CHD2 variants are a risk factor for photosensitivity in epilepsy

Photosensitivity in epilepsy is common and has high heritability, but its genetic basis remains uncertain. Galizia et al. reveal an overrepresentation of unique variants of CHD2 — which encodes the transcriptional regulator ‘chromodomain helicase DNA-binding protein 2’ — in photosensitive epilepsies, and show that chd2 knockdown in zebrafish causes photosensitivity.


Introduction
Photosensitivity is a heritable abnormal cortical response to flickering light, often manifesting as EEG changes called a photoparoxysmal response (Walter et al., 1946). Photoparoxysmal response may occur with seizures, and in normal subjects, or with neuropsychiatric disorders (So et al., 1993). The photoparoxysmal response is age-dependent: prevalence in healthy children is between 1.4 and 8.3%, dropping to 51% in adults (Gregory et al., 1993;Quirk et al., 1995;Kasteleijn-Nolst Trenite et al., 2003;Verrotti et al., 2012). Photosensitive epilepsy is a reflex epilepsy, with seizures triggered by visual stimuli. A population-based study in Great Britain determined that the annual incidence of epilepsy with photoparoxysmal response was 1.1 per 100 000 in the overall population, and 5.7 per 100 000 between 7 and 19 years of age (Quirk et al., 1995). About 40% of people with photosensitive epilepsy only have seizures on exposure to visual stimuli. Photosensitive seizures also feature in specific epilepsy syndromes, with other seizure types, and in nonsyndromic epilepsies. Examples include juvenile myoclonic epilepsy (Tauer et al., 2005;Koeleman et al., 2013;Taylor et al., 2013), other genetic generalized epilepsies (GGE) (Taylor et al., 2013), idiopathic photosensitive occipital epilepsy, and other focal (Taylor et al., 2004;Lu et al., 2008), symptomatic occipital, and progressive myoclonic, epilepsies. The archetypal photosensitive syndrome is eyelid myoclonia with absences (EMA), a GGE characterized by rapid eyelid jerks and upward eyeball deviation on eye closure: photosensitivity is an essential feature (Sadleir et al., 2012).
Photosensitivity occurs in some epileptic encephalopathies, such as Dravet syndrome due to mutation in SCN1A and encephalopathy associated with mutation in CHD2 . Published data do not allow determination of whether the photosensitivity in these conditions is due to the underlying gene mutation or to the epileptic encephalopathy per se. CHD2 encodes chromodomain helicase DNA-binding protein 2, involved in transcriptional regulation. Additional attention was drawn to CHD2 as a candidate photosensitive epilepsy gene as the only shared gene within several reported overlapping copy number variants of the chromosome 15q26.1 region associated with complex phenotypes including epilepsy with photosensitivity. Eight patients with de novo deletions of 15q26 encompassing part or all of CHD2 have been reported (Veredice et al., 2009;Dhamija et al., 2011;Capelli et al., 2012;Lund et al., 2013;Mullen et al., 2013;Chénier et al., 2014). We and others subsequently showed 6/500 epileptic encephalopathy cases had de novo CHD2 mutations Epi4K Consortium et al., 2013;Suls et al., 2013;Lund et al., 2014), and recently showed that clinical photosensitivity was prominent in the rare CHD2-associated myoclonic encephalopathy (Thomas et al., 2015).
These findings led us to hypothesize that CHD2 disruption would be associated with common forms of photosensitive epilepsy or photosensitivity manifesting as a photoparoxysmal response alone.

Materials and methods
Written informed consent was obtained from patients or parents/guardians for minors or those with intellectual disability. The study was approved by relevant institutional ethics committees.
We defined photosensitive epilepsy as the presence of a photoparoxysmal response (Kasteleijn-Nolst Trenité et al., 2012) with a history of epilepsy, or seizures reproducibly induced by flickering light. The photoparoxysmal response per se was not an essential inclusion requirement in every patient with epilepsy because age, state (e.g. sleep deprivation) and antiepileptic medication affect its detectability. To test the effect of CHD2 variation beyond the epileptic encephalopathies alone, we included a broad range of epilepsy types. Recruitment was from nine countries (see Supplementary material for details) (Tauer et al., 2005;Lu et al., 2008;Taylor et al., 2013). The cohort included 36 patients with EMA: all had photoparoxysmal response. We sequenced CHD2 in 580 people with photosensitive epilepsy and 55 people with photoparoxysmal response but no history of seizures. All patients were of European ancestry. The phenotypic distribution is given in Table 1.
We evaluated data from two additional exome-sequenced cohorts of GGE patients, to determine the role of CHD2 variation in GGE per se, independent of photoparoxysmal response. Not all patients in these cohorts had been formally assessed for photoparoxysmal response. These two groups were the Complex Genetics of Idiopathic Epilepsies Consortium (CoGIE) cohort of 238 probands with familial GGE (Supplementary material), and a published cohort of 118 patients with GGE (Heinzen et al., 2012).
Targeted sequencing of CHD2 was undertaken either using Illumina TruSeq Custom Amplicon TM (TSCA) or molecular inversion probes (see Supplementary material for details). Whole exome sequencing (Supplementary material) was performed on five EMA samples. Coverage data for all experiments are provided in the Supplementary material. Only variants confirmed by a second method (Sanger sequencing or a second independent molecular inversion probe capture, see Supplementary material) were used in analyses.
The Exome Aggregation Consortium (ExAC) formed a large control population of disease and population genetic studies (ExAC, Cambridge, USA; URL: http://exac.broadinstitute.org accessed October 2014; non-Finnish European samples only used), giving the best available population frequency of CHD2 variants of interest. Detailed phenotypic data are not available for these individuals; some might, if tested, have or have had photoparoxysmal response or a history of photosensitive seizures. These unselected cases are unlikely to harbour more than the best estimates of photoparoxysmal response prevalence in the general population (1.4%) (Kasteleijn-Nolst Trenite et al., 2003).
We focused on unique variants, in our cohort and in ExAC: this is a well-established approach Cnossen et al., 2014;Wain et al., 2014). We hypothesized an over-representation of unique variants in our cohort compared with the phenotypically-unselected ExAC cohort. We defined unique variants as those that occurred in one individual only, in cases and controls (from ExAC) considered together, that were non-synonymous, splice-site or frameshift. We used several methods for prediction of the functional consequences of unique variants in cases (Supplementary material). We defined 'rare' variants as those with a minor allele frequency 51% in the non-Finnish European ExAC samples.
We undertook functional studies. To test functional consequences of Chd2 loss in zebrafish, we used the chd2 E2I2 morpholino reported previously (Suls et al., 2013). Briefly, morpholino (12 ng) was microinjected into 1-to 2-cell-stage embryos of the AB (wild-type) strain. Embryos were raised in a dark incubator. At 1 day post-fertilization (dpf), embryos were prepared using the least possible amount of light. In parallel, control non-injected embryos from the same clutch of eggs were processed in the same manner. At 4 dpf, optic tectal field recordings were performed (Suls et al., 2013) (Supplementary material). The first 10 s of recording were performed in minimal light in order to place the needle. Immediately following these first 10 s, recordings were performed in the dark for five minutes. At the end of this 5min period, a very bright light was switched on ('light ON' state; six times the standard brightness level used for needle placement), and recording continued for 5 min. A paroxysm of high-frequency activity (200-500 Hz) with amplitude 43 times background level, either spontaneous or evoked by light, was defined as a polyspiking episode.

Statistics
We performed a two-tailed Fisher's exact test to determine whether the burden of unique variants in our case cohorts was greater than expected compared to ExAC controls. We examined the frequency of all rare variants in the entire cohort, and the frequency of unique variants only separately in patients with EMA, patients with GGE excluding EMA, and patients with focal epilepsies. The threshold for significance was set at P 5 0Á01, applying Bonferroni correction for these five comparisons. For the single separate comparison of cases with photoparoxysmal response without epilepsy and ExAC, significance was set at P 5 0Á05. For zebrafish data, comparison of the parameters of spiking activity (dark versus light condition) for each treatment group was performed using the Mann-Whitney test.
We investigated the predicted deleteriousness of the 11 unique variants in the cases ( Table 2). Eight of 11 unique variants (73%) had scaled CADD scores 410, placing them in the top 10% most deleterious single nucleotide variants; as a group, the 11 variants had a mean scaled CADD score of 32.6, ranking higher than 99.95% of all possible human single nucleotide variants (Kircher et al., 2014).
Next, we analysed variation by epilepsy type. The archetypal photosensitive GGE syndrome EMA had the highest frequency of unique variants, found in 3/36 patients, more than expected compared to ExAC controls (3/72 alleles versus 128/68 854 alleles) (P = 3.50 Â 10 À4 ). As a post hoc comparison, the frequency of unique variants (4.2%) in the small EMA group is considerably greater than in our overall cohort excluding EMA (0.74%) (P = 0.026).
Notably, two of three EMA variants were frameshift, compared to 9/128 unique variants in ExAC. One EMA variant was shown to have arisen de novo, strengthening its role in causation of the phenotype.
For all GGE excluding EMA, we found no significant difference compared to ExAC (4/888 alleles versus 128/ 68 854 alleles, P = 0.089). We also did not find significant over-representation in focal epilepsies compared with ExAC (2/118 alleles versus 128/68 854 alleles; P = 0.021). One case was included in both GGE and focal epilepsy cohorts, as the phenotype evolved from early-onset absence epilepsy to idiopathic photosensitive occipital epilepsy (Patient 11, Table 2). One of 55 (1.82%) individuals with photoparoxysmal response but no seizures had a unique CHD2 variant (Table 2 and Fig. 1): this did not represent overrepresentation compared to ExAC (1/110 alleles versus 128/68 854 alleles; P = 0.186). This case has not developed epilepsy by the age of 18 years. We provide 99% confidence intervals (CI) (accounting for multiple comparisons) for all these comparisons in Table 3.
To investigate whether CHD2 may be associated with the broader phenotype of GGE rather than photosensitive epilepsies specifically, we tested whether rare variants in CHD2 were enriched in patients with GGE, with or To test functional consequences of Chd2 loss in zebrafish, we used the chd2 E2I2 morpholino reported previously (Suls et al., 2013). As described, chd2 morpholinoinjected larvae displayed body curvature, excessive body pigmentation, and developmental delay (Suls et al., 2013). This phenotype was observed after 50% knockdown of chd2. All non-treated larvae appeared normal. Recordings were obtained from 15 morpholino-injected larvae and 10 sibling controls. In comparison to 7 dpf larvae (Afrikanova et al., 2013), spikes from 4 dpf larvae were shorter in duration and displayed a higher frequency of oscillations in polyspike complexes. Due to these differences, spontaneous spiking in controls was not excluded, but also quantified. We analysed duration of discharges, number of discharges under light conditions, cumulative duration of spiking activity, and cumulative discharge frequency distribution. Representative recordings are shown in Fig. 2.
In line with the previous findings (Suls et al., 2013), the morpholino-injected larvae showed spontaneous abnormal burst discharges. There was a preferential occurrence during the light ON state (17 discharges in the dark versus 59 in the light). In the morpholino-injected group, 14/15 larvae had discharges during the light ON state; 7/15 larvae had spiking only during the 5-min light ON state, and 10/15 showed spiking activity within the first 3-5 s after the light ON. The average duration of any event (spike or polyspike discharge) in the morpholino-injected group fell during the light ON state (Fig. 3A), attributable to the fact that morpholino-injected larvae also displayed spontaneous polyspike discharges in the dark: the events under light conditions were more heterogeneous (i.e. spontaneous polyspikes plus light-induced spiking), explaining reduced average duration. The average number of events/ larva significantly increased in the morpholino-injected group in the light opposed to the dark period; this was not seen in the control group (Fig. 3B). A similar pattern was observed for cumulative duration of spiking activity (Fig. 3C): morpholino-injected larvae showed a steep increase in polyspike discharges in the light ON state, not observed for controls. The larvae from the non-injected control group also reacted to the light ON state by displaying an initial locomotor response, with 7/10 displaying short spontaneous burst activity within 2-13 s after the light was switched on. However, the overall distribution of event duration is different from that of morpholinoinjected larvae (Fig. 3D): the controls' curve lies to the left of the morpholino-injected curve, indicating that the proportion of longer discharges is higher in the morpholino-injected group.  The associations with photosensitive epilepsy overall and with EMA alone are significant, as documented in the text. PPR = photoparoxysmal response.

Discussion
We show an enrichment of unique variants in CHD2 with photosensitivity in the common epilepsies overall, identifying CHD2 as a photosensitive epilepsy gene. We also examined the distribution of unique variants by syndrome. CHD2 is also the first gene to be discovered for EMA, the archetypal photosensitive epilepsy syndrome. In CHD2 encephalopathy, though published phenotypes can be difficult to interpret, the seizure type of absence seizures with eyelid myoclonia, rather than the epilepsy syndrome, is seen in as many as 8/23 (35%) patients with de novo CHD2 mutation or deletion (Veredice et al., 2009;Dhamija et al., 2011;Capelli et al., 2012;Carvill et al., 2013;Chénier et al., 2014;Lund et al., 2014). Together, these results suggest that CHD2 is an important contributor to both the absence seizures with eyelid myoclonia seizure type and EMA epilepsy syndrome. For other epilepsy syndromes, CHD2 variation over-representation in the photosensitive GGE or the mixed cohort of photosensitive focal epilepsies failed to meet the corrected threshold for significance. A single unique CHD2 variant was found in one patient with photoparoxysmal response without seizures.
In view of the comparatively small sizes of these syndrome cohorts, we can only confidently exclude effects with odds ratios greater than the upper limit for the 99% confidence intervals given in Table 3. Further studies in larger cohorts of these phenotypes would seem warranted.
Previous studies of photoparoxysmal response support a model of significant genetic heterogeneity and an overall complex genetic architecture (Sadleir et al., 2012;Verrotti et al., 2012;Taylor et al., 2013): indeed, none of the several linkage regions contain CHD2. Our findings confirm heterogeneity and complexity in the genetics of photosensitivity, but also suggest a single gene may contribute to photosensitivity in some cases. Two mutations we detected are recurrent: p.Glu1412Glyfs*64, previously reported in epileptic encephalopathy with marked photosensitivity ; and p.Gln1392Thrfs*17, in Lennox-Gastaut syndrome with photosensitivity (Lund et al., 2014). The unique variants detected are, as a group, predicted to be amongst the most deleterious variants possible (Kircher et al., 2014) and CHD2 is amongst the genes least tolerant of functional variation (Petrovski et al., 2013;Residual Variation Intolerance Score 2.37).
CHD2 does not encode an ion channel, opening up new avenues for research into cortical excitability. CHD2 is one of nine genes from a highly-conserved protein family with a unique domain combination: two N-terminal chromatinorganization modifier (chromo), SNF2-related helicase/ ATPase and DNA-binding domains (Woodage et al., 1997;Schuster et al., 2002;Kulkarni et al., 2008). Chd2 knockdown zebrafish have multiple developmental abnormalities, abnormal movements and epileptiform discharges (Suls et al., 2013). Disruption of Chd2 in mice causes embryonic death in some heterozygote pups and a complex phenotype including growth retardation and lordokyphosis (Marfella et al., 2006;Kulkarni et al., 2008): epilepsy has not yet been described. Interestingly, the reported human mutations do not cluster to accessory domains of the protein and no obvious pattern has emerged. Recent data demonstrated that the N-terminal region of CHD2 plays an inhibitory role, reducing DNA affinity and ATPase activity which may confer specificity, while the C-terminus enhances DNA binding and stimulates ATPase activity Figure 2 Representative tectal field recordings of 4-dpf zebrafish larvae. Background fragment of non-treated wild-type control in the dark (A); reaction of a non-injected fish to light ON -movement artefacts (wavy background) and a very short spike were observed (B); response to light ON of the morpholino-injected larvae: significantly more spiking activity is seen (C). The scale is the same for all three fragments. (Liu et al., 2015). Additional studies investigating protein interacting partners and post-translational modifications of CHD2 will be necessary to understand how abnormal CHD2 leads to photosensitive epilepsy.
Our zebrafish data show that partial (50%) loss of chd2 function causes photosensitivity. Although Suls et al. (2013) showed chd2 knockdown could cause seizures, photosensitivity was not studied. Although normal zebrafish show complex sensitivity to light (Moore and Whitmore, 2014), and untreated larvae show minor sensitivity to sudden exposure to light, morpholino-injected larvae show significantly more spiking activity on sudden light exposure. Photosensitivity on constant, rather than only flickering, light exposure has been described in humans (Oguni et al., 2001). The functional consequences of each of the human mutations we detected is not known, but some at least very probably lead to loss of function, as caused by partial chd2 knockdown that results in markedly enhanced photosensitivity in zebrafish. Together, these data strongly suggest that some human CHD2 mutations cause photosensitivity.
There are potential limitations of our work. Different sequencing platforms were used for the various studied groups. However, we note that all unique variants in cases were confirmed by a second method, whereas for ExAC controls we used a liberal threshold to maximize sensitivity to unique variants, such that a proportion of variants selected from ExAC will be false positive: the net result of this overall conservative approach is only to reduce study power. The ExAC cohort is also the biggest relevant control data set available, and the most likely of any existing data set to provide an accurate estimate of the Figure 3 Electrographic activity of zebrafish larvae with chd2 knockdown and light ON stimulus. Zebrafish larvae (4 dpf) were kept in the dark (or darkened environment, if not possible otherwise) for all groups in Danieau's medium. Tectal field recordings were performed for the first 5 min in the dark and subsequently in light ON state for the following 5 min in morpholino-injected larvae (n = 15) and non-injected larvae (n = 10). A spiking episode, either spontaneous or evoked by light, was defined as a paroxysm of high-frequency (200-500 Hz) activity with the amplitude exceeding three times the background. Average duration of spiking events AE SEM detected per condition is shown in A. Average number of events per fish AE SEM is shown in B. Cumulative duration of spiking activity per fish as seconds AE SEM is shown in C. Cumulative frequency distribution of spiking episodes is shown in D: morpholino-injected larvae show more activity than any of the non-injected controls, and a higher photosensitivity (curve shift to the right in the light compared to the dark recordings). *P 5 0.05 and **P 5 0.01 Mann-Whitney test.
true frequency of unique variation in CHD2 in a population not enriched for photosensitive epilepsy. Taking all these factors into account, the use of different platforms is very unlikely to have generated false positive resultsindeed, we are more likely to have underestimated unique variant numbers in cases. It is also possible that our choice of statistical test may have missed a true association between rare variation in CHD2 and GGE (irrespective of photoparoxysmal response or photosensitivity), and we did not test whether CHD2 variation contributes to epilepsy more broadly: we therefore cannot exclude the possibility that rare CHD2 variation contributes to epilepsy per se. Lack of parental samples meant we could only confirm variants were de novo in two patients. Family samples were only available in one other case (Case 9): the variant was inherited from a clinically-unaffected mother in whom no EEG studies had been carried out.
Our results provide evidence for a specific gene in a particular trait in epilepsy. Understanding the genetic basis of the photosensitivity trait is a first step to elucidating the biology that underlies photoparoxysmal response and its relation to epilepsy. Human photosensitive epilepsy paradigms have facilitated epilepsy treatment discoveries (French et al., 2014): understanding photoparoxysmal response biology may increase the value of these paradigms. Our findings may also provide new directions for understanding human cortical excitability.