Models of α-synuclein aggregation in Parkinson’s disease

SNCA

Several dominant mutations have been described in the gene encoding α-syn, SNCA, with varying penetrance. In 1997, the A53T missense mutation was reported in a large Italian family associated with familial PD [4]. Subsequently, two additional missense mutations in SNCA were also seen to be associated to familial PD, A30P [46] and E46K [47], although these point mutations are extremely rare. Duplicate and triplicate loci of SNCA of different sizes (from 0.4 to 4.5 Mb) have been seen to give rise to PD, which has been related to the overexpression of the wild-type protein [43],[48], influencing the age of onset and severity of the disorder [49]. Indeed, SNCA duplications have also been reported in apparently sporadic PD patients [50] and interestingly, patients carrying these mutations present a broad clinical phenotype, even within a given family, suggesting an influence of genetic modifiers [25],[51],[52]. In addition, a dinucleotide repeat polymorphism (Rep1) has been described in the promoter region, 10 Kb upstream of the start codon, which induces SNCA overexpression and that may account for increased risk in 3% of non-familiar PD [53],[54]. Moreover, specific haplotypes, primarily in the 3’ UTR region, may also be associated to sporadic PD [55],[56].

Although SNCA mutations are very rare, their identification led to the association of α-syn with LBs. This protein is predominantly localized in presynaptic nerve terminals [57],[58] and missense mutations may reduce its affinity for lipids, enhancing its propensity to adopt a β-sheet conformation and promoting self-assemble into oligomers and fibril formation [59]. In this sense, the A53T mutation of SNCA is associated with the predominantly neuritic aggregation of α-syn in the human brain [60],[61] and in a mouse model [62], as well as with severe motor impairments. Alternatively, SNCA overexpression may decrease the density of dopaminergic vesicles and of synaptic contacts [63]. Neuropathological diagnosis of PD requires both dopaminergic neurodegeneration in the SNpc and the presence of LBs, and as described previously, the severity of the clinical manifestation has been correlated with the broad distribution of LBs [31], which may be related to the number of alleles. Both triplication and point mutations in SNCA are associated with the formation of cortical and subcortical LBs, and a clinical diagnosis of PD with dementia [47],[64].

The SNCA gene and that encoding the microtubule-associated protein tau (MAPT) have consistently been associated in different populations of sporadic PD [65], mainly those of European origin [66]. However, such an associated was not identified in Japanese population [67], suggesting that population-specific differences may exist in the genetics of PD.

Leucine-rich repeat kinase 2 (LRRK2)

Point mutations in the LRRK2 gene are frequently associated to PD, and they are found in both late onset familial and sporadic PD. LRRK2 mutations have the highest prevalence rate in PD patients discovered to date, having been found in 10% of cases with autosomal dominant familial PD, in 3.6% of sporadic PD cases and even in 1.8% of healthy controls [44],[45],[68]. The presence of mutations in healthy controls may suggest a reduced and incomplete age-dependent penetrance. However, while over 80 missense variants have been identified, only seven mutations are considered pathogenic (N1437H, R1441G/C/H, Y1699C, G2019S and I2020T), most of them lying in the C-terminal half of the protein, while two others are considered as risk factors (G2385R and R1628P: [69]. Of these, G2019S is the most frequent mutation in the Caucasian population, which explains 1% of PD cases [70]. The prevalence of this mutation is strongly influenced by ethnicity and as such, it is more frequent in patients of Southern European or North African origin, and in Ashkenazi Jews (18-40% PD cases), yet it is very rare in Asians or Northern Europeans [40],[69]. The risk of PD for a person who inherits this mutation is 28% at age 59 years, 51% at 69 years and 74% at 79 years [71]. The G2019S mutation gives rise to a uniform clinical phenotype that resembles sporadic PD, both in homozygous and heterozygous carriers [71].

LRRK2 is a gene with 51 exons and it encodes a large protein with two different enzymatic activities in the same molecule: that of a kinase and a GTPase. LRRK2 also has multiple protein interaction domains suggesting that it may serve as a scaffold for the assembly of other proteins. The G2019S mutation affects the kinase domain of the protein, with other common mutations lying in the GTPase domain or in the protein interaction domains. The function of LRRK2 remains largely elusive but G2019S mutations enhance its kinase activity, which may mediate neural toxicity [72]. LRRK2 is widely expressed in the healthy adult brain, mainly in the endoplasmic reticulum. Interestingly, in sporadic PD patients LRRK2 is found in the core of LBs in the SN and locus coeruleus, suggesting it may contribute to LB formation [25]. Indeed, some PD patients with LRRK2 mutations have LB pathology [73] but this is not always the case, even within a family [40]. How LRRK2 might participate in LB formation remains unknown, although dysfunctional autophagy has been proposed [74], as well as altered solubility and aggregation of α-syn [75].

Recessive mutations

Homozygous or compound heterozygous mutations in the recessive genes Parkin, PINK and DJ-1 are associated to relatively rare forms of familial PD, and these result in early onset PD and nigral dopaminergic neuronal loss. Parkin mutations account for around 50% of familial juvenile and early onset PD, reaching 80% in patients with onset before the age of 20 and decreasing with increasing age at onset, becoming very rare when onset occurs after the age of 50 [42]. At least 170 mutations have been described in Parkin, including point mutations, exon rearrangements, “indels” and duplications. Homozygous mutations with loss-of-function predominantly induce neuronal loss in the SN and locus coeruleus, in the absence of LBs. However, compound heterozygous mutations with LB pathology have been described in a minority of patients [76]. This variability may indicate some mutation-specific effect, as it has been described loss-of-function, inactivating and activating mutations [77]. Parkin is a cytosolic E3 ubiquitin ligase that transfers ubiquitin to specific protein substrates for proteasomal and autophagic degradation. Mutations may impair Parkin activity, which might reduce the clearance and aggregation of proteins. As such, these mutations could participate in decreased α-synuclein clearance [78] or alternatively, Parkin mutations might decrease its solubility and lead to aggregate formation [79].

Other recessive mutations are less common, such as PINK-1[80] and DJ-1[81], producing clinical manifestations broadly similar to Parkin mutations and in the absence of LBs.

GBA mutations

Glucocerebrosidase (GBA) mutations were first described in Gaucher’s disease, a recessive lysosomal storage disorder that may develop into parkinsonism, and PD patients may have an increased frequency of GBA mutations [82]. Indeed, carriers of GBA mutations exhibit clinical features related to early-onset, levodopa-responsive PD, experiencing hallucinations and symptoms of cognitive decline or dementia, with the abundant presence of LBs in the neocortex [83]. GBA mutations in PD patients confer increased susceptibility to an earlier disease onset, to have affected relatives and to develop atypical clinical manifestations [84].

MPTP in mice

In mice, the MPTP model must be induced in specific strains for it to be a consistent model, such as C57 and Swiss Webster mice [99]. The loss of dopaminergic neurons depends on the administration regime (from acute to chronic), ranging from 60% to 90% [88]. Acute or sub-chronic administration does not lead to the formation of inclusions and while α-syn inclusions have been detected in the study of some chronic models, this is not always replicated. In this sense, α-syn-ir was detected in dopaminergic neurons when chronic MPTP treatment was coupled with probenecid administration to block the rapid clearance of MPTP and its derivates. On examination by electron microscopy, the neurons that accumulate α-syn appeared to contain lipid droplets or secondary lysosomes covered by proteins. Although this is quite a distinct morphology to the straight filaments detected in human LBs, it was suggested that lipofuscins may be important for the development of LBs [100],[101]. When MPTP was administered continuously using an osmotic minipump, ubiquitin-ir and α-syn-ir nigral inclusions formed, although their co-localization in the same cell or aggregate was not studied. Ultrastructurally, these inclusions appeared as concentric membranes containing α-syn that was transformed into fibrillar morphology [102]. However, other studies failed to find such inclusions following acute, semi-chronic or chronic MPTP treatment with probenecid or using osmotic minipumps [103]-[105]. Thus, further work is required before this model can be used to study aggregate formation.

Rotenone in rats

Rotenone is an insecticide that accumulates in the mitochondria, inhibiting complex I and promoting oxidative stress [106], as well as inhibiting the ubiquitin-proteasome system in vitro[107]. While this model is one of the most promising models for PD studies, it has several drawbacks since rotenone is not a selective dopaminergic neurotoxin, it produces high rates of mortality (~30%) and there is also strong variability in neurodegeneration (only observed in 50% of rats). Interestingly, peripheral treatment with rotenone induces the formation of α-syn inclusions, with a dense core and fibrillar surrounding that resembles those observed in PD [108],[109]. The aggregation of α-syn may be enhanced by neonatal lipopolysaccharide (LPS) treatment, suggesting some cooperation between perinatal inflammatory processes and exposure to this pesticide [110]. In vitro studies have shown that rotenone induces a conformational change in α-syn and that it accelerates the rate of fibril formation [111],[112]. Indeed, like the human pathology α-syn deposits are observed in the myenteric plexus [113]. In order to avoid peripheral rotenone toxicity and hence, the high mortality rate in rats, intracerebral administration of this toxin has been used, showing significant nigro-striatal neurodegeneration but not the appearance of α-syn inclusions [114]. Indeed, variability in the results obtained following peripheral administration (subcutaneous versus intravenous infusion) has limited the use of this model as a reliable tool to study PD [115],[116].

Combined administration of other herbicides and fungicides, such as paraquat and Maneb also produces a high mortality rate and modest nigro-striatal neurodegeneration without LB deposition [88].

Proteasome inhibitors

The discovery of mutations in parkin and ubiquitin carboxy terminal hydrolase-1 (UCH-L1) in familial PD focused attention on the ubiquitin-proteasome system, suggesting that the toxicity of α-syn was associated with misfolding of the protein, and impaired proteasomal and lysosomal degradation [117]. Indeed, the combined systemic administration of the proteasome inhibitors PSI/epoxomicin produces a loss of nigro-striatal neurons and progressive motor disabilities. Moreover, inclusion bodies were observed not only in the SN but also, in the locus coeruleus, raphe and substantia innominata [118]. However, this model has proved very difficult to reproduce and the data has not been replicated in mice, rats or primates, in which only a partial response to proteasomal inhibitors is obtained [88]. This variability frustrates the use of this model in PD research. More recently, conditional genetic deletion of a proteasomal subunit in mice that disrupts 26S proteasome degradation was shown to induce the loss of dopaminergic neurons and the appearance of inclusions in SN neurons that contain ubiquitin and α-syn. Although no biochemical studies of aggregation or double-immunolabeling have been performed to clarify the nature of these inclusions, the inclusions appear to contain mitochondria and double-membraned autophagolysomes at the ultrastructural level. However, there was no resemblance of the radially oriented, straight filaments observed in LBs [119]. Indeed, in this model the accumulation of aberrant mitochondria into inclusions seems to be independent of α-syn [120].

Mitochondrial dysfunction

A loss of mitochondrial complex I is observed in sporadic PD that may lead to selective dopaminergic neurodegeneration due to oxidative stress [121]. Midbrain dopaminergic neurons of PD patients and elderly humans carry high levels of somatic mtDNA mutations [122]. Indeed, neurotoxins like MPTP and rotenone inhibit this mitochondrial complex [123]. Transgenic mice overexpressing wild-type α-syn induces mitochondrial fragmentation, which may predispose to neural degeneration [124], and transgenic mice overexpressing human A53T α-syn mutation also display mitochondrial abnormalities that may explain some aspects of the aggregated α-syn toxicity [125]. Disruption of the respiratory chain has been modeled in mice by deleting the gene encoding the mitochondrial transcription factor A (TFAM) in dopaminergic neurons, thereby inhibiting the transcription of mtDNA genes. Although these mice have an interesting phenotype, with impaired motor function and neurodegeneration, α-syn is not needed to express this phenotype. Indeed, the inclusions found in dopaminergic neurons have no α-syn but rather, abnormal mitochondrial membranes [126].

α-syn transgenesis

Different transgenic models have been developed that overexpress wild type, A53T, A30P and truncated α-syn. However, despite initial results with mutated α-syn in rodents, they have failed to translate into truly effective transgenic models of PD [127]. Attempts to overexpress or to knock-out α-syn in rodents have produced a variety of pathological abnormalities, including aggregate formation. However, no clear loss of dopaminergic neurons is observed. On the other hand, viral transduction induces rapid neurodegeneration but it is limited to the region targeted [125]. The use of cell-type promoters, such as tyrosine hydroxylase, is a limitation to transgenic overexpression that prevents the induction of the broad brain pathology observed in humans. However, this localized overexpression of α-syn (or truncated forms) may be useful to address specific features of PD. A variety of transgenic models using different promoters that drive broad expression have induced the graded appearance of α-syn aggregations with no fibrillar morphology [125],[128],[129]. One particularly interesting model is the transgenic A53T α-syn under the control of the mouse prion promoter, which overexpresses the mutant protein in neurons. These mice develop progressive motor failure at 8 months of age leading to paralysis and death [62]. Inclusions have been detected in different brain areas like the SN, raphe, pons, pontine reticular nuclei, locus coeruleus and deep cerebellar nuclei. The biochemical characterization of these inclusions showed SDS-insoluble α-syn aggregation into dimers, trimers and multimers, and these LB-like inclusions have a filamentous structure like human LBs [62],[130]. However, as indicated, a major drawback of this model is that no dopaminergic neurodegeneration is detected in the SN. Conversely, there is neurodegeneration in areas not affected in PD, such as among motor neurons [127],[131],[132].

LRRK2 transgenesis

LRRK2 null mice have no overt dopaminergic deficit or clear pathology in the brain, although age-dependent renal atrophy is observed that is associated with the aggregation of α-syn in renal tubules of the aged kidney [133],[134]. The R1441C mutation in LRRK2 generated by a knock-in strategy in mice does not induce dopaminergic neurodegeneration but rather, DA neurotransmission and D2 receptor dysfunction. However, no obvious accumulation of α-syn is detected [135]. Mice overexpressing the G2019S mutation in LRRK2 do not reproduce any obvious gross neuropathological phenotypes nor is α-syn aggregation observed, although this LRRK2 mutation may accelerate the pathogenic phenotype of A53T α-syn mice [136]-[138].

Autophagy-lysosome system

Several studies have suggested a role for autophagy in PD. The autophagy-lysosome system is a catabolic pathway involved in protein and organelle degradation. Several types of autophagy have been described, including: microautophagy; macroautophagy, engulfing large structures; and chaperone-mediated autophagy that degrades only soluble proteins in a selective manner. Cellular homeostasis of autophagy is crucial to maintain the balance between healthy and unhealthy cells [139]. In PD and related LB diseases an accumulation of autophagosomes has been described, coupled with a reduction of lysosomal markers in nigral dopaminergic neurons, suggesting a defect in lysosome-mediated clearance of α-syn aggregates [140],[141]. Indeed, α-syn is physiologically degraded by both the ubiquitin-proteasome and the autophagy-lysosome system (even by the chaperone-mediated mechanism), although the mutant forms of α-syn appear to inhibit their own degradation. Hence, these pathways would appear to participate in LB formation [142],[143].

In mice, deletion of the gene essential for macroautophagy (Atg7) in dopaminergic neurons induces progressive moderate dopaminergic loss in the SN and striatal DA depletion at an age of 9 months, although most dopaminergic neurons are resistant to the long-term stress induced by impairing autophagy. These mice have ubiquitinated-SQSTM1 inclusions in dopaminergic neurons but no abnormal aggregation of α-syn [74]. In PD patients, the translocation to the nucleus of the TFEB transcription factor (a regulator of autophagy), is dampened in midbrain dopaminergic neurons and it co-localizes with α-syn in LBs. Using adeno-associated viral vectors to model α-syn transgenesis in rats, under the control of the synapsin-1 promoter and the WPRE enhancer, stimulation of TFEB and Beclin (an activator of autophagy) was shown to rescue nigral DA neurons from α-syn toxicity [144]. These results identify interesting new mechanisms, however no clear α-syn aggregates have yet been formed by manipulating the autophagy-lysosome system.

Neurotrophic factors: Smad3 deficiency

GDNF and its close relative Neurturin provide functional rescue of nigro-striatal dopaminergic neurons after 6-OHDA or MPTP treatment. These results led to clinical trials into the use of these molecules in patients with PD, although the use of exogenous GDNF in clinical trials produced inconclusive results. The neuroprotective effect of this neurotrophic factor may not be translated to the clinic as GDNF fails to protect against α-syn-induced toxicity, probably due to the blockade of GDNF signaling by α-syn [145]. Indeed, transgenesis of α-syn induced by viral overexpression in the nigro-striatum, induces dystrophic terminals in the striatum containing α-syn aggregates, which are not modified by exogenous administration of GDNF [146],[147].

Recently, intracellular TGF-β1 signaling has been implicated in different pathological events related to PD in a mouse model, including LB-like formation. In humans, TGF-β1 is up-regulated in the striatum and in the ventricular cerebrospinal fluid of patients with PD [148],[149]. Moreover, active TGF-β1 overexpression in the nigro-striatal system of MPTP-treated mice using adenoviral vectors produces poorer survival of dopaminergic neurons and higher levels of striatal DA depletion [150],[151]. However, since the effects of TGF-β1 are dose- and context-dependent [152], its overexpression may introduce a bias in studies on animal models. Smad3 deficiency, a molecule involved in the intracellular TGF-β1 signaling cascade, provides us with a new and interesting model of PD, as it promotes selective postnatal neurodegeneration of dopaminergic midbrain neurons, strong MAO-mediated catabolism of DA in the striatum and oxidative stress, as well as dampening the trophic and astrocytic support to dopaminergic neurons. Interestingly, Smad3 deficiency induces the formation of α-syn inclusions in selected brain areas, which accumulate with age in a progressive and gene dosage dependent manner. These α-syn inclusions appear in both the perikaryon (SN and paralemniscal nucleus) and in neurites (in the motor and cingulate cortices, striatum, corpus callosum and spinal cord). Indeed, these α-syn deposits are phosphorylated at Ser¹²⁹ and ubiquitinated, and they form a core/halo distribution that resembles the deposits observed in human LBs. In other brain areas α-syn expression is associated with an irregular morphology, increased α-syn-ir staining and neurite thickness (pontine nuclei, cerebellar white matter, cerebral peduncle, diencephalic nuclei and internal capsule). Moreover, α-syn inclusions are also detected in glial cells in the cerebellum and spinal cord. Biochemical analyses show the presence of detergent-insoluble dimers, trimers and oligomers of α-syn in the ventral midbrain, motor cortex and spinal cord [153].

There is currently increasing attention being paid to the non-motor symptoms of PD, such as cognitive impairment and behavioral disorders. The hippocampus is implicated in physiological learning and memory, as well as in the cognitive dysfunction seen in some PD patients where there is an interaction with the dopaminergic system [154]. The hippocampus of PD patients is also affected by the presence of LBs [31] and in mouse models, α-syn modulates adult neurogenesis in the dentate gyrus [155]. Indeed, the accumulation of extracellular α-syn oligomers impairs hippocampal LTP and it enhances basal synaptic transmission [156]. Both triplication and point mutations of SNCA are associated with cortical and subcortical LB accumulation, and such patients are clinically diagnosed as PD with dementia [47],[64]. These data suggest a role for the α-syn aggregates in the cognitive impairment observed in demented PD patients. Similarly, Smad3 deficient mice have α-syn inclusions in the neuronal layers of the hippocampus. Smad3 is strongly expressed in hippocampal neurons and Smad3 deficiency induces a strong decrease in adult neurogenesis in the dentate gyrus, inducing the apoptosis of early stage and highly proliferative intermediate precursor cells (IPCs). Indeed, Smad3 deficiency abolishes the induction of LTP in the dentate gyrus but not in the CA1, highlighting the specificity of this effect. Both neurogenesis and LTP induction in the adult hippocampus are two aspects of hippocampal brain plasticity related to learning and memory that decline with age, and as a result of neurological disorders [157]. Smad3 deficiency sheds light on a new interesting pathological mechanism and provides a new model of PD to be explored in which dopaminergic dysfunction, widespread α-syn inclusions and cognitive impairment co-exist.