and colleagues studied mice lacking Atf4 in the cortex.[2] Suzuki and colleagues tested five patient variants in human cells and fruit flies.[6] In the 11 reports reviewed, none tested a CUX2 patient variant in a mouse model.

4. The Variants Reported

E590K

E590K (c.1768G>A, p.Glu590Lys) changes one amino acid in the first CUT domain. The affected position is fully conserved across 19 species, and a negative charge at this position is conserved across all human CUT domains.[8] The variant was not found in the ExAC or gnomAD population databases and was predicted to be damaging by several computational tools.[7,8] Rauch and colleagues list their patient's CUX2 change as g.111748354G>A (hg19) without a protein change; Chatron and colleagues give the same genomic change, Chr12:111748354G>A (hg19), for E590K and state that E590K was first reported in single cases from the Rauch and Epi4K studies.[8,11]

Chatron and colleagues report that protein stability prediction tools suggested the change could lead to destabilization of CUX2, while computer modeling suggested no major change in the alpha helix; in a related protein crystallized with DNA, the equivalent position faces the DNA. They state that the variant may alter how CUX2 binds DNA and may cause a partial loss of function, with gain of function or a dominant negative effect as alternative possibilities.[8] Barington and colleagues state that it is not known whether the variant causes loss or gain of function, and that its recurrence makes it highly likely that the affected nucleotide, c.1768G, is a mutational hot spot.[7]

Chatron and colleagues report that, of their nine patients, two were identified through clinical diagnostic testing, three of 1,222 through research studies of patients with DEE, and four of 2,207 through screening of patients with intellectual disability with or without epilepsy.[8]

p.Thr945Met

This variant is in the second CUT domain and changes a highly conserved threonine. It was found de novo in one boy and classified as likely pathogenic. Modeling predicted that it could alter DNA-binding specificities, and the authors state that functional studies are required to confirm the mechanism.[5]

Frameshift variant c.633_636del, p.(His211Glnfs*3) (preprint)

This de novo variant was found in one girl. The authors state it is predicted to produce a truncated, nonfunctional protein lacking all annotated DNA-binding domains, classify it as likely pathogenic, and note that only one frameshift variant is listed as likely pathogenic for DEE67 in ClinVar.[3] Earlier, Chatron and colleagues had stated that no truncating variants had been observed in patients or controls,[8] while Barington and colleagues cited a de novo nonsense variant, c.1897C>T, p.(Arg633*), reported in a cohort study of patients with autism spectrum disorder, for which no other clinical data were available.[7]

Variants in temporal lobe epilepsy

Suzuki and colleagues found five CUX2 missense variants in nine of 271 Japanese patients with epilepsy, eight of whom had temporal lobe epilepsy. Three variants were absent or rare (less than 0.5%) in Japanese controls and databases and were predicted to be damaging. One, p.E1283K, which the authors describe as a frequent variant predicted to be less damaging, was found at a significantly higher rate in patients with temporal lobe epilepsy than in controls, and the authors hypothesized that it may also contribute to temporal lobe epilepsy. In human cells, four of the five variants (all except p.R34W) showed abnormal localization of the protein. In fruit fly neurons, three of the five (all except p.R34W and p.D337N) had a reduced ability to increase dendrite branching. The authors state that these observations suggest the variants cause loss of function.[6]

5. Treatment Experience

Treatments reported

The following treatments were given to patients with CUX2 variants in the reports reviewed. The list does not show how many patients received each treatment or whether it was effective. Each patient's treatments, where reported, are given in Appendix A. The type labels give general categories added by this review; they are not drawn from the sources.

• Acetazolamide[8]: Antiepileptic drug
• Carbamazepine[8]: Antiepileptic drug
• Clobazam[8]: Antiepileptic drug
• Clonazepam[8]: Antiepileptic drug
• Diazepam[8]: Antiepileptic drug
• Ethosuximide[8]: Antiepileptic drug
• Felbamate[8]: Antiepileptic drug
• Lacosamide[8]: Antiepileptic drug
• Lamotrigine[7,8]: Antiepileptic drug
• Levetiracetam[5,8]: Antiepileptic drug
• Nitrazepam[8]: Antiepileptic drug
• Oxcarbazepine[8]: Antiepileptic drug
• Perampanel[8]: Antiepileptic drug
• Phenobarbitone[8]: Antiepileptic drug
• Phenytoin[8]: Antiepileptic drug
• Rufinamide[8]: Antiepileptic drug
• Stiripentol[8]: Antiepileptic drug
• Sulthiame[8]: Antiepileptic drug
• Topiramate[8]: Antiepileptic drug
• Valproate (valproic acid)[3,7,8]: Antiepileptic drug
• Vigabatrin[8]: Antiepileptic drug
• Zonisamide[8]: Antiepileptic drug
• Synacthen[8]: Hormone treatment
• Prednisolone[8]: Steroid
• Ketogenic diet[8]: Dietary therapy
• Vagal nerve stimulation[8]: Device therapy

What the reports say about response

In the Chatron report, valproate fully controlled seizures in two patients; one patient's absence seizures were initially well controlled on valproate until an episode of nonconvulsive status epilepticus at age 3, with rare episodes every few years; and five patients were refractory to multiple antiepileptic drugs, including valproate. Table 2 of the report lists 24 different medicines and hormone treatments across the nine patients, as well as vagal nerve stimulation and the ketogenic diet; each patient's treatments are listed in Appendix A.[8]

In the Barington patient, lamotrigine and valproate reduced seizure frequency, and treatment was stopped at age 13 with occasional relapses.[7] The boy with p.Thr945Met had no further seizures after starting levetiracetam, which was later stopped.[5] The girl with the frameshift variant has been seizure-free on valproic acid.[3]

In the 11 reports reviewed, none describes a controlled treatment trial in people with CUX2 variants.

6. How CUX2 Came to Be Studied

2012. Rauch and colleagues sequenced the exomes of 51 children with severe nonsyndromic intellectual disability, and their parents, from 10 centers in Germany and Switzerland. One boy, patient 07E0967, had a de novo CUX2 missense change, which the authors listed among the probable disease-causing de novo variants. They named CUX2 among the genes that "might also cause intellectual disability."[11]

2013. The Epi4K Consortium sequenced the exomes of 264 patients with infantile spasms (149) or Lennox-Gastaut syndrome (115), and their parents, and confirmed 329 de novo mutations.[10] The supplied text does not name CUX2. Later reports cite this study as the source of one patient with the E590K variant.[7,8]

2017. Geisheker and colleagues sequenced 20 recurrently mutated sites, including CUX2 codon 590, in 17,689 patients with autism spectrum disorder or developmental delay. They found one additional patient with E590K, whose inheritance could not be determined. Their Table 1 also lists two earlier de novo mutations at this codon in the denovo-db database.[9]

2018. Chatron and colleagues described nine patients with E590K, seven new patients collated through the Matchmaker Exchange network or personal communication with colleagues and two previously reported, and characterized the associated epilepsy spectrum. Gemma L. Carvill of Northwestern University and Gaetan Lesca of Lyon University Hospital are the corresponding authors, and the acknowledgments state that Carvill was supported by National Institutes of Health grant NS089858 and a CURE Taking Flight Award.[8] CURE Epilepsy published a summary of the study for the public, which states that Carvill began her research into the genetic causes of childhood epileptic encephalopathy in 2015 with a CURE Taking Flight Award.[12] In the same year, Barington and colleagues described a Danish girl with E590K, whom they reported as the third known case.[7]

2022. Suzuki and colleagues sequenced CUX2 in 271 Japanese patients with epilepsy and studied Cux2 knockout mice.[6] Zhang and colleagues described a boy with a different CUX2 variant, p.Thr945Met.[5]

2023. Oppermann and colleagues described 34 individuals with variants in the related gene CUX1.[4]

2025. Romano and colleagues posted a preprint describing a girl with a CUX2 frameshift variant.[3]

2026. Morcom, Xia and colleagues published two related studies on the role of CUX2-expressing neurons and DNA repair.[1,2]

7. The Individuals Studied

Seven of the 11 reports describe patients with CUX2 variants. The Epi4K study is also included in the list below because later reports cite it as the source of one patient. The list shows, for each report, the patients it describes, any earlier patients it cites, and the nationality or ethnicity the report states. Each report's patients are described in full in Appendix A, in the form that report published them. Counts are not added across reports, for the reasons given in About This Report.

Rauch 2012[11]. Patients described: 1 (patient 07E0967). Earlier patients cited: None. Variant: CUX2 missense, g.111748354G>A (hg19). Nationality or ethnicity stated: Not stated. 

Epi4K 2013[10]. Patients described: Not identified individually in the supplied text. Earlier patients cited: None. Variant: Not named. Nationality or ethnicity stated: Not stated. 

Geisheker 2017[9]. Patients described: 1 with E590K. Earlier patients cited: 2 de novo mutations at codon 590 listed in the denovo-db database; individual reports not named. Variant: E590K. Nationality or ethnicity stated: Not stated. 

Chatron 2018[8]. Patients described: 9 (the authors describe 7 as new and 2 as previously reported). Earlier patients cited: Cites earlier reports beside Patients 1, 6 and 9. Variant: E590K. Nationality or ethnicity stated: Not stated. 

Barington 2018[7]. Patients described: 1. Earlier patients cited: 3: 2 with E590K, cited to Rauch and Epi4K; 1 with a nonsense variant, cited to an autism study. Variant: E590K (own patient). Nationality or ethnicity stated: Danish (own patient); German (earlier patient cited to Rauch). 

Suzuki 2022[6]. Patients described: 9. Earlier patients cited: Refers to E590K patients reported by Chatron and Barington, without a count. Variant: Five missense variants. Nationality or ethnicity stated: Japanese. 

Zhang 2022[5]. Patients described: 1. Earlier patients cited: 10, attributed to Chatron, Barington, Rauch, Epi4K and Geisheker. Variant: p.Thr945Met. Nationality or ethnicity stated: Chinese family. 

Romano 2025[3]. Patients described: 1. Earlier patients cited: Refers to 10 earlier E590K patients as a count and to 1 patient with p.(Thr945Met); Figure 1 marks five missense variants previously reported in patients with epilepsy. Variant: Frameshift. Nationality or ethnicity stated: Not stated. 

Oppermann 2023[4], Morcom 2026[1] and Xia 2026[2] do not describe individual patients with CUX2 variants. The supplied Epi4K text does not name CUX2.[10]

The Rauch study enrolled patients from centers in Germany and Switzerland.[11] The Epi4K study enrolled patients at 23 clinical sites whose countries are not listed.[10] The Geisheker study was led from the University of Washington, with samples from sites in Australia, Belgium, China, the Czech Republic, Italy, the Netherlands, Sweden and the United States.[9] The Chatron study was a collaboration of authors in Australia, Denmark, France, Germany, Italy, the Netherlands, Sweden and the United States.[8] The Suzuki study was conducted in Japan,[6] the Zhang study in Ganzhou, China, with sequencing in Beijing,[5] and the Romano study in Genoa, Italy.[3] The supplied Barington text does not state where that study was conducted.[7]

8. CUX2 and Its Related Gene CUX1

Oppermann and colleagues describe 34 individuals with variants in CUX1. Speech delay was present in 31 of 32, motor delay in 24 of 32, seizures in eight of 32 (all but two became seizure-free), and autism spectrum disorder in six of 26. Six of 27 had IQ in the normal range, and seven of 18 with assessable information showed developmental catch-up. After kainic acid (kainate), mice carrying one Cux1 null allele developed severe seizures that wild-type mice did not, with no difference in time to the intermediate seizure stage.[4] The report notes high sequence conservation between CUX1 and CUX2 and cites the Suzuki report, but does not compare the severity of the two conditions.[4]

9. What Remains Unknown

The authors identify the following open questions: the functional effect of E590K, to be determined through laboratory and animal studies[8]; how often E590K causes intellectual disability and whether other CUX2 variants cause a similar condition[7]; the relationship between p.Thr945Met and the condition, which requires more case reports[5]; whether the variants found in temporal lobe epilepsy are true loss-of-function changes, which requires mouse models carrying those variants[6]; genotype and phenotype correlations, especially for variants other than E590K[3]; and whether the vulnerability of CUX2-expressing cells is due to heightened sensitivity to reactive oxygen species, to the cells producing more of them, or both.[2]

Part 2: Appendices

Appendix A. Patient Descriptions by Report

Reports are listed in order of publication. Each patient is described as the report published them. Information a report gives about all of its patients appears once at the start of that report's section. Where a report cites an earlier patient, the citation is reported as the authors made it; this review does not determine whether descriptions in different reports refer to the same person.

Rauch et al., 2012[11]

The study enrolled 51 children with severe nonsyndromic intellectual disability from 10 centers in Germany and Switzerland and sequenced their exomes and those of their parents. Inclusion criteria were severe intellectual disability with grossly preserved motor function, absence of malformations, absence of syndrome-specific minor anomalies, absence of specific neurological signs, absence of causative copy number variants by high-resolution molecular karyotyping, nonconsanguineous parents and sporadic disease. All de novo variants detected by exome sequencing were validated by Sanger sequencing.

Patient 07E0967

• Sex: Male
• Variant: De novo CUX2 missense change, chromosome 12: g.111748354G>A (hg19); PolyPhen probability 0.999. The protein change is not stated. Chatron et al. give the same genomic change for E590K and state that E590K was first reported in single cases from this study and the Epi4K study.[8]
• Classification: Listed in Table 4, "Probable disease-causing de-novo variants in each patient-parent trio"
• Gene information: Haploinsufficiency index 40.0%; smallest region of overlap: deletion, 54 genes, 4.8 Mb; very high brain expression; mouse phenotype: MGI:107321, homozygotes for a targeted null mutation have various neural defects; description: "Belongs to the CUT homoeobox family, thought to be a transcription factor involved in neural specification"
• Clinical details: Not reported in the main text. The online appendix was not supplied.

Source: Summary; Methods; Table 4

Epi4K Consortium, 2013[10]

The study sequenced the exomes of 264 patients with infantile spasms (149) or Lennox-Gastaut syndrome (115), and their parents. Patients were enrolled through the Epilepsy Phenome/Genome Project across 23 clinical sites. The supplied text does not identify a patient with a CUX2 variant individually. The full list of de novo mutations is in Supplementary Table 2, which was not supplied.

Geisheker et al., 2017[9]

The study sequenced 20 recurrent missense sites in 17,689 patients with a primary diagnosis of autism spectrum disorder or developmental delay, including intellectual disability, and 3,023 unaffected siblings.

Patient with E590K

• Identified through: Targeted sequencing, Troina site
• Variant: CUX2 p.Glu590Lys
• Inheritance: Unknown. The authors state inheritance could be determined for only eight variants because of missing parental DNA.
• Table 1 values for this codon: Protein ID NP_056082.2. De novo mutations at this codon in denovo-db v.0.9 (N = 8,477): 2. Codon de novo P = 5.38E-07 (genome-wide correction: 1). Total de novo: 2; codon de novo P = 5.12E-06 (genome-wide correction: 1). ExAC v.0.3 allele count (N = 45,376): 0.
• Clinical details: Not reported

Source: Results; Table 1

Table 2 of the report ("Rare clustered missense mutations identified by targeted sequencing (CADD >20)") lists, at CUX2 p.Glu590, one missense variant among 3,023 controls, three among 17,689 cases, five among 26,166 cases in denovo-db and the targeted sequencing combined, and two known de novo. These individuals are not described further.

Chatron et al., 2018[8]

The authors describe nine patients with E590K: "seven additional patients" and "the two individuals previously reported." They report the variant as de novo, and state that the pregnancy and perinatal period were unremarkable in all patients. Nationality, ethnicity and family history are not reported for individual patients. The entries below are reproduced from the report's Tables 1 to 3, with medicine abbreviations written out in full; head circumference is given in centiles or percentiles as printed. CURE Epilepsy's summary of this report describes the same nine patients.[12]

Patient 1

The authors cite Rauch et al., 2012, beside this patient.

• Sex; age at study: Male; 8 years
• First seizures: 6 months
• Seizure type at onset: Focal spasms
• Course: Seizure-free
• EEG at onset; later: Hypsarrhythmia; not done
• Medicines tried: Sulthiame, valproate, Synacthen for 3 months with improvement
• Current treatment: Valproate
• Epilepsy syndrome: DEE: West syndrome
• Brain MRI: Normal (8 months)
• Head circumference: 50 cm (7 years, 50th centile)
• Intellectual disability: Severe
• Other features: Nonverbal; movement disorder: stereotypies (hand flapping)

Source: Tables 1 to 3

Patient 2

• Sex; age at study: Male; 19 years
• First seizures: 7 months
• Seizure type at onset: Myoclonic seizures and right occipital seizures with apnea
• Course: 30 months: 1 febrile tonic seizure. 3 years: 2 focal seizures. 4 years: myoclonic seizures. 7 years: generalized tonic-clonic and myoclonic seizures
• EEG at onset; later: Generalized spike-wave, sometimes with myoclonic seizures; right occipital seizure recorded. Later: 3 Hz generalized spike-wave, absence seizures with myoclonic components
• Medicines tried: Valproate, clobazam
• Current treatment: Valproate
• Epilepsy syndrome: Myoclonic DEE
• Brain MRI: Normal (8 months)
• Head circumference: 51 cm (13 years, 25th centile)
• Intellectual disability: Never normal; regression at 7 months and at 12 years; profound
• Other features: Nonverbal; movement disorder: athetosis, dystonia, stereotypies; hypotonic at 1 year, then spastic tetraparesis from 12 years with loss of ambulation

Source: Tables 1 to 3

Patient 3

• Sex; age at study: Male; 21 years
• First seizures: 6 months
• Seizure type at onset: Atypical absences with myoclonus
• Course: Atypical absences, tonic seizures with myoclonic component, atonic seizures, generalized tonic-clonic seizures
• EEG at onset; later: 3 to 4 Hz generalized spike-wave; generalized polyspike-wave. Later: background slowing; frequent trains of 1.5 to 2 Hz and 3 to 4 Hz generalized spike-wave; myoclonic seizures with generalized spike-wave
• Medicines tried: Valproate, oxcarbazepine, carbamazepine, lamotrigine, vigabatrin, zonisamide, phenytoin, nitrazepam
• Current treatment: Levetiracetam, clobazam, valproate, zonisamide; vagal nerve stimulation (beneficial)
• Epilepsy syndrome: DEE: Lennox-Gastaut syndrome
• Brain MRI: Cerebellar atrophy (6 months)
• Head circumference: 56 cm (21 years, 73rd percentile)
• Intellectual disability: Severe
• Other features: Nonverbal; no eye contact; inappropriate laughter episodes

Source: Tables 1 to 3

Patient 4

• Sex; age at study: Female; 9 years
• First seizures: 6 months
• Seizure type at onset: Myoclonic
• Course: Myoclonic seizures, tonic seizures, generalized tonic-clonic seizures
• EEG at onset; later: Generalized spike-wave and polyspike-wave. Later: myoclonic seizures with generalized polyspike-wave; tonic seizures with generalized polyspike
• Medicines tried: Levetiracetam, clonazepam, ethosuximide, carbamazepine, lacosamide, rufinamide
• Current treatment: Ethosuximide, clobazam, lacosamide, rufinamide; ketogenic diet
• Epilepsy syndrome: DEE: Lennox-Gastaut syndrome
• Brain MRI: Hippocampal asymmetry (6 years 3 months)
• Head circumference: 52 cm (9 years, 44th percentile)
• Intellectual disability: Never normal; severe regression at 8 years; severe
• Other features: Nonverbal; decreased reflexes; ataxic gait

Source: Tables 1 to 3

Patient 5

• Sex; age at study: Female; 14 years
• First seizures: 12 months
• Seizure type at onset: Absence seizures
• Course: Myoclonic seizures; nonconvulsive status epilepticus. The text adds that absence seizures were initially well controlled on valproate until nonconvulsive status epilepticus at age 3, with rare episodes every few years.
• EEG at onset; later: 3 Hz generalized spike-wave. Later: slow background; absence seizure with 3 Hz generalized spike-wave; fast irregular generalized spike-wave; generalized polyspike-wave
• Medicines tried: Valproate
• Current treatment: Valproate, lamotrigine
• Epilepsy syndrome: Severe developmental encephalopathy with genetic generalized epilepsy (early-onset absence epilepsy and myoclonic seizures)
• Brain MRI: Normal (5 years)
• Head circumference: 54 cm (14 years, 50th centile)
• Intellectual disability: Severe
• Other features: Movement disorder: stereotypies (hand flapping); obsessional features

Source: Tables 1 to 3; Results

Patient 6

The authors cite the Epi4K study beside this patient, and state that Geisheker et al. included molecular data from Patients 6 and 9.

• Sex; age at study: Male; 12 years
• First seizures: 2 months
• Seizure type at onset: Myoclonic seizures
• Course: Atonic seizures; myoclonic and absence seizures with eye fluttering; generalized tonic-clonic seizures
• EEG at onset; later: Left temporal spikes and sharp waves. Later: slow background with multifocal epileptiform discharges
• Medicines and therapies tried: Carbamazepine, valproate, clobazam, acetazolamide, lamotrigine, felbamate, diazepam, levetiracetam, clonazepam, topiramate, prednisolone, vagal nerve stimulation, ketogenic diet
• Current treatment: Valproate, clonazepam, topiramate
• Epilepsy syndrome: Myoclonic DEE
• Brain MRI: Normal (6 months)
• Head circumference: 51 cm (7 years, 50th centile)
• Intellectual disability: Severe
• Other features: Ataxic gait; movement disorder: stereotypies (hand flapping); inappropriate laughter episodes

Source: Tables 1 to 3

Patient 7

• Sex; age at study: Male; 16 years
• First seizures: 5 months
• Seizure type at onset: Myoclonic seizures
• Course: Myoclonic seizures, especially in the mornings; myoclonic drop attacks
• EEG at onset; later: Generalized spike-wave and polyspike-wave. Later: generalized spike-wave with myoclonic seizure; during sleep, continuous 1 to 2 Hz generalized spike-wave
• Medicines tried: Carbamazepine, valproate, acetazolamide, lamotrigine, topiramate, clonazepam, levetiracetam, rufinamide, lacosamide, clobazam, perampanel
• Current treatment: Valproate, clobazam, stiripentol
• Epilepsy syndrome: Myoclonic DEE
• Brain MRI: Normal (16 months and 15 years)
• Head circumference: 49.5 cm (3.5 years, 25th centile)
• Intellectual disability: Severe
• Other features: Nonverbal; autistic features

Source: Tables 1 to 3

Patient 8

The Results state: "Patient 8 had focal epilepsy but, as he is only 6 months old, he is too young to predict his developmental outcome." Table 2 lists multifocal epilepsy.

• Sex; age at study: Male; 6 months
• First seizures: 2 months
• Seizure type at onset: Generalized tonic-clonic seizures; atypical absences with apnea and myoclonus; right hemiclonic seizures
• Course: Generalized tonic-clonic seizures; focal impaired awareness seizures; 1 febrile seizure after vaccination
• EEG at onset; later: Multifocal discharges, mainly independent bi-parieto-temporal. Later: slowing and sharp waves in the left temporal region
• Medicines tried: Valproate, diazepam, phenobarbitone, levetiracetam
• Current treatment: Valproate
• Epilepsy syndrome: Multifocal epilepsy, too early to diagnose DEE
• Brain MRI: Normal (age printed as 15 years)
• Head circumference: 41 cm (6 months, 15th centile)
• Intellectual disability: Too young to assess
• Other features: Movement disorder: dyskinesia; inappropriate laughter spells; mild hypotonia

Source: Tables 1 to 3; Results

Patient 9

The authors cite Geisheker et al., 2017, beside this patient, and note that work on Patient 9 was supported by the Italian Ministry of Health and "5 per mille" funding. The Results state that this patient's genetic generalized epilepsy comprised absence seizures, and that "The EEG was normal in patient 9 who was seizure free."

• Sex; age at study: Male; 14 years
• First seizures: 9 years (108 months)
• Seizure type at onset: Absence seizures
• Course: Seizure-free from age 12
• EEG at onset; later: Left fronto-central slowing; later normal
• Medicines tried: Valproate
• Current treatment: Valproate
• Epilepsy syndrome: Severe developmental encephalopathy with absence epilepsy
• Brain MRI: Thin posterior corpus callosum (5 years 8 months)
• Head circumference: 54 cm (14 years, 50th centile)
• Intellectual disability: Severe
• Other features: Nonverbal; autistic features

Source: Tables 1 to 3; Results

The Results also state that Patients 2 and 4 "experienced marked developmental regression during childhood, not related to frequent seizures or epileptiform activity."

Barington et al., 2018[7]

Patient 3 (labeled 2017 in Table 1, "this study")

• Sex; age: Female; 17 years at last follow-up
• Ethnicity: Danish
• Variant: c.1768G>A; p.(Glu590Lys) (NM_015267.3), heterozygous, by whole-exome sequencing. The methods state that selected variants were validated by PCR and Sanger sequencing. Not identified in DNA from blood samples from her parents (de novo). Absent from 245,000 gnomAD alleles. Predicted damaging by SIFT (score 0), probably damaging by PolyPhen2 (score 0.999) and disease-causing by MutationTaster. Submitted to the LOVD database.
• Birth history: Second child of healthy nonconsanguineous parents (mother 32, father 39). Uneventful pregnancy; born at term by uncomplicated vaginal delivery; no neonatal complications; no craniofacial dysmorphology noted.
• Family history: Two relatives with intellectual disability, not considered related to her phenotype: a paternal grandmother's cousin's son, intellectually disabled and dysmorphic, attributed to a chromosome 1/18 translocation; and a maternal grandfather's cousin's daughter's child, with intellectual disability and infantile autism of unknown cause.
• Early development: At 6 months, her mother observed delayed development, especially lack of eye contact and head control.
• Diagnoses: Moderate to severe intellectual disability and infantile autism, at age 4. Files described a "Rett-like" phenotype, but she fulfilled only one Rett criterion (stereotypic movements).
• Seizures: From 12 months: generalized and myoclonic seizures, apparently provoked by stress and lack of sleep. Generalized seizures of 2 to 3 minutes, day and night, with unconsciousness, jerks and twitching of the upper limbs. Myoclonic seizures: intensifying axial spasms up to several times daily.
• Treatment: Lamotrigine and valproate reduced seizure frequency; discontinued at age 13 with only occasional relapses.
• Motor: Delayed; rigid movement pattern, stereotypic movements, toe walking, difficulty getting up from lying.
• Language: None by late adolescence; deeply dependent on her parents.
• Other features: Hypersalivation; chronic constipation
• EEG; MRI: Not reported; not available
• Other testing: Normal: urinalysis, lysosomal enzymes, long-chain fatty acids, phytanic acid, array comparative genomic hybridization, MECP2 and CDKL5 sequencing

Source: Case presentation; Results; Discussion; Table 1

Earlier patients cited in this report

Table 1 of the report also lists two earlier patients with E590K. The authors note that "Only sparse clinical information was available" for them. The Discussion cites a third earlier patient, with a different variant.

• Patient 1, cited to Epi4K (labeled 2012 in Table 1): Gender unknown. Epileptic encephalopathy; intellectual disability or autism spectrum disorder. De novo. Other fields unknown. The authors state that the original report did not state that the variant caused the epileptic encephalopathy, but that none of the patient's other de novo variants were predicted to affect function.
• Patient 2, cited to Rauch (labeled 2013 in Table 1): Male; German. Family history uneventful. Uneventful pregnancy, birth at term, premature rupture of membranes. Moderate/severe intellectual disability (IQ 20 to 49); few words; autistic traits. "BNS" (benign neonatal seizures), onset 6 months. "Motor retardation, at age 15 months, walked at age 3 years." Hypersalivation, hypermetropia, astigmatism, talipes. 7 years at last follow-up. MRI normal at 8 months. De novo.
• Patient with a nonsense variant, cited to an autism cohort study (De Rubeis et al., 2014): De novo nonsense CUX2 variant c.1897C>T, p.(Arg633*). The authors state that no other clinical data were available.

Source: Table 1; Discussion

Suzuki et al., 2022[6]

The study sequenced CUX2 in 271 Japanese patients with epilepsy: 116 with genetic generalized epilepsies and 155 with structural or metabolic epilepsies. The structural or metabolic group included 68 with temporal lobe epilepsy (57 mesial, 11 lateral). Controls were 311 healthy Japanese volunteers; 69 additional patients with temporal lobe epilepsy were used for the p.E1283K frequency comparison. None of the patients with genetic generalized epilepsies had CUX2 variants other than silent variants.

Nine unrelated patients

• Nationality: Japanese
• Variants: Five heterozygous CUX2 missense variants: p.R34W, p.D337N, p.P454L, p.W958R, p.E1283K. Which patient carries which variant is given in the report's Table 1 and Supplemental Note, which were not in the supplied file.
• Diagnosis: Eight with temporal lobe epilepsy (one lateral, seven mesial); all nine in the structural or metabolic epilepsy group. The authors state: "All of the mTLE patients showed hippocampal sclerosis."
• Seizures: The authors state that all patients with CUX family variants showed symptoms of epileptic seizures.
• Variant notes: p.R34W, p.P454L and p.W958R absent or rare (less than 0.5%) in Japanese in-house controls and databases, and predicted to be damaging. p.E1283K (c.3847G>A) described as a frequent variant predicted to be less damaging; found at a significantly higher ratio in temporal lobe epilepsy (P = 5.93 x 10^-3, OR = 6.94, 95% CI 1.39 to 34.61; 137 patients with temporal lobe epilepsy vs. 311 controls); the authors hypothesized it may contribute to temporal lobe epilepsy. p.D337N in one patient with temporal lobe epilepsy and in controls at a similar allele frequency. In Figure 1A, p.W958R lies within the second CUT domain; p.R34W, p.D337N and p.P454L lie before the first CUT domain, and p.E1283K lies after the homeodomain.
• Other details: Age, sex and development are not reported in the main text. Table 1 and the Supplemental Note, which may contain further patient details, were not in the supplied file.

Source: Methods; Results; Discussion

Zhang et al., 2022[5]

Individual 1 ("this study")

• Sex; age: Male; 5.5 years
• Background: From a nonconsanguineous Chinese family; third child; two unaffected sisters
• Variant: NM_015267.4: c.2834C>T, p.Thr945Met, heterozygous, by trio whole-exome sequencing, confirmed by Sanger sequencing. Not carried by the parents or sisters (de novo). Not in 1000 Genomes or ExAC. CADD 26.9. Predicted deleterious by PolyPhen (probably damaging, 1.0), Provean (deleterious, -4.16), SIFT (damaging, 0.0), MutationTaster (disease causing, 1) and REVEL (deleterious, 0.582). Likely pathogenic (ACMG/AMP PS2, PM2, PP2, PP3); the Introduction describes it as pathogenic. ClinVar SCV001946791 (variation 1296981).
• Diagnosis: Presented by the authors as a case diagnosed with DEE67; global developmental delay diagnosed at 2 years 9 months
• Seizures: Febrile seizure at 2 years with unconsciousness, up-rolling of the eyes, facial twitching, limb stretch and no visible limb shaking. Second 6 months later, third 3 months after that. Each lasted 3 minutes with spontaneous remission.
• Treatment: Levetiracetam 40 mg/kg/day from 2 years 9 months; no further seizures. Stopped after 2 years without clinical or electrographic seizures. At last follow-up (5.5 years), seizure-free for 6 months off treatment.
• EEG: At 2 years 9 months: somewhat slower background; a small number of bilateral spikes and spike-wave discharges in central, parietal and posterior temporal regions during sleep. At 5 years: normal.
• Brain MRI: Normal (1.5T)
• Development: Walked at 18 months. At 2 years 9 months, could not form three-word sentences and could say few words with correct meaning; developmental quotient 56 on the Gesell Development Diagnosis Scale (mild delay). After 3 years of home-based rehabilitation, 75 (borderline) at 5.5 years, in kindergarten.
• Other features: Speech delay. Table 1: intellectual disability NA; autism, nonverbal status and movement disorder: no; seizure outcome: seizure-free; EEG: bi-parieto-temporal discharges.

Source: Abstract; Methods; Results; Table 1

Earlier patients cited in this report

Table 1 also lists 10 earlier patients with E590K, attributed by the authors to Chatron et al. (six), Barington et al., Rauch et al., Allen et al. (Epi4K) and Geisheker et al. The clinical values in these rows repeat values published in the Chatron and Barington reports. The Rauch, Epi4K and Geisheker texts do not give these clinical details. The rows are not repeated here. The authors' entries for the patient attributed to Barington differ from the Barington report; see Appendix C.

Romano et al., 2025 (preprint)[3]

"The patient"

• Sex; age: Female; 8 years
• Variant: c.633_636del, p.(His211Glnfs*3) (NM_015267.4; hg38/GRCh38), by trio exome sequencing and Sanger sequencing. Not found in the parents or siblings (de novo). Not in ClinVar or gnomAD. Phred-scaled quality score 1801.29, read depth 8306, mapping quality 60. Likely pathogenic (classified using Franklin and ACMG criteria). The authors note CUX2 has a gnomAD LOEUF of 0.383 and is predicted deleterious by MutationTaster2.
• Diagnosis: Clinical features compatible with DEE67; moderate intellectual disability
• Birth history: Nonconsanguineous, healthy parents. Pregnancy unremarkable except mild maternal hyperglycemia managed with diet. Perinatal period uneventful.
• Family history: Younger sister: epileptic spasms from 6 months, treated with ACTH, resolved at age 4. Older sister: isolated intellectual disability (IQ 49); does not carry the variant.
• Seizures: One seizure at 4 years 7 months, during sleep, about one minute. Described as generalized tonic in the case presentation and generalized brief tonic-clonic in the abstract.
• Treatment: Valproic acid; seizure-free since
• EEG: Multifocal bilateral anomalies, activated by sleep
• Brain MRI: Submillimetric areas of gray matter heterotopia in the right centrum semiovale
• Development: Sat at 5 months; walked at 16 months; first words at 3 years. IQ 40 (WISC-IV) at 8 years.
• Other features: Long palpebral fissures, thin upper lip, dental malocclusion with widely spaced incisors; motor clumsiness, poor fine motor coordination, mild hypotonia, slight oromotor dyspraxia; repeated bronchospasm and lower airway infections requiring hospitalization in her first months

Source: Abstract; Case presentation; Exome sequencing and variant analysis; Discussion

Earlier patients cited in this report

"To date, ten individuals have been reported to carry the same recurrent de novo missense variant, p.(Glu590Lys)." The authors cite Chatron et al. and Barington et al. for these individuals and state that some were first reported in Rauch et al., 2012, the Epi4K Consortium, 2013, and Geisheker et al., 2017. They also refer to one patient with later-onset epilepsy and p.(Thr945Met). Figure 1 of the report marks five missense variants as "previously reported variants appeared in patients with epilepsy": p.R34W, p.D337N, p.P454L, p.W958R and p.E1283K. The authors do not give a source for these variants. No individual data are given.

Appendix B. Summary of Each Source

Rauch et al., 2012, The Lancet.[11] Exome sequencing of 51 children with severe nonsyndromic intellectual disability and their parents led the authors to estimate that, after exclusion of copy number variations, severe intellectual disability could be caused by de novo variants in approximately 45% to 55% of patients. One child had a de novo CUX2 missense change, and the authors named CUX2 as a candidate gene.

Epi4K Consortium, 2013, Nature.[10] Exome sequencing of 264 patients with infantile spasms or Lennox-Gastaut syndrome and their parents confirmed 329 de novo mutations and identified GABRB3 and ALG13 as associated genes. CUX2 is not named in the supplied text.

Geisheker et al., 2017, Nature Neuroscience.[9] An analysis of de novo missense mutations in patients with neurodevelopmental disorders found significant clustering in 200 genes, and targeted sequencing of 20 recurrent sites in 17,689 patients identified 21 new patients with identical missense mutations, one of them with CUX2 E590K.

Chatron et al., 2018, Annals of Neurology.[8] A clinical description of nine patients with E590K, most with developmental and epileptic encephalopathy, with seizure, EEG, treatment and development data.

Barington et al., 2018, European Journal of Human Genetics.[7] A case report of a Danish girl with E590K, intellectual disability, seizures and autism, with a summary of two earlier patients with E590K and a reference to one earlier patient with a de novo nonsense variant.

Suzuki et al., 2022, Scientific Reports.[6] CUX2 sequencing in 271 Japanese patients with epilepsy found missense variants mainly in temporal lobe epilepsy. Mice lacking Cux2 showed increased susceptibility to seizures induced by kainate and increased excitatory signaling to the hippocampus.

Zhang et al., 2022, Frontiers in Genetics.[5] A report of one Chinese boy with a new CUX2 variant, p.Thr945Met, febrile seizures controlled with levetiracetam and developmental delay, which the authors describe as a less severe phenotype than previously reported in patients with E590K.

Oppermann et al., 2023, European Journal of Human Genetics.[4] A description of 34 individuals with variants in the related gene CUX1, with developmental delay, intellectual disability ranging from borderline to severe, seizures in a quarter, and catch-up development in some.

Romano et al., 2025, preprint.[3] A case report of a girl with a de novo CUX2 frameshift variant, a single seizure controlled with valproic acid and moderate intellectual disability. Not peer reviewed.

Morcom et al., 2026, Nature.[1] Layer 2 and 3 neurons that express CUX2 have more DNA damage within multiple sclerosis brain lesions and are selectively lost, with DNA damage, in mouse models of demyelination and brain inflammation. Cux2 and Atf4 were essential for these neurons' resilience during postnatal neuroinflammation in mice, acting in neurons to enhance DNA double-strand break repair. The inflammatory signal interferon-gamma caused DNA damage and neuron death in laboratory-grown neurons and selective loss of these neurons in mice.

Xia et al., 2026, Nature.[2] During brain development in mice, the protein ATF4 is required for the progenitor cells that become CUX2-expressing neurons in layers 2 and 3 to repair DNA damage and survive. The authors suggest that recruitment of additional DNA repair proteins was crucial to the expansion of layer 2 and 3 excitatory neurons.

CURE Epilepsy, research summary, undated.[12] A summary for the public of the Chatron report, describing its nine patients and the work of Gemma Carvill with Dr. Gaetan Lesca. It states that most of the patients had severe treatment-resistant epilepsy that began early in life, severe intellectual disability, and speech not appropriate for their age. It adds no patient data.

Appendix C. Where the Reports Differ

This list shows places where reports, or parts of one report, give different information. Clinical differences are listed first. Where two reports cite the same earlier report beside a patient, the list shows each report's description and does not determine whether they describe the same person or which description is correct.

• Autistic features, Chatron cohort: Abstract: 6 patients. Results and Table 3: 2 patients.[8]
• Drug resistance, Chatron cohort: Abstract: 7 drug resistant, 2 controlled with valproate. Results: 2 controlled, 1 initially controlled with rare later episodes, 5 refractory.[8]
• Seizure type at onset, Chatron cohort: Discussion: myoclonic or myoclonic-absence onset in 5 of 9. Results: myoclonic in 4 and atypical absence with myoclonic components in 2.[8]
• Nonverbal status, Chatron cohort: Results: 7 with severe intellectual disability and nonverbal. Table 3: nonverbal listed for Patients 1, 2, 3, 4, 7 and 9.[8]
• Epilepsy type, Chatron Patient 8: Results: focal epilepsy. Abstract and Table 2: multifocal epilepsy.[8]
• EEG, Chatron cohort: Abstract: generalized polyspike-wave in 6. Results: generalized spike-wave or polyspike-wave at onset in 5.[8]
• Age at study, Chatron cohort: Abstract: mean 13 years. Results: median 13 years.[8]
• Chatron Patient 8 MRI: Age at MRI printed as 15 years; age at study 6 months.[8]
• Earlier patient cited to Rauch: Chatron, Patient 1: focal spasms, West syndrome, severe intellectual disability, age 8.[8] Barington, Patient 2: "BNS" (benign neonatal seizures), moderate/severe intellectual disability, age 7.[7]
• Earlier patient cited to Epi4K: Chatron, Patient 6: male.[8] Barington, Patient 1: gender unknown.[7]
• Barington patient: Barington: occasional relapses after treatment ended; moderate to severe intellectual disability.[7] Zhang, Table 1: "Seizure free"; "Severe."[5]
• Zhang patient EEG: Results: spikes and spike-wave discharges in central, parietal and posterior temporal regions during sleep. Discussion and Table 1: bi-parieto-temporal discharges.[5]
• Romano seizure type: Abstract: generalized brief tonic-clonic. Case presentation: generalized tonic.[3]
• Severity of E590K: Romano describes the E590K phenotype as typically including profound intellectual disability.[3] Chatron, Table 3: severe in 7, profound in 1.[8]
• Previously reported patients, Chatron cohort: Text: "the two individuals previously reported," citing Rauch and Epi4K. Tables: earlier reports cited beside Patients 1, 6 and 9.[8]
• Inheritance, E590K data shared between Geisheker and Chatron: Geisheker reports the inheritance of the E590K variant found at the Troina site as unknown.[9] Chatron states that Geisheker included molecular data from Patients 6 and 9, and reports the variant as de novo in all nine patients.[8]
• Sequencing method, Chatron cohort: Abstract: exome in 5, gene panel in 4. Methods: exome in 2, gene panel in 5 (seven in total).[8] CURE Epilepsy summary: whole exome sequencing, "among other techniques."[12]
• Truncating variants: Chatron: none observed in patients or controls.[8] Barington cites a de novo nonsense variant, p.(Arg633*), in an autism study.[7] Romano reports a frameshift variant.[3]
• Zhang variant classification: Introduction: pathogenic. Results: likely pathogenic (ACMG/AMP).[5]
• Position of E590K: Chatron Results: third alpha helix. Chatron Discussion: first alpha helix of CUT domain 1.[8]
• Name of the changed residue: Chatron and Barington describe it as glutamine; Chatron Results also as glutamate. "Glu" denotes glutamic acid (glutamate); "Gln" denotes glutamine.[7,8]
• CUT domain positions: Zhang text: CUT domain 1 at amino acids 549 to 627, CUT domain 2 at 892 to 967.[5] Suzuki Figure 1A: first CUT domain printed as 631-547, then 888 to 970 and 1039 to 1125; homeodomain 1168 to 1230.[6] Romano Figure 1: 544 to 631, 887 to 974 and 1083 to 1125; homeodomain 1168 to 1230.[3]
• Protein length: Zhang Introduction: 1,426 amino acids. Zhang Discussion, Barington, Romano: 1,486.[3,5,7]
• Chromosomal location: Barington: CUX2 is located on 12q23.13.[7] Suzuki: CUX2 lies within an epilepsy-associated region at 12q24.11 to 12q24.13.[6]
• Years in Barington Table 1: Epi4K patient labeled 2012; Rauch patient labeled 2013; Barington's own patient labeled 2017. Rauch was published in 2012, Epi4K in 2013 and Barington in 2018.[7]
• Geisheker cohort size: Text and Table 2: 17,689 patients. Table 1 header: 17,850.[9]
• Geisheker controls at codon 590: Text: "No variants were observed at any of the 20 sites in controls." Table 2: one rare missense variant at CUX2 p.Glu590 among 3,023 controls.[9]
• How the finding is described: CURE Epilepsy: "Dr. Gemma Carvill has identified a new cause of epilepsy: a mutation in the gene CUX2."[12] Chatron abstract: "We identify a recurrent de novo CUX2 p.Glu590Lys as a novel genetic cause for developmental and epileptic encephalopathy (DEE)." Chatron introduction: the variant "was first reported in single cases from two large-scale whole-exome sequencing (WES) studies," Rauch and Epi4K.[8]
• Effect of E590K: CURE Epilepsy: "Mutations in CUX2 cause errors in this process that can lead to epilepsy."[12] Chatron: the results "suggest" the variant alters the stability or specificity of CUX2 binding to DNA, which "may" cause a partial loss of function; functional studies "need to be performed."[8]

Appendix D. Terms Used in This Report

General definitions for readers. Except where a source number is shown, they are not drawn from the sources.

• Gene variant: A change in the DNA sequence of a gene. Reports may also use the terms mutation or pathogenic variant.
• De novo: A variant found in a child but not in either parent.
• Missense variant: A change that replaces one amino acid in the protein with another. In E590K, amino acid 590 changes from glutamic acid (Glu, E) to lysine (Lys, K).
• Frameshift and nonsense variants: Changes expected to stop the protein from being made in full length.
• Transcription factor: A protein that controls when other genes are switched on or off.
• CUT domain; homeodomain: Regions of the CUX2 protein that bind DNA.[3,8]
• DEE: Developmental and epileptic encephalopathy. Chatron and colleagues describe these as severe infantile and childhood onset epilepsies characterized by developmental slowing or regression in the context of recurrent seizures and frequent interictal epileptiform discharges, often on a background of developmental delay.[8]
• DEE67: Developmental and epileptic encephalopathy 67 (OMIM 618141), the autosomal dominant disorder associated with CUX2 variants.[3,5]
• Myoclonic seizure: A brief, shock-like jerk of a muscle or group of muscles.
• Absence seizure: A brief lapse in awareness, sometimes with staring.
• Lennox-Gastaut syndrome; West syndrome: Severe childhood epilepsy syndromes. West syndrome involves infantile spasms.
• EEG: Electroencephalogram, a recording of the brain's electrical activity.
• MRI: Magnetic resonance imaging, a scan of brain structure.
• Knockout mouse: A mouse bred without a working copy of a specific gene.
• Preprint: A research paper shared publicly before peer review.

Appendix E. Dedication

This review recognizes the individuals with CUX2 variants whose participation made these studies possible, and their families. Each person is described in this report with the information the researchers published, using the identifiers the researchers assigned where they assigned one. The reports do not use names. The researchers acknowledged the patients and families in their reports:

"We thank all patients and their families for participating in this study."[11]

"We are deeply grateful to the patients, their families, clinical research coordinators and referring physicians for participating in the Epilepsy Phenome/Genome Project (EPGP) and providing the phenotype data and DNA samples used in this study."[10]

"We thank the individuals and their families for participation in this study."[9]

"The authors would like to thank the families for their participation."[8]

"We thank the family for their participation in the study."[7]

"We thank the patients, families, volunteers..."[6]

"We are incredibly thankful to the participants in our study."[5]

The acknowledgments of the Romano report do not include a statement thanking the patient or family.[3]

Sources

Numbers in brackets in the text refer to these sources. Sources 1 to 11 are research reports, listed in reverse chronological order. Source 12 is a research summary of source 8; the page shows no publication date.

1. Morcom, Laura, Wenlong Xia, Zhaoyang Xu, et al. "DNA Damage Burden Causes Selective CUX2 Neuron Loss in Neuroinflammation." Nature, April 1, 2026. https://www.nature.com/articles/s41586-026-10310-3

2. Xia, Wenlong, Laura Morcom, Zhaoyang Xu, et al. "Expansion of Outer Cortical CUX2 Neurons Requires Adaptations for DNA Repair." Nature, April 1, 2026. https://www.nature.com/articles/s41586-026-10290-4

3. Romano, Ferruccio, Mohammad Sadegh Shams Nosrati, Francesca Madia, et al. "A Novel Frameshift CUX2 Variant in a Patient With Epilepsy and Global Developmental Delay: Phenotypic and Genotypic Expansion." Preprint, SSRN, 2025. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5517855

4. Oppermann, Henry, Elia Marcos-Grañeda, Linnea A. Weiss, et al. "CUX1-Related Neurodevelopmental Disorder: Deep Insights Into Phenotype-Genotype Spectrum and Underlying Pathology." European Journal of Human Genetics, Aug. 30, 2023. https://www.nature.com/articles/s41431-023-01445-2

5. Zhang, Feng, Fuwei Li, Fujian Chen, et al. "Novel Variant Expands the Clinical Spectrum of CUX2-Associated Developmental and Epileptic Encephalopathies." Frontiers in Genetics, July 1, 2022. https://www.frontiersin.org/journals/genetics/articles/10.3389/fgene.2022.808181/full Note: Zhang et al. report original data only for their own patient, a boy with p.Thr945Met. Their Table 1 rows 2 to 11 and Discussion summaries present earlier patients with E590K, attributed by the authors to Chatron, Barington, Rauch, Allen (Epi4K) and Geisheker; the clinical values repeat those published by Chatron et al. and Barington et al. This review cites Zhang et al. for their own patient and statements, and in Appendix C where their versions of earlier data differ from the originals.

6. Suzuki, Toshimitsu, Tetsuya Tatsukawa, Genki Sudo, et al. "CUX2 Deficiency Causes Facilitation of Excitatory Synaptic Transmission Onto Hippocampus and Increased Seizure Susceptibility to Kainate." Scientific Reports, May 17, 2022. https://www.nature.com/articles/s41598-022-10715-w

7. Barington, Maria, Lotte Risom, Jakob Ek, et al. "A Recurrent De Novo CUX2 Missense Variant Associated With Intellectual Disability, Seizures, and Autism Spectrum Disorder." European Journal of Human Genetics, May 24, 2018. https://pmc.ncbi.nlm.nih.gov/articles/PMC6117349/

8. Chatron, Nicolas, Rikke S. Møller, Neena L. Champaigne, et al. "The Epilepsy Phenotypic Spectrum Associated With a Recurrent CUX2 Variant." Annals of Neurology, April 30, 2018. https://findresearcher.sdu.dk/ws/portalfiles/portal/142981525/The_epilepsy_phenotypic_spectrum_associated_with_a_recurrent_CUX2_variant.pdf

9. Geisheker, Madeleine R., Gabriel Heymann, Tianyun Wang, et al. "Hotspots of Missense Mutation Identify Neurodevelopmental Disorder Genes and Functional Domains." Nature Neuroscience, August 2017. https://repository.uantwerpen.be/docman/irua/5ab892/145173.pdf

10. Epi4K Consortium and Epilepsy Phenome/Genome Project (Andrew S. Allen, Samuel F. Berkovic, Patrick Cossette, et al.). "De Novo Mutations in Epileptic Encephalopathies." Nature, Aug. 11, 2013. https://www.nature.com/articles/nature12439

11. Rauch, Anita, Dagmar Wieczorek, Elisabeth Graf, et al. "Range of Genetic Mutations Associated With Severe Non-Syndromic Sporadic Intellectual Disability: An Exome Sequencing Study." The Lancet, Sept. 27, 2012. https://www.academia.edu/24047713/Range_of_genetic_mutations_associated_with_severe_non_syndromic_sporadic_intellectual_disability_an_exome_sequencing_study

12. CURE Epilepsy. "New Cause of Severe Childhood Epilepsy Found: Genetic Mutation in the CUX2 Gene." Research summary, undated; accessed Oct. 10, 2026. https://www.cureepilepsy.org/research-discoveries/new-cause-of-severe-childhood-epilepsy-found-genetic-mutation-in-the-cux2-gene/

Artificial Intelligence Disclaimer:

This report was prepared by the CUX2 Foundation with the assistance of artificial intelligence. The full text of each source was collected from its original location and provided to Claude Opus 5.5, an AI model developed by Anthropic, with strict parameters for research and consolidation. Every statement had to come from these 12 sources and be referenced to its source, except for the general definitions in Appendix D and the treatment categories in Section 5. Interpretations and hypotheses had to be attributed to the authors who made them, and no subjective or editorial statements could be added. The model then checked the draft against the full text of each source. Numbers in brackets refer to the numbered list of sources at the end of the document.

Despite these parameters, the information in this report may not be fully accurate or complete. Errors can occur in AI-assisted analysis. Four sources were reviewed in a form other than the final published version, and they may differ from it. The Romano report[3] is a preprint that has not been peer reviewed. The Chatron report[8] was reviewed as an accepted manuscript, and the Geisheker[9] and Epi4K[10] reports as author manuscripts. The supplied Epi4K manuscript is titled “De novo mutations in the classic epileptic encephalopathies.”[10] The CURE Epilepsy summary[12] is not a research report, adds no patient data and shows no publication date. Readers should refer to the original reports, listed in the Sources section, for exact findings.