Journal of Molecular Neuroscience

, Volume 53, Issue 2, pp 183–188

Sequence Variants in SLC6A3, DRD2, and BDNF Genes and Time to Levodopa-Induced Dyskinesias in Parkinson’s Disease


  • Natalie Kaplan
    • The Parkinson Disease and Movement Disorders Clinic, Department of Neurology and Sagol Neuroscience CenterChaim Sheba Medical Center
    • The Sackler Faculty of MedicineTel Aviv University
  • Aya Vituri
    • The School of Mathematical SciencesTel Aviv University
  • Amos D. Korczyn
    • The Sieratzki Chair of NeurologyTel Aviv University
  • Oren S. Cohen
    • The Parkinson Disease and Movement Disorders Clinic, Department of Neurology and Sagol Neuroscience CenterChaim Sheba Medical Center
    • The Sackler Faculty of MedicineTel Aviv University
  • Rivka Inzelberg
    • The Parkinson Disease and Movement Disorders Clinic, Department of Neurology and Sagol Neuroscience CenterChaim Sheba Medical Center
    • The Sackler Faculty of MedicineTel Aviv University
  • Gilad Yahalom
    • The Parkinson Disease and Movement Disorders Clinic, Department of Neurology and Sagol Neuroscience CenterChaim Sheba Medical Center
  • Evgenia Kozlova
    • The Parkinson Disease and Movement Disorders Clinic, Department of Neurology and Sagol Neuroscience CenterChaim Sheba Medical Center
  • Roni Milgrom
    • The Susanne-Levy Gertner Oncogenetics Unit, The Institute of Human GeneticsChaim Sheba Medical Center
  • Yael Laitman
    • The Susanne-Levy Gertner Oncogenetics Unit, The Institute of Human GeneticsChaim Sheba Medical Center
  • Eitan Friedman
    • The Susanne-Levy Gertner Oncogenetics Unit, The Institute of Human GeneticsChaim Sheba Medical Center
    • The Sackler Faculty of MedicineTel Aviv University
  • Saharon Rosset
    • The School of Mathematical SciencesTel Aviv University
    • The Parkinson Disease and Movement Disorders Clinic, Department of Neurology and Sagol Neuroscience CenterChaim Sheba Medical Center
    • The Sackler Faculty of MedicineTel Aviv University

DOI: 10.1007/s12031-014-0276-9

Cite this article as:
Kaplan, N., Vituri, A., Korczyn, A.D. et al. J Mol Neurosci (2014) 53: 183. doi:10.1007/s12031-014-0276-9


Levodopa-induced dyskinesias (LID) present a common but elusive complication of levodopa therapy in Parkinson’s disease (PD). In order to identify genetic factors associated with LID, 352 (213 males) levodopa-treated Israeli PD patients were genotyped for 34 polymorphisms within three candidate genes affecting dopaminergic activity and synaptic plasticity: dopamine transporter gene (DAT1 or SLC6A3) [14 single nucleotide polymorphisms (SNPs) and 40-bp variable number tandem repeat (VNTR)], DRD2 [11 SNPs and dinucleotide CA short tandem repeat (STR)], and BDNF (7 SNPs). A comparison of patients with and without LID was performed by applying a time-oriented approach, with survival analyses evaluating LID development hazard rate over time [Cox proportional hazards and accelerated failure time (AFT) lognormal models]. Overall, 192 (54.5 %) participants developed LID, with a mean latency of 5.0 (±4.5) years. After adjusting for gender, age at PD onset, duration of symptoms prior to levodopa exposure, and multiple testing correction, one SNP in SLC6A3 (with 81 % genotyping success) was significantly associated with LID latency: the C allele of the rs393795 extended the time to LID onset, time ratio = 4.96 (95 % CI, 2.3–10.9; p = 4.1 × 10−5). This finding should be validated in larger, ethnically diverse PD populations, and the biological mechanism should be explored.


Levodopa-induced dyskinesias (LID)Parkinson’s diseaseSingle nucleotide polymorphism (SNP)Dopamine transporter gene (SLC6A3)DRD2BDNF


For almost half a century, treatment with levodopa remains the “gold standard” for alleviating the motor symptoms of Parkinson’s disease (PD) (Khor and Hsu 2007). However, inevitable and, at times, severe side effects, combined with the temporary and gradually diminishing beneficial effect, are major limiting factors in the successful long-term levodopa treatment of PD patients (Muenter and Tyce 1971). The most clinically significant long-term side effects are the motor complications: response fluctuations and levodopa-induced dyskinesias (LID) (Nutt 2001).

LID are hallmarked by involuntary choreiform movements noted in 40–60 % of patients following 5 years of levodopa treatment (Rascol et al. 2000; Nutt 2001). While the precise pathophysiology of LID still remains to be established, several mechanisms have been proposed, focusing on both pre- and postsynaptic processes: denervation supersensitivity of postsynaptic dopamine receptors, altered accumulation and release ability of dopamine (DA) by presynaptic terminals, and aberrant plasticity in response to nonphysiological pulsatile stimulation (Bezard et al. 2001; Morgante et al. 2006).

The reported variability in LID prevalence may be attributed in part to methodological issues (e.g., study population, diagnostic criteria, and duration of follow-up), but may also be taken as an indirect evidence for LID being a multifactorial phenomenon, with contributions from both environmental and genetic factors. The risk for developing LID has been found to increase with the extent of nigrostriatal denervation (i.e., disease severity), younger age at disease onset, negative smoking history, female gender, low body weight, higher cumulative levodopa dosage, faster rate of dose escalation, pattern of drug administration, and duration of levodopa therapy (Jankovic 2005; Hauser et al. 2006).

Several candidate genes have been proposed as putatively contributing to the propensity of developing LID: D2 dopamine receptor (DRD2) gene (Oliveri et al. 1999; Kaiser et al. 2003; Strong et al. 2006), solute carrier family 6 member 3 (SLC6A3) dopamine transporter (DAT1) gene (Fuke et al. 2001; Kaiser et al. 2003), brain-derived neurotrophic factor (BDNF) gene (Egan et al. 2003; Foltynie et al. 2009), catechol-O-methyltransferase (COMT) gene (Bialecka et al. 2004), D2-like dopamine 3 receptor (DRD3) gene (Lee et al. 2011), monoamine oxidase B (MAO-B) gene (Bialecka et al. 2004), and opioid μ receptor (OPRM1) gene (Strong et al. 2006). However, most of the previous reported studies evaluated genetic associations with LID development per se, without considering the latency period to LID as an end point. The latter has been our subject of interest in the following analysis.

We previously evaluated the effects of APOE polymorphism (no association found) (Molchadski et al. 2011) and gender (increased hazard in female patients) (Hassin-Baer et al. 2011) as well as the effect of the LRKK2 G2019S mutation presence/absence (no association found) (Yahalom et al. 2012) on time to LID development in Israeli PD patients, and this approach seems to represent a clinically meaningful analysis scheme pertaining to the LID phenotype. The main objective of the present study was to evaluate the contribution of genetic polymorphisms to LID in Israeli PD patients, by analyzing the impact of specific alleles from three genes (DRD2, SLC6A3, and BDNF), which have shown more convincing evidence for association in previous studies, on the latency period to LID.



The study is a historical prospective study. Israeli PD patients followed up at the Movement Disorders Clinic at the Chaim Sheba Medical Center were consecutively recruited between January 2005 and September 2010. Inclusion criteria were (1) clinical diagnosis of PD according to the UK Parkinson’s Disease Society Brain Bank criteria (Hughes et al. 1992) (however, patients with a positive family history and symmetrical onset were also included) (Korczyn 2011), (2) treatment with levodopa, (3) reliable data concerning time of levodopa treatment initiation and time of dyskinesia presentation (retrieved from patients’ medical charts or patients’/caregivers’ reports), and (4) no history of stroke or of significant head trauma. The study was approved by both Sheba Medical Center and the Israeli Ministry of Health ethics committees, and all participants signed an informed consent.

All participants were examined by a movement disorder specialist, and their medical charts were reviewed for demographic and clinical data, including gender, age at disease onset, year of levodopa therapy initiation, and time of LID initial appearance (if it occurred). Noteworthy, LID were not differentiated by type or severity. All participants were unrelated to each other.

Genomic DNA was isolated from the patients’ blood samples, using Promega’s (Madison, WI) Wizard® Genomic DNA purification kit, following the manufacturer’s recommended protocol.

SLC6A3, BDNF, and DRD2 Genes: Single Nucleotide Polymorphism genotyping

Single nucleotide polymorphism (SNP) information within and around the three candidate genes (DRD2, BDNF, and SLC6A3) was obtained from dbSNP ( for intronic and functional SNPs and from ( (freeze October 9), for tagging SNPs. SNP genotyping was performed using the iPLEXTM platform of Sequenom MassARRAY® System technology, allowing multiplexing of assays up to the 29-plex level (Oeth et al. 2005). A total of 37 putative SNPs flanked by ∼60 bp on each side (121 bp total) were submitted to Sequenom’s primer design software (MassARRAY Assay Design 3.0). These were further grouped into two multiplex assays containing 24 and 13 SNPs each. Designed oligonucleotide sequences were obtained from IDT Syntezza (Rehovot, Israel). Experiments were conducted following the Sequenom iPLEX Assay application note (Oeth et al. 2005). Genotyping data were acquired using the Sequenom MassARRAY and processed using Sequenom Typer 3.4 software.

SLC6A3 Gene: 40-bp Variable Number Tandem Repeat Genotyping

Analysis of the 40-bp variable number tandem repeat (VNTR) polymorphism in the 3′ untranslated region of the SLC6A3 gene was performed by PCR of the designated genomic region, using the following set of primers: DAT1_F: 5′-TCCTTGAAACCAGCTCAGGCTACT-3′ and DAT1_R: 5′-TGTGGCACGCACCTGAGAGAAATA-3′. A touchdown PCR protocol (Korbie and Mattick 2008) was used, with an initial denaturation step at 94 °C for 5 min, followed by 35 cycles of denaturation at 94 °C for 30 s, annealing at 66–56 °C for 2 min and extension at 72 °C for 1.5 min, and a final 10-min extension at 72 °C. PCR products (75 %) were run against pBR322 DNA/Alu I, Marker 20 on 3 % agarose gel stained with ethidium bromide and visualized under ultraviolet light. The differentiation by allele size was further confirmed using sequence analysis.

DRD2 Gene: (CA)n Dinucleotide Short Tandem Repeat Genotyping

Analysis of intron 2 dinucleotide short tandem repeat polymorphism in the DRD2 gene was performed by PCR amplification using the following set of primers: DRD2_F: 5′-TGAGTGTGTGCCTGTGTGTGAATG-3′ and DRD2_R: 5′-ACCCAGAGAACGAAAGAAGGCAGA-3′. PCR products were then 3′-end fluorescent labeled by TAMRA dideoxynucleotides (Applera Corporation, Foster City, USA) and analyzed using the ABI Prism 3100 DNA Sequencer (Applied Biosystems, Foster City, USA). Allele size was established using ABI Prism GeneScan and Genotyper Analysis software. Sequence analysis was used to confirm differentiation of alleles by size.

Statistical Analysis

We performed survival analysis to examine the effect of the 32 genotyped SNPs and two types of tandem repeats (dinucleotide short tandem repeat (STR) and 40-bp VNTR) on the hazard rate of dyskinesia. SNP effects were examined in three different modes: additive, dominant, and recessive, whereas for the repeat data, we tested the association with an average number of repeats. The tests performed were controlled for gender, age at disease onset, and latency to levodopa treatment initiation.

The survival analysis was performed using two models: Cox proportional hazards and accelerated failure time (AFT) (Wei 1992) analysis using the lognormal model. The tests were performed on all the observations and then on the Ashkenazi patients alone (one or both parents of Ashkenazi origin), in order to screen for effects in a more genetically homogenous group. Three hundred ninety-two tests performed set a significance threshold p value = 1.28 × 10−4 as determined by the Bonferroni correction.

Following the analyses of the effect of each individual SNP, we tested all pairs of SNPs/mean number of tandem repeats together with additional covariates, in order to examine a possible inter-gene interaction (e.g., synergistic effect of combined genotypes), using a configured Cox model incorporating coupled polymorphisms and an interaction (significance threshold p value = 8.4 × 10−5).

The statistical analysis was performed using an up-to-date version of the “R data analysis software” package (


Clinical Characteristics

A total of 352 Israeli levodopa-treated PD patients participated in the study: 224 of Ashkenazi origin and 116 non-Ashkenazi Jews, 11 Israeli Arabs, and 1 non-Jewish Russian/Ukrainian. Of these, 192 (54.5 %) patients developed LID in the course of levodopa treatment. Demographic and relevant clinical features of all participants, with and without LID, are presented in Table 1. Patients who developed LID were significantly younger at PD symptom onset, 55.2 ± 13.5 vs. 64.6 ± 11.4 years (p < 0.001), and were treated with levodopa for a longer time period, 5.0 ± 4.5 vs. 3.7 ± 3.5 years (p < 0.05).
Table 1

Demographic and clinical features of study participants


Patients with LID

Patients without LID

p value

n = 192 (54.5 %)

n = 160 (45.5 %)

Male (%)

116 (60.4)

97 (60.6)


Ashkenazi (%)

124 (64.6)

100 (62.5)


Age at PD symptom onset (mean ± SD), years

55.2 ± 13.5

64.6 ± 11.4


Time from PD symptom onset to levodopa treatment initiation (mean ± SD), years

2.7 ± 3.7

2.5 ± 2.7


Duration of levodopa treatment (mean ± SD), yearsa

5 ± 4.5

3.7 ± 3.5


Duration of PD (mean ± SD), yearsa

7.7 ± 5.1

6.2 ± 3.9


p values from t-tests for continuous variables and Fisher’s exact test for dichotomous features comparing the two groups (patients with and without dyskinesia) are presented. Statistically significant difference is p < 0.05

PD Parkinson’s disease, SD standard deviation, LID levodopa-induced dyskinesias

aFor subjects with dyskinesia, the number of years until onset of LID was used for calculating disease duration and treatment duration. For subjects without dyskinesia, the number of years till last observation was used for this calculation

Genotyping Data

Thirty-four of the 39 initially selected polymorphisms were included in the statistical analysis: 11 SNPs (rs4245147, rs6275, rs6276, rs6277, rs4630328, rs17529477, rs1079597, rs4938017, rs4245148, rs1079594, rs1800497) and CA STR in the DRD2 gene, 7 SNPs (rs6265, rs11030104, rs2049045, rs7103411, rs7127507, rs2030324, rs988748) in the BDNF gene and 14 SNPs (rs393795, rs3756450, rs2550956, rs2937639, rs6347, rs2617605, rs3863145, rs250686, rs464049, rs4975646, rs1048953, rs11133767, rs27048, rs40184), and 40-bp VNTR in the SLC6A3 gene. Genotyping technically failed for four SNPs (rs1967554, rs2075654, rs2652510, rs7124442), and one SNP (rs11564752) was excluded as it was not polymorphic in our study population. Of the remaining 32 genotyped SNPs, four did not display Hardy–Weinberg equilibrium (HWE) (rs2550956, rs2937639, rs4630328, rs11030104). However, these SNPs were included in the subsequent analysis, as the allele distributions of each were similar to those described for these SNPs in European samples in HAP-MAP CEU (

Genetic LID Association Analysis

After adjusting for time from onset of symptoms to levodopa treatment initiation, gender, and age at onset of PD symptoms, and upon correction for multiple testing, among the 34 polymorphisms examined, only one SNP in the SLC6A3 gene remained significant, showing an association with latency to LID: rs393795, p = 4.1 × 10−5 (n = 285, 81 % successfully genotyped).

In the AFT log normal model analysis, rs393795 C allele presence was found to be significantly associated with longer latency period to LID, manifested in extension of the expected time to LID in comparison with the AA genotype, with a LID-free survival time ratio of 4.96 (95 % CI, 2.3–10.9). The likelihood of not developing LID over time for the possible rs393795 variants (AA genotype vs. CA/CC genotypes) is shown in Fig. 1a, b. Figure 1a depicts the LID-free survival curves for the two groups of genotypes, in an unadjusted model with the genotype being the sole effector analyzed. Figure 1b, on the other hand, exhibits the pattern of behavior for the same survival curves function in the combined/adjusted model, with time from onset of symptoms to levodopa treatment initiation, gender, and age at onset of PD symptoms as covariates. The illustration provides the model’s general behavior, exemplified by the group of patients who had initiated levodopa treatment within a year of onset of PD motor symptoms (31 % of the whole cohort).
Fig. 1

Survival curves for possible rs393795 genotypes. a Unadjusted model: LID-free over time survival curves of the unadjusted model for the AA genotype variant (n = 24) vs. the CA/CC genotype variant (n = 261) of the rs393795 SNP [time ratio = 4.74 (95 % CI, 2.1–10.6; p = 1.61 × 10−4)]; the C allele carriers’ rate of LID development is slower: at 5 years of treatment ∼78 % of the AA genotyped patients had developed dyskinesias compared to ∼45 % of the C allele carriers. b Adjusted model: LID-free over time survival curves of the combined/adjusted model for the AA genotype variant (n = 10) vs. the CA/CC genotypes variant (n = 78) of the rs393795 SNP [exhibiting the model’s general behavior using the set of data pertaining to the group of patients that had initiated levodopa treatment within a year of onset of PD motor symptoms]. Time ratio = 4.96 (95 % CI, 2.3–10.9; p = 4.1 × 10−5); The C allele carriers are characterized by a slower rate of LID development: at 5 years of treatment, ∼68 % of the AA genotyped patients had developed dyskinesias compared to ∼35 % of the C allele carriers

There was no statistical difference in time from onset of PD symptoms to levodopa treatment initiation (2.5 ± 3.4 vs. 2.7 ± 3.4 years), gender (75 vs. 61.7 % males), and age at onset of PD symptoms (58 ± 16.2 vs. 59 ± 13.2 years) between the AA genotype and the CC/CA genotype groups of the rs393795 SNP (p = 0.78, p = 0.2, and p = 0.72, respectively).

Inter-polymorphism Interaction Analysis

An analysis of combined pairs of polymorphisms, adjusted for time to levodopa exposure, age at symptom onset, and gender, in an attempt to detect possible interactions, yielded no statistically significant results (data not shown).


In the present study conducted on Israeli PD patients, a statistically significant association between an intronic SNP (rs393795) in the SLC6A3 gene, encoding for the dopamine transporter (DAT), and time to LID was evident. Specifically, the C allele of the rs393795 SNP was associated with a later occurrence of LID (time ratio = 4.96). Interestingly, it is the AA genotype of the same SNP that was previously reported to be protective against delirium in the elderly (van Munster et al. 2010). As DAT is a vital determinant of DA function, mainly via spatial and temporal buffering of released DA (Uhl 2003), and LID has been frequently correlated with the pattern of dopaminergic stimulation (Bezard et al. 2001; Morgante et al. 2006), one plausible explanation for this association with LID, is an altered rate of DA reuptake from the synapse: a phenotype exhibiting a more balanced DA buffering and sustained reuptake pattern, with less prominent fluctuations in DA concentration. However, without any functional analysis, this proposed mechanism remains theoretical and speculative at best.

An alternative splicing of a novel cassette exon within intron 3 of the SLC6A3 gene, which was confirmed in postmortem human substantia nigra (Talkowski et al. 2010), possibly implies that certain genetic variants lead to reduced translation and thus decreased DAT protein availability.

Lower DAT expression, although potentially serving as a compensatory mechanism in early PD, was associated with higher oscillations in synaptic DA concentration. Clinical observations have shown that greater DAT levels are directly associated with lower DA turnover and lower changes in synaptic DA concentration in PD patients (de la Fuente-Fernández et al. 2004; Sossi et al. 2007).

The rs393795 SNP, which is located on intron 4, falls within the same linkage disequilibrium cluster with two of the SNPs (rs458609, rs457702) that affect alternative splicing (Talkowski et al. 2010) and with rs460000 SNP, located within the exon 4–intron 3 boundary, also possibly affecting RNA splicing (Cartegni et al. 2002). The latter SNP has recently been associated with behavioral and neural measures of response inhibition, affecting frontostriatal response inhibition circuits (Cummins et al. 2012). Further analysis is required to substantiate any speculation regarding the functional role of these SNPs at the expression level of the transporter, its availability or activity and consequent impact on the LID latency period.

Hereditary DAT deficiency syndrome, an infantile onset neurogenetic disorder characterized by elevated DA metabolites in the cerebrospinal fluid, is the first identified genetic movement disorder caused by mutations in the SLC6A3 gene (Kurian et al. 2011). The mutations described to date result in the loss of DAT function expressed by a combination of hypokinetic and hyperkinetic symptoms—dystonia and parkinsonism—manifesting in infancy (Kurian et al. 2011). Some of these patients were described to develop dyskinesia after levodopa exposure, suggesting that genetically determined malfunction of DAT activity may serve as a risk factor for LID.

Previously reported associations of LID with sequence variants that were genotyped in the present study were not validated or replicated (Fuke et al. 2001; Kaiser et al. 2003; Strong et al. 2006; Foltynie et al. 2009). For some polymorphisms (i.e., DAT1 40-bp VNTR, DRD2 (CA)n dinucleotide STR, and DRD2 rs1800497 SNP), this lack of validation could in part be explained by the application of a different design (survival plot of time to LID, rather than prevalence or a grouping using an arbitrary definition of early vs. late LID onset). Diverse statistical analyses (e.g., statistical models, confounding factors adjusted for, etc.), study population, and number of patients with LID (188 patients successfully genotyped in our study vs. 47 patients in the study of Foltynie et al. 2009) could have additionally contributed to the lack of validation of the reported association of BDNF rs6265 SNP and the time to LID appearance (Foltynie et al. 2009). It is also possible that Israeli patients, namely Jewish and Arab subjects examined in the present study, represent ethnical populations different from those that were examined in previous studies reporting the above mentioned polymorphisms.

Our study has several limitations that need to be pointed out: Although in most cases, LID were observed in the context of routine follow-up of patients in the clinic and thus noted when initially occurred, for some patients, the data were collected retrospectively, as reported by the patient/caregiver at the clinic visits. Data regarding several important factors known to impact LID, e.g., the daily dose of levodopa or its equivalents, the cumulative dosage of levodopa, and valid assessment of disease severity at the time of LID onset, could not be obtained in most cases and were therefore not controlled for. We also did not have an accurate description of the type of dyskinesias in all patients and thus grouped all LID together, as done in previous studies (Oliveri et al. 1999; Kaiser et al. 2003). Finally, as this was a retrospective collection of data, we have no information for those who have died or otherwise not followed up.

In conclusion, we report a significant association noted between a single SNP in SLC6A3, the DAT-encoding gene, and LID latency in Israeli PD patients. Though practically assessed only for 81 % of the patients enrolled, if confirmed in additional, larger, and ethnically diverse populations, this finding may possibly shed light on the complex mechanism of LID and could potentially pave the road to a pharmacogenetic “tailor-made” intervention plan for PD patients providing a refined profile of adverse effects related to levodopa treatment.


Special appreciation is due to Ms. Julia Feiler and Prof. Gideon Rechavi’s research laboratories, at Sheba Cancer Research Center laboratories, for the use of their facilities and equipment and assistance with operating the Sequenom MassARRAY® System technology. This study was supported by the Bharier Medical Fund, in memory of Nat and Sophie Bharier, and by the Mariana and George Saia Foundation (Tel Aviv University).

Conflict of interest

The authors have no conflicts of interest to declare.

Copyright information

© Springer Science+Business Media New York 2014