Tyrosine Hydroxylase Gene: Another Piece of the Genetic Puzzle of Parkinson’s Disease

Abstract

The tyrosine hydroxylase (TH) gene encodes a monoxygenase that catalyzes the rate limiting step in dopamine biosynthesis. A hallmark of Parkinson’s disease (PD) is the loss of dopaminergic neurons in the substantia nigra. Consistent with the essential role of TH in dopamine homeostasis, missense mutations in both alleles of TH have been associated with severe Parkinsonism-related phenotypes including infantile Parkinsonism. It has been speculated for a long time that genetic variants in the TH gene modify adult-onset PD susceptibility but the answer has not been clear. Genetic variants (both sequence variations and structural variations) can be classified into three categories based on their relative frequency in population: common variants (polymorphisms), rare variants and mutations. Each of these factors has a different mode in influencing the genetic risk and often requires different approaches to decipher their contributions to the disease. In the past few years, the revolutionary advances in genomic technology have allowed systematic evaluations of these genetic variants in PD, such as the genome-wide association study (GWAS, to survey common variants), copy number variation analysis (to detect structural variations), and massive parallel next generation sequencing (to detect rare variants and mutations). In this review, we have summarized the latest evidence on TH genetic variants in PD, including our ongoing effort of using whole exome sequencing to search for rare variants in PD patients.

1. INTRODUCTION

Parkinson’s disease (PD) is a major neurodegenerative disorder [1]. Although the molecular mechanism underlying the development of PD remains largely unknown, it is clear that PD is under strong genetic influences. For the Mendelian forms of PD, exonic mutations in the genes PARK2, SNCA, PARK7(DJ-1), PINK1, ATP 13A2 and LRRK2 have been identified as the genetic causes. However, the Mendelian forms of PD comprise less than 10% of all PD cases. The majority of PD cases are believed to result from a complex interplay between environmental and genetic factors. Several environmental and genetic risk factors, as well as gene-environment interactions in PD have been reported by us and others [2–14]. Environmental-wise, cigarette smoking, coffee drinking, use of on-steroid anti-inflammatory drugs and pesticide exposure have been identified as environmental modifiers of PD [15–17]. Genetic-wise, genome-wide linkage and association studies have identified susceptibility loci for the common non-Mendelian form of PD. These loci include genomic regions containing SNCA, MAPT, BST1, HLA, GAK and LRRK2 [18–22]. Despite the progress that has been made, the known environmental and genetic risk factors cannot account for all the PD cases, warranting further efforts to uncover additional contributing factors missed in previous studies.

PD is characterized by the development of a combination of classic cardinal features including tremor at rest, rigidity, bradykinesia, and postural instability. The disease is hallmarked by the loss of dopaminergic neurons in the substantia nigra pars compacta [23, 24]. Indeed, a dramatic response to levodopa, the metabolic precursor of dopamine is an important diagnostic indicator for PD. The tyrosine hydroxylase (TH; EC 1.14.16.2) gene encodes a monoxygenase that catalyzes the conversion of L-tyrosine to L-dihydroxyphenylalanine (L-DOPA), which is the rate limiting step in dopamine biosynthesis. Consistent with the essential role of TH in dopamine homeostasis, missense mutations in TH have been associated with severe Parkinsonism-related phenotypes, such as Segawa’s syndrome, L-DOPA responsive infantile Parkinsonism, or L-DOPA responsive dystonia (DRD), in recessive form, i.e. both alleles are affected [25, 26]. It has been speculated for a long time that heterozygous missense mutations, i.e. one allele is affected, or genetic variants with less damaging effects than the missense mutations in the TH gene could modify susceptibility for the common form of PD. The answer, however, has not been clear. In this review, we will discuss the current understandings on the genetic variants in TH and their correlations with PD.

2. GENETIC STRUCTURE AND REGULATION THE TH GENE

Human TH has 14 exons and is located in a gene-rich region on the short arm of chromosome 11 in band 15.5 [27, 28]. TH is physically close to the genes encoding insulin (INS) and insulin-like growth factor 2 (IGF-2). This genomic region is also known as the IGF2-INS-TH gene cluster [29]. The TH transcript is subject to alternative splicing at the 5′ end producing four different isoforms of the protein (hTH1, hTH2, hTH3, hTH4) with different N-terminal ends [30]. All four isoforms of TH are expressed in healthy human brain and all are found to be markedly decreased in substantia nigra in the postmortem brain of PD patients comparing to age-matched non-affected people [31]. Among the four isoforms, hTH1 is the most conserved isoform across species and is the most abundant isoform in the brain. Given that hTH1 is the most studied isoform in the literature, all the amino acid positions referred in this review are based upon the codon in hTH1, which encodes 498 amino acids [31–33]. Through alternative splicing, other TH isoforms have inframe insertion between the first and second exon, which results in slightly larger TH proteins (502, 525, 529 amino acids for hTH2, hTH3, hTH4, respectively) [34]. The first 164 amino acids from the N-terminus constitute the regulatory domain modifying enzyme activity through phosphorylation while the rest of the protein constitutes a catalytic domain which starts from the 165th amino acid toward the C-terminus [30]. The last 20 amino acids at the C-terminus are thought to facilitate tetramerization of the enzyme [35]. The variations among the four TH isoforms are all located within the regulatory domain.

Cellular TH enzyme activity is regulated by a wide variety of mechanisms ranging from DNA methylation, transcriptional regulation, to numerous posttranslational events. In short-term regulations, the catalytic activity of TH protein is inhibited by catecholamines (CA). This CA-mediated inhibition can be relieved by phosphorylation of the serine residues (ser19, ser31 and ser 40) in the regulatory domain of TH [31, 36]. In the short-term regulation, only the balance within the pool of existing CA and TH enzyme is altered, respectively through degradation of CA or phosphorylation of TH [37, 38]. In long term regulations, prolonged or repetitive impulses in the dopaminergic cells activate TH gene transcription, increase stability of the TH mRNA and augment translation; together leading to increased levels of TH protein [38, 39]. As such, genetic variations in the regulatory elements in TH promoter, such as response element for cAMP-responsive element binding protein and Nurr1, as well as genetic variations in the 3′UTR, which is essential for mRNA stability, could contribute to the inter-individual variation on TH level.

3. GENOMIC VARIATIONS IN THE TH GENE

Genetic variants (both sequence variations and structural variations) can be classified into three categories based on their relative frequency in population: common variants (polymorphisms), rare variants and private mutations. The different frequencies of these variants in the population require different experimental strategies to detect them. In addition, each of them has different modes on influencing the genetic risk and often demands different analytic approaches to decipher its contributions to diseases [40, 41]. In the past a few years, the revolutionary advances in genomic technology has allowed systematic evaluations of these genetic variants in complex diseases, such as the GWAS to survey common variants, high throughput array-based copy number variation analysis to detect structural variations [42], and massive parallel next generation sequencing to identify rare variants and mutations. In the following sections, we will discuss studies on TH genetic variants in each of the three classes.

3.1. Disease Causing Mutations in the TH Gene and Infantile Parkinsonism

TH has been established as a recessive gene causing TH deficiency, which is associated with a broad phenotypic spectrum ranging from the milder DRD to infantile Parkinsonism or progressive infantile encephalopathy at the severe end [43]. In total, 41 disease causing point mutations in TH have been reported and recorded in the Human Genome Mutation Database [44]. Among them, 33 are missense mutations leading to an amino acid change, 3 are nonsense mutations leading to a premature stop codon, 2 are frame shifting single base deletion, and 3 are promoter mutations affecting TH transcription (Fig. 1). Of interest is that all these disease causing missense mutations are clustered in the catalytic domain and none is found in the regulatory domain of the enzyme (Fig. 1b). On the contrary, a few missense mutations have been reported in the regulatory domain but none of them has been demonstrated a high penetrance on disease phenotype (Table 1). This observation is consistent with the importance of the catalytic domain in maintaining essential functions of tyrosine dydroxylase. In addition, 30 of the 33 amino acid changes caused by these missense mutations are predicated to have damaging effects on the protein product by the Polyphen2 program [45] (Table 1).

Fig. 1.

Genomic variations in the TH gene. a) Display of the genomic variations in TH using the UCSC [50] genome browser (NCBI37/hg19). The “TH mutations” track displays the genomic position of disease causing missense, nonsense, frameshift and promoter region mutations as the vertical red line. The red horizontal bar displays the start and end position of the genomic deletions in TH reported by Bademci et al., 2010 and Ormazabal et al., 2011; b). Protein structure and missense/nonsense mutations in the TH protein. Point mutations are depicted with red arrows, phosphorylation sites of the serine residues activating the enzyme in the regulatory domain are depicted in blue ovals. The nomenclature of point mutation is based on the TH isoform 1 with 498 amino acids (NM_199292.2). Fig. (1b) is modified from Hufton et al. 1995[30]. (IBR=Intersubunit binding region)

Table 1.

Missense and Nonsense Variations in TH

Variant	Description	Predicated Function#	Population Frequency*
A6T	Rare variant	Benign	0.46%
V81M	Common variant	Benign	34%
R122X	Disease causing mutation	Nonsense	N/A
V136M	Rare variant	Benign	0.02%
R138X	Disease causing mutation	Nonsense	N/A
C176Y	Disease causing mutation	Probably damaging	N/A
D196G	Disease causing mutation	Probably damaging	N/A
R202H	Disease causing mutation	Probably damaging	0.01%
L205P	Disease causing mutation	Benign	N/A
A210T	Disease causing mutation	Probably damaging	N/A
H215Y	Disease causing mutation	Benign	N/A
G216S	Disease causing mutation	Probably damaging	N/A
E228G	Disease causing mutation	Probably damaging	N/A
T245P	Disease causing mutation	Possibly damaging	N/A
G263R	Disease causing mutation	Probably damaging	N/A
P270A	Disease causing mutation	Probably damaging	N/A
F278S	Disease causing mutation	Probably damaging	N/A
T283M	Disease causing mutation	Probably damaging	N/A
G284S	Disease causing mutation	Probably damaging	N/A
R288P	Disease causing mutation	Probably damaging	N/A
R297W	Disease causing mutation	Probably damaging	N/A
R306H	Disease causing mutation	Probably damaging	N/A
C328F	Disease causing mutation	Probably damaging	N/A
F344L	Disease causing mutation	Probably damaging	N/A
A345V	Disease causing mutation	Probably damaging	N/A
A354V	Disease causing mutation	Possibly damaging	N/A
L356M	Disease causing mutation	Probably damaging	N/A
I363T	Disease causing mutation	Probably damaging	N/A
T368M	Disease causing mutation	Probably damaging	N/A
G377R	Disease causing mutation	Probably damaging	N/A
Q381K	Disease causing mutation	Possibly damaging	N/A
G383R	Disease causing mutation	Probably damaging	N/A
Q428X	Disease causing mutation	Nonsense	N/A
S436G	Disease causing mutation	Possibly damaging	N/A
P461R	Disease causing mutation	Probably damaging	N/A
P461L	Disease causing mutation	Probably damaging	N/A
T463M	Disease causing mutation	Probably damaging	N/A
D467G	Disease causing mutation	Benign	N/A
L479Q	Disease causing mutation	Probably damaging	N/A

^#Predicatedby the Polyphen2 program⁴⁷;

^*Based on theEsome Variation Server (EVS)⁷³;

N/A=not observed.

The first mutation in TH was found in the study of the Segawa syndrome characterized by progressive dystonia with diurnal fluctuation [46]. Through linkage analysis of microsatellite markers and partial sequencing of TH exons (exon 1–8 and 11–12), the Q381K mutation was identified in one Caucasian family with two affected and one unaffected children. Both affected children are homozygotes while the unaffected sibling and both parents are heterozygotes for the Q381K. Subsequently, seven missense mutations were reported in four families with infantile Parkinsonism. The group who identified the Q381K mutation first reported the L205P mutation in a Greek family with a female whose symptoms of Parkinsonism manifested at 3 months of age [47]. DNA sequencing of TH exons revealed a homozygous L205P mutation in the aforementioned female. Both parents and the brother of the affected female carry the heterozygous L205P mutation and were non-symptomatic at the time of examination. In addition to severe Parkinsonian symptoms in early infancy, the female also had symptoms of a dysfunction of specific sympathetic neurons in the peripheral nervous system (ptosis). Both dopaminergic and noradrenergic neurons also seemed to be affected in this patient. Unlike the Segawa syndrome patient, there was no diurnal variability in symptoms in the patient with the homozygous L205P mutation. Although the L205P mutation is predicted to be a benign amino acid substitution by Polyphen2, in vitro expression of the mutant TH with L205P demonstrated that the mutant protein has very low homospecific activity as compared to wild type TH [47]. Later, Swaans et al. reported four TH missense mutations in two families with infantile Parkinsonism [48]. In family A, two affected brothers had displayed motor symptoms at the age of 2 and 5 years old, and were not able to walk by the age of 5 and 9 years old. Sequencing TH exons revealed that the unaffected mother and the two affected brothers have heterozygous R306H mutation whereas the unaffected father and the two affected brothers have heterozygous T463M mutations, suggesting that the two affected children are compound heterozygote of the TH missense mutations. In family B, a patient had impaired motor development which began at age of 2. The patient carries heterozygous T245P and T283M mutations in the TH gene. It is not clear whether the two mutations are in cis- or trans- state as none of the parents or siblings were available for DNA testing. Given that TH is a recessive gene for TH deficiency, it is likely that the patient is a compound heterozygote of the two mutations [48]. The fourth family was reported by Diepold and colleagues. A female patient of 14 months of age started refusing to walk, development of intention tremor, fisting of hands, and the inability to hold her head or sit. after experiencing a brief viral infection [49]. The symptoms were somewhat alleviated by 19 months of age, but were later exacerbated after another viral infection. By the age of 2 years and 4 months, she suffered severe bradykinesia, hypertonic limbs with cogwheel rigidity, and dystonic posturing of her hands. Sequencing analysis found that the patient carries heterozygous H215Y and D467G mutations, which were transmitted from her mother and father, respectively [49]. In contrast to other mutations identified in other patients with infantile Parkinsonism, both H215Y and D467G are predicted coding tolerable amino acid change in the protein, i.e. having less deleterious effect on the enzyme activity. It is worth noting that the symptoms in this case were induced and worsen after viral infections suggesting a gene-environmental interaction.

In all above patients with TH mutations, administering a low to moderate dose of levodopa, sometimes in combination with a decarboxylase inhibitor eliminated the symptoms including Segawa syndrome and infantile Parkinsonism promptly. Interestingly, the dosage of levodopa treatment seemed to be correlated with the severity of mutations. For example, the Q381K mutation and the L205P mutation led to a decrease of TH activity to 15% and 1.5% of the wild type TH, respectively, in an in vitro model [46, 47]. The dosage of levodopa treatment administrated to relieve symptoms in patients carrying the Q381K mutation and the L205P mutation were 2mg/kg body weight one time and five times daily, respectively [46, 47]. Therefore, accurate genetic diagnostic will facilitate the management of these devastating symptoms effectively.

3.2. Common Variants in the TH Gene and Parkinson’s Disease

It is clear that individuals carrying deleterious TH mutations in both alleles can develop Parkinsonism symptoms at very young age. These mutations, however, are extremely rare; they have only been seen in a few families. Thus, several attempts have been made to examine if common variants with less deleterious effects modulate the susceptibility to adult onset PD. There are mainly two approaches to study association between common genetic variant and disease: GWAS and candidate gene study.

GWAS

Propelled by the international HapMap project and ultra high-throughput genotyping technology, GWAS has become a powerful approach to dissect the genetic basis of complex diseases. This approach uses a dense array of up to 1,000,000 single nucleotide polymorphisms (SNPs) that are printed or synthesized on a microarray (chip) to genotype hundreds and thousands of individuals. Estimates suggest that a well-designed array with 600,000 SNPs will capture ~93% of the common genetic variations in the human genome at an r² of 0.7 within a Caucasian population. During the past a few years, GWAS has identified hundreds of “loci of interest” for many common complex diseases, including obesity, type I and type II diabetes, inflammatory bowel disease, prostate cancer, breast cancer, colorectal cancer, coronary artery disease, Alzheimer’s disease and PD [51–64]. Five independent GWAS on the common form of PD have been reported; four were conducted on non-Hispanic Caucasians and one was conducted on Japanese individuals. After correcting for multiple testing, genome-wide significant results were found in the genomic regions near SNCA, MAPT, BST1, HLA, GAK and LRRK2 [18–22, 65].

Within the TH gene, three SNPs (rs2072056, rs6356, and rs10743152) are either included in the GWAS SNP array or can be imputed using the HapMap reference panel. None of the GWAS, including the one conducted by us at the Hussman Institute for Human Genomics (HIHG) at the University of Miami, reported significant association with PD susceptibility at these SNPs. The negative finding, however, could be due to a lack of statistical power to detect a mild genetic effect in any one particular GWAS. To increase the power in the association analysis, we performed a joint analysis of our GWAS data set with two additional PD GWAS that were deposited in the database of Genotypes and Phenotypes (dbGAP), which is publically accessible; the National Institute of Neurological Disorders and Stroke (NINDS) and the joint dataset from the Progeni/GenePD studies that was genotyped at the Center for Inherited Disease Research (CIDR) [18, 65]. The final data set for the joint analysis include a combined sample size of 1752 cases and 1745 controls. Table 2 displays the allele frequency of the three TH SNPs in each of the three GWAS data sets and the p-value for association with PD risk in the combined data set. Even with improved power in the joint analysis, no evidence for association was found in any GWAS SNPs in TH.

Table 2.

The TH Gene and PD Susceptibility in GWAS and Joint Analysis

GWAS Data Set	Minor Allele Frequency		P-Value in Combined Dataset	Major Allele	Minor Allele
	Case	Control
rs2070762			0.87	A	G
HIHG	0.483	0.504
CIDR	0.494	0.497
NINDS	0.521	0.448
rs6356			0.84	T	C
HIHG	0.383	0.374
CIDR	0.360	0.353
NINDS	0.356	0.375
rs10743152			0.1	T	C
HIHG	0.401	0.384
CIDR	0.381	0.376
NINDS	0.386	0.314

Minor allele frequency of each TH SNP that has been examined in three GWAS, i.e. HIHG, CEDR and NINDS, were listed in cases and controls separately. The P value is generated using an additive model in the combined GWAS dataset of HIHG, CIDR, and NINDS. The Major and minor allele for each SNP is also listed.

Candidate gene study

Contrary to the “null-hypothesis” strategy in a GWAS, candidate gene study selects one or a few biological plausible genes to examine based on a priori knowledge. Since a candidate gene study involves much less genotyping compared to a GWAS, it is more affordable to researchers with limited resources. In addition, the threshold to claim statistical significance in a candidate gene study is much lower than that used in a GWAS due to less statistical tests performed. Given the pivotal role of TH in dopamine biosynthesis, several candidate gene studies have examined the relationship between common genetic variations in TH and adult onset PD susceptibility (Table 3).

Table 3.

The TH Gene and PD Susceptibility in Candidate Gene Study

Study	Polymorphism	Description of Data Set			Results
		Case	Control	Population
Plante-Bordeneuve, 1994	HUMTH01	44	50	England	Negative
Kunugi, 1998	rs6356	99	161	Japan	Negative
Mizuta, 2006	rs2070762	190	190	Japan	Negative
Sutherland, 2008	HUMTH01, rs680 (IGF2), rs689 (INS)	215	215	Australia	Positive (mainly driven by rs680)
Punia, 2010	10 tagSNPs including rs6356	339	334	North India	Negative

Summary of candidate gene studies on the association between TH SNPs and PD susceptibility. The table lists the polymorphism(s) examined, number of cases and controls, countries where the study was carried out in each study. No study found a significant association between TH polymorphism and PD susceptibility.

The first PD genetic study on TH was conducted in England. One polymorphism, a tetra-nucleotide repeat marker HUMTH01 in the first intron of TH was examined in 44 individuals with sporadic PD, 48 individuals with familial PD and 89 of their unaffected relatives, and 50 population control subjects. No evidence for association or linkage at HUMTH01 was found in this small sample set [66]. To expand the study on TH and PD, Kunugi et al. examined the HUMTH01, the L205P mutation (causative TH variant for infantile PD) and rs6356 (a common missense mutation examined in GWAS) in 90 Japanese PD cases and 161 population controls [67]. The L205P mutation was not present in this sample set. The HUMTH01 and rs6356 were found to be highly correlated with each other (in high degree of linkage disequilibrium). Again, this study failed to show an association between TH and PD susceptibility. Later, Mizuta et al. conducted a large-scale candidate gene study in PD involving 1800 Japanese subjects (sample set contained approximately equal number of sporadic PD cases and controls) [68]. In total, 268 SNPs in 121 candidate genes, with one intronic SNP (rs2070762) in TH, were surveyed. The strongest association with PD was found in SNCA (P value =5.0 × 10⁻¹⁰), which is consistent with the GWAS findings. The TH SNP, however, is not significantly associated with PD in this data set. Finally, Sutherland and colleagues took a haplotyping approach to determine whether common genetic variation in the IGF2-INS-TH cluster influences the risk of idiopathic PD [69]. The subjects included in this study are Caucasian recruited from Brisbane, Australia. SNPs rs680 in IGF2, rs689in INS, and the tetra nucleotide repeat marker HUMTH01 in TH, were genotyped in 215 cases and 215 age- and gender-matched controls. The three polymorphisms together define 20 haplotypes. Among them, 8 common haplotypes account for >90% of the genetic variation in the population. The IGF2-INS-TH*6 haplotype was significantly less frequent in the PD cases (P value = 0.001). Additional analysis of the each individual polymorphisms, however, revealed that the A allele of rs680 in IGF2 is responsible for the protective effect of the haplotype, dwindling the importance of the TH variant on PD susceptibility. Recently, a study conducted in a North India PD cohort evaluated six genes that are involved in dopamine synthesis and metabolism, including TH. The study design includes 339 sporadic PD patients and 334 age and gender-matched controls. Within TH, only the missense SNP rs6356 was tested. A highly significant haplotype was found within the dopamine β-hydroxylase (DBH), but no evidence for association was found at rs6356 [70].

Within TH, rs6356 is the most examined SNP in the association studies. Nucleotide variation at this SNP leads to amino acid change between Met and Val at codon 81. As a missense SNP, rs6356 has been postulated to be a variant with functional consequence and therefore more likely to confer genetic risk. Closer examination at rs6356 reveals that the SNP is located in the regulatory domain of TH, a region that is not as functionally critical as the catalytic domain where all disease causing mutations cluster. In addition, the amino acid at codon 81 is not well conserved and the amino acid substitution is predicted to be benign according to PolyPhen2 (Table 1). The well-tolerated change at rs6356 is consistent with its high frequency in the population (Table 1). Given the little function impact at rs6356, may be it is not so surprising that no study found evidence for association at rs6356.

In summary, neither GWAS nor the candidate gene study approach provides convincing evidence for association between common TH polymorphisms and adult-onset PD. Both GWAS and candidate gene study, however, have limitations. Both of them are based on a tag SNP approach using representative markers across the genome or candidate gene to capture the majority of common genetic variations in the population. Such approaches are cost-efficient for screening large numbers of genes and individuals but could miss the actual functional variants underlying the phenotypic variations. In fact, although GWAS has led to novel findings in a number of complex diseases, the associated SNPs only account for a small fraction of the phenotypic variations, which have been referred to as the “missing heritability”. One of the theories for this phenomenon is that rare variants, in contrast to common variants, are not well-represented in the GWAS and they could account for, at least partially, the “missing heritability” in GWAS of complex diseases such as PD [40, 71]. Given the limitations embedded in the tagSNP approach, other strategies need to be taken to fully characterize the contribution of TH variants to the genetic susceptibility of PD.

3.3. Rare Variants in the TH Gene and Parkinson’s Disease

Rare variants have a population frequency less than 1–2%. The low frequency of rare variants makes them hard to be detected and catalogued. In addition, despite that rare variants are believed to have bigger effect size than the common variants, single-locus association test does not provide sufficient statistical power in analyzing rare variants due to its low allele frequency. The recent advance in next-generation sequencing technologies enables us to move beyond the “tagSNP” approaches implemented in GWAS and most candidate gene studies to exhaustively catalogue sequence variants and identify the causal variants leading to disease phenotypes. To accelerate genetic understandings on heart, lung and blood disorders, the National Heart, Lung and Blood Institute has funded a large Exome Sequencing Project aiming to extensively investigate exonic variants that could have been overlooked by GWAS. An internet based Exome Variant Server (EVS) has been developed to release whole exome data derived from 5379 individuals included in ESP cohorts [72]. Accompanied with the rapid advances in sequencing technology, new statistical methods have also been developed to analyze rare variants in complex diseases. Some of them simply examine if rare variants identified in cases are also seen in controls. Recently, the importance of evaluating cumulative effect of rare variants within a gene or a biological pathway has been recognized and several statistical programs have been developed to evaluate the multiple rare variants in aggregate.

In an effort to systematically investigate the contribution of exonic rare variants in PD, we have initiated a whole exome sequencing project in a cohort of well-characterized PD patients and unrelated controls. We have used the Agilent SureSelect exome kit to enrich genomic sequences of coding exons as listed in the Consensus Coding Sequences collection, exon-intron boundaries, and nearly all microRNA genes. The Agilent kit covers ~50Mb of genomic sequence comprising ~200,000 exons and exon-intron junctures. After the initial screening of 164 PD cases and 50 controls, we identified a heterozygous missense variation in TH in one case but not in any controls. The variation leads to an amino acid change from valine to metionine at codon 136 (V136M). The PD patient has an age-at-onset (AAO) of 54 and is from a multiplex family with 5 affected within one generation. DNA samples are available for testing in 5 affected and 2 unaffected siblings/cousins. Sanger sequencing of the 7 family members confirmed the V136M variation in the index patient screened in the whole exome sequencing but not in any other family members. We noted that all the five affected and one unaffected family members also carry one allele of the LRRK2 L1795F missense variation. Interestingly, the patient carrying the V136M has the youngest age-at-onset. In another candidate gene study on rare variants, Hertz et al. examined exonic variations in PARK2, TH, and GTP cyclohydrolase I (GCH1) in 87 Danish patients with earlyonsetPD (AAO <40, or AAO<50 if first degree relative is affected too) [26]. Within TH, only a heterozygous missense variant (A6T) was found in one patient. Interestingly again, the patient also has a heterozygous duplication of exon 11 in PARK2 [26].

With the limited data, it is not clear if A6T and V136M can significantly increase the risk of developing PD. Although both of them cause amino acid change in TH, there are several notable differences between the two rare variants and the previously discussed disease causing mutations in TH. First, both A6T and V136M are located within the regulatory domain where variations are believed to have a milder effect on the enzyme activity while all the disease causing mutations are located within the catalytic domain. Secondly, both rare variants are predicted to cause benign amino acid substitution while most of the diseases causing mutations are predicated to cause damaging amino acid substitution or have been demonstrated to have functional consequences on enzyme activity in vitro (Table 1). Thirdly, none of the disease causing missense mutations, except for the R202H, are found in more than 10,000 chromosomes recorded in the EVS database, with tR202H being found only once. In contrast, A6T and V136M have slight higher allele frequency (0.46% and 0.02%, respectively in the EVS database, Table1). Therefore, it is likely that the two rare variants themselves are not sufficient to significantly increase the risk for developing PD. It remains unclear, however, if they could serve as a modifier for genetic risk conferred by heterozygous variations in other genes, such as LRRK2 and PARK2, or have cumulative effect with other genetic variants on PD susceptibility. Future large scale, systematic studies are needed to fully understand the question.

3.4. Copy Number Variations in the TH Gene and Parkinson’s Disease

In recent years, it has been recognized that DNA copy number variations (CNVs) represent a major source of genome variations, accounting for more nucleotide variation per human genome than SNPs [73–75]. CNVs have been associated with several complex phenotypes, including autism [76, 77] osteoporosis [78], schizophrenia [79], psoriasis [80], autoimmune diseases [81, 82], bipolar disorder [83] and body mass index [84]. In PD, whole gene copy number gain in SNCA and duplication/deletion of one or multiple exons in PARK2 have been identified as causal genetic variations [85]. CNVs, however, in general are underrepresented in genetic studies because it has been cost and labor prohibitive to characterize CNV in large number of samples. CNV genotyping is technically more challenging than SNP genotyping, partially due to the quantitative rather than the qualitative nature of CNV assays. In the past, CNV is genotyped by PCR-based methods, such as semiquantitative PCR, quantitative PCR, and multiple ligation-dependent probe amplification (MLPA). For large recurrent deletions, direct PCR over the fixed breakpoints could be developed for genotyping. Many CNVs, however, are non-recurrent (as seen in TH and PARK2). They vary in size and breakpoint location even when they involve the same gene, thus requiring multiple probes (reactions) to comprehensively evaluate CNVs in one particular gene. As a result, identifying and characterizing previously unknown CNVs are even more challenging. Lately, SNP-array used in the GWAS and high resolution comparative genomic hybridization -array offer a high-throughput approach for genome-wide interrogation of CNVs.

A commercially available MLPA kit (P099-B2, MRC-Holland, Amsterdam, Netherlands) has been designed to survey CNVs in genes related to Dopa-responsive dystonia, including TH [86]. Limited by the MLPA technology, probes can only be designed to target exon 1, 3, 4, 8, 12 and 14 of TH, missing the other eight exons of TH. Using this kit, Liu et al. has screened 16 DRD patients (10 spordic patients and two pedigrees including six patients)⁸⁷. No CNV was found in the TH gene in this small study [87]. Using the same MLPA kit, Ormazabal et al. recently reported a DRD patient with a deletion in the TH gene that encompasses several exons [88]. A heterozygous exon 12 deletion was first identified by MLPA. Additional analysis with long range PCR over breakpoints confirmed the deletion encompassing a segment of 716 bp (c.1197+25_1391del) including exon 12 and part of exon 13. The patient also has a heterozygous mutation (c.1 - 70G>A) at promoter region in another allele. The promoter mutation is believed to be the underlying cause for DRD in other patients [89].

In our GWAS on PD, we have used the Illumina BeadChips (Illumina, Inc) for a genome-wide SNP genotyping of 635 PD cases and 642 unrelated controls. The probe signal intensity and b-allele frequency data generated from the SNP-array were used for a genome-wide CNV analysis. The SNP-array data indicated a 34 kilobase deletion in the genomic region harboring TH in one PD patient but not in any controls. The deletion was defined by seven SNPs, starting at rs2070762 and ending at rs3922756. Using multiple qPCR assays targeting the deleted region delimited by the GWAS, we confirmed the deletion and narrowed the deletion to 17 kb over the entire TH gene. Re-sequencing all the exons of the remaining copy of the TH gene in the patient revealed no exonic mutation. The patient had an age-at-onset of 54 years, with no evidence for dystonia, and was responsive to L-DOPA. It is intuitive to propose that the hemizygous deletion of TH leads to a lower level of TH activity and thus increases PD risk [42]. Neither parent has DNA available for testing but both were reported to have been without any clinical symptoms of PD, well into their 70’s and 80’s. We cannot exclude the possibility that the entire TH deletion is a de nova mutation.

Our analysis of GWAS data suggests that the entire TH deletion is a rare event. It is possible, however, that smaller deletions in TH exist but were missed in the GWAS study due to the limited resolution of SNP-array for CNV. The probe density in a SNP-Array is a few kilobases on average and multiple probes are required to define a CNV. Therefore, SNP-array is useful in identifying large CNVs such as the 17 Kb entire TH deletion but will miss smaller deletion such as the 766 bp TH deletion involving only two exons. High resolution study is needed to comprehensively evaluate the contribution of CNVs in TH to PD.

4. INTERACTION BETWEEN TH AND OTHER PARKINSON’S DISEASE RELATED GENES

As discussed earlier, multiple genes have been identified as genetic risk factors for PD. The molecular mechanisms underlying the association between these genes and PD are not completely understood. Interestingly, the protein products of some of these genes interact with the TH protein and modulate the enzyme activity, thus providing a plausible mechanism for these risk genes to be involved in PD pathogenesis. These biological interactions also raise the question whether variants in TH could modify PD risk via increasing the genetic burden through an accumulative or synergistic effect with other risk genes.

In SNCA, gain-of-function misssense mutations (A30P, E46K and A53T) and copy number gain of the entire gene have been identified as the cause for early-onset familial PD [90–93]. Despite that aggregates of α-synuclein is a major component of Lewy bodies (a neuropathological hallmark of PD); the function of α-synuclein is unclear. Immunoprecipitation and immunoelectron microscopy have demonstrated a direct interaction between α-synuclein and tyrosine hydroxylase in brain homogenates and MN9D dopaminergic cells. Importantly, over-expressing α-synuclein in MN9D cells reduced the tyrosine hydroxylase enzyme activity which was accompanied with diminished dopamine synthesis and lower dopamine levels [94, 95]. In addition, it was shown that α-synuclein suppresses tyrosine hydroxylase enzyme activity without affecting the TH protein level, probably via inhibiting TH phosphorylation. Thus, SNCA could contribute to PD development by serving as a negative regulator of dopamine biosynthesis, which is consistent with the gain-of-function mechanism of identified SNCA mutations.

In DJ-1, multiple loss-of-function missense mutations have been identified as being responsible for early-onset familial PD [96]. DJ-1 plays a role in cellular protection against oxidative stress and regulates mitochondrial functions [97]. Recently, it has been reported that DJ-1 binds directly to the TH protein and activates the enzyme [98]. Mutant DJ-1 with missense mutations identified in PD patients, such as M26I, E64D and L166P, binds to TH protein but does not increase enzyme activity. Rather, the mutant DJ-1 blocks the TH activity induced by wild type DJ-1. Furthermore, several studies have shown that DJ-1 is involved in the transcriptional regulation of TH gene through different mechanisms in different species. In the mouse cells, DJ-1 increases TH gene expression by activating the transcription factor Nurr1 [99], while it has been shown that TH expression in human neuronal progenitor cells does not require Nurr1 [100]. Interestingly, in human cells, it has been found that DJ-1 serves as a positive regulator of TH by suppressing a transcriptional co-repressor polypyrimidine tract-binding protein-associated splicing factor [101, 102]. Nonetheless, DJ-1 could contribute to PD risk by acting as a positive regulator of dopamine biosynthesis, which is consistent with the loss-of-function mechanism of DJ-1 mutations.

5. SUMMARY

Encoding the enzyme catalyzing the rate limiting step of dopamine synthesis, TH is a theoretical susceptibility gene for PD pathogenesis. Indeed, recessive missense mutations in TH have been identified as a genetic cause for infantile Parkinsonism. The implication of these mutations in the adult-onset PD, however, is trivial due to the extreme rare presence of them. These disease-causing mutations have only been seen in a few families with infantile Parkinsonism but not in the population. On the other hand, several studies have examined association between common variants (minor allele frequency is greater than 5% in the population) in TH and adult onset PD. No evidence for association was found supporting a role of common TH variants in PD susceptibility. In comparison to disease-causing mutations, these common variants do not seem to have significant functional consequences, including a missense SNP rs6356. Recently, there are several cases in which rare genetic variants (minor allele frequency lesser than 1% in the population), such as a whole gene deletion and missense substitutions, are speculated to increase the risk of adult onset PD.

To fully understand the genetic risk conferred by TH, systematic study on rare variants, including point sequence variations and CNVs, are desired. With the rapid advances in genomic technology, such as next generation sequencing technology for sequence variations and Nanostring® for CNV, comprehensive genomic study can be performed to thoroughly evaluate rare variants. In particular, large scale whole exome sequencing and whole genome sequencing will exhaustively survey all the functional variants in TH as well as other genes through the human genome. This will not only allow for evaluation of the collective effects of multiple rare variants in TH, but also the epistasis between TH and other genes, especially the ones that interact with TH.

ABBREVIATIONS

AAO

Age-at-onset

CA

Catecholamines

CGH

Comparative genomic hybridization

CIDR

Center for Inherited Disease Research

CNVs

Copy number variations

dbGAP

Database of Genotypes and Phenotypes

DRD

L-DOPA responsive dystonia

EVS

Exome Variant Server

GWAS

Genome-wide association study

IGF-2

Insulin-like growth factor 2

IND

Insulin

L-DOPA

L-dihydroxyphenylalanine

MLPA

Multiple ligation-dependent probe amplification ()

NINDS

National Institute of Neurological Disorders and Stroke

PD

Parkinson’s disease

SNPs

Single nucleotide polymorphisms

TH

Tyrosine Hydroxylase