Association analysis of PARP1 polymorphisms with Parkinson's disease.

Alpha-synuclein accumulation in intracellular inclusions, oxidative stress and microglia-mediated inflammation in the substantia nigra are crucial events in the pathogenesis of Parkinson's disease (PD). Poly (ADP-ribose) polymerase-1 (PARP1), a DNA-binding enzyme and transcriptional regulator, plays an important role in modulating the cellular response to oxidative stress, inflammatory stimuli, and in apoptotic cell death. Inhibition of PARP1 results in significant neuroprotection in PD animal models; moreover PARP1 has a physiological role in the regulation of alpha-synuclein expression. A previous study had demonstrated that variants located within the PARP1 gene promoter reduce the risk of PD and delay the disease age at onset. In light of these data, we carried out an association study to investigate whether variability within this gene is associated with PD risk and disease age at onset in an Italian cohort composed of 600 PD patients and 592 healthy controls. To this purpose, we used a comprehensive tag SNP approach spanning the entire gene and the upstream and downstream regions. We did not detect any significant association of the PARP1 gene with PD either at genotypic or haplotypic level; none of the 11 genotyped SNPs was significantly associated with PD age at onset. We conclude that, despite previous evidence, PARP1 is not a susceptibility gene for PD in our population.

Introduction

Parkinson’s disease (PD) is a debilitating neurological disorder characterized clinically by bradykinesia, resting tremor, rigidity, and postural instability with a therapeutic response to levodopa. The disease is pathologically characterized by the loss of dopaminergic neurons in the substantia nigra pars compacta (SNpc) and the presence of intracellular inclusions known as Lewy bodies, which are composed primarily of alpha-synuclein protein aggregates. There is accumulating evidences that hereditary factors play a significant role in the onset of PD [1]. In addition to the rare causal mutations, which have provided us with many molecular clues to elucidate the etiology of PD, also gene variants may promote pathophysiological mechanisms, thereby affecting the risk to develop sporadic PD. For instance, variability in the length of a complex repeat sequence within the SNCA gene promoter (NACP-Rep1) has been shown to affect risk of PD across populations worldwide [2], presumably via mechanisms of modulation of alpha-synuclein expression.

Poly (ADP-ribose) polymerase-1 (PARP1) is widely known as part of the cellular defense against oxidative stress and repair program due to its DNA-binding and ADP-ribosylation properties. Poly(ADP-ribose) formation can influence also the action of transcription factors, notably nuclear factor kappa B, which regulates the expression of several inflammatory mediators such as cytokines, chemokines, and adhesion molecules. However, excessive PARP1 activation can promote neuronal death through mechanisms involving NAD+ and ATP depletion and release of apoptosis inducing factor (AIF) from the mitochondria to the nucleus. The involvement of PARP1 in PD pathogenesis is supported by the presence of abundant PARP-containing nuclei in the dopaminergic neurons of the substantia nigra in PD [3]. Furthermore, PARP1 null mice are resistant to the loss of dopaminergic neurons in the SNpc following MPTP-induced neurotoxicity [4]. Finally, in vitro studies have shown that PARP1 participates in the regulation of alpha-synuclein expression by binding to the Rep1 polymorphic site upstream of the SNCA gene [5].

In light of these data, we carried out an association study to evaluate the role of the PARP1 gene as a susceptibility factor for PD. To this purpose, we performed an association study of 11 single nucleotide polymorphisms (SNPs) spanning the entire PARP1 gene at the single locus and haplotype level in a series of Italian PD patients and unrelated controls.

Subjects and methods

Subjects

The study included 600 PD patients consecutively recruited from the outpatient clinic for movement disorders at the Department of Neurology of San Gerardo Hospital in Monza and at the Parkinson Institute in Milan. PD patients were diagnosed according to standard criteria [6]. Among patients, 45 (7.5%) reported a family history of PD in at least a first-degree or second-degree relative. Control population consisted of 592 unrelated individuals, comprising healthy spouses and blood donors. None of the controls had clinical evidence of neurological disease (as assessed by neurological examination) or familial history of neurodegenerative disease. Both patients and controls were Caucasians and the controls were geographically matched to the cases. Informed consent for participation was obtained from all subjects. Full ethical committee approval for this investigation was obtained.

Markers selection and genotyping

We used the LD-select algorithm, as implemented on the Seattle SNPs Genome Variation Server (http://gvs.gs.washington.edu/GVS/) to identify tag SNPs at the PARP1 locus based on data from the International HapMap Project Centre d’Etude du Polymorphism Humain collection (CEU) population (http://www.hapmap.org). The linkage disequilibrium (LD) r2 threshold for bins was 0.80, the minor allele frequency cut-off >5%, and the region covered included 10 kb of upstream sequence (5′end) and 2 kb of downstream sequence (3’end). Eight tag SNPs were selected (rs8679, rs3219142, rs1136410, rs3219123, rs3219110, rs1805410, rs1805414, rs3219053) which captured a minimum of 85% of the genetic variation across the gene. Moreover, we genotyped one SNP flanking the 3′UTR (rs12567614), one intra-genic SNP (rs1805405) shown to be haplotype tagging in a previous study [7] and one SNP in the 5′UTR region (rs907187). Table 1 highlights the genomic positions and type of the genotyped markers.

Table 1

Overview of location and type of the genotyped SNPs.

SNP id	Genome position (bp)a	Allele 1	Allele 2	Type of variant
rs12567614	224611043	C	T	Downstream
rs8679	224615177	T	C	3′UTR SNP
rs3219142	224618691	C	T	Intronic SNP
rs1136410	224621925	T	C	Coding-nonsynonymous
rs3219123	224621971	C	T	Intronic SNP
rs3219110	224624501	A	G	Intronic SNP
rs1805410	224635288	A	G	Intronic SNP
rs1805414	224639987	T	C	Coding-synonymous
rs3219053	224641177	G	A	Intronic SNP
rs1805405	224646644	C	A	Intronic SNP
rs907187	224662270	G	C	5′UTR SNP

Allele 1 = major frequency allele; allele 2 = minor frequency allele.

Base position according to NCBI genome build 36.

Genomic DNA was extracted from peripheral venous blood using a Qiagen Blood kit (Qiagen, Milan, Italy). SNPs were typed using the GoldenGate Assay on VeraCode microbeads (Illumina San Diego, CA, USA). The genotyping call rate for the studied SNPs was in the range of 97–100%. Goodness of fit of the genotype frequencies to Hardy–Weinberg expected proportions in control subjects was examined.

Data analysis

Chi-square tests were used to compare gender distribution, and genotype frequencies between patients and controls and to test for deviation from Hardy-Weinberg equilibrium. Student’s t-test was used to assess differences in age between the two groups. We used logistic regression analysis to test for association between the PARP1 gene selected variants and PD risk. The estimated odds ratios (ORs) and relative 95% confidence intervals (95% CI) were adjusted for gender and age at enrollment. We performed analyses assuming autosomal dominant and autosomal recessive models of inheritance. We also performed genetic analyses testing a log-additive model in which genotype was categorized by three levels (0, 1 and 2, representing number of variant alleles), in order to assess the impact of each additional copy of the risk allele on the odds of PD.

Analyses were performed for subjects overall and stratified by family history of PD, age at study (divided in quartiles), and gender. The association of each genetic variant with age at onset of PD was assessed using Cox proportional hazard models, adjusted for gender. For each genetic variant, we calculated a hazard ratio (HR), a 95% CI, and a two-tailed P-value. We performed similar analyses of age at onset of PD in men and women separately. All the analyses were performed using SPSS v. 16.

PARP-1 inter-SNPs LD measures (D′, r2) were calculated using Haploview software version 4.1 (http://www.broad.mit.edu/mpg/haploview/) for the set of control samples.

We also used Haploview to reconstruct haplotypes and to assess the association of the inferred haplotypes with PD. Haplotypes with inferred frequencies <0.01 were excluded from the analysis. The level of significance was set at 0.05. Power calculations were carried out using the Quanto software version 1.2.

Results

The demographic data for the 600 PD patients were as follows: mean age at study was 65.15 ± 9.11 years (range 32–89), mean age at onset of symptoms was 57.28 ± 10 years (range 27–87). There were 319 men and 281 women (male-female ratio 1.13:1). The 592 control subjects consisted of 308 males and 284 females; their mean age at examination was 64.15 ± 8.5 years (range 37–90). There was no statistically difference in the age at study and gender between patients and controls (p > 0.05).

The control genotype distributions were in agreement with the Hardy–Weinberg (HW) equilibrium (HW p-value >0.05) for all except one marker (rs3219123, HW p-value = 0.005), which was therefore excluded from further analyses. The allele frequencies of the remaining 10 polymorphisms showed close similarity to HapMap CEU data. All the considered SNPs showed a minor allele frequency (MAF) over 5%. Fig. S1 shows the LD map for the PARP1 gene generated using Haploview. Some markers throughout the gene showed evidence for linkage disequilibrium (LD); in particular, two SNPs pairs (rs3219053–rs1805405 and rs907187–rs1136410) proved to be in complete or almost complete LD (r2 = 1 and r2 = 0.95 respectively), therefore the non-tagging SNP of each couple (rs1805405 and rs907187) was removed from the subsequent association analysis.

Table 2 lists the estimated ORs and associated 95% CIs from the association analysis between each SNP and PD, assuming autosomal dominant, autosomal recessive and log additive models of inheritance. Neither SNP was significantly associated with PD overall or in strata defined by age at study, gender and family history (data not shown).

Table 2

Single markers frequencies and association analyses with risk of PD

	Genotypea frequency, n (%)			Trend modelb		Dominant model		Recessive model
	1/1	1/2	2/2	OR (95% CI)	p-value	OR (95% CI)	p-value	OR (95% CI)	p-value
rs12567614
Cases	211 (35.3)	295 (49.3)	92 (15.4)	1.02 (0.86–1.20)	0.83	1.09 (0.86–1.39)	0.46	0.91(0.67–1.25)	0.58
Controls	221 (37.5)	272 (46.1)	97 (16.4)
rs8679
Cases	322 (53.7)	234 (39.1)	43 (7.2)	0.97 (0.81–1.16)	0.75	0.95 (0.75–1.19)	0.65	1.02 (0.65–1.59)	0.92
Controls	311 (52.7)	238 (40.3)	41 (7)
rs3219142
Cases	404 (67.5)	180 (30)	15 (2.5)	1.02 (0.83–1.26)	0.86	1.09 (0.86–1.40)	0.47	0.64 (0.33–1.23)	0.17
Controls	410 (69.4)	158 (26.7)	23 (3.9)
rs1136410
Cases	456 (76)	135 (22.5)	9 (1.5)	0.87 (0.69–1.11)	0.26	0.87 (0.67–1.13)	0.29	0.77 (0.32–1.87)	0.56
Controls	433 (73.2)	147 (25)	11 (1.8)
rs3219110
Cases	213 (35.5)	275 (45.8)	112 (18.7)	0.98 (0.84–1.15)	0.81	0.96 (0.75–1.22)	0.73	1 (0.74–1.34)	0.98
Controls	202 (34.3)	278 (47.1)	110 (18.6)
rs1805410
Cases	434 (72.3)	145 (24.2)	21 (3.5)	1.01 (0.82–1.26)	0.9	1.01 (0.78–1.3)	0.96	1.09 (0.58–2.04)	0.8
Controls	426 (72)	147 (24.8)	19 (3.2)
rs1805414
Cases	275 (46)	262 (43.8)	61 (10.2)	1.07 (0.90–1.27)	0.45	1.15 (0.92–1.45)	0.22	0.93 (0.64–1.36)	0.72
Controls	292 (49.4)	236 (40)	63 (10.6)
rs3219053
Cases	395 (66)	186 (31)	19 (3)	1.20 (0.97–1.49)	0.1	1.23 (0.97–1.57)	0.09	1.24 (0.62–2.47)	0.54
Controls	416 (70.3)	161 (27.2)	15 (2.5)

The estimated odds ratios (ORs) and relative 95% confidence intervals (95% CI) were adjusted for gender and age at enrollment.

Allele 1 and allele 2 for each marker are specified in Table 1.

The OR is computed for an increase of 1 minor allele.

Furthermore, we tested the possible association between all the different combinations of alleles at the eight polymorphic loci with PD. Six common haplotypes (CCCTAATG, CTTTGATG, TTCTAACA, CTCTGGTG, TTCCAACG, TTCTGATG) and several rare ones (with frequency < 5%) were inferred, and rare haplotypes were excluded in the association analyses. None of the common haplotypes was associated with overall risk for PD (global haplotype association p-value = 0.36). Finally, in analyses restricted to PD cases, neither SNP was significantly associated with PD age at onset (data not shown).

Discussion

Our results suggest that, despite the strong biological plausibility, PARP1 is not a susceptibility locus for PD in our population. The human PARP1 gene consists of 23 exons spanning 43 kb and has been localized to chromosome 1q41–q42. To date, no other association studies using a comprehensive tag SNP (or comparable) approach have been published on this gene in PD. The recently launched website PDGene (http://www.pdgene.org/), which systematically reviews genetic association studies of PD, confirms that information regarding the association of PARP1 with PD is still lacking, with the single exception of a case–control study comprising 146 PD patients and 161 controls which showed that variants located within the PARP1 gene promoter reduce risk of PD and delay the disease age at onset [8]. However, the area covered by the tagging SNPs genotyped in the present study included a 10-kb region upstream of the 5’end of the gene. The presence of potential promoter activities for the PARP1 gene has been demonstrated up to 1790 bp from the transcription initiation site by chloramphenicol acetyltransferase assay [9].

Our negative findings are consistent with those from genome-wide association studies (GWAs). Using the NEI/NCBI dbGAP database (http://www.ncbi.nlm.nih.gov/), we spotted that some of the SNPs genotyped in our study were contained in two previous GWAs [10,11], which included three (rs1136410, rs1805410, rs1805414) and two (rs1136410, rs1805410) PARP1 variants respectively. However, both studies failed to identify any significant association between those SNPs and PD after correction for multiple testing (p-values > 0.30).

Our study has some strengths: patients and controls came from the same geographical area; diagnosis of PD was made by experienced neurologists specialized in movement disorders; genotyping errors were avoided using duplicate samples; many markers were tested to assure a true association. For the association testing, we also considered multiple genetic models, adjusted our analyses for possible confounders (gender and age at study), and stratified our sample for multiple variables (age at study, gender, familiarity) to explore possible effect modifications. We genotyped haplotype tagging SNPs, which ensures that all of the common variability within a region of LD is surveyed in a disease association study. Finally, the sample size employed and the minor allele frequency of the selected SNPs provided sufficient statistical power for the main effect analyses. Assuming the population susceptibility allele frequency to be the values observed in controls and a population prevalence of 0.02, our study had 80% power (alpha = 0.05) to detect ORs as small as 1.3 (or 0.77 or smaller) for the SNP with the highest MAF, and as small as 1.42 (or 0.67 or smaller) for the SNP with the lowest MAF under an additive model of inheritance. While the current study has sufficient power to detect ORs greater than 1.3, we cannot, however, exclude the possibility that low frequency alleles within PARP1 may exert an effect on lifetime risk for PD or disease age at onset. Moreover, these data cannot rule out the presence of a genetic association between PARP1 and PD in other populations; replication of the present study is required to clarify this issue.

Despite our data argue for a lack of association between PARP1 and PD in our population, it is still possible that this gene plays a minor role in the susceptibility to the disease. Additional functional as well as association studies investigating gene-gene interactions are required to elucidate this issue, and it remains possible that PARP1 variability may affect disease progression or the susceptibility to develop PD non-motor symptoms.