Genetic Association Studies and the Search for Causal Variants

The first genome-wide association study (GWAS) was conducted with 306 RLS cases (with PLMS) and 15,664 controls from Iceland [242]: The study found a genome-wide significant intronic single nucleotide polymorphism (SNP) at the BTBD9 locus (OR = 1.8, rs3923809, chromosome 6), which could be replicated in 123 cases (with PLMS) and 1,233 controls from Iceland and in 188 cases (with PLMS) and 662 controls from the US (OR = 1.5), and the population attributable risk was estimated to be 57% with a multiplicative model in the combined Icelandic cohort [242]. The study also demonstrated a significant decrease of serum ferritin levels in RLS cases by the identified risk allele [242].

The second GWAS was done with 401 familial German RLS cases and 1,644 population based controls [243]. Using an Armitage test of trend, an association of rs2300478 was genome-wide significant, which mapped to an intronic region of MEIS1, and subsignificant signals were observed on chromosome 6 (BTBD9 locus), chromosome 9, chromosome 15, and chromosome 16 [243]. A replication study could partially confirm four associated loci in 903 familial and sporadic RLS cases and 891 population based controls from Germany as well as in 255 RLS cases and 287 controls from Canada, and the LD blocks of the associated SNPs were in proximity to four genes: MEIS1 (containing exon 9, chromosome 2), BTBD9 (in intron 5, chromosome 6), MAP2K5 (3’ end, chromosome 15) to SKOR1 and also PTPRD (linkage region RLS3, chromosome 9, nominally significant) [243]. Of note, a risk haplotype could be found for MEIS1 (allele A of rs6710341 to allele G of rs12469063), which gave a stronger association signal in both the German and the Canadian samples (OR = 2.75 and 2.36, respectively) [243]. Furthermore, estimations hinted to a recessive model for the MAP2K5/SKOR1 related signal [243]. The three loci at MEIS1, BTBD9 and MAP2K5/SKOR1 had a population attributable risk of 68.6% in the German and 74.2% in the Canadian population [243].

The odds ratio for the MEIS1 region was higher in familial RLS compared to sporadic RLS (1.82 vs 1.59, respectively) [243].

The next GWAS could confirm the previously observed PTPRD locus by a significant association [244].

Using 2,458 RLS cases and 4,749 controls (from Germany, Czech, Canada, Austria), two independent SNPs were associated with RLS: rs4626664 (OR = 1.44) and rs1975197 (OR = 1.31) [244]. Both SNPs were located in the 5’ UTR of PTPRD and were splice variants, but an interaction between the two SNPs (epistasis) could not be observed [244]. Of note, this genomic locus was observed before in the linkage studies, but sequencing could not detect obviously causal mutations in the PTPRD exons of affected family members from a pedigree linked to RLS3 [244].

1 Introduction

The last genome-wide association study on common variants revealed two more loci using 954 German/Austrian RLS cases and 1,814 German controls [245]. The replication was conducted with 3,935 cases and 5,754 controls (of European origin and from the US of European origin) [245]. The new loci were found in an intergenic region on chromosome 2 (rs6747972, OR = 1.23, approx 1.3 Mb downstream of MEIS1) and in a LD block of chromosome 16 that intersects the 5’ end of TOX3 and is in proximity to the non-coding RNA CASC16 (LOC643714, chromosome 16, rs3104767, OR = 1.35) [245]. All previously observed loci could be replicated: MEIS1 locus (rs2300478, OR = 1.68), BTBD9 locus (r29357271), PTPRD locus (rs1975197, OR = 1.29) and MAP2K5/SKOR1 locus (rs12593812, OR = 1.41) [245]. Additionally, the genome-wide data could explain 6.8% of the genetic variance [245].

1.9.3.2 Examination of GWAS SNPs in Diverse Populations and Phenotypes

Many regional association studies were conducted to replicate or examine the findings from the different GWAS in different populations or diverse RLS associated conditions. The BTBD9 locus was confirmed in a Korean population, and an attempt to show an association with serum iron levels failed [246]. The PTPRD locus could be validated in a Taiwanese study with RLS in ESRD patients but not the BTBD9, MAP2K5/SKOR1, PTPRD or TOX3/CASC16 locus [247]. In contrast, a study on Greek/German ESRD patients could confirm the BTBD9 locus, but it observed only a trend for the MEIS1 (German subset) and MAP2K5 (Greek subset) locus, and the other loci were not tested [248].

Only the MAP2K5/SKOR1 locus could be confirmed in Czechian MS patients stratified by RLS but not the MEIS1, BTBD9 and PTPRD locus [249]. A US study tested the MEIS1, BTBD9 and MAP2K5 locus with a family-based and a case-control based association study with participants of European descent, and all loci were significantly associated with RLS [250]. The PTRPD locus could also be confirmed in the US with Caucasian individuals, but no coding variant segregated in an RLS3 family [251]. Another project was conducted with the Wisconsin Sleep Cohort, where PLM was used as a quantitative trait and tested successfully for association with the BTBD9 (top signal), TOX3/CASC16, MEIS1, MAP2K5/SKOR1 and PTRPD locus [252]. Another study could confirm the BTBD9 locus as being associated with both familial and sporadic RLS, and the other tested loci of the MEIS1, BTBD9 and MAP2K5 locus were also associated with RLS in familial RLS cases [253].

A study asked the question whether the RLS GWAS loci could be associated with comorbidity in RLS patients (or the general population) from German population surveys as multimorbidity was shown to be a risk factor of RLS, but no such association could be found [254]. Similarly, another study asked whether the GWAS loci might be associated with the serum iron levels in the general population as iron deficiency was shown to be a RLS risk factor, but no association could be found in a general German population [255]. But the first GWAS did find such an association in Icelandic RLS cases [242].

1.9.3.3 Regional Association Studies Based on Pathophysiological and Epidemiological Considerations

Many regional association tests were performed based on hypothesis mainly driven by epidemiological observations and pathophysiological considerations. Often only a small set of selected genetic variants was tested and some studies could report associations, but none would have been genome-wide significant. Some examples were the following genes: VDR (vitamin D (1,25-dihydroxyvitamin D3 receptor [2]), Spanish Caucasian population, 2 common SNPs, p < 0.001 and p < 0.01) [256], CLOCK (Korean schizophrenia patients, haplotypes in CLOCK and NPAS2, p = 0.021 for the CLOCK haplotype) [257], HMOX1 (heme-oxygenase 1 [2] for iron metabolism, case-control cohort of Spanish descent, also tested HMOX2) [258], IL1B and IL17A (HIV cohort stratified by RLS,

1.9 Genetics of Restless Legs Syndrome

tested SNPs in NOS1 and NOS2 for dopamine metabolism, in HFE for iron metabolism, in IFN, IL, NFKB and TNFA for inflammation related genes) [259], SNCA (synuclein alpha [2], German RLS case-control cohort) [260], NOS1 (screening of more than 1,500 SNPs in RLS1 locus, RLS case-control cohort of European descent) [240] (a replication failed [261]), TH (tyrosine hydroxylase [2], Val81Met polymorphism, study of antipsychotic-induced RLS in Korean female schizophrenic patients) [262], and MAOA (dopamine metabolism, RLS case-control cohort of European descent) [263]. Many more studies were done that could not find an association at all. One of those studies focused on 111 iron related genes, e.g. HFE, which common SNPs were regressed against the phenotypes of a case-control cohort (in total 2,425 German Austrian cases and 3,285 German case-controls) but with negative results, and the authors discussed that an intermediate phenotype might exist that underlies the RLS phenotype and rendered the study powerless, and that this putative intermediate phenotype might be influenced by additional factors [255]. As another example, a recent study could not show an association of the BTBD9 locus with ferritin levels in 14,126 blood donors [264].

1.9.3.4 Search for Causal Variants in GWAS Loci

Some studies were done to find obviously causal variants in the GWAS loci. A recent study sequenced the exomes of 4 to up to 8 RLS patients from 7 French-Canadian families with an autosomal dominant pattern of RLS inheritance [265]. It analyzed the published RLS associated loci and identified common GLO1 variants in putative functional regions (E111A and in the promotor region) that co-segregated with the phenotype in 4 families and that were associated with RLS in a case-control cohort (191 French-Canadian cases, 191 French-Canadian controls, 255 US cases, 291 US controls) [265]. But a conditional haplotype analysis could not confirm the association after correcting for the published BTBD9 GWAS lead SNP rs9357271 [265]. Other putative functional variants were found in the GWAS loci, but none segregated with the phenotype in the families [265].

As another example, a pedigree showed a co-segregation of the phenotype with a MEIS1 coding variant (R272H) (3 affected members, 3 unaffected family members) [266]. In another project, PTPRD was sequenced in an RLS3 family, but no variant could be found that segregated with the phenotype [251].

Another study used a multistage approach: Initial sets of 188 cases and 182 controls were selected for each GWAS locus separately and screened for variation using a high resolution melting curve genotyping method (HRM) [267]. Then the resulting candidate variants were genotyped in 3,262 German cases and 2,944 German controls using the Sequenom MALDI-TOF technology [267].

Furthermore, the MEIS1 exons were screened in 3,760 German cases and 3,542 German controls using the HRM technology [267]. As a result, a significant excess of rare variants was found in MEIS1 exons, especially non-synonymous coding variants of the transcript ENST00000398506, in the 5’ UTR and in the 3’ UTR (minimal empirical p = 0.0001, obtained by 10,000 permutations) [267]. The strongest signal was mainly driven by the rare SNP rs11693221 [267]. The observed rare nonsynonymous variants were screened in an in vivo complementation assay with zebrafish embryos, which showed a knock out phenotype in the neuronal development (optic tectum) mainly with variants that surrounded the TALE box region of MEIS1 [267]. Of note, the R272H variant, which was formerly reported as segregating in an RLS family [266], showed a neurological phenotype in the in vivo assay [267]. A further study asked a similar question with a focus on MEIS1: Long range PCRs were used to screen the region of MEIS1 where the risk haplotype was located (3 RLS cases homozygous for the MEIS1 RLS risk haplotype and 2 RLS cases homozygous for the MEIS1 RLS non-risk haplotype) [70]. No insertions/deletions were detected being larger than 250 bp [70]. A

1 Introduction

sequencing approach could not find structural variants in the MEIS1 exons or in some conserved intronic regions between exon 8 and exon 9 in 570 French-Canadian cases/controls, and no loss of heterozygosity was detected for carriers of the MEIS1 RLS risk haplotype [70]. Furthermore, only one coding variant was found, but it was non-synonymous and not associated with the phenotype [70].

But a non-coding variant was found (rs113851554) in a conserved region and this marker was significantly associated with RLS [70]. The SNP was in high LD with rs11693221 (r² = 0.83) that was observed in the previously described study [267].