Genetics in Parkinson disease: Mendelian vs. non-Mendelian inheritance

Abstract

Parkinson’s disease is a common, progressive neurodegenerative disorder, affecting 3% of those older than 75 years of age. Clinically PD is associated with resting tremor, postural instability, rigidity, bradykinesia and a good response to levodopa therapy. Over the last fifteen years, numerous studies have confirmed that genetic factors contribute to the complex pathogenesis of PD. Highly penetrant mutations producing rare, monogenic forms of the disease have been discovered in singular genes such as SNCA, Parkin, DJ-1, PINK 1, LRRK2 and VPS35. Unique variants with incomplete penetrance in LRRK2 and GBA have been shown to be strong risk factors for PD in certain populations. Additionally, over 20 common variants with small effect sizes are now recognized to modulate the risk for PD. Investigating Mendelian forms of PD has provided precious insight into the pathophysiology that underlies the more common idiopathic form of disease; however, no treatment methodologies have developed. Furthermore, for identified common risk alleles, the functional basis underlying risk principally remains unknown. The challenge over the next decade will be to strengthen the findings delivered through genetic discovery by assessing the direct, biological consequences of risk variants in tandem with additional high-content, integrated datasets.

Introduction

Substantial progress has been made in understanding the genetics of PD. Highly penetrant mutations producing rare, monogenic forms of the disease have been discovered in singular genes such as SNCA, Parkin, DJ-1, PINK1, LRRK2 and VPS35. Unique variants with incomplete penetrance in LRRK2 and GBA have been shown to be strong risk factors for PD in certain populations. In addition, over 20 common variants with small effect sizes have been shown to modulate the risk for PD. Whole-genome approaches have clearly aided our quest to comprehend the role genetics plays in the pathogenesis of PD. We now detect Mendelian mutations in segregating families in a straightforward manner and we also have the ability to assess how common variability may play a role in disease.

Parkinson’s disease was long thought to be a sporadic disorder without genetic causation. However, in 1997 mutations responsible for the disease were identified in the alpha-synuclein gene (SNCA) [1, 2]. This landmark discovery revealed the first indisputable, heritable component of PD and launched years of significant research into the genetics of PD. Quickly following detection of the first mutations in SNCA, additional genetic links were identified at two novel chromosomal regions and linkage of SNCA was excluded in >200 PD families [3–7].

Therefore, by 1998, it was evident that PD was a genetically heterogeneous disease. Several genes have since been linked to inherited forms of parkinsonism and several monogenic forms of the disease and numerous genetic risk factors have been identified. However, it has taken many years of study to begin understanding the underlying gene functions and molecular mechanisms that lead to disease. In this review, we first provide a brief overview of the monogenic forms of disease and then move forward to discuss how our view of PD etiology has matured since 1997 to now include risk alleles.

Monogenic Forms of Parkinson’s Disease

Understanding the monogenic forms of PD provides insight more broadly into the genetic architecture of this disease, and as described later, there appears to be overlap in the genes that contain disease causing mutations and those that contain risk variants (Table 1). Mutations in three genes, SNCA (PARK1; encoding α-synuclein), LRRK2 (PARK8; encoding dardarin) and VPS35 (encoding vacuolar protein sorting 35) have been shown to cause autosomal dominant forms of PD. Mutations in six other genes, PINK1 (PARK6; PTEN induced kinase 1), DJ-1 (PARK7), Parkin (PARK2), ATP13A2 (PARK9), FBXO7 and PLA2GB have been shown to cause autosomal recessive PD and/or parkinsonism. The mutations in these genes, with the exception of LRRK2, cause PD in a small subset of patients. All known monogenic forms of PD combined explain only about 30% of familial and 3–5% of sporadic cases [8].

Table 1.

Loci involved in monogenic forms of PD and risk loci identified prior to the advent of genome wide association studies.

Grey – loci for which only questionable evidence exists.

Locus	Gene	Protein	Model
Park1	SNCA	α-synuclein	Autosomal Dominant
Park2	PARK2	Parkin	Autosomal Recessive
Park3	unknown	unknown	Autosomal Dominant
Park4	SNCA	α-synuclein	Autosomal Dominant
Park5	UCHL1	Ubiquitin c terminal hydrolase	Autosomal Dominant
Park6	PINK1	Pten-induced putative kinase 1	Autosomal Recessive
Park7	PARK7	DJ-1	Autosomal Recessive
Park8	LRRK2	Leucine rich repeat kinase 2	Autosomal Dominant
Park9	ATP13A2	lysosomal type 5 ATPase	Autosomal Recessive
Park10	unknown	unknown	Risk locus
Park11	GIGYF2	GRB interacting GYF protein 2	Autosomal Dominant
Park12	unknown	unknown	X-linked
Park13	HTRA2	HTRA serine peptidase 2	Autosomal Dominant
Park14	PLA2G6	Phospholipase A2	Autosomal Recessive
Park15	FBXO7	F-box only protein 7	Autosomal Recessive
Park17	VPS35	Vacuolar protein sorting 35	Autosomal Dominant
Park18	EIF4G1	Eukaryotic translation initiation factor 4 gamma 1	Autosomal Dominant
Park19	DNAJC16	DNAJ/HSP40 homolog subfamily C member 6	Autosomal Recessive
-	SNCA	α-synuclein	Risk locus
-	LRRK2	Leucine rich repeat kinase 2	Risk locus
-	GBA	Glucocerebrocidase	Risk locus

Autosomal Dominant Parkinson’s Disease

Alpha-Synuclein

The first mutation underlying PD in SNCA (Ala53Thr in exon 4) was discovered in a large Italian family and subsequently identified in three Greek families with familial PD using traditional linkage mapping [2]. The primary Greek families found to harbor the p.A53T mutation originated from a very small geographical area in southern Greece. Eight additional families, located in central and southwestern Greece, were also confirmed to have mutations in α-synuclein, suggesting the presence of a founder mutation [9, 10]. A decade later, two Korean, and one Swedish family were shown to have the same mutation [11–13].

Shortly following the discovery of SNCA mutations causing a rare familial form of PD, Spillantini and colleagues determined that α-synuclein was a major constituent of Lewy bodies, the pathological hallmark of PD [14].

This was a profound finding in the brains of typical sporadic PD patients that distinctly tied together the etiology and pathogenesis of rare familial forms of PD with sporadic cases. Understanding the crucial link between the two disease phenotypes ultimately established that examination of rare forms of familial PD, even those that differed clinically and neuropathologically from typical PD, was pertinent to the study of the common form.

Mutations in SNCA are rare. As yet, only five (p.G51D, p.G50D) different autosomal dominant, missense mutations have been discovered in α-synuclein along with duplications and triplications of the complete gene [15]. The first identified missense mutation, p.A53T is the most frequent and has been found in seven families throughout the world. The remaining two missense mutations were found in only one family each: p.A30P was reported in a German family with autosomal dominant PD and p.E46K in a Spanish family from the Basque country [16, 17]. Triplication of the entire genomic region containing SNCA was first discovered in 2003 and has since been reported as a cause of disease in several families [18–20]. However, duplications of the entire coding region of SNCA have been reported as a more common cause of disease in families and apparently sporadic cases [19, 21–27].

The clinical phenotype associated with SNCA mutations consists of progressive L-DOPA responsive parkinsonism with cognitive decline, autonomic dysfunction and dementia. The average age of onset for those patients with the p.A53T mutation is 46 years of age, which is significantly younger than typical sporadic PD and the disease is fully penetrant [28]. In contrast, families with the A30P mutation have an age of onset that is slightly later (age 52) and the disease is not fully penetrant [16], while the p.E46K mutation causes dominant PD and Lewy body dementia with symptoms beginning between the ages of 50 and 65 years with dementia presenting within 2 years of diagnosis [17].

Genomic duplications and triplications at the SNCA locus cause early-onset PD with the age of onset and severity of the disease phenotype correlating with the SNCA copy number, suggesting a gene-dose effect. Triplication of α-synuclein causes a fully penetrant early onset Parkinson’s disease (EOPD) that is rapidly progressive, dopa-responsive parkinsonism, and is accompanied or followed by dementia. Clinical presentation ranges widely from severe idiopathic PD to PD with dementia or diffuse Lewy body disease. PD patients with duplication of SNCA, therefore generating three copies of the gene, develop the disease about a decade later than those with four copies of synuclein and the disease course, while still aggressive, is generally more benign [18, 20, 23, 29, 30].

Although mutations in SNCA are a clear cause of PD and the presence of α-synuclein aggregates is a hallmark of PD, the normal function of α-synuclein remains poorly understood. SNCA encodes a small 140 amino acid protein that exists in a range of quaternary states from monomeric, to low molecular weight oligomers, to the high molecular weight amyloid fibrils that are found in Lewy bodies. Alpha-synuclein is mainly found in the cytosol where it is bound to lipid rafts in an interaction that is required for it association with the synapse [31]. It has been reported that α-synuclein interacts with members of both the Rab and SNARE families, pointing to a role in vesicular trafficking [32–34].

Consistent with its localization at the synapse, Nemani et al. showed that overexpression of human SNCA in mice inhibits synaptic transmission [35]. Interestingly, they also observed synaptic inhibition when overexpressing the PD-linked mutations p.A53T and p.E46K but not with the p.A30P mutation. It has been suggested through experiments using overexpression of human SNCA mutants in rats, that mutants favoring low molecular weight oligomers that bind membranes (p.E35K and pE57K) cause more dopaminergic neuron death in the substantia nigra than those that quickly form fibrils (WT, p.A30P, and p.A53T) [36]. This mechanism is somewhat controversial, as the mutants that increase cell death are not found in familial PD cases, and it does not account for neurodegeneration caused by duplication or triplication of the SNCA locus or for disease caused by familial mutants that quickly form fibrils. These studies point to an important role for SNCA in intracellular trafficking, membrane interaction, and synaptic activity; but more work must be done to fully understand the mechanisms underlying disease and normal gene function.

Leucine Rich Repeat Kinase 2

In addition to SNCA, autosomal-dominant PD-causing mutations have been found in the gene encoding Leucine-rich repeat kinase 2 (LRRK2). Linkage of PD to a region on chromosome 12 was originally mapped in a large Japanese family with autosomal-dominant, late-onset PD showing incomplete penetrance [37]. Within two years, the locus was verified and further delineated in several European families [38]. In 2004, two groups performed positional cloning identifying mutations in LRRK2 as the root cause of chromosome 12-linked PD [38, 39]. Mutations in LRRK2 are now recognized as the most common known cause of familial PD.

To date, more than 100 distinct missense and nonsense mutations have been reported in LRRK2 [40]; however, only for a small minority is there overwhelming proof of pathogenicity (p.R1441C/G/H, p.Y1699C, p.S1761R, p.I2012T, p.G2019S, and p.I2020T) [41–45]. These pathogenic modifications are clustered in exons encoding the Ras of complex proteins (ROC), C-terminus of ROC (COR) or kinase domains of the protein.

The most well studied mutation, p.G2019S, is common across many populations and has been identified in up to 42% of familial cases, depending on the ethnic background [43, 46]. It is frequent in North African, Middle Eastern and Ashkenazi Jewish PD patients and it is believed that most LRRK2 p.G2019S mutation carriers are from a common founder originating in North Africa and spreading with the Jewish diaspora [43, 47–52]. However, pG2019S remains rare in Asian populations, where it accounts for less than 1% of LRRK2 mutations [53, 54]. Importantly, p.G2019S is also detected in sporadic PD cases. The mutation is seen in approximately 2% of sporadic cases in Northern European and U.S. populations and up to 10% of sporadic cases worldwide [42]. The penetrance of the p.G2019S mutation is age dependent and varies from 28% at 59 years, to 51% at 60 years, to 74% at 79 years of age.

Overall, mutations in LRRK2 are the most common known genetic cause of late-onset PD and are found in both autosomal dominant and sporadic cases. LRRK2 mutations are found in ~10% of patients with autosomal dominant familial PD [55–58], 3.6% of patients with sporadic PD and 1.8% of healthy controls [50]. Phenotypically, LRRK2 mutation carriers are essentially indistinguishable from sporadic PD [59] demonstrating mid to late onset of disease around 60 years of age, with a slow progression and a good response to levodopa therapy. Dementia in individuals with LRRK2 mutations is rare. Neuropathologic features are consistent with typical PD showing Lewy bodies (LB) in the brainstem and loss of dopaminergic neurons in the substantia nigra [59].

LRRK2 encodes a large, multi-domain protein of 2527 amino acids in the ROCO protein family [60, 61]. It is widely expressed in brain tissue and localized to LBs in the brainstem where it is associated with the endoplasmic reticulum of dopaminergic neurons [62, 63]. The GTPase ROC and COR domains as well as the Ser/Thr kinase domain forms an enzymatic core that characterizes the LRRK2 protein and contains the proven autosomal dominant pathogenic mutations. The most common p.G2019S mutation has been shown to consistently increase kinase activity [64].

The flanking ankryn, leucine rich repeat and WD40 domains allow for interaction with numerous other proteins including 14-3-3, BAG5, GAK and Rab7l1 [65–67]. LRRK2 has been implicated in numerous cellular processes including autophagy, cytoskeletal dynamics, kinase cascades, mitochondrial function and vesicular trafficking [66, 68–75]. The many diverse activities of LRRK2 prohibit a clear view of its exact physiological functions, but investigations are ongoing to create a framework in which its role in PD can be better understood. The identification of LRRK2 mutations has proven to be a landmark discovery that has profoundly impacted our understanding of Parkinson’s disease.

Vacuolar Protein Sorting 35

In 2011, Zimprich and colleagues were the first to use next-generation sequencing methods to detect a PD causing gene. They identified a mutation in vacuolar protein sorting 35 homolog gene (VPS35, encoding vacuolar protein sorting 35) in a family from Austria with 16 affected individuals, as a cause of late-onset autosomal dominantly inherited parkinsonism. Exomes from a pair of affected second cousins were compared to generate a list of rare, shared heterozygous coding mutations. Subsequently, only one mutation, p.D620N, was validated by Sanger sequencing, and was not found in a group of more than 650 controls. This mutation was found in every affected individual in the family and segregates with late-onset PD in a Mendelian autosomal dominant manner [59].

Another group, also using exome sequencing, simultaneously identified the p.D620N mutation as a cause of late-onset PD in a large Swiss family [76]. Sequencing of the entire coding portion of VPS35 revealed a handful of other mutations in these studies (p.G51S, p.M57I, p.T82R, p.I241M, p.P316S, p.R524W and p.L774M), however the mutations have not been proven to be pathogenically relevant. Several thousand PD and control subjects have since been screened for the c.1858G>A (p.D620N) mutation, identifying numerous families with this form of PD. Overall VPS35 mutations are a rare cause of PD accounting for only about 1% of familial parkinsonism and 0.2% of sporadic PD [77–79].

VPS35-linked PD resembles typical idiopathic disease with a mean age of onset at 53 years, bradykinesia, resting tremor, and good response to levodopa therapy [80]. VPS35 encodes a highly conserved 796 amino acid protein. The homologous yeast gene, vps35, has been well characterized by the discovery of mutants showing abnormal vacuole sorting and secretion [81]. These early experiments in yeast showed that Vps35p is a core member of the retromer complex, which also contains Vps26p and Vps29p, and is responsible for the retrograde transport of proteins in endosomes to the trans-Golgi network (TGN) [82]. Vps35p is located at the center of the complex and is required for the recognition and binding of the cytosolic domains of cargo for retrograde transport [83]. The human homologs of the retromer complex were later cloned and shown to function in the same endosome-TGN pathway [84].

A recent report links the PD associated p.D620N mutation to dysfunction of the retromer complex by redistributing retromer bound endosomes to the perinuclear region in both cell lines and PD patient derived fibroblasts. The study goes on to show that the mutant VPS35 protein alters the trafficking of cathepsin D, a protein that is implicated in the degradation of α-synuclein [85, 86]. Although mutations in VPS35 are rare, its high level of conservation has made it one of the best understood genes associated with Mendelian PD. VPS35 is linked to the other autosomal dominant PD genes SNCA and LRRK2 through endosomes and vesicular trafficking and underscores the importance of studying these pathways in health and disease.

Autosomal Recessive Parkinsonism

Mutations in six genes: PARK2 (encoding parkin), PINK1 (PARK6; PTEN induced kinase 1), DJ-1 (PARK7), ATP13A2 (PARK9; ATPase type 13A2), PLA2G6 (PARK14; phospholipase A2, group VI) and FBX07 (PARK15; F-box only protein 7) have been shown to cause autosomal recessive (AR) PD/parkinsonism. The mutations in these genes cause early onset PD in a small subset of patients. All known monogenic forms of PD combined explain only about 20% of early-onset PD and less than 3% of late-onset PD, although as will be discussed below, this proportion varies across ethnic groups.

Parkin

Parkin was the second gene identified to cause parkinsonism and the first gene decisively shown to be inherited in an autosomal recessive (AR) manner. A homozygous deletion of exons 3–7 in the Parkin gene was first reported by Kitada and colleagues in Japanese families with autosomal-recessive juvenile on-onset parkinsonism (ARJP); a severe, early-onset form of disease with onset often occurring before twenty years of age [4]. This study also identified four families with ARJP showing homozygous deletions of exon 4 only.

Mutations in Parkin are the primary cause of ARJP and early onset, recessive parkinsonism. Numerous unique mutations in all 12 exons of Parkin have been identified throughout various ethnic populations. These mutations consist of point mutations and exon rearrangements, including both deletions and duplications [45, 87–91]. To date, approximately 147 different exonic mutations have been described of which a third are single-nucleotide changes, 13% are minor deletions and 54% are larger deletions or duplications comprised of one or more exons [92].

The number of disease causing exon rearrangements in Parkin is likely still to increase as many exon rearrangements were often missed due to the labor intensive and expensive early methods required for their identification. Mutations are present in approximately 50% of patients with recessive, EOPD in the age range of 7–58 years of age and present in up to 77% of sporadic cases with disease onset younger than 20 years [93].

Key clinical features of Parkin-linked disease have been reported to include age at onset <40 years, foot dystonia, psychiatric symptoms and a dramatic response to treatment [95]. However, these symptoms can mirror those of typical EOPD cases without Parkin mutations [96].

The pathology of Parkin disease consists of severe neuronal loss in the substantia nigra, occasional tau pathology and a distinct lack of postmortem LBs in most cases [97–99]. A possible explanation for the lack of LBs, generally a pathological hallmark of PD, may be the young age of Parkin disease onset [100]. It is notable that no cases of juvenile-onset PD have been reported with postmortem LBs [101, 102] and the rare patients with Parkin mutations and LBs have a significantly older age of disease onset (mean age of onset: without postmortem LBs = 27 years; with postmortem LBs = 46 years) [100, 103].

Parkin is one of the largest genes in the human genome and codes for a 465-amino acid E3 ubiquitin ligase made up of an ubiquitin-like domain (Ubl) and two RING domains [94]. The Ubl domain of E3 ubiquitin ligases is required for recognizing substrates and the RING-box is needed for interaction with its specific E2 ubiquitin-conjugating enzyme. Shimura and colleagues have demonstrated that the E2 enzyme associated with Parkin is UbcH7 and that a PD associated missense mutation (p.T240R) in the first RING domain is sufficient to disrupt this interaction [104]. Narendra and colleagues demonstrated that Parkin has a generally cytosolic localization except in the context of mitochondrial damage when it is phosphorylated by PINK1, another gene associated with ARJP and translocated to the surface of unhealthy mitochondria to ubiquitinate mitochondrial membrane proteins [105, 106]. These ubiquitinated proteins signal to the cell that the damaged mitochondria must undergo mitophagy. The prevalence of Parkin mutations has implicated mitochondrial quality control in the pathogenesis of PD indicating that multiple pathways lead to different parkinsonian phenotypes.

PTEN-Induced Putative Kinase 1

The PARK6 locus was first mapped to a 12.5-centimorgan (cM) region on chromosome 1p35-p36 in a large consanguineous family from Sicily [107]. In 2004, Valente and colleagues identified two homozygous mutations in the PTEN-induced putative kinase 1 (PINK1) gene. A p.G309D missense mutation and a p.W437X truncating mutation were found in a Spanish family and two Italian families respectively. Both Italian families share a common haplotype, demonstrating a shared ancestry [108].

Since the discovery of PINK1 in PD more than ten different mutations have been associated with EOPD. The combination of homozygous and compound heterozygous loss of function mutations in PINK1 is the second most common cause of autosomal-recessive EOPD, predicted to be present in about 3.7% of patients [108–113]. The clinical phenotype of PINK1-related PD strongly resembles levodopa responsive, classic idiopathic PD with no reports of dementia [114].

PINK1 encodes a 581 amino acid Ser/Thr kinase that localizes to the mitochondria [108]. As previously mentioned, PINK1 phosphorylates Parkin to regulate mitophagy of damaged mitochondria [106, 115]. Additionally, overexpression of wild type PINK1 but not the disease associated mutation p.G309D has been demonstrated to rescue stress induced apoptotic death via the mitophagy pathway [108]. The convergence of two genes in the mitophagy pathway confirms that it is essential to neuronal health and survival and may be a good therapeutic target in a subset of EOPD patients.

DJ-1

The PARK7 locus on chromosome 1p was established through the discovery of a consanguineous pedigree from the Netherlands with autosomal recessive PD. In 2003, Bonifati and colleagues identified recessively inherited missense and exonic deletions in DJ-1 in two European families making it the third gene associated with ARPD [116]. Homozygosity mapping and positional cloning was performed on consanguineous pedigrees from a genetically isolated population in the Northern Netherlands revealing a homozygous deletion of several exons in DJ-1 causing disease. Subsequently, a missense mutation in a highly conserved residue (p.L166P) of DJ-1 was found to cause disease in an Italian ARPD family. Mutations in DJ-1 are extremely rare, identified in <1% of early-onset PD cases. The mutations are found in both homozygous and compound heterozygous states, resulting in loss of protein function. Phenotypically, DJ-1 mutations cause levodopa responsive disease onset in the mid-twenties, resembling Parkin and PINK1 linked forms [116–122].

The DJ-1 gene spans 24 kb in length and includes 8 exons encoding a protein of 189 amino acids that was first identified in cancer [123]. DJ-1 belongs to the pepidase C56 family of proteins and has been reported to protect cells against oxidative stress and to play a role in maintaining normal dopaminergic function in the nigrostriatal pathway [118, 124–129]. Besides the importance of DJ-1 in dopamine neurotransmission and signaling, it has been reported to have multiple functions associated with PD pathogenesis, such as chaperone activity and the ability to inhibit α-synuclein aggregation, which is thought to be a key event in Lewy body formation [130]. It has also been suggested that DJ-1 may be involved in transcriptional regulation of neuroprotective or anti-apoptotic genes [131]. Discovery of two sibs with EOPD that have inactivating heterozygous mutations in DJ-1 and PINK1 led to the suggestion that these two proteins interact to protect the cell from stress induced apoptosis [132]. More recent work has linked DJ-1 to the Parkin/PINK1 pathway again through transcriptional regulation of PINK1 [133].

ATP13A2, PLA2G6 and FBXO7

More rare, recessively inherited forms of PD are caused by mutations in three genes: ATP13A2 (ATPase type 13A2), PLA2G6 (phospholipase A2, group VI) and FBXO7 (F-box only protein 7). Mutations in ATP13A2 cause the rare, juvenile-onset disorder Kufor-Rakeb syndrome, which is characterized by a lower response to levodopa and additional atypical features of disease such as dystonia and supranuclear palsy [134, 135]. At least eleven families with numerous different missense, nonsense and deletion mutations in ATP13A2 have been identified to date. Phenotypic severity is highly variable between patients and appears to be related to the type of mutation inherited [136].

ATP13A2 encodes 1175 amino acid member of the P5 family of ATPases. ATP13A2 is a multifunctional protein with ten transmembrane domains that is hypothesized to play a role in both of the previously described PD pathways, endosome-lysosome dynamics [137, 138] and mitochondrial health [139, 140], as well as in protecting cells from metal (Mn2+ and Zn2+) induced toxicity [141, 142]. These studies also indicate that ATP13A2 is present in Lewy bodies of PD patients [137] and overexpression of ATP13A2 rescues neurons from α-synuclein accumulation [141, 142]. Mutations in ATP13A2 are a very rare cause of parkinsonism but these mutations represent an important link between autophagy-lysosome function, mitochondrial health and neurodegeneration.

Mutations in the gene PLA2G6 cause autosomal recessive, levodopa-responsive parkinsonism with dystonia [103]. Brain iron accumulation is found in most but not all affected individuals [143–145]. PLA2G6 mutations have also been associated with neurodegeneration with brain iron accumulation and Karak syndrome, both of which are forms of severe infantile or childhood neurodegeneration [146, 147]. PD associated with PLA2G6 is caused by the homozygous or compound heterozygous inheritance of various missense mutations [144, 145, 148]. PLA2G6 encodes an 806 amino acid protein called phospholipase a2 group VI with seven ankryn repeats, a lipase domain and a calmodulin binding domain [103]. This enzyme hydrolyzes glycerophospholipids to produce free fatty acids and 2-lysophospholipids [149]. Recent studies in Drosophila have shown that knockout of the PLA2G6 homolog results in mitochondrial dysfunction and neurodegeneration. Fibroblasts from a patient with the PD-associated mutation p.R747W also showed similarly impaired mitochondria [150]. Although mutations in PLA2G6 are rare, they present an interesting link of a PD gene to other more severe forms of neurodegeneration that merits further study.

Shojaee and colleagues first identified mutations in FBXO7 through linkage mapping followed by gene sequencing in an Iranian family with AR recessive juvenile-onset parkinsonian-pyramidal syndrome [151]. The affected individuals in this family first demonstrated early onset spastic paraplegia and then later displayed dopa-responsive parkinsonism. Individuals from two additional families with similar symptoms were later shown to have three different mutations in FBXO7 in either the homozygous or compound heterozygous state [152]. FBXO7 encodes F-box only protein 7 and is made up of 443 amino acids. It has been shown to directly interact with Parkin and PINK1 in mitochondrial maintenance and mitophagy [153, 154]. These studies further demonstrate that FBXO7 aids in translocation of Parkin to the mitochondria in response to cell stress and a PD associated mutation (p.T22M) leads to mislocalization of FBXO7 to the cytosol. Mutations in FBXO7 are a rare cause of ARJP but are found in the heterozygous state across many populations.

Genetic Risk

Complex diseases such as PD likely arise due to a combination of many genetic, environmental, and lifestyle factors. There are two principle theories regarding the genetic component of complex disease: the common disease rare variant (CDRV) hypothesis and the common disease common variant (CDCV) hypothesis. At the inception of these theories, there was much discussion regarding the likely validity of these hypotheses, and in particular much argument in favor of one over the other; however it is important to note that these two hypotheses are not mutually exclusive, and it is likely that any complex disease will contain both genetic components.

The CDRV hypothesis speculates that a contributing risk component for complex disease will be rare genetic variants; these are strictly defined as any genetic allele with a frequency of 1% or less. The CDRV hypothesis suggests that because low frequency variants are abundant in human populations and because there has been little opportunity for the removal of these functional but rare variants through natural selection (as a consequence of recent population growth) deleterious rare alleles are likely to exist. This phenomenon may be particularly pronounced in late onset diseases, where selective pressures, mainly driven through traits that occur before reproductive age is reached, do not apply. A primary limitation in the investigation of the CDRV hypothesis has been a technical one; it has traditionally been extremely difficult to identify rare variants in suitably powered sample series. However, there is now a great deal of interest in the investigation of the CDRV hypothesis due to the affordability of second generation sequencing methods to identify these rare alleles in large groups of individuals. This is one of the fastest growing fields of disease genetics and has had some initial success [155, 156].

However, the rarity of these alleles requires very large sample numbers to detect significant effects and therefore, it is expensive to execute well-powered studies. As a consequence, the CDRV hypothesis remains largely untested in most common disorders, a situation that will inevitably change over the next few years as next generation sequencing technologies become cheaper and more accessible to labs across the world.

The CDCV hypothesis posits that a significant proportion of risk for common diseases is mediated through common genetic variants (i.e. variants present at greater than 1% allele frequency within a population) [157, 158]. By definition these variants are common and therefore will have been within the population for a significant amount of time. Thus, conversely to rare variants, highly functional, deleterious alleles are more likely to have been selected out of the population. Therefore, the CDCV hypothesis accepts that the effect of individual common alleles on any deleterious trait is likely to be quite small, but numerous such common alleles may contribute to that trait, so collectively the contribution of common alleles to a trait may be substantial. Testing of the CDCV has been well established, primarily because the technology available to genotype a large number of common variants has been available, and relatively affordable since ~2005. This has led to an extremely large number of identified common risk loci as evidenced by the large catalog of positive genome wide association studies (http://www.genome.gov/gwastudies/).

Common Genetic Risk Factors in Parkinson’s disease

During recent years several susceptibility genes and numerous risk loci associated with PD have been identified. While success in this regard has been driven by largely unbiased genome wide studies, some limited success was achieved through candidate gene-based assessments in three genes: SNCA, LRRK2 and GBA.

Alpha-synuclein

As previously noted, families with rare mutations in SNCA enabled the novel discovery of a genetic link to PD [2]. Following this hallmark finding, Kruger and colleagues examined common variability within SNCA to establish whether the gene was also associated with risk for the sporadic form of PD [159]. Kruger’s study initially reported that APOE genotype, a major risk factor for late-onset Alzheimer’s disease [160], interacted with a variable dinucleotide repeat within SNCA. The combination of the APOE4 allele and NACP allele 1 of the SNCA promoter polymorphism were shown to be significantly different between sporadic PD patients and controls. PD patients presenting this genotype had a 12.8-fold increased relative risk for developing PD over the course of their lives. Unfortunately, this interaction between SNCA and APOE genotypes has not been replicated; however, Maraganore and colleagues later demonstrated variability of risk for PD between SNCA promoter alleles using meta-analysis of existing dinucleotide repeat sequence alleles at the SNCA promoter (REP1) genotype data [161]. Maraganore showed an unequivocal association between genetic variability within the SNCA locus and PD [161]. Since then, association of Parkinson’s disease with SNCA has been overwhelmingly established in GWA studies identifying a handful of significant SNPs and revealing more about the architecture of genetic risk at this locus.

Leucine Rich Repeat Kinase 2

As with SNCA, subsequent to the identification of LRRK2 mutations as a cause of monogenic PD [32, 33] common variability across LRRK2 was examined in several populations. Within Asian populations, the variant p.G2385R was first identified as a cause of PD [162]. However, this variant was present in 5% of the population and later shown to be a risk allele that doubled the risk of PD in individuals [163]. This finding was replicated extensively in Asian populations including those from Singapore, Taiwan, China, Korea and Japan [11, 164–174]. An additional variant described in 2008, p.R1628P was also shown to be associated with a ~2 fold increase risk for developing PD and has been replicated in several Asian populations including Thai, Chinese and Taiwanese populations [165, 175–178]. Several additional variants within LRRK2 have been assessed and have varying levels of support for association with risk for PD [176, 179]. The LRRK2 locus is repeatedly identified by GWA studies in Caucasian populations, but the exact causative variants remain elusive, requiring a deeper analysis of the locus.

Glucocerebrosidase

Remarkably, thorough clinical observation rather than previously known genetic association, lead to the discovery of PD risk variants within the gene encoding glucocerebrocidase (GBA); a gene long tied to the autosomal recessive lysosomal storage disorder, Gaucher’s disease [180]. Tayebi and colleagues observed that a portion of Gaucher’s disease patients manifested with parkinsonism, compelling an early hypothesis that GBA deficiency may lead to a predisposition to parkinsonism [181, 182]. A year later Aharon-Peretz and colleagues were able to show that inheritance of a single mutation in GBA increased the risk for PD [183]. Further, meta-analysis of existing data was later used to show that in Ashkenazi Jewish populations the frequency of two common mutations in GBA (p.N370S and p.L444P ) was 15% in PD and 3% in controls; whereas non-Ashkenazi Jewish populations demonstrated a much lower 3% frequency of these mutations in cases and <1% in controls [184]. Overall, these data indicate that a single heterozygous mutation in GBA escalates the risk for PD ~5 fold, while remaining inadequate to cause Gaucher’s disease. These two variants have also been linked to risk for dementia with Lewy bodies and PD with dementia [185].

The function of GBA is well established as an integral enzyme required for the breakdown of glucocerebroside to ceramide. After the association of GBA with PD, Mazzulli and colleagues showed that the knockdown of GBA in cortical neurons leads to the accumulation of α-synuclein. They further demonstrated an increase in α-synuclein in Gaucher patient iPS-derived dopaminergic neurons; however, they did not see significant accumulation of huntingtin or tau, indicating that GBA has a preference for α-synuclein [186]. Multiple studies have shown that glucocerebrosidase activity is lower in patients carrying GBA mutations and in sporadic PD cases, suggesting a broader role for GBA in pathogenesis [187, 188].

Identifying Risk Through Genome Wide Association

GWA studies have been applied as a means of identifying risk loci since the first successful published GWA study in 2005, identifying CFH (complement factor H) polymorphisms as a significant risk factor for age related macular degeneration [189]. PD is no exception to this trend, with a long history of investigation using GWA [2, 190–197]. Early GWA efforts in PD failed to convincingly identify risk loci as these studies were, like many other studies at the time, of low power, only examining ~300 cases [192, 198]. However, in 2009 two collaborative studies examining Caucasian and Asian subjects were the first to reveal genome wide significant risk alleles for PD [195, 196]. The Caucasian study identified risk loci at SNCA and MAPT (encoding microtubule associated protein tau) and provided supporting evidence for association at LRRK2 and PARK16, a locus in a large LD block that includes NUCKS1 (nuclear casein kinase and cyclin-dependent kinase substrate 1), RAB7L1 (RAB7, member RAS oncogene family-like 1) and SLC41A1 (solute carrier family 41 (magnesium transporter), member 1) [196]. The study in Asian subjects revealed association at SNCA, LRRK2, PARK16, and BST1 (bone marrow stromal cell antigen 1) [195].

Over the next several years these loci were replicated and two additional risk loci were nominated at HLA-DRB5 (major histocompatibility complex class II, DR beta 5) and GAK (cyclin G associated kinase) [193, 194]. A clear trend then followed, which was the use of meta analysis of extant data sets to provide greater resolution in the identification of risk loci. The most recent work identified or confirmed 28 independent disease associated risk loci (table 2) [199]. Notably, this includes GCH1 (encoding the enzyme GTP cyclohydrolase 1) that catalyzes the first step in the synthesis of tetrahydrobiopterin. This chemical is in turn a required cofactor for tyrosine hydroxylase, the rate-limiting enzyme in dopamine synthesis.

Table 2.

Risk loci for PD identified through GWA. Loci in grey indicate a second risk allele that confers an effect independent of a primary risk allele at the same locus.

SNP	Chr	Base Pair	Nominated Gene	OR
rs35749011	1	155,135,036	GBA/SYT11	1.824
rs114138760	1	154,898,185	GBA/SYT11	1.586
rs823118	1	205,723,572	RAB7L1/NUCKS1	1.122
rs10797576	1	232,664,611	SIPA1L2	1.131
rs6430538	2	135,539,967	ACMSD/TMEM163	0.875
rs1474055	2	169,110,394	STK39	1.214
rs12637471	3	182,762,437	MCCC1	0.842
rs34311866	4	951,947	TMEM175/GAK/DGKQ	0.786
rs34884217	4	944,210	TMEM175/GAK/DGKQ	1.105
rs11724635	4	15,737,101	BST1	1.126
rs6812193	4	77,198,986	FAM47E/SCARB2	0.907
rs356182	4	90,626,111	SNCA	0.76
rs7681154	4	90,763,703	SNCA	0.934
rs9275326	6	32,666,660	HLA-DQB1	0.826
rs13201101	6	32,343,604	HLA-DQB1	1.217
rs199347	7	23,293,746	GPNMB	1.11
rs591323	8	16,697,091	FGF20	0.916
rs117896735	10	121,536,327	INPP5F	1.624
rs329648	11	133,765,367	MIR4697	1.105
rs76904798	12	40,614,434	LRRK2	1.155
rs11060180	12	123,303,586	CCDC62	1.105
rs11158026	14	55,348,869	GCH1	0.904
rs2414739	15	61,994,134	VPS13C	1.113
rs14235	16	31,121,793	BCKDK/STX1B	1.103
rs11868035	17	17,715,101	SREBF/RAI1	0.939
rs17649553	17	43,994,648	MAPT	0.769
rs12456492	18	40,673,380	RIT2	0.904
rs8118008	20	3,168,166	DDRGK1	1.111

In the context of PD there are loci which appear to contain more than one risk allele and this variation in risk extends not only to the identification of independent risk alleles, but also to different types of risk allele. The loci containing SNCA, HLA-DRB5, GAK, and SYT11 (synaptotagmin 11) each contain two identified common risk alleles that act independently of one another [199]. In addition, the SNCA and LRRK2 loci contain multiple types of risk allele. Both are genes that contain missense mutations that cause disease and both also contain non-coding variability that imparts moderate risk for disease, presumably through modulating expression levels. In addition, LRRK2 contains common protein coding changes that increase risk for disease a moderate amount and SNCA gene multiplications can be causal for PD. The notion of multiple types of risk allele at the same locus, or the pleomorphic risk locus hypothesis, suggests that genes containing mutations that cause disease are involved in the pathogenesis of typical, apparently sporadic disease, presumably through an overlapping mechanism [200].

As with GWA detected loci in other complex diseases, the effect sizes for each of these loci are individually modest [2, 190–197] (table 2). Notably however, the risk conferred by these alleles, when summed in an individual can be considerable.

The functional and mechanistic underpinnings of these risk associations remain to be unraveled; however, early reports are beginning to appear in the literature. For example, mass spectrometry analysis has shown that the risk loci GAK and RAB7L1 interact with LRRK2 implicating these risk loci in an already known PD pathway [66]. Additionally, MAPT aggregates have been reported to be present in both neurons and glia of patients with various forms of parkinsonism [201]. Genetics has given us a great starting point to target future studies of PD; however, there is much work to be done to truly understand the molecular mechanisms and present therapeutic strategies.

The Future

Substantial progress has been made in understanding the genetics of PD. Highly penetrant mutations producing rare, monogenic forms of the disease have been discovered in singular genes such as SNCA, LRRK2, VPS35, Parkin, PINK1, and DJ-1. Unique variants with incomplete penetrance in LRRK2 and GBA have been shown to be strong risk factors for PD in certain populations. In addition, over 20 common variants with small effect sizes have been shown to modulate the risk for PD. Whole-genome approaches have clearly aided our quest to comprehend the role genetics plays in the pathogenesis of PD. We now detect Mendelian mutations segregating in families in a straightforward manner and we also have the ability to assess how common variability may play a role in disease. Investigating Mendelian forms of PD has provided precious insight into the pathophysiology that also underlies the more common idiopathic/sporadic form. However, Mendelian forms of disease are rare and as yet no treatment methodologies have developed. Furthermore, for identified common risk alleles, the functional basis underlying risk principally remains unknown. Since identified PD mutations and risk variants explain only a small percentage of disease burden, additional genetic determinants of PD remain to be discovered. In fact, it is anticipated that future GWA studies in complex diseases such as PD will identify an increasingly larger number of risk alleles with correspondingly smaller individual contributions to disease heritability [202, 203]. Likewise, forthcoming results from sequencing studies should further expand our ability to predict PD through genetic contributions [204]. As the genomics of PD rapidly advance, there is anticipation that full resolution of the genetic architecture underlying disease will help stimulate the design of treatments that may alter the course of the disease.

To this end, assessing the direct, biological consequences of known risk variants and incorporating those biological measures into a clinically useful (discriminatory) genetic risk model is crucial. One approach has been to integrate risk alleles with a quantitative trait, such as gene expression and/or DNA methylation levels, in a manner in which biological meaning can be derived from the association between the two datasets. Further investigation of individual loci, prioritizing those with the most biologically interesting signal/target is one route that may provide pertinent information. Additionally, as technologies such as proteomics, transcriptomics, whole-genome sequencing and chromatin immunoprecipitation provide insight into the functional role of non-coding DNA regions, interpreting and annotating these regions with reference to regulatory elements will support the recognition of plausible candidate susceptibility genes. In the short term, genetic risk models centered on functional variants rather than index tag SNPs may provide a track to more precise translation from genetic association into functional target.

However, a more comprehensive approach integrating further high content data to create high dimensional datasets is in order. Studying gene expression within gene networks where risk factors are operating and modeling transcriptional networks in iPS lines derived from patients are approaches that would integrate genetic, functional and clinical data. Furthermore, identification of biomarkers in blood and/or cerebrospinal fluid that could help classify individuals at risk for developing PD, before chief clinical symptoms and dopamine loss are present, would not only aid in early diagnosis but may allow tracking of disease progression and testing of treatments. At this point in time, open access to high-content, integrated datasets encompassing basic information from age and sex through clinical measures, such as MRI, DaT scan and olfactory tests all the way to molecular biomarkers and genetics will do the most to support the discovery and development of next generation therapies for PD.

Abbreviations

ARJP

Autosomal recessive juvenile parkinsonism

AR

Autosomal recessive

CDCV

Common disease common variant

CDRV

Common disease rare variant

COR

C terminus of ROC

DaT

Dopamine transporters

DNA

Deoxyribonucleic acid

EOPD

Early onset Parkinson’s disease

GWA

Genome wide association

iPS

induced pluripotent stem cells

LB

Lewy body

MRI

Magnetic resonance imaging

OR

Odds Ratio

PD

Parkinson’s disease

ROC

Ras of Complex proteins

SNP

Single nucleotide polymorphism

TGN

Trans-Golgi Network

Ubl

Ubiquitin like domain

Genes

SNCA

α-synuclein

LRRK2

Leucine-rich repeat kinase 2

VPS35

Vacuolar protein sorting 35

PINK1

PTEN induced kinase 1

PARK7

protein deglycase DJ-1

ATP13A2

ATPase type 13A2

PARK2

Parkin

FBXO7

F-box only protein 7

PLA2GB

phospholipase A2, group VI

14-3-3

tyrosine 3-monooxygenase/tryptophan 5-monooxygenase activation protein

BAG5

BCL2 associated athanogene 5

GAK

Cyclin G associated kinase

RAB7L1/ RAB29

Rab-7-like protein 1

CTSD

Cathepsin D

APOE

Apolipoprotein E

GBA

Glucocerebrocidase

CFH

Complement factor H

NUCKS1

Nuclear casein kinase and cyclin dependent kinase substrate 1

SLC41A1

solute carrier family 41 (magnesium transporter), member 1

BST1

Bone marrow stromal cell antigen 1

MAPT

Microtubule associated protein tau

HLA-DRB5

Major histocompatibility complex class II, DR beta 5

GCH1

GTP Cyclohydrolase 1

TH

Tyrosine hydroxylase

SYT11

Synaptotagmin 11