Epileptic encephalopathies are a phenotypically and genetically heterogeneous group of severe epilepsies accompanied by intellectual disability and other neurodevelopmental features –6. Using next-generation sequencing, we identified four different de novo mutations in KCNA2, encoding the potassium channel K V .2, in six isolated patients with epileptic encephalopathy (one mutation recurred three times independently). Four individuals presented with febrile and multiple afebrile, often focal seizure types, multifocal epileptiform discharges strongly activated by sleep, mild to moderate intellectual disability, delayed speech development and sometimes ataxia. Functional studies of the two mutations associated with this phenotype showed almost complete loss of function with a dominant-negative effect. Two further individuals presented with a different and more severe epileptic encephalopathy phenotype. They carried mutations inducing a drastic gain-of-function effect leading to permanently open channels. These results establish KCNA2 as a new gene involved in human neurodevelopmental disorders through two different mechanisms, predicting either hyperexcitability or electrical silencing of K V .2-expressing neurons. Many of the voltage-gated potassium channels (K V 1–K V 12) are expressed in the central nervous system (CNS), having an important role in neuronal excitability and neurotransmitter release 7. Mutations in potassium channel–encoding genes cause different neurological diseases, including benign familial neonatal seizures (KCNQ2, encoding K V 7.2; KCNQ3, encoding K V 7.3) 8–10 , neonatal epileptic encephalopathy (KCNQ2) 11,12 , episodic ataxia type 1 (EA1) (KCNA1, encoding K V 1.1) 13 and peripheral nerve hyperexcitability (KCNA1, KCNQ2) 13–15. In addition, antibodies against K V 1.1 or associated proteins such as contactin-associated protein 2 (CASPR2) or leucine-rich, glioma-inactivated 1 (LGI1) cause limbic encephalitis or neuromyotonia 16. Therefore, potassium channel genes represent interesting candidates for neurodevelopmental disorders. To identify mutations in presumed genetic forms of epilepsy, we designed a targeted resequencing panel 17 comprising 265 known and 220 candidate genes for epilepsy (Supplementary Table 1). Screening a pilot cohort of 33 patients, we identified mutations in known epilepsy-related genes in 16 cases 17. We evaluated the remaining 17 cases for mutations in candidate genes (Supplementary Note), which led to the detection of a heterozygous de novo mutation in KCNA2, c.1214C>T (encoding p.Pro405Leu), affecting the highly conserved pore domain of the voltage-gated potassium channel K V 1.2 (NM_004974, CCDS827). This mutation was not present in control databases (1000 Genomes Project, Exome Variant Server (EVS), dbSNP138 or the Exome Aggregation Consortium (ExAC) database). The affected female (patient 1) carrying this mutation had unre-markable early development until the onset of epilepsy at 17 months of age. The phenotype included febrile and afebrile alternating hemiclonic seizures and status epilepticus, reminiscent of Dravet syndrome. The electroencephalogram (EEG) showed multifocal spikes with marked activation during sleep. After seizure onset, ataxia and delay of psychomotor and language development became apparent. She had postnatal short stature, growth hormone deficiency and hypothyroidism. Seizures and ataxia responded poorly to antiepileptic drugs (topiramate, oxcarbazepine, valproic acid and bromide), including acetazolamide (known to be effective in EA1 caused by mutations in KCNA1; ref. 18). At last follow-up at 8 years of age, she had remained seizure free for the past 6 months without previous change of medication. We identified further KCNA2 mutations in several parallel studies (Supplementary Fig. 1). First, we performed whole-exome sequenc-ing in 86 parent-offspring trios with epileptic encephalopathy (31 with Dravet syndrome negative for mutations in SCN1A, 39 with myoclonic-atonic epilepsy (MAE) and 16 with electrical status epi-lepticus in slow-wave sleep (ESES)). Second, we performed panel sequencing (Supplementary Note) in 147 adults with a broad spectrum of epilepsy phenotypes associated with intellectual disability. Third, we performed whole-exome sequencing in an adult cohort of 10 independent trios with severe epilepsy and intellectual disability and whole-exome sequencing in another cohort of 12 independent, isolated index cases with early-onset ataxia and epilepsy. We identified six additional independent cases with previously unre-ported heterozygous KCNA2 variants (Table 1, Supplementary Fig. 2 and Supplementary Note). Patient 2 (initially classified as having MAE) carried the de novo mutation c.788T>C (encoding p.Ile263Thr). Patient 3 (intellectual disability with neonatal-onset focal epilepsy and cerebel-lar hypoplasia) carried the variant c.440G>A (encoding p.Arg147Lys), of unknown inheritance. We considered p.Arg147Lys to be a variant of unknown relevance because (i) it could not be confirmed as de novo, (ii) it was predicted to be benign using seven of nine prediction tools, (iii) a lysine occurs naturally at this position in Drosophila melanogaster and zebrafish, and (iv) the change did not show functional consequences (Supplementary Fig. 3, Supplementary Tables 1 and 2, and Supplementary Note). Patient 4 (initially classified as having Dravet