Case–Control and Family‐Based Association Study of Specific PTPRD Variants in Restless Legs Syndrome

Abstract

Background

The exact genetic causes within each of the known restless legs syndrome (RLS) loci are still unknown. Recently, it was suggested that an intronic protein tyrosine phosphatase, receptor type δ (PTPRD) single‐nucleotide polymorphism (SNP) (reference SNP no. rs2381970) is associated with its expression, which may lead to RLS and other related phenotypes. Another study identified 3 nonsynonymous PTPRD variants in familial RLS cases: p.Q447E (a residue change from glutamine to glutamic acid at position 447), p.T781A (a residue change from threonine to alanine at position 781), and p.R995C (a residue change from arginine to cysteine at position 995).

Methods

Two cohorts of sporadic RLS, a French‐Canadian cohort and a cohort from the United States, with a total of 577 patients and 455 controls, and an additional familial RLS cohort with a total of 635 individuals (140 families) were genotyped for these 4 variants (rs2381970, p.Q447E, p.T781A, and p.R995C) by using specific TaqMan probes, and the effects of each variant as well as haplotypes were analyzed.

Results

None of the 4 PTPRD‐specific variants or haplotypes that were tested were associated with RLS in the case–control cohorts or in the familial cohort. The frequencies of the rs2381970 variant in the French‐Canadian and US cohorts were 0.07 and 0.04, respectively, and their frequencies in the respective control populations were 0.06 and 0.04, respectively (P > 0.4 for both). Similar results were obtained for the 3 nonsynonymous variants.

Conclusions

Although the PTPRD gene is well established as an RLS‐associated locus, the rs2381970 SNP and the 3 nonsynonymous PTPRD variants are not likely to cause or affect the risk for developing RLS in the study population. More studies in other populations are needed to determine their potential role in RLS.

Several genetic loci are associated with restless legs syndrome (RLS), a common sensory‐motor disorder with a prevalence of up to 15% in different populations.1, 2 Genome‐wide association studies (GWAS) have identified 6 loci around the protein tyrosine phosphatase, receptor type δ (PTPRD), Meis homeobox 1 (MEIS1), BTB domain containing 9 (BTBD9), mitogen‐activated protein kinase kinase 5/SKI family transcriptional corepressor 1 (MAP2K5/SKOR1), and TOX high‐mobility group box family member 3 (TOX3) genes and in an intergenic region on chromosome 2p14, all of which affect the risk for RLS.3, 4, 5, 6 However, it is still not clear which specific genetic variants within each locus cause the increased susceptibility for RLS. It was previously reported that the strongest intronic variants of MEIS1 associated with RLS led to decreased expression of its messenger RNA (mRNA) and protein,7 which may lead to RLS.8 Similarly, a recent study focusing on PTPRD suggested that an intronic single nucleotide polymorphism (SNP), rs2381970, has a strong effect on the expression of the PTPRD gene, whereas the known risk markers at this locus, rs1975197 and rs4626664, had a milder effect.9 However, unlike the MEIS1 SNPs, it is not clear whether the PTPRD rs2381970 SNP is associated with the risk for RLS.

Furthermore, an exome‐sequencing study of 7 large RLS families demonstrated that no nonsynonymous, splicing, stop, or frame‐shift variants in these 6 putative RLS genes co‐segregated with the disorder.10 In that study, 3 PTPRD nonsynonymous variants were identified in 5 of the 7 families. In 1 family, the PTPRD p.T781A (a residue change from threonine to alanine at position 781) and p.Q447E (a residue change from glutamine to glutamic acid at position 447) variants were found in 6 and 5 of 8 patients, respectively. In another family, the p.Q447E variant was found in 3 of 5 patients, and 4 unrelated individuals from 4 different families carried the p.R995C variant (a residue change from arginine to cysteine at position 995).10 This lack of co‐segregation, however, does not rule out the possibility that these variants cause RLS. This can be explained by the high likelihood of phenocopies of RLS within families; because RLS is a very common disorder and because the GWAS risk factors for RLS are also very common, with frequency ranging from 0.16 (PTPRD locus) up to 0.76 (BTBD9 locus) in the general population,5, 6 it is likely that individuals from the same family may carry different genetic risk factors for RLS, hence preventing the detection of co‐segregation of a specific genetic variant with the disease status.

To examine whether the PTPRD rs2381970 SNP and the 3 PTPRD nonsynonymous variants are associated with susceptibility for RLS, we studied their effects in 2 case–control cohorts and 1 familial cohort.

Patients and Methods

Population

Two case–control cohorts of patients with RLS were analyzed: a cohort of unrelated French‐Canadian (FC) patients (n = 350; 62.6% women; average age, 53.6 ± 12.3 years) and controls (n = 238; 48.3% women; average age, 40.3 ± 14.9; P < 0.05 for sex and age) and a cohort of unrelated US patients (n = 255; 60.0% women; average age, 61.8 ± 14.1 years) and controls (n = 196; 61.2% women; average age 63.3 ± 9.5 years; P = 0.80 for sex;, P = 0.20 for age). An additional familial RLS cohort of 653 FC individuals from 140 families, including 458 patients and 177 unaffected individuals, was used for the family‐based association test. All patients were diagnosed according to International RLS Study Group (IRLSSG) criteria.11 All participants signed an informed consent form before entering the study, and the study protocols were approved by the respective institutional review boards.

Genotyping

DNA was isolated from peripheral blood leukocytes using a standard salting‐out method. Three PTPRD nonsynonymous variants—p.Q447E (rs10977171), p.T781A (rs72694737), and p.R995C (rs35929428)—and the intronic SNP rs2381970 were genotyped using specific TaqMan assays (C__25472373_20, C__97531567_10, C___1634674_20 and C__16226413_10, respectively; Applied Biosystems, Foster City, CA). The categorization of genotypes was done using the QuantStudio 7 Flex Real‐Time polymerase chain reaction system and software (Applied Biosystems).

Statistical analysis

Categorical variables are presented as percentages or frequencies between 0 and 1, and continuous variables are presented as an averages ± standard deviation. Differences in sex and age between patients and controls were analyzed using χ² and t tests, respectively. The χ² test was used to examine whether the distribution of genotypes in the controls deviated from Hardy–Weinberg Equilibrium (HWE). To analyze the association between PTPRD variants and RLS in the 2 case–control cohorts, a logistic regression model was used with the status of the individual (affected or unaffected) as the dependent variable and the genotypes as covariates. When the cases and controls did not match for sex and age, these variables were used as covariates to adjust for their effects. To analyze the familial cohort, a family‐based association test (FBAT) was performed for each variant separately and for the inferred haplotypes. Power analysis demonstrated that our case–control cohort had a power of 80% to detect minor allele frequency changes from 0.02 to 0.0412 (representing the lowest allele frequency identified in the study) and from 0.09 to 0.129 (representing the highest allele frequency identified in this cohort). Furthermore, the addition of a familial cohort of 635 individuals further increased the power of the study. The software package SPSS v. 22 (IBM, Somers, NY) was used for statistical analyses except for the FBAT, which that was performed using the FBAT v2.0.4 software package.12 In the haplotypes analysis only haplotypes with frequency >0.03 were included to avoid unstable estimates.

Results

Table 1 details the results from the association study between the 4 PTPRD variants and RLS in the 2 case–control cohorts. None of the variants deviated from HWE, and none were associated with risk for RLS in either cohort (Table 1) or in a combined analysis of both populations together adjusted for sex, age, and ethnicity (data not shown). To examine whether carriage of more than 1 nonsynonymous variant may have a cumulative effect on the risk for RLS, we compared the rates of individuals who carried 1, 2, or 3 of the variants in patients and controls. Among the FC RLS patients, 26.3% carried 1 variant, 4.9% carried 2 variants, and 0% carried 3 variants, compared with 23.9%, 4.6%, and 0.4%, respectively, of controls (χ² = 1.89; degrees of freedom [df], 3; P = 0.60); and, in the US cohort, the rates were 23.3%, 2.2% and 0%, respectively, among patients and 30%, 1.8%, and 0%, respectively, among controls (χ² = 2.50; df, 2; P = 0.29). Similarly, in the familial cohort, FBAT revealed that none of the variants or haplotypes were associated with RLS after Bonferroni correction for multiple comparisons (Table 2).

Table 1.

Association between Protein Tyrosine Phosphatase, Receptor Type δ Variants and Restless Legs Syndrome in 2 Case‐Control Cohorts

ID	Variant	Genotype/Allele	French‐Canadian Cohorta					US Cohort
			No. (%)					No. (%)
			Patients, n = 350	Controls, n = 238	OR	95% CI	P	Patients, n = 227	Controls, n = 217	OR	95% CI	P
rs10977171	p.Q447E	G/G	303 (86.6)	212 (89.1)	Ref	—	—	211 (93.0)	200 (92.2)	Ref	—	—
		G/C	46(13.1)	24 (10.1)	1.42	0.78‐2.60	0.26	15 (6.6)	17 (7.8)	0.85	0.41‐1.77	0.66
		C/C	1 (0.3)	2 (0.8)	0.47	0.04‐6.25	0.57	1 (0.4)	0 (0.0)	—	—	—
		G, no. [freq]	652 [0.93]	448 [0.94]	Ref	—	—	437 (0.96)	417 (0.96)	Ref	—	—
		C, no. [freq]	48 [0.07]	28 [0.06]	1.28	0.73‐2.22	0.39	17 (0.04)	17 (0.04)	0.97	0.49‐1.94	0.93
rs72694737	p.T781A	A/A	333 (95.1)	225 (94.5)	Ref	—	—	215 (94.7)	206 (94.9)	Ref	—	—
		A/G	17 (4.9)	12 (5.0)	0.69	0.28‐1.71	0.42	12 (5.3)	11 (5.1)	0.97	0.42‐2.25	0.94
		G/G	0 (0.0)	1(0.4)	—	—	—	0 (0.0)	0 (0.0)	—	—	—
		A, no. [freq]	683 [0.98]	462 [0.97]	Ref	—	—	442 [0.97]	423 [0.98]	Ref	—	—
		G, no. [freq]	17 [0.02]	14 [0.03]	0.59	0.25‐1.38	0.26	12 [0.03]	11 [0.02]	1.02	0.45‐2.35	0.96
rs35929428	p.R995C	G/G	294 (84.0)	201 (84.5)	Ref	—	—	195 (85.9)	175 (80.6)	Ref	—	—
		G/A	51 (14.6)	34 (14.3)	1.07	0.63‐1.81	0.81	30 (13.2)	39 (18.0)	0.68	0.40‐1.14	0.15
		A/A	5 (1.4)	3 (1.3)	1.14	0.24‐5.38	0.87	2 (0.9)	3 (1.4)	0.59	0.10‐3.61	0.57
		G, no. [freq]	639 [0.91]	436 [0.92]	Ref	—	—	420 [0.93]	389 [0.90]	Ref	—	—
		A, no. [freq]	61 [0.09]	40 [0.08]	1.07	0.67‐1.70	0.79	34 [0.07]	45 [0.10]	0.71	0.45‐1.13	0.15
rs2381970	Intronic	T/T	311 (88.9)	212 (89.1)	Ref	‐	‐	206 (90.7)	192 (88.5)	Ref	‐	‐
		T/C	38 (10.9)	25 (10.5)	1.01	0.55‐1.85	0.99	21 (9.3)	23 (10.6)	0.86	0.46‐1.62	0.65
		C/C	1 (0.3)	1 (0.4)	0.14	0.01‐2.42	0.18	0 (0.0)	2 (0.9)	—	—	—
		T, no. [freq]	660 [0.94]	449 [0.94]	Ref	—		433 [0.95]	407 [0.94]	Ref	—	—
		C, no. [freq]	40 [0.06]	27 [0.06]	0.90	0.50‐1.60	0.71	21 [0.05]	26 [0.06]	0.75	0.42‐1.36	0.35

^aThe regression model for the French‐Canadian cohort was adjusted for sex and age.

ID, reference single nucleotide polymorphism identifier; OR, odds ratio; CI, confidence interval; p.Q447E, residue change from glutamine to glutamic acid at position 447; G, guanine; Ref, referent category; C, cytosine; freq, frequency; p.T781A, residue change from threonine to alanine at position 781; A, adenine; p.R995C, residue change from arginine to cysteine at position 995; T, thymine.

Table 2.

Association of Single Nucleotide Polymorphisms and Haplotypes in a Familial Cohort of Restless Legs Syndrome

SNP	Varianta	AF	Familiesb	Z	P
					Uncorrectedc	Correctedd
ID
rs10977171	p.Q447E	0.056	15	0.31	0.76	1.00
rs72694737	p.T781A	0.023	3	1.40	0.16	0.64
rs35929428	p.R995C	0.077	20	−2.15	0.03	0.12
rs2381970	Intronic	0.066	20	0.72	0.47	1.00
Haplotype
A	G‐A‐G‐Td	0.814	42	0.18	0.86	1.00
B	G‐A‐A‐Td	0.050	19	−2.19	0.03	0.12
C	C‐A‐G‐Td	0.049	16	0.77	0.44	1.00
D	G‐A‐G‐Cd	0.046	16	1.07	0.28	1.00

^aFor variants, p.Q447E indicates a residue change from glutamine to glutamic acid at position 447; p.T781A, a residue change from threonine to alanine at position 781; p.R995C, a residue change from arginine to cysteine at position 995; G, guanine A, adenine; T, thymine; C, adenosine

^bValues indicate the number of families with carriers of the allele; 85 individuals were excluded from the analysis because they were from singleton families.

^cThese are uncorrected P values (Bonferroni‐corrected p values put the cutoff for statistical significance at P < 0.0125).

^dThese are Bonferroni‐corrected P values.

The order of the nucleotides corresponds to rs10977171, rs72694737, rs35929428, and rs2381970.

SNP, single nucleotide polymorphism; AF, allele frequency; ID, reference single nucleotide polymorphism identifier.

Discussion

The aim of the current study was to determine whether specific PTPRD variants, including 3 nonsynonymous PTPRD variants that were identified in exome‐sequencing study10 and 1 PTPRD intronic SNP that was suggested to strongly affect the expression of PTPRD,9 are associated with RLS. Although PTPRD is a well‐established susceptibility locus for RLS, our results suggest that none of these specific variants are associated with susceptibility for RLS in our populations. We note that part of our population was included in the original description of PTPRD as a risk factor for RLS, which identified 2 SNPs associated with RLS.6 The allele frequencies of these SNPs in a Canadian population in that study were 0.203 and 0.156 for rs1975197 in patients and controls, respectively, and 0.159 and 0.117, respectively, for rs4626664. These SNPs are not in linkage disequilibrium (LD) with the 4 variants that were analyzed here.

Some of the genetic risk loci for RLS encompass more than 1 gene. For example, the BTBD9 locus includes 2 more genes, glyoxalase I (GLO1) and dynein, axoneamal, heavy chain 8 (DNAH8),3 and either of genes may be the cause for the association with RLS, although functional studies suggest that it is most likely the BTBD9 gene.13, 14 However, the PTPRD locus seems to include solely PTPRD, suggesting that variants that affect either the function or the regulation of the gene are the cause of the increased risk for RLS. The recent observation that the PTPRD intronic SNP rs2381970 is associated with the expression of PTPRD—more so than the 2 known risk markers in this locus (rs1975197 and rs4626664)9—suggests that it may affect RLS susceptibility and that the expression of the gene has a role in the pathogenic mechanism leading to RLS. This was supported by knockdown experiments in mice, which resulted in behavioral effects on sleep patterns.9 If this SNP was the direct cause for PTPRD reduced expression, and if reduced expression did have a role in RLS pathogenesis, then this SNP should have been associated with the risk for RLS. However, although this SNP was not associated with RLS in our cohorts, it does not rule out the possibility that PTPRD expression affects the risk for RLS. It is possible that, in the original study, this SNP was in LD with another variant that affected the expression of PTPRD, whereas, in our populations, this SNP was not in LD with a variant that affected the expression.

Alternatively, it is possible that other regulatory effects are responsible for the association between PTPRD and RLS rather than levels of expression. Intronic variants or synonymous coding variants may affect the splicing of PTPRD and modify the balance between the different isoforms of PTPRD. Such changes will not necessarily be identified when examining the total expression of the gene based on RNA levels. DNA variants may also affect other regulatory elements, such as the secondary structure of mRNA, RNA editing and methylation, binding of microRNAs, etc. Therefore, it is still not clear which PTPRD variants directly affect the risk for RLS and by which mechanism they influence the function or regulation of PTPRD. Identifying these variants and characterizing their effects on PTPRD may be of great importance for future therapeutic interventions. It is possible that such identification will only be possible when whole‐genome sequencing or targeted whole‐genome sequencing is available.

Author Roles

1. Research Project: A. Conception, B. Organization, C. Execution; 2. Statistical Analysis: A. Design, B. Execution, C. Review and Critique; 3. Manuscript Preparation: A. Writing the First Draft, B. Review and Critique.

Z.G.‐O.: 1A, 1B, 2A, 2B, 3A

S.Z.: 1C, 2B, 3B

A.J.: 1B, 1C, 3B

J.Y.M.: 1A, 1B, 3B

R.P.A.: 1A, 1B, 3B

C.J.E.: 1A, 1B, 3B

A.D.: 1A, 1B, 3B

P.A.D.: 1A, 1B, 3B

L.X.: 1A, 1B, 1C, 2C, 3B

G.A.R.: 1A, 1B, 2C, 3B

Disclosures

Funding Sources and Conflicts of Interest: This work was financially supported by CIHR (grant MOP‐82900). The authors report no conflicts of interest.

Financial Disclosures for the previous 12 months: Ziv Gan‐Or received grants from the Michael J. Fox Foundation and a fellowship from the CIHR. Jacques Y. Montplaisir received grants from the CIHR, the Canadian Research Chair Program, the W. Garfield Weston Foundation, and Parkinson Society Canada. Richard P. Allen received grants from the US National Institutes of Health‐National Institute of Neurological Disorders and Stroke and AMAG Pharmaceuticals and fees for consultant services from Luitpold Pharmaceuticals. Alex Desautels received grants from Novartis and Jazz Pharmaceuticals and honoraria for conferences from UCB and Paladin Laboratories. Lan Xiong received grants from the CIHR, and from the Canadian Foundation for Innovation. Patrick A. Dion received a grant from the ALS Society of Canada. Guy A. Rouleau received grants from the CIHR, Genome Quebec, the ALS Society of Canada, the ALS Association, the Michael J. Fox Foundation, and from Fonds de recherche du Québec.