Family studies to find rare high risk variants in migraine

With a lifetime prevalence of 16%, migraine affects 75 million Europeans. It can be very disabling for the individual and is a large economic burden to society [1]. Unraveling the genetics of migraine is therefore highly relevant. Migraine is a complex disorder caused by several genes and environmental factors [2, 3]. A higher concordance of migraine in monozygotic than in dizygotic twins, and the 1.9-3.8 fold higher risk of migraine among first degree relatives of affected individuals, indicates an important genetic component [2‐5] Twin studies show a heritability of 34%–65% [5, 6]. Heritability of migraine with typical aura (MA) that affects approximately one third of migraineurs is higher than for migraine without aura (MO), and evidence supports that MA and MO have different, though somewhat overlapping, etiology [7, 8]. The diagnosis of migraine is based solely on patients’ history in the absence of validated biomarkers.

Causal mutations in three genes have been identified for familial hemiplegic migraine (FHM), a rare and severe, autosomal dominantly inherited subtype of migraine with aura [9‐11]. However, these genes are apparently not involved in the prevalent types of migraine [12, 13]. Several linkage studies and association studies using a candidate-gene approach have failed to identify any robust association between genetic variants and the prevalent types of migraine (see Table 1 for explanation of terms and methods mentioned) [14]. Recently, a large meta-analysis including almost 60,000 affected subjects demonstrated 44 independent common single nucleotide polymorphisms (SNPs) associated with migraine [15]. Odds ratios (ORs) ranged between 0.85 and 1.11, and the mechanisms of action in migraine are unknown. It is now clear that no common variants are associated with a medium or high risk of migraine. Thus, common variants cannot explain much of the observed heritability of migraine. In general, only 1.5%–50% of heritability of common complex diseases and traits can be explained by common variants [16]. In other words, variants with medium or high risk must be rare and therefore not possible to capture by genome-wide association studies (GWAS) in unrelated case–control samples [17]. Migraine sometimes clusters in families with an inheritance pattern that often looks dominant. MA in particular seems to aggregate in large families [2, 3]. It is reasonable to think that rare genetic variants or mutations with medium to high effect size play an important role in these large families [18]. The same could be the case for families with multiple individuals affected by MO. A search for rare variants conferring a medium to high risk of migraine should therefore focus on families with many affected with migraine. Thus, it is hypothesized that the prevalent types of migraine can be oligogenic or even monogenic inherited, as is also hypothesized for many other common complex diseases [18]. If oligogenetic inherited, susceptibility is due to a specific combination of rare variants and unrelated cases represent a wide range of different combinations. These would be impossible to capture by GWAS alone. In a family, the combination of variants is probably specific to that one family, and because of this clustering of variants it will increase the chance to find it. Linkage analysis is the method of choice when studying monogenic disorders, but linkage studies are difficult in case of oligogenic inheritance. Previous linkage analyses in migraine families have shown association between several loci and MA and MO with LOD > 3 [19‐25]. None of these associations have been consistently replicated and the causal genetic variants remain to be identified, but it shows that the family approach is promising for migraine. Based on an a-priori hypothesis about the involvement of the TRESK potassium channel (encoded by the KCNK18 gene) in MA, Lafreniére et al. 2010 sequenced the entire gene region (i.e. a candidate-gene approach) and identified five variants in a case–control sample, of which one rare variant was subsequently shown to segregate perfectly with MA in a large multigenerational family (eight affected and 8 unaffected members) [26]. However, this association has later become questioned [26‐28]. The technology of next-generation sequencing (NGS) now provides an affordable tool to investigate genetic variation in the entire exome or genome [29]. In theory, a family based study design and genome sequencing can embrace the search for rare variants with both high and medium effect size, whether monogenic or oligogenic. The family approach using NGS has not yet been used in migraine genetic research on a larger scale, but we and probably several other groups now have ongoing studies. We therefore judge it timely to investigate how the family approach and NGS has been used successfully in other common complex disorders, and to deduce from that the possibilities for its application in migraine.

This is a focused review and by no means exhaustive on the topic of NGS in common complex diseases.

PubMed was searched to find studies on rare genetic variants using NGS in common complex diseases with positive findings.. There is no clear definition of the term “common disease”. The definition of a “rare disease” is not clear either, but it is defined by the European Commission as “prevalence of less than five per 10,000 in the Community” [30]. We therefore defined diseases not fitting this definition as common. Successful studies with a family-based design using a NGS method, not looking for de novo mutations or already known candidate genes were included. Studies focusing on cancer were also excluded. We searched the following terms: “Exome AND sequencing AND pedigree AND rare AND variants” which resulted in 192 articles of which 29 abstracts were read, 15 were read in full and 14 were included. Also the terms “family-based AND “exome-sequencing” AND rare AND variants” were searched and yielded 17 articles (three already included) out of which nine abstracts were read and eight articles were read in full resulting in three included studies.

“Migraine AND “whole genome sequencing ”and “migraine AND” “whole exome sequencing” only yielded three and nine hits respectively and one article were read in full and included.

We especially wanted to include studies of bipolar disorder and schizophrenia, because the diagnosis, like for migraine, relies on clinical characteristics in lack of a biomarker. Therefore, we also searched the terms “Schizophrenia AND “whole genome sequencing” AND families” yielding six articles whereof two were included, “Schizophrenia AND “whole exome sequencing” AND families” yielding 13 articles (one already included) of which 8 abstracts were read, 5 of them in full resulting in 3 included. The terms ““bipolar disorder” AND “whole exome sequencing” AND families” resulted in one article which was not included and ““bipolar disorder” AND “whole genome sequencing” AND families” resulted in two articles of which one was included.

The last search was performed the 12^th of June 2016.

We expected highly relevant articles on the topic would be published in Nature Genetics and PLOS genetics. Therefore every issue of the two journals published in 2014, 2015 and 2016 (UTAI June), were screened for relevant papers. This and reading through reference lists of reviews and other relevant papers lead to inclusion of another 10 studies. A total of 32 studies were included (Fig. 1).

In the following we shall describe a promising design for a family-study to find rare high risk variants in complex diseases used in our ongoing studies. For explanation of the different methods and terms mentioned, see table 1.

Family studies using different variants of the approach described above or some of its elements have already provided interesting results in common complex diseases other than migraine.

Thirty-two studies that fulfilled the criteria were included in this review [36‐38, 41‐69]. Six of them studied bipolar disorder, five of them studied schizophrenia (SCZ), three of them autism spectrum disorder (ASD) and two of them type II diabetes, while four studies focused on late onset Alzheimer’s disease (LOAD). There was one study focusing on each of the following: Parkinson’s disease, age-related macular degeneration (AMD), adiponectin level, atrial fibrillation, intracranial aneurysms, nonsyndromic cleft lip and palate (NSCLP), nonsyndromic hearing impairment, otitis media, preeclampsia, primary open angle glaucoma, inflammatory bowel disease and reumathoid arthritis. The study groups ranged from one to eighty families of varying size. They all used a combination of genetic methods like WGS, WES, sanger sequncing, GWAS and linkage-analysis. Some also used identity by descent analysis (IBD). Only three studies did WGS and three studies made use of family trios. One study searched for rare CNVs instead of SNPs [41]. 23 of the studies found one or a few variants (<5) specific for a family and the rest found several variants without finding one causal variant. Study details on disease, sample, techniques and findings are listed in Additional file 1: Table S1.

Four studies performed a follow-up analysis in unrelated cases and controls and presented an OR. Wetzel-smith et al. detected a missense variant (rs137875858 in UNC5C) in a large family with 8 LOAD cases using WGS, WES, linkage-analysis and assay genotyping. The same variant was found in four other pedigrees. Further genotyping of LOAD cases (8,050) and controls (98,194) resulted in an OR = 2.15 (95% CI = [1.21; 3.84], P = 0.0095) [42]. By WES and genotyping Cruchaga et al. detected a rare variant (rs145999145 in PLD3) segregating in 2 pedigrees with multiple LOAD cases. When testing independent cohorts of sporadic cases (4,998) and controls (6,356) they found an OR = 2.10 (95% CI = [1.47; 2,99] P=2.39×10⁻¹⁰-). For familial LOAD cases (1,106) vs. unrelated controls (6,356) an OR of 3.39 (95% CI = [2.14; 5.39] P = 1.18×10⁻⁶) was calculated [36]. Kohli et al. made use of WES, linkage analysis, genotyping and Sanger sequencing to study one pedigree counting 15 individuals. They found a rare variant (rs377155188 in TTC3) which segregated perfectly with LOAD in the pedigree. They calculated an OR for LOAD of 3.35 for sporadic cases (6,669) vs. controls (5,585) [43]. The result did not reach statistical significance (CI not specified). Goes et al. exome sequenced 36 individuals from 8 families and calculated ORs for variants in three genes with association to bipolar disorder in a case–control follow-up study (3,541 cases, 4,774 controls). It resulted in ORs of 2.73 (P = 0.016), 6.7 (P = 0.0039) and 2.78 (P = 0.045) for variants in MLK4, APPL2 and HSP90AA1 respectively [44].

Some studies found complete segregation between a variant and the studied disease and a few of them will be highlighted in the following. Cruceanu et al. exome-sequenced DNA from caucasian individuals affected with a highly heritable subtype of bipolar disorder in multigenerational families with three to seven affected individuals. Focus was on variants with MAF<1% in the general population. A missense variant was detected on chromosome 11 (position116652892) that leads to substitution of an aminoacid in the encoded protein. The variant segregated with affected family members in a family (five individuals) and was not present in unaffected family members selected as controls (six individuals). Whether this variant contributes to bipolar disorder is not known because of lack of knowledge about the involved gene [37].

Nyegaard et al. studied a five generational family containing 17 individuals with nonsyndromic hearing impairment (two of them dead). Seven already known hearing loss genes were not involved. 11 individuals were selected for SNP genotyping. 11.034 detected SNPs were included in a parametric linkage analysis which resulted in a significant linkage peak at chromosome six. To narrow down the size of the detected locus, 26 family members were genotyped for seven microsatellite markers. Then DNA from an affected individual was selected to undergo next generation sequencing (NGS) at the locus site by a costume designed sequence array. This identified 28,300 variants. One variant found in a coding region and predicted to have a functional effect was identified after excluding common variants. The mutation c.574C>T in CD164 was found in all affected individuals including a young girl with signs of a beginning hearing loss. The variant was absent in 12 unaffected family members, in one with unknown phenotype and in 1200 unrelated controls [45].

The studies reviewed here largely followed the steps we have described, but with many variations. In our description of a family based study design, we suggest to use family trios for WGS. Only three studies [46‐48], one of them still ongoing, did this. Also, only three studies [42, 46, 49] made use of WGS, which probably reflects that the cost of WGS has only been manageable very recently. It is highly likely that the non-coding regulatory areas play an important role in common migraine [70]. Georgi et al. chose to study a pedigree in an isolated population which has also been suggested as a possible approach to find rare variants [16, 46, 71, 72]. 23 studies [36‐38, 41‐45, 48‐62] reported the finding of less than five family specific variants (not all significant). Some studies could not find one specific causal variant probably because disease susceptibility is caused by more than one rare variant specific to a family (oligogenic inheritance). Detection of variants only makes sense if it is followed by further studies to clarify the causality of the variant. An et al. found several variants associated with autism spectrum disorder. They found enrichment of rare causal variants in key neurobiological processes, and overrepresentation of the rare causal variants in functions involving neuronal development, signal transduction and synapse development [48].

In the future, combinations of gene variants might be analyzed by “omics” approach, where bioinformatics integrate genomics, epigenomics, transcriptomics, proteomics and metabolomics [73, 74]. We excluded Ratnapriya et al. [75] from the reviewing part of this study because they studied a rare subtype of macular degeneration. They found a rare variant in a family with early onset macular degeneration in FBN2, but also a common variant in the same gene with a modest association to AMD cases. It is an excellent example of how both rare and common variants in a single gene can contribute to complex forms of a disease phenotype and the understanding of its pathophysiology.

Few studies focused on CNV’s, and only Van Den Bossche et al. [41], studying schizophrenia, succeeded in finding a CNV associated with disease. Rare inherited CNVs were more frequent in familial schizophrenia than in an unaffected control cohort [76]. This supports CNVs as an area of interest when searching for rare disease variants in migraine. In the future, NGS methods will be able to capture CNVs [77].

As mentioned, ORs for SNPs associated with migraine found through a GWAS ranged between 0.85 and 1.24. In the studies using a family approach reviewed here the ORs in follow-up case–control studies ranged between 2.10 and 6.7, the last one for the association between a variant in APPL2 and bipolar disorder. Like migraine, bipolar disorder and schizophrenia are common and complex neurological disorders with a clearly heritable factor and a diagnosis based on history in the absence of a biomarker. Success in these disorders raises hope to find specific rare variants with high relative risk in migraine families. Jiang et al. reported the preliminary finding of six novel rare non-synonymous mutations in a Chinese family with clustering of migraine without aura using WES. They included four cases (a father and three children) and four unrelated controls. However, the study had several limitations [78]. As far as we know, Jiang et al. is the only study using a family approach and NGS in migraine that has been published. F. Michael Cutrer, Mayo Clinic, Rochester has carried out WES in two large migraine families, according to a published grant description [79]. Five candidate genes were found to segregate with MA in one family. In the other family, including individuals with varying phenotypes, a single variant was detected. Whether the variants are rare was not stated. These results have unfortunately not been published. A similar study is ongoing on at the Danish Headache Center. This project aims to find rare genetic variants conferring a high risk of migraine using a family approach exactly as described previously. Extended Danish families with MA or MO are included. The study is still collecting data and biological material for sequencing and genotyping. A family approach in migraine will encounter obstacles. Correct phenotyping cannot avoid that unaffected controls may develop migraine later. The probability that some affected family members do not carry a family specific variant is high due to the high prevalence of migraine in the general population. Also, unaffected carriers are a possibility due to low level of penetrance. This will complicate the analysis. Many families contain a mix of individuals with MO, MA or so called MAMO (co-occurring MA and MO). It is not known whether the two phenotypes are part of a spectrum of the same disease or different diseases. Taking these problems into account, we still believe that a family approach is the best way to find variants with a high relative risk. Such variants can be the key to understand the pathophysiological mechanisms of migraine, and much more so than the common variants discovered by GWAS. It is obvious that migraine is highly genetically heterogeneous. Pathophysiology as well as response to prophylactic drugs vary considerably [80, 81]. On the other hand, 80% of patients respond to injection of Sumatriptan suggesting the existence of a final common pathway [82, 83]. If the etiology of just a few sub-phenotypes can be identified with certainty, it seems possible to identify one or more migraine pathwaysthat may be relevant for many patients, even if the genetic cause that lead to the discovery is rare. New targets for better and more specific treatments may then be discovered.

It has proven possible to find rare high risk variants for common complex diseases through a family based approach. One study using a family approach and NGS to find rare variants in migraine has already been published but it has strong limitations. More studies are under way.

Future family approach studies could be advanced by choosing isolated populations or individuals with severe phenotypes as study groups and include analysis of mitochondrial DNA and “omics”.