High-density SNP screen of sodium channel genes by haplotype tagging and DNA pooling for association with idiopathic generalized epilepsy

Summary

We have investigated seven voltage-gated sodium channel genes for association with idiopathic generalized epilepsy (IGE). Probands and control DNA were grouped into pools and used to screen 85 single-nucleotide polymorphisms (SNPs), mostly HapMap SNPs tagging the common variation in these genes. Twelve SNPs exhibiting an allele frequency difference between pools were genotyped individually in our sample of 232 probands, 313 controls, and 95 parent–proband trios. Two SNPs, in SCN1A and SCN8A, were associated by allele and genotype at nominal level of significance, but were not significant after Bonferroni correction. Two SCN2A SNPs (rs3943809 and rs16850331) were associated by case–control with a subgroup with IGE and history of febrile seizures and also by transmission disequilibrium test (TDT) in parent–proband trios. Both SNPs are part of a linkage disequilibrium (LD) cluster of 38 SNPs, but none are obvious functional variants. The association of rs3943809 with the febrile seizure subgroup (p = 0.0004) remains significant after the conservative Bonferroni correction for multiple testing.

Idiopathic generalized epilepsy (IGE) is a common neurologic disorder involving generalized seizures, with generalized spike–wave on electroencephalography (EEG). It consists of many overlapping age-related syndromes. Twin and family studies demonstrate strong genetic heritability. Causative mutations in several genes have been identified, following linkage studies in families with uncommon Mendelian forms of IGE (Heron et al., 2007). The more common forms, however, are likely to involve variants in several genes, where association studies have greater power to identify susceptibility variants (Risch & Merikangas, 1996).

Voltage-gated sodium channel genes constitute an important group of candidate genes for IGE. Many antiepileptic drugs act by inhibiting sodium channel activity. Mutations in SCN1A, SCN2A, and SCN1B are implicated in families with generalized epilepsy with febrile seizures plus (GEFS+) (Heron et al., 2007). Many SCN1A mutations have also been reported in severe myoclonic epilepsy of infancy (SMEI) and SCN2A mutations in benign familial neonatal infantile seizures (BFNIS). We hypothesize that sodium channel genes also contain more frequent variants that predispose to common forms of IGE. In this study we examined a high density of single-nucleotide polymorphisms (SNPs) from the seven sodium channel genes expressed mainly in the central nervous system (CNS). We used a cost-effective approach: haplotype-tagging (Weale et al., 2003) in combination with DNA pooling (Sham et al., 2002) to screen for promising SNPs, which are then individually genotyped.

Methods

Clinical sample

Our sample comprised 232 IGE probands and 313 controls identified from white Caucasians at East Kent Hospitals. Probands had a consistent clinical history and generalized spike–wave on EEG. Both parents were recruited where possible, giving 95 family trios. Controls were unrelated, with no known history of epilepsy/blackouts. DNA was extracted from blood or cheek swabs using standard methods. Classification of IGE syndromes was as described previously (Barratt et al., 2006). Twenty-eight of our IGE cohort previously had febrile seizures (IGEFS), which have a higher a priori probability for association because of sodium channel mutations in GEFS+.

SNP selection

Linkage disequilibrium (LD) data for SNPs with minor allele frequency (MAF) ≥0.1 from HapMap April 2007 European dataset (http://hapmap.ncbi.nlm.nih.gov/cgi-perl/gbrowse/hapmap24_B36/) were used with Tagger (HAPLOVIEW 4.0, http://www.broadinstitute.org/haploview) to identify SNPs tagging clusters with LDs of r² > 0.8. These were supplemented with other SNPs with MAF ≥0.1 (http://www.ncbi.nlm.nih.gov/projects/SNP) that covered exon, exon/intron boundary, or promoter regions.

DNA pools

DNA was quantified by fluorometry using PicoGreen (Invitrogen Ltd, Paisley, UK) and combined into pools. Each pool had DNA entirely from blood or cheek swabs to exclude the possibility of unequal contributions of amplifiable DNA from different sources. We made three proband and three control pools, in duplicate, to generate 12 pools. SNaPshot primer extension assay (Applied Biosystems Inc., Foster City, CA, U.S.A.) was used to determine allelic frequencies for each SNP, with relative efficiencies of both alleles (coefficient k) assumed to be 1. Because this might underestimate some allele frequency differences, we applied a lenient level of significance (p < 0.15) for selecting SNPs for individual genotyping. Significance of allele frequency differences between proband and control pools were estimated by meta-regression (Knight & Sham, 2006).

Individual genotyping

SNPs were individually genotyped (Prevention Genetics, Marshfield, WI, U.S.A.) using invader assay in preamplified DNA from probands, parents, and controls. Preamplification was performed using 50-μm random primer N15, 10 mm dNTPs, 25 mm MgCl₂, cycled 50× 94 degrees 1 min, 37 degrees 2 min, 55 degrees 4 min, 72 degrees 30 s and combining four independent reactions.

Significance of case–control associations and deviation from Hardy-Weinberg equilibrium (HWE) were estimated by chi-square tests. Significance of within-family associations in trios were estimated by transmission disequilibrium test (TDT) using Transmit (Clayton, 1999). LD was estimated using Gene Counting (Curtis et al., 2006).

Results

Pooling screen

Figure S1 shows the number and density of HapMap SNPs with MAF ≥0.1 for the seven sodium channel genes examined. We screened at least the translated regions of these genes and most of their transcribed regions, but SCN1B was poorly covered. At least one SNP per LD cluster was screened by genotyping in DNA pools.

Sixteen of 85 SNPs (Table S1) exceeded the lenient significance threshold of p < 0.15. Twelve were individually genotyped, excluding four in strong LD with others. Unsurprisingly, because coefficient k was not determined, allele frequencies showed poor agreement with the pooling data (Table 1). However, allele frequency differences showed good agreement and exhibited similar levels of significance. Nine of the 12 SNPs achieved p < 0.15 predicted by pooling.

Table 1.   Comparison of allele frequency differences determined in pools with those determined by individual genotyping

Gene	SNP	Genotyping of DNA pools			Individual genotyping			LD^c
		IGE (n = 226)	Controls (n = 223)	Sig^a p-value	IGE (n = 228)	Controls (n = 296)	Sig^b p-value
^aSignificances of allele frequency differences determined by meta-regression. ^bSignificances of allele frequency differences determined by chi-square tests. ^cStrong LD (r^2 > 0.8) between certain SNPs within each gene indicated by 1A, 2A, or 8A, from HapMap or our genotyping data; slightly weaker LD (r^2 = 0.76) indicated by brackets. ^dr^2 between these two SNPs is 0.80 according to HapMap but 0.67 according to our genotyping data. IGE, idiopathic generalized epilepsy; LD, linkage disequilibrium; SNP, single-nucleotide polymorphism.
SCN1A	rs16851381	0.33	0.24	0.05	0.16	0.11	0.05	1A
	rs8191987	0.33	0.24	0.06	0.16	0.12	0.03	1A
	rs16851382	0.2	0.15	0.1	–	–	–	(1A)
	rs10497276	0.07	0.04	0.09	–	–	–	1A
	rs2298771	0.22	0.30	0.06	0.25	0.34	0.002
SCN2A	rs935403	0.23	0.31	0.1	0.22	0.28	0.04
	rs2060199	0.39	0.50	0.04	0.38	0.45	0.03
	rs2075704	0.16	0.21	0.14	0.13	0.15	0.26
	rs3943809	0.22	0.15	0.07	0.29	0.23	0.04	2A
	rs16850331	0.22	0.15	0.09	0.29	0.23	0.06	2A
	rs1947114	0.2	0.14	0.08	–	–	–	2A
SCN8A	rs303778	0.29	0.21	0.11	0.15	0.10	0.007	8A^d
	rs303777	0.27	0.17	0.04	0.14	0.12	0.48	8A^d
	rs7309995	0.23	0.15	0.09	–	–	–	8A
	rs11169883	0.36	0.45	0.1	0.39	0.41	0.46
SCN3B	rs1783901	0.27	0.16	0.02	0.22	0.18	0.14

There is redundancy, as some clusters had more than one SNP screened, some clusters overlap, and two non–HapMap SNPs are in strong LD with HapMap SNPs. The r² values for LD between SNPs used in the pooling screen (Table S2, Figure S2) were generally in good agreement with HapMap values.

SCN1A

All three individually genotyped SCN1A SNPs were in HWE and had nominally significant allele frequency differences between probands and controls (Table 1), with only rs2298771 also significant by genotype (p = 0.006, Table 2). Within-family associations for rs2298771 reached significance (p = 0.05) by TDT (Table S3). This SNP tags 40 other SNPs in HapMap, but only rs2298771 leads to an amino acid change (Thr1056Ala).

Table 2.   Genotype counts and frequencies in some SNPs selected by pooling

Gene	SNP	Genotype	IGE	IGEFS^a	Controls
Chi-square (Chi-sq) significance shown in bold exceeds Bonferroni-corrected significance of p = 0.0007. ^aSubset of IGE patients with a history of febrile seizures. ^bChi-sq, with respect to controls. ^cChi-sq with respect to IGE patients with no known history of febrile seizures. FS, febrile seizures; IGE, idiopathic generalized epilepsy; SNP, single-nucleotide polymorphism.
SCN1A	rs2298771	AA	116 (0.56)		125 (0.44)
		AG	81 (0.39)		124 (0.44)
		GG	11 (0.05)		35 (0.12)
		Chi-sq^b	10.39 (p = 0.006)
SCN2A	rs3943809	AA	109 (0.52)	11 (0.44)	170 (0.58)
		AG	82 (0.39)	8 (0.32)	108 (0.37)
		GG	20 (0.09)	6 (0.24)	13 (0.04)
		Chi-sq^b	5.77 (p = 0.06)	15.61 (p = 0.0004)
		Chi-sq^c		6.40 (p = 0.04)
	rs16850331	CC	107 (0.50)	8 (0.35)	161 (0.57)
		CT	88 (0.42)	10 (0.43)	108 (0.38)
		TT	17 (0.08)	5 (0.22)	12 (0.04)
		Chi-sq^b	4.21 (p = 0.12)	13.68 (p = 0.001)
				7.24 (p = 0.03)
	rs12614399	GG	104 (0.48)	10 (0.40)	164 (0.57)
		GC	99 (0.46)	11 (0.44)	110 (0.38)
		CC	12 (0.06)	4 (0.16)	12 (0.04)
		Chi-sq^b	4.03 (p = 0.13)	7.64 (p = 0.02)
		Chi-sq^c		6.47 (p = 0.04)
SCN8A	rs303778	TT	146 (0.71)		234 (0.82)
		TC	58 (0.28)		50 (0.17)
		CC	3 (0.01)		3 (0.01)
		Chi-sq^b	8.23 (p = 0.02)

SCN2A

All individually genotyped SCN2A SNPs were in HWE, of which three showed nominally significant association by allele frequency only (Table 1). In the IGEFS subgroup, rs3943809 and rs16850331 were associated by genotype versus controls (p = 0.0004 and p = 0.001, respectively) and were associated versus non-IGEFS IGE patients (Table 2). TDT analysis in trios showed significant association for rs3943809 (p = 0.008) and rs1680331 (p = 0.02) with IGEFS (Table S3).

SNP rs3943809 tags 36 other SNPs in HapMap (Mar 2008), with none exonic. We examined rs12614399 from this large LD cluster, as its location 41-bp upstream of exon 1c suggested that it might affect promoter function. This SNP was more weakly associated with IGEFS (Table 2).

SCN8A

One of three SCN8A SNPs (rs303778) had nominally significant allele frequency difference between probands and controls (Table 1), also significant by genotype (Table 2). There was no evidence for association by TDT (Table S3).

Discussion

We used DNA pooling and haplotype tagging to investigate association with IGE in sodium channel genes strongly expressed in the CNS. Of 441 common HapMap SNPs, we screened 80 tagging the entire dataset. Sixteen SNPs exceeded a lenient significance threshold (p < 0.15). Nine of 12 individually genotyped exceeded this threshold, yielding two SNPs nominally significant by allele and genotype: rs2298771 (SCN1A) and rs303778 (SCN8A). SNP rs2298771 was also associated by TDT, suggesting that the case–control association was unlikely to be due to population stratification. This SNP has been previously investigated, with no association found in two studies (Escayg et al., 2001; Chou et al., 2003). In our study, multiple testing needs to be considered. We estimate that we screened the equivalent of approximately 70 independent SNPs, giving a Bonferroni-corrected significance threshold of p < 0.0007. Neither nominally significant association exceeds this level of significance. Therefore, our extensive examination of the common variation in CNS-expressed sodium channel genes did not identify significant association with IGE.

We have identified, however, associations of three SCN2A SNPs with IGE patients with a history of febrile seizures. One of these (rs3943809) is significant at the Bonferroni-corrected threshold, with another (rs16850331) just failing to reach this conservative value. Because these SNPs are in strong LD, it is unlikely that these associations were inflated by genotyping error. They are further supported by within-family studies, suggesting that the case–control associations were unlikely to have resulted from hidden population stratification. They are in a cluster of 38 known SNPs in strong LD with each other. None are exonic, but six are 5′ of exon 1c, one of two common alternative 5′ untranslated exons (Martin et al., 2007; Figure S1). SNP rs12614399 is 41-bp upstream of exon 1c, but with the weakest association is unlikely to be directly responsible. SNP rs16850331 is 1,088-bp upstream of exon 1c, close enough to the promoter region to potentially influence its function. The other SNP associated with IGEFS (rs3943809) is in intron 15, where it is unlikely to be functional. It is probable that there are many unidentified SNPs in this LD cluster, which may include the SNP responsible for this association.

The role of SCN2A mutations in FS is uncertain. One rare SCN2A mutation (R188W) was described in a Japanese family with both febrile and afebrile seizures, but another rare SCN2A mutation R524Q segregating with R188W is probably involved (Ito et al., 2006). A second family with partial seizures and FS also had this R524Q mutation plus a rare SCN1A mutation. Other rare SCN2A mutations have been found in BFNIS (Heron et al., 2007), including two families in which FS also occurred. These examples suggest that SCN2A mutations may play a modifier role for the presence of FS with other idiopathic epilepsies. Our finding of an association of an SCN2A SNP with IGEFS is consistent with this view, but clearly needs to be replicated. We are currently further investigating these associations.

Acknowledgment

We would like to acknowledge the support of Epilepsy Research U.K. for funding this study and the Sir Jules Thorn Charitable Trust and Wellcome Trust in funding the work to set up DNA pooling. We would also like to thank Dave Vaske of Prevention Genetics for his patience in dealing with our many requests and queries and Dr Neeti Hindocha for helpful comments on the paper.

None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal’s position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Disclosure

None of the authors has any conflict of interest to disclose.