Role of the sodium channel SCN9A in genetic epilepsy with febrile seizures plus and Dravet syndrome

Authors


Summary

Mutations of the SCN1A subunit of the sodium channel is a cause of genetic epilepsy with febrile seizures plus (GEFS+) in multiplex families and accounts for 70–80% of Dravet syndrome (DS). DS cases without SCN1A mutation inherited have predicted SCN9A susceptibility variants, which may contribute to complex inheritance for these unexplained cases of DS. Compared with controls, DS cases were significantly enriched for rare SCN9A genetic variants. None of the multiplex febrile seizure or GEFS+ families could be explained by highly penetrant SCN9A mutations.

Dravet syndrome (DS) is a debilitating infantile-onset encephalopathy. Mutations in the sodium channel subunit gene SCN1A account for 70–80% of cases (Marini et al., 2007) and mutations in GABRG2 and PCDH19 and homozygous SCN1B mutations can cause a clinical picture suggestive of DS (Harkin et al., 2002; Depienne et al., 2009; Patino et al., 2009). However, the cause in approximately 20–30% of cases remains elusive. Possibilities include de novo dominant mutation in unknown genes or a multifactorial etiology.

Genetic epilepsy with febrile seizures plus (GEFS+) has heterogeneous phenotypes, with both complex and familial autosomal dominant inheritance. A sodium channel SCN9A mutation is known in a family variously diagnosed as autosomal dominant febrile seizures (FS) or GEFS+ (Peiffer et al., 1999; Scheffer et al., 2000; Singh et al., 2009). SCN9A variation has also been suggested as a genetic modifier that exacerbates mild SCN1A mutation–associated GEFS+ and as a susceptibility gene for DS (Singh et al., 2009).

We examined additional multiplex FS and GEFS+ families for highly penetrant SCN9A mutations and explored the possibility of SCN9A contributing to a multifactorial etiology for Dravet syndrome.

Materials and Methods

Clinical sample

SCN9A was genotyped in 13 and 46 multiplex families with FS and GEFS+, respectively, and 125 sporadic patients with DS. One DS patient had a previously reported GABRG2 mutation, 106 had a mutation in SCN1A, and 18 had no SCN1A mutation or large pathogenic copy number variant. Controls for checking population frequency for variants detected in our screen were 96 anonymous blood donors and controls for comparing overall frequency of rare variants in cases and controls were 6,515 exomes (http://evs.gs.washington.edu/EVS/). Ethical approval was provided by the Austin Health Human Research Ethics Committee and the Research Ethics Committee of the Child, Youth and Women's Health Service.

Molecular methods

Mutation screening for SCN9A, GABRG2, PCDH19, and SCN1B and large pathogenic copy number variants was carried out as described in Appendix S1.

Prediction of pathogenic effect of low frequency SCN9A genetic variants

PolyPhen-2 (Adzhubei et al., 2010) and SIFT (Kumar et al., 2009) were applied as predictors of pathogenic impact of SCN9A amino acid substitutions. The efficacy of this approach was tested on established pathogenic mutations responsible for SCN9A-related pain disorders, and then translated to SCN9A low frequency genetic variants detected from our patient group. Statistical analyses were carried out using a two-tailed Fisher's exact test.

Results and Discussion

Large pathogenic copy number variants (Mefford et al., 2011), SCN1B homozygous mutations (Kim et al., 2013), and GABRG2 and PCDH19 mutations were excluded as alternative causes for the 18 DS cases with no SCN1A mutation. The genetic modifier hypothesis for SCN9A was not testable, since none of the GEFS+ families included a case of DS where SCN9A variation might exacerbate effects of a mild GEFS+ related SCN1A mutation.

The SCN9A variation detected in the 125 DS patients with and without causative SCN1A mutations was then analyzed (Table 1 and Tables S1 and S2). Rare SCN9A variants were inherited from an unaffected parent, apart from one de novo E582K variant in a sporadic DS case. The same patient had a causative SCN1A frameshift mutation (E1032fsX1045). Status of the SCN9A E582K variant was therefore indeterminate and could not be resolved by prediction with PolyPhen-2 (benign) and SIFT (damaging) (Table S2).

Table 1. SCN9A genetic variation for cases with Dravet syndrome and GEFS+ (present study)
Exon/AmpliconPositionaAmino acidbAmino acidcAllele frequency (%)d
SCN9A SCN9ASCN1ADravet (N = 125)Controls (N = 96)
  1. Dravet syndrome is structured by frequency of the SCN9A variant (rare <1%; polymorphic = or >1%) and on presence or absence of a known de novo SCN1A mutation.

  2. a

    Base change in the cDNA sequence for SCN9A (NCBI Accession Number NM_002977).

  3. b

    The corresponding amino acid substitution in SCN9A (NCBI Accession Number NP 002968).

  4. c

    The known SCN1A mutation from the same patient (NCBI Accession Number AB093548).

  5. d

    Refers to SCN9A allele frequency (as a percentage) across all 125 Dravet syndrome cases. In the case of the rare L266M variant present in one GEFS+ family, members within each family are related; therefore, the allele frequency was calculated as occurrence in one family over the 59 families studied. Allele frequencies (as percentages) for P610T and N1245S are based on numbers summed across sections (the two polymorphic sections with, and without, SCN1A mutations).

  6. e

    De novo SCN9A mutation.

  7. f

    The two W1538R patients also carried the M932L, V991L, and D1908G variants, which are present at low frequency in controls.

  8. g

    Times (X) is the number of cases with that genetic variant.

  9. h

    SCN1A-related F14fsX91, A98P, R1645Q, T166fsX170, V944E, entire SCN1A gene deleted.

  10. i

    SCN1A-related Q142X, del exons 21–26, IVS1-1G>A, F1707V.

Rare (<1%) SCN9A variants in Dravet syndrome with known de novo SCN1A mutation
12c.1675G>AG559SD1416H0.400.00
12c.1744G>AE582KeE1032fsX10450.400.00
26c.4612T>CW1538Rf (2X)gY413N & T226M0.800.00
27–4c.5723A>GD1908GA1441P & I1922T0.800.52
Low frequency SCN9A polymorphism (>1%) in Dravet syndrome with known de novo SCN1A mutation
12c.1828C>AP610Td (6X)gVarioush2.803.13
18c.3329G>AR1110Q (4X)gVariousi1.601.04
20c.3734A>GN1245SdL1660P1.201.56
Rare (<1%) SCN9A variants in Dravet syndrome but SCN1A mutation negative
3c.360T>GI120MNegative0.400.00
5c.554G>AR185HNegative0.400.00
26c.4664T>CL1555PNegative0.400.52
Low frequency SCN9A polymorphism (>1%) in Dravet syndrome but SCN1A mutation negative
12c.1828C>AP610TdNegative2.803.13
20c.3734A>GN1245Sf (2X)gNegative1.201.56
Rare (<1%) SCN9A variants in Dravet syndrome as part of a GEFS+ family with a known GABRG2 mutation
21c.3799C>GL1267VNegative0.400.00
Rare SCN9A variants in GEFS+/FS families
7c.796C>AL266MNegative0.85d0.00

The other low frequency SCN9A variants detected are summarized in Table 1. The distinction between rare variants and polymorphisms was based on a minor allele frequency (MAF) of 1%, which defines polymorphism (rare variant MAF <1%; polymorphic MAF ≥1%). The presence of low frequency subclinical susceptibility variants is not unexpected in the normal population (Klassen et al., 2011).

The degree of protein sequence conservation of SCN9A among vertebrates and among other α-subunits of human sodium channels for the variants described in Table 1 is shown in Table S1. Conserved protein sequence suggests functional importance and is taken into account by PolyPhen-2 and SIFT scores (Table S2).

We compared predicted pathogenicity of SCN9A variants detected in patients with DS to the predictions associated with known deleterious SCN9A mutations in the pain disorders (Table S2). The PolyPhen-2 scores for this analysis were viewed in the context of cases grouped as “like” disorders. Therefore, a “pathogenic pattern” of variation is used to analyze the overall significance of a cluster of SCN9A genetic variants within clinically distinct disorders. PolyPhen-2 prediction of pathogenicity for SCN9A mutations associated with the pain disorders erythromelalgia/primary erythermalgia (IEM) and paroxysmal extreme pain disorder (PEPD) is consistent with their established status as causative with 77% and 100% concordance, respectively (Table 2). Consistency for IEM rises from 77% to 100% using SIFT, a result supported by functional studies as explained in footnote “e” of Table S2, and consistency for PEPD drops from 100% to 89% using SIFT. Therefore, these bioinformatic analyses are useful predictors for overall pathogenic patterns in both IEM and PEPD, and these two pain disorders were combined for subsequent statistical comparisons with DS.

Table 2. Pathogenic patterns predicted by PolyPhen-2 for SCN9A low frequency variants in cases with IEM, PEPD, and Dravet syndrome, compiled from Table S2 but excluding R1150W, which is a common variant
DisorderSCN9A PolyPhen-2 prediction [SCN9A SIFT prediction]
Benign [Tolerated]Possibly damagingProbably damaging [Damaging]
  1. SIFT predictions are given in square parentheses for comparison.

  2. a

    See Table S2 footnote “e” for evidence of true pathogenic effect from functional studies for the two IEM variants SCN9A I136V and Q10R, which negates the PolyPhen-2 benign prediction for these two variants. SIFT predicted both as damaging, as well as the third variant P610T.

  3. b

    SIFT predicted R996C as tolerated (see Table S2 footnote “k”) reducing the 100% predicted by PolyPhen-2 down to 89%.

  4. c

    The two cases of Dravet syndrome without myoclonic seizures and ataxia (SMEB-MA) listed in Table 1 of Singh et al. (2009) with de novo SCN1A mutations were grouped with the remainder of the SCN1A mutation positive Dravet syndrome mutations in that report. Four of the five SCN1A mutation–positive cases predicted as benign by PolyPhen-2 were predicted as damaging by SIFT.

  5. d

    Includes SCN9A variants W1538R, M932L, and V991L, which are rare variants all present in each of two patients. See footnote “f” of Table 1.

  6. e

    Includes SCN9A L1267V present in the patient with the GABRG2 Q351X mutation. Five of the six cases were predicted as damaging by SIFT.

IEMa3/13 [0/13]0/13

10/13

[13/13]

(77%)

[100%]

PEPDb0/9 [1/9]0/9

9/9

[8/9]

(100%)

[89%]

Dravet syndrome SCN1A positivec (Singh et al., 2009)5/7 [2/7]1/7

1/7

[5/7]

 
Dravet syndrome SCN1A positived (present study)6/9 [5/9]1/9

2/9

[4/9]

 
Dravet syndrome SCN1A negativec (Singh et al., 2009)2/3 [2/3]0/3

1/3

[1/3]

 
Dravet syndrome SCN1A negativee (present study)1/6 [1/6]1/6

4/6

[5/6]

 
Dravet syndrome (All combined)14/25 [10/25]3/25 (12%)

8/25

[15/25]

(32%)

[60%]

Combined Dravet SCN1A positive only11/16 [7/16]2/16

3/16

[9/16]

(19%)

[56%]

Combined Dravet SCN1A negative only3/9 [3/9]1/9

5/9

[6/9]

(56%)

[67%]

PolyPhen-2 and SIFT analyses of Dravet syndrome

Analyses were translated to SCN9A genetic variants detected in our patient group and it can be seen that the PolyPhen-2 tool is a more conservative predictor than SIFT (Table 2). We therefore regarded PolyPhen-2 as our primary tool but compared those results with the corresponding SIFT result. Ignoring the indeterminate category of “possibly damaging” variants, we compared known SCN9A pathogenic mutations in IEM and PEPD with those of unknown effect detected in DS (all cases combined). Thirty-two percent (8/25; Table 2) of these low frequency SCN9A genetic variants detected in DS combined from this study and a previously published study were predicted as probably damaging. The difference between the proportion of pathogenic mutations in IEM and PEPD and low frequency variants found in DS was highly significant (p = 0.0016). Using SIFT data (Table 2) the difference remained very significant (p = 0.0054). Therefore, SCN9A variants in all DS do not present with the same pathogenic pattern as for the pain disorders.

DS may be divided into cases with an established SCN1A mutation and cases negative for an SCN1A mutation (Table 2). Differences in numbers of patients with a predicted damaging SCN9A variant in these two groups (3/16 with SCN1A mutation vs. 5/9 without SCN1A mutation) is approaching statistical significance (p = 0.08), suggestive of a difference between the two groups of DS with or without SCN1A mutations. But this was not confirmed using SIFT results from Table 2 (p = 0.6913) owing to more of the SCN9A variants (9/16 as opposed to 3/16) in the SCN1A-positive group predicted as damaging by SIFT.

Considering those cases of DS with a causative SCN1A mutation and damaging SCN9A variant (3/16 for Polyphen-2; 9/16 for SIFT), compared with damaging SCN9A mutations in IEM and PEPD (19/22), the difference is significant for both PolyPhen-2 (p = 0.0002) and SIFT (p = 0.0054). Statistically the pathogenic patterns are different and do not support a role for SCN9A variants in cases already explained by SCN1A mutation.

With respect to those cases of DS with no SCN1A mutation and with a damaging SCN9A variant (5/9 for PolyPhen-2; 6/9 for SIFT), compared with damaging mutations in IEM and PEPD (19/22), the difference is NOT significant for both PolyPhen-2 (p = 0.30) and SIFT (p = 0.3204), indicating that there is no statistically significant difference in the role of SCN9A variation between DS and pain disorders. That is, SCN9A variation may contribute to genetic susceptibility in DS where there is no SCN1A mutation.

Comparison with controls

There are 143 rare coding variants in SCN9A from 6,515 exomes, a frequency of 2.2% (http://evs.gs.washington.edu/EVS/). We detected 15 rare coding variants in SCN9A from 125 cases of DS (Table 2), a frequency of 12%. Singh et al. (2009) detected 10 rare coding variants in SCN9A from 109 cases of DS, a frequency of 9.2%. The enrichment for rare SCN9A variants in DS compared with controls is extremely significant in both studies (p < 0.0001 and p = 0.0002).

PolyPhen-2 predicted 49 of the 143 EVS variants (34%) as probably damaging. When combining DS results from our study and Singh et al. (2009) (Table 2) we determined that 8/25 (32%) are probably damaging. Therefore, frequencies of damaging variants are similar in cases and controls, but rare variants are greatly enriched in DS.

Febrile seizures and GEFS+ families

Of the 59 FS and GEFS+ families analyzed, only one GEFS+ family had a potentially pathogenic SCN9A variant, L266M (in exon 7), which was found only in one mother and child in one branch of the pedigree. The SCN9A L266M substitution affected a highly conserved amino acid position (Table S1) and was predicted as probably damaging (Table S2) by both PolyPhen-2 and SIFT.

Conclusion

The well-accepted IEM and PEPD pain-related SCN9A mutations show the expected pathogenic pattern derived from the computational prediction analyses used. The pathogenic pattern detected for DS without SCN1A mutation was consistent with a role for SCN9A in genetic susceptibility to DS. Genetic heterogeneity is now well established for the monogenic epilepsies so that polygenic heterogeneity for epilepsy with complex inheritance sometimes including SCN9A could be an explanation for unexplained Dravet syndrome.

Acknowledgments

We are indebted to the patients and their families for contributing to our research. We thank the National Health and Medical Research Council of Australia, Thyne-Reid Charitable Trusts, MS McLeod Foundation, and SA Pathology for support; Robert Schultz and Beverley Johns for technical assistance; Simon Harvey for clinical information on patients; and two anonymous referees for their rigorous comments and constructive suggestions.

Disclosures

None of the authors have any conflicts 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.

Ancillary