Vitamin D Receptor Gene as a Candidate Gene for Parkinson Disease

Authors

  • Megan W. Butler,

    1. Department of Pediatrics, Duke University Medical Center, Duke University School of Medicine, Durham, NC
    Search for more papers by this author
  • Amber Burt,

    1. John P. Hussman Institute for Human Genomics and Dr. John T. Macdonald Foundation Department of Human Genetics, Miller School of Medicine, University of Miami, Miami, FL
    Search for more papers by this author
  • Todd L. Edwards,

    1. John P. Hussman Institute for Human Genomics and Dr. John T. Macdonald Foundation Department of Human Genetics, Miller School of Medicine, University of Miami, Miami, FL
    2. Current address: Institute of Medicine and Public Health, Division of Epidemiology, Department of Medicine, Vanderbilt Epidemiology Center, Vanderbilt University School of Medicine, Nashville, TN 37203
    Search for more papers by this author
  • Stephan Zuchner,

    1. John P. Hussman Institute for Human Genomics and Dr. John T. Macdonald Foundation Department of Human Genetics, Miller School of Medicine, University of Miami, Miami, FL
    Search for more papers by this author
  • William K. Scott,

    1. John P. Hussman Institute for Human Genomics and Dr. John T. Macdonald Foundation Department of Human Genetics, Miller School of Medicine, University of Miami, Miami, FL
    Search for more papers by this author
  • Eden R. Martin,

    1. John P. Hussman Institute for Human Genomics and Dr. John T. Macdonald Foundation Department of Human Genetics, Miller School of Medicine, University of Miami, Miami, FL
    Search for more papers by this author
  • Jeffery M. Vance,

    1. John P. Hussman Institute for Human Genomics and Dr. John T. Macdonald Foundation Department of Human Genetics, Miller School of Medicine, University of Miami, Miami, FL
    Search for more papers by this author
  • Liyong Wang

    Corresponding author
    1. John P. Hussman Institute for Human Genomics and Dr. John T. Macdonald Foundation Department of Human Genetics, Miller School of Medicine, University of Miami, Miami, FL
    Search for more papers by this author

Corresponding author: Liyong Wang, University of Miami, 1501 NW 10th Ave, BRB 609, Miami, FL 33136. Tel.: 305-243-2377; Fax: 305-243-2704; E-mail: lwang1@med.miami.edu

Summary

Vitamin D and vitamin D receptor (VDR) have been postulated as environmental and genetic factors in neurodegeneration disorders including multiple sclerosis (MS), Alzheimer disease (AD), and recently Parkinson disease (PD). Given the sparse data on PD, we conducted a two-stage study to evaluate the genetic effects of VDR in PD. In the discovery stage, 30 tagSNPs in VDR were tested for association with risk as a discrete trait and age-at-onset (AAO) as a quantitative trait in 770 Caucasian PD families. In the validation stage, 18 VDR SNPs were tested in an independent Caucasian cohort (267 cases and 267 controls) constructed from a genome-wide association study (GWAS). In the discovery dataset, SNPs in the 5′ end of VDR were associated with both risk and AAO with more significant evidence of association with AAO (P= 0.0008–0.02). These 5′ SNPs were also associated with AD in another study. In the validation dataset, SNPs in the 3′ end of VDR were associated with AAO (P= 0.003) but not risk. The 3′ end SNP has been associated with both MS and AD in previous studies. Our findings suggest VDR as a potential susceptibility gene and support an essential role of vitamin D in PD.

Introduction

Parkinson disease (PD) is a progressive neurodegenerative disorder that affects roughly 1.5 million people in the US. The Mendelian forms of PD comprise <10% of all cases and involve well-defined exonic mutations in the genes PARK2, SNCA, PARK7, PINK1, ATP13A2, and LRRK2. The majority of PD cases result from complex interplay between environmental and genetic factors (Thomas & Beal, 2007). Several genetic risk factors and gene–environment interactions for PD have been reported by us and others (Li et al., 2003; van der Walt et al., 2004; Oliveira et al., 2005; Hancock et al., 2006; Maraganore et al., 2006; Mizuta et al., 2006; Winkler et al., 2007; Hancock et al., 2008; McCulloch et al., 2008; Mizuta et al., 2008). Vitamin D level is an environmentally modifiable factor as it is largely determined by diet and sunlight exposure. The effects of vitamin D and genetic variants in the vitamin D receptor (VDR) gene have recently gained interest in PD and neurodegenerative research in general (Evatt et al., 2008; Smolders et al., 2009; Knekt et al., 2010). It was first noted in Japan that PD patients had lower serum vitamin D levels and higher prevalence of vitamin D deficiency than their age-matched controls (Sato et al., 2005). This observation was recently confirmed in a European Caucasian population (Evatt et al., 2008). In a longitudinal study of more than 3000 participants in Finland, higher serum vitamin D level was associated with reduced risk for PD: people in the highest quartile had one-third of the risk compared to people in the lowest quartile (Knekt et al., 2010). The serum vitamin D level was also negatively correlated with PD severity as measured by Hoehn and Yahr stages and motor symptom as measured by the motor part of the Unified Parkinson's Disease Rating Scale III (UPDRS III) (Sato et al., 1997, 2005).

To exert its biological functions, vitamin D is first converted to the active metabolite 1,25-dihydroxy vitamin D3. Upon binding to 1,25-dihydroxy vitamin D3, VDR is activated and interacts with vitamin D responsive elements in the promoters of vitamin D target genes to regulate their expression (Garcion et al., 2002). VDR is the primary mediator of vitamin D's biological actions. VDR and 1α-hydroxylase, an essential enzyme to convert vitamin D to 1,25-dihydroxy vitamin D3, are expressed in human brain. Within the brain, the highest expression was found in the hypothalamus and in the large neurons (likely to be the dopaminergic neurons) of the substantia nigra (SN) (Eyles et al., 2005). The expression pattern of VDR and 1α-hydroxylase supports an important role of vitamin D in the etiology of PD. In addition, VDR null mice have impaired locomotor activity (Burne et al., 2005). Recently, the expression level of VDR mRNA was identified as a potential blood biomarker for PD (Scherzer et al., 2007).

Genetic studies on VDR polymorphisms also have suggested involvement of VDR with several neurodegenerative diseases, including multiple sclerosis (MS) and Alzheimer disease (AD). Motivated by the striking correlation between MS incidence rates and latitude, vitamin D status and VDR polymorphisms have been an active research topic in MS. SNPs in VDR have been associated with both MS risk and degree of disability after disease onset (Smolders et al., 2009). In AD, one candidate gene study reported a significant association at the 3′ end of VDR in a Turkish population (Gezen-Ak et al., 2007). A recent genome-wide association study (GWAS) of late-onset AD has reported strong evidence for association in a cluster of SNPs near the 5′ end of VDR in a sample of non-Hispanic Caucasians. (Beecham et al., 2009). In PD, Kim et al. (2005) examined the BsmI polymorphism (rs1544410) in a Korean population and found that the b allele was over-represented in 85 PD patients compared to 231 controls. Given the growing body of evidence for vitamin D and VDR's involvement in several neurodegenerative disorders including PD, and the lack of any published study evaluating association between VDR and PD in Caucasians, we sought to thoroughly examine VDR polymorphisms in a large Caucasian sample of PD families and validate the association using publicly available GWAS data.

Materials and Methods

Subjects

Affected individuals and family members were collected by the Morris K. Udall Parkinson Disease Research Center of Excellence (PDRCE) at Duke University and then at the University of Miami (J.M. Vance, PI), and the 13 centers of the Parkinson Disease Genetics Collaboration. A standard clinical evaluation has been described elsewhere (Hancock et al., 2006). Individuals were classified as affected if he or she had at least two cardinal signs of PD (resting tremors, rigidity, and bradykinesia), asymmetry of symptom onset, and absence of atypical signs. Unaffected individuals had no signs of PD. Individuals who showed only one cardinal sign and/or atypical signs were classified as unclear and were not used in the final analysis. Age-at-onset (AAO) was defined as the age at which affected individuals recalled the development of one of the cardinal signs. We used denature HPLC, sequencing and genotyping to screen probands from all families for exonic mutations in PARK2. Forty-two families carrying PARK2 mutation were excluded from the analysis. We also screened the proband of all families for the G2019S mutation in LRRK2 and the N370S and 84 GG mutations in GBA. Given the low prevalence of those mutations in our dataset (2% for G2019S, and less than 1% for GBA mutations) and therefore the small impact of those mutations, we did not exclude additional families based on LRRK2 and GBA screening. To reduce population substructure, only non-Hispanic Caucasian families (by self-report) were used in the statistical analysis. The final family dataset used in the analysis included 770 families comprising 1068 affected individuals, 1282 unaffected relatives and 436 individuals with unclear PD status. These 436 individuals were used to infer genotype for ungenotyped parents but were not used directly in the association analyses. All individuals or their legal representative gave informed consent prior to this study. All ascertaining protocols were approved by the institutional review board of PDRCE and each of 13 centers of the Parkinson Disease Genetics Collaboration (Oliveira et al., 2003).

SNP Selection and Genotyping

We used UCSC genome browser (http://www.genome.ucsc.edu), and the HapMap project (http://www.hapmap.org) to identify SNPs in and around VDR. TagSNPs (r2 threshold of 0.8) with a minor allele frequency ≥5% in Caucasians and all validated exonic SNPs were examined. Genomic DNA was extracted from peripheral blood using the PureGene system (Gentra systems, Minneapolis, MN, USA). We genotyped all SNPs using the Taqman Allelic Discrimination® assay (Applied Biosystems, Carlsbad, CA, USA) as described before (Hancock et al., 2006). All SNPs examined were successfully genotyped for 95% or more of the individuals in the study.

GWAS Dataset

To further validate the association between VDR SNPs and PD AAO (the most significant finding in our family study), we examined publicly available GWAS datasets. Three GWAS on PD have been deposited in the db genotypes and phenotypes (dbGAP) database (http://www.ncbi.nlm.nih.gov/gap): the National Institute of Neurological Disorders and Stroke (NINDS) study, the Progeni/GenePD study and the Linked Efforts to Accelerate Parkinson's Solutions (LEAPs). The Progeni/GenePD study has recently published a genome-wide analysis on AAO (Latourelle et al., 2009). No evidence for association was found in or near VDR (Supplemental Materials of Latourelle et al., 2009). The SNPs used in the LEAPs study are sparse (<200,000 SNP) and less informative (uniformly spaced). Therefore, the LEAPs GWAS data are not suitable for reliable imputation and comparison with our dataset. Genotype and phenotype data for the NINDS study were downloaded from the dbGAP database. Detailed data cleaning has been reported elsewhere (Edwards et al., 2010). Briefly, imputation of SNP genotypes was performed using Impute (Marchini et al., 2007). Samples with genotyping efficiency <0.98 and SNPs with genotyping call rate <0.98 were removed from analysis. In addition, SNPs with MAF < 0.01 or HWE P < 10−7, were excluded (Edwards et al., 2010).

Statistical Analysis

Tests for deviations from HWE were performed using Genetic Data Analysis software (http://hydrodictyon.eeb.uconn.edu/people/plewis/software.php). HaploView software (Barrett et al., 2005) was used to estimate linkage disequilibrium (LD) between SNPs. Family-based association analyses were performed using the test for association in the presence of linkage (APL), which infers genotypes of ungenotyped parents and therefore utilizes all families, including singleton and multiplex families, in the analysis (Martin et al., 2000). The Monks–Kaplan method was used to evaluate association with AAO as a quantitative trait (Monks & Kaplan, 2000). To explore potential genetic heterogeneity introduced by different AAO, we also stratified the dataset based on AAO for analysis of PD risk using APL. In order to have well-powered subsets, we started the stratification analysis by using the mean AAO of PD, that is, AAO of 60, to define earlier and later onset subsets (Destefano et al., 2002, Maher et al., 2002). A different AAO cutoff was explored to further examine the effect of AAO after the initial analysis. A similar strategy and AAO cutoff was implemented by Rocca and colleagues in their efforts to evaluate possible genetic heterogeneity in PD (McDonnell et al., 2006). For the GWAS dataset, linear regression analysis was used to test association between SNP genotypes and AAO using PLINK version 1.06 (Purcell et al., 2007).

To correct for multiple testing of SNPs in the gene, we applied SimpleM (Gao et al., 2008). This method takes into account the nonindependence between SNPs due to LD and infers the number of independent tests such that a standard Bonferroni correction can been applied to correct for multiple testing.

Results

The VDR gene has nine exons and multiple alternative first exons (Fig. 1). Thirty tagSNPs within the VDR exonic boundaries and the 5-kilobase (kb) flanking regions were genotyped to cover most of the common genetic variations surrounding VDR (Fig. 1). Using the SimpleM test, the effective number of independence tests was reduced from 30 SNPs to 24. One coding SNP in exon 9 (rs731236) is in perfect LD (r2= 1) with the BsmI polymorphism previously reported by Kim et al. (2005) Therefore, rs731236 was used as the surrogate for the BsmI polymorphism. LD between the 30 SNPs is shown in Figure 2.

Figure 1.

Diagram of VDR gene structure with relative positions of genotyped SNPs. The VDR gene is located on the bottom strand of genomic DNA and therefore is displayed 3′ to 5′, left to right. Coding exons are shown as higher blocks while noncoding exons are shown as shorter blocks, introns are shown as straight lines, and the dotted line suggests alternative splicing of exons. Thirty tagSNPs were chosen and genotyped in the family dataset. The locations of them are displayed in the context of VDR gene structure. SNP shown in bold was the proxy marker for the most significant SNP (P= 0.003) in NINDS dataset using linear regression; underlined SNP was the proxy marker for the previously reported significant SNP in the Korean study; SNP shown in bold and underlined is the most significant SNP (P= 0.0008) in the family dataset using Monks–Kaplan analysis.

Figure 2.

Pairwise LD plot for examined VDR SNPs. Pairwise r2 was estimated in unrelated individuals with no PD and is displayed by gray-shaded square with the hundredths of r2 value inside (HaploView) (Barrett et al., 2005). The shades of gray are proportional to the r2 value with darker color representing higher r2 value.

SNP associations with AAO as a quantitative trait (Table 1) and with risk as a discrete trait (Fig. 3) were evaluated using family-based association tests. The AAO analysis revealed strong evidence of association at several SNPs with the most significant association at rs4334089 (nominal P= 0.0008, corrected P= 0.02, Table 1). The risk analysis found marginal evidence for association at rs2853559 (nominal P= 0.02), which was not significant after multiple testing correction (Fig. 3). Given the strong effect of AAO, we further examined SNP association with risk in subsets with different AAO. Using AAO of 60 as the cutoff value, the overall family dataset was stratified into 459 earlier onset families and 311 later onset families. Five SNPs were significantly associated with PD risk in the earlier onset subset (nominal P < 0.05) but none in the later onset subset (Fig. 3). All five SNPs overlapped with the significant SNPs in AAO analysis including rs4434089. Further stratification using younger AAO (AAO < 40) did not reveal any significant results (data not shown), probably due to the small sample size (78 families with AAO < 40). The previously reported association with PD at BsmI polymorphism was not significant in any analysis as indicated by association tests at rs731236.

Table 1.  Monks–Kaplan SNP association tests with PD AAO in the family dataset.
SNPLocation (bp)Minor allele frequency*P values
  1. *Minor allele frequency was calculated in the overall dataset. SNPs with corrected P values less than 0.05 are in bold.

rs2853563465220050.030.4019
rs3858733465242340.030.6453
rs3847987465243350.120.5563
rs11574119465243590.040.1939
rs739837465244880.450.5951
rs731236465250240.400.5117
rs2239182465416780.490.6099
rs2107301465418370.260.1168
rs1540339465435930.350.9943
rs2239179465440330.430.7076
rs2189480465500950.340.0439
rs3819545465512730.380.5050
rs3782905465524340.330.9517
rs2239186465556770.180.8133
rs10735810465591620.380.1861
rs2254210465599810.380.1358
rs2238136465639800.270.4991
rs4760648465669320.420.0634
rs2853559465690720.400.2514
rs11168287465716810.490.4486
rs4334089465722820.250.0008
rs4237855465734700.390.1321
rs10783219465817550.370.4302
rs10083198465822320.280.0046
rs7299460465825350.280.0016
rs4760658465827530.340.0175
rs7976091465908190.210.0229
rs10875708466200170.200.0072
rs1015390466303050.150.1413
rs4760673466423760.160.3049
Figure 3.

SNP association tests with PD risk as a discrete trait. Each bar represents an association test with PD risk in different datasets: overall (white), earlier onset (gray), and later onset (black). The dotted line depicts the P value of 0.05. The association tests were performed using APL. AAO of 60 was used to stratify the overall families (N= 770) into earlier (N= 459) and later (N= 311) onset family subsets.

All significant SNPs in the family dataset are located in the 5′ end of the gene. The majority of them are located in intron 1A_1B and intron 1B_1C, between the alternative exon 1A, 1B, 1C that generate alternative splicing isoforms of VDR (Fig. 1). This cluster of SNPs is in moderate LD between each other (pairwise r2= 0.14–0.74 to rs4334089, Fig. 2). Haplotype analysis of these 5′ end SNPs did not define a haplotype that could explain the association better than single SNP analysis (data not shown).

Given the association between VDR and AAO in our family dataset, we further tested the association using the NINDS GWAS data deposited in dbGAP.

Eighteen SNPs in and around VDR were either genotyped or imputed in the NINDS dataset (N= 534). Sixteen of them overlapped with the SNPs studied in the family dataset, either by the SNP itself or by a tagSNP. The effective number of independence tests as estimated by SimpleM for the 18 SNPs is 15. In the NINDS dataset, three SNPs were significantly associated with AAO (Table 2) but none was associated with PD risk (data not shown). The strongest association was found at rs7968585 (nominal P= 0.003, corrected P= 0.045), which remained significant after correction for multiple testing in the NINDS dataset. None of the 5′ end SNPs or SNPs in LD with them was significant in the NINDS dataset.

Table 2.  Linear regression-based SNP association tests with PD AAO in the NINDS GWAS dataset.
SNPLocation (bp)BetaLower 95% confidence intervalUpper 95% confidence intervalP value
  1. Corrected P-value less than 0.05 is in bold.

rs7968585465183601.99400.69053.29700.0030
rs75734346525942−2.3860−4.2990−0.47280.0152
rs154441046526102−1.0920−2.50300.31920.1305
rs2239182465416781.3080−0.02282.63900.0551
rs210730146541837−0.0969−1.58301.39000.8984
rs223918146542216−1.7170−3.70600.27080.0916
rs1540339465435930.5490−0.86701.96500.4480
rs223917946544033−1.7180−3.1140−0.32210.0165
rs886441465492311.3690−0.32183.05900.1137
rs218948046550095−1.1150−2.48900.25970.1131
rs3819545465512730.0683−1.39901.53500.9273
rs2239186465556770.5753−1.21502.36600.5294
rs2254210465599810.1515−1.18701.49000.8245
rs223813646563980−0.3045−1.81701.20800.6936
rs285356446564754−0.0372−1.39201.31800.9571
rs4760648465669320.0858−1.23001.40100.8984
rs4334089465722820.5517−0.90622.01000.4589
rs389073346575640−0.2609−1.64701.12600.7126

Discussion

We conducted a comprehensive genetic analysis of VDR in PD. Previous studies of VDR in general have focused mainly on four restriction fragment length polymorphisms (RFLPs): TaqI (rs731236), ApaI (rs7975232), BsmI (rs1544410), and FokI (rs10735810) (Niino et al., 2000; Partridge et al., 2004; Uitterlinden et al., 2004; Tajouri et al., 2005; Mamutse et al., 2008). Herein, we systematically chose tagSNPs that cover all common variants in and around VDR, including the four previously described RFLPs. We examined these SNPs in a large dataset of Udall PD families and validated the association in the NINDS GWAS case-control cohort. Since a genetic component for the age-dependent penetrance of PD has been demonstrated (Destefano et al., 2002; Li et al., 2002; Maher et al., 2002), we also analyzed SNP association with AAO in addition to risk. Evidence for association between PD AAO and VDR SNPs were found in both family dataset and case-control cohort.

The significant SNPs in the NINDS dataset are not in LD with the significant SNPs in the Udall family dataset, suggesting allelic heterogeneity, which is not unexpected for complex diseases (Horne et al., 2007; Schrauwen et al., 2009). The associations at different markers could be related to the same one or more underlying causative variants yet to be identified, although the significant markers themselves are not in strong LD with each other. Alternatively, these findings could suggest that each of the significant markers or polymorphisms in LD with them has a distinct effect on the VDR gene expression/function (as discussed below), but these functional changes lead to the same phenotype.

The most significant SNP in our family study, rs4334089, is also associated with increased AD risk in a recent GWAS of AD (Beecham et al., 2009). The most significant SNP in the validation NINDS dataset, rs7968585, is in high LD with the ApaI polymorphism (r2= 0.92), which has been associated with AD risk in a Turkish cohort (Gezen-Ak et al., 2007) and MS risk in multiple studies (Smolders et al., 2009). The mechanisms underlying the association between VDR SNPs and multiple neurodegenerative diseases are not clear. Given that vitamin D is a well-documented anti-inflammatory molecule (Garcion et al., 2002), one explanation could be the inflammatory pathways that have been reported to be important in all three disorders. Another possibility is related to VDR's role as a transcription factor. In that role, VDR regulates the transcription of many genes that are implicated in different diseases. The AD and MS studies corroborate the association we found in PD and suggest that the significant SNPs in VDR have functional consequences contributing to interindividual variations of disease susceptibility.

In our family dataset, all the significant SNPs located in introns are between the alternatively spliced first exons, which generate alternative transcripts with different 5′ untranslated region (UTR) (Fig. 1). Although the functions of alternative transcripts of VDR in brain have not been extensively studied, a potential mechanism underlying the intronic SNP association is the SNP-mediated allelic-specific alternative splicing. Indeed, we have observed this phenomenon in a previous study on coronary artery disease. In the limbic system associated membrane protein (LSAMP) gene, we have found that a risk haplotype located with the intron between alternative exon 1A and 1B was correlated with preferred usage of exon 1A, whose expression was associated with higher atherosclerosis burden (Wang et al., 2008a). In the NINDS dataset, the significant SNP is in strong LD with the 3′UTR SNP rs739837. 3′UTR polymorphism, when located within miRNA target site, can modulate miRNA binding efficiency and affects translation/transcription of the gene. We have shown this is the case for the FGF20 gene (Wang et al., 2008b). We did not find any predicted miRNA binding site around rs739837 but we cannot exclude the existence of such a site unless the entire 3′UTR of VDR is carefully characterized by functional tests.

Although we did not replicate the BsmI polymorphism previously reported in a Korean cohort (Kim et al., 2005) and no evidence for association was reported for PD AAO near VDR in the Progeni/GenePD GWAS study (Latourelle et al., 2009), the strength of our study lies in providing evidence for association in the VDR gene in two independent datasets. It has been shown that any replications considerably increase the positive predictive value of a research finding despite negative results in replication efforts (Moonesinghe et al., 2007). An important factor that could influence the effect size of VDR polymorphism and therefore the association result is the vitamin D status. It has been found that vitamin D status modulated the association between VDR polymorphisms and type I diabetes mellitus incidence. The odds ratio of the risk allele increased with higher vitamin D level in the studied population (Ponsonby et al., 2008). It has been speculated that a poor vitamin D status would outpace the effect of VDR polymorphism and that the VDR polymorphisms only manifest phenotypic variations in the presence of certain vitamin D level. This notion seems to be supported by the genetic studies in MS: positive association between VDR and MS was found in the studies conducted in regions with lower latitude but not in regions with higher latitude (Smolders et al., 2009). Therefore, incorporating vitamin D level in any future study is crucial in order to fully characterize the genetic effects of VDR polymorphisms in PD.

In the family dataset, we found evidence for association with AAO. However, the association was not detected in our case-control cohort constructed using probands and unrelated controls from the family dataset (Edwards et al., 2009). This might be attributed to the different study design or properties of the test statistics: family-based associations versus case-control association. It is worth pointing out that many of our initial reports of association with AAO in our family dataset have been replicated by other groups as risk associations, such as the ELAVL4, GSTO1, and USP24 genes (Li et al., 2003; MA et al., 2005; Oliveira et al., 2005; Li et al., 2006; Wahner et al., 2007; Destefano et al., 2008).

Many studies have demonstrated a beneficial role of vitamin D in PD in animal and cell culture studies (Wang et al., 2001, Smith et al., 2006). There are several plausible mechanisms for the VDR-mediated vitamin D effects on PD. In addition to its anti-inflammatory property, which plays a role in neurodegenerative diseases in general (Deluca & Cantorna, 2001; Garcion et al., 2002), vitamin D regulates the expression of glial cell line-derived neurotrophic factor (GDNF) (Evans 1988; Naveilhan et al., 1996; Ferrari et al., 1998; Garcion et al., 2002; Smith et al., 2006). Administration of GDNF has been shown to alleviate PD symptoms in a nonhuman primate model of PD and in PD patients (Gash et al., 1996; Kordower et al., 2000; Gill et al., 2003). Lastly, vitamin D is well known for its role in Ca2+ homeostasis. Chan et al. (2009) have suggested that the SN dopaminergic neuron is particularly vulnerable to cell death due to its sustained opening of L-type Ca2+ channels for its pacemaker activity. Maintaining proper intracellular Ca2+ concentration in the face of extended Ca2+ influx puts SN dopaminergic neurons under substantial cellular stress. It is intuitive that any genetic or environment changes tampering with Ca2+ homeostasis will accelerate SN dopaminergic neuron loss.

Conclusion

Our data suggest that common genetic variations in VDR modulate AAO of PD. The evidence of VDR as a genetic risk factor for PD supports the potential importance of vitamin D in PD. Given that vitamin D status is an environmentally modifiable factor, vitamin D presents a potential preventive/therapeutic strategy for this debilitating neurological disorder. Our findings warrant the need for future studies of VDR as well as its interaction with vitamin D levels in PD.

Acknowledgements

We are grateful to the families and staff who participated in this study. We thank the members of the PD Genetics Collaboration who contributed families to the study. Some of the samples used in this study were collected while the Udall PDRCE was based at Duke University. This work was supported by National Institutes of Health grant NS39764.

Ancillary