Parkinson’s Disease (PD) is the second most common neurodegenerative disorder after Alzheimer’s Disease, affecting 0.3% of the population in industrialized countries – 1% of the population over 60 years old [4, 5], and its growth surpasses that of Alzheimer's disease [6]. PD is caused by the dysfunction and/or degeneration of dopaminergic neurons (DN), resulting in difficulty with both motor and non-motor function. PD is a progressive disorder, typically considered idiopathic, with some known genetic and toxicological factors, including common herbicides/fungicides such as paraquat and maneb [7-9], and trichloroethylene (TCE) [10]. Current drug therapies act to suppress PD symptoms, however, significant cellular damage may have already occurred once motor deficits are evident and no neuroprotectant therapies are currently available [11, 12]. In addition, many of these drugs (such as Pramipexole, Ropinrole, Rotigotine and Apomorphine) operate by increasing dopamine signaling in the body and may have lessening effectiveness over time. Not all clinical manifestations of PD are thought to be due to DN degeneration [13]. This may explain recent data demonstrating why simply targeting a single neurotransmitter may eventually lose its therapeutic efficacy [14].

The initial PD models were pharmacologically induced. Although these models did not encompass the neurodegenerative aspects of the disease, their use led to a major breakthrough in 1957 when L-DOPA administrations were noted to lessen PD-like movements in reserpine-treated mice [15]. Despite the lack of PD neuropathology, experiments using reserpine-treated mice as PD models have proven to be predictive of clinical success; all drugs currently on the market were successful in alleviating symptoms in these mice [16]. While this serves to underscore how a predictive model can be useful for screening therapies aimed at alleviating disease-associated symptoms, these models may be less useful for determining curative therapies as the disease mechanism may be dissimilar.

One of the most widely used pharmacological agents to create rodent PD models is 6-hydroxydopamine (6-OHDA) [17-19] ; it has been utilized in nonhuman primates as well but to a lesser degree [20, 21]. 6-OHDA is useful in that it provides a way to create an array of lesion sizes that are dose-dependent. This methodology allows for partial lesions mimicking early stage PD to much larger lesioned areas (greater than 90%). One major drawback of this approach is that 6-OHDA does not cross the blood-brain barrier and must be physically injected into the brain, increasing experimental variables. For example, animals with partial lesions (50–70%) demonstrated motor function deficits in some studies while in other experiments, the lesion had to be greater than 80% [22, 23]. Additionally, behavioral test outcomes tend to be highly variable and dependent on the chosen screening methodology [16]. Another problem these animals share with most other PD models is that they do not form the pathological PD hallmark, Lewy bodies. Classically, Lewy bodies are eosinophilic inclusions that contain ubiquitinated proteins such as α-synuclein [24]. Different conformational strains of α-synuclein are proposed to underscore various synucleinopathies like Parkinson’s disease, dementia with Lewy bodies, and multiple system atrophy [25]. Protein misfolding cyclic amplification (PMCA) assay for α-synuclein is being considered to discriminate them [25]. The exact role of Lewy bodies is unknown but as the aggregation of these bodies appears to be a critical step in the development of the disease, it is a major focus for PD therapeutics [26]. Other pesticide-induced models, such as rotenone, are problematic in that they require surgical exposure or have issues with systemic toxicity and resistance to treatment [27]. Nevertheless, linagliptin, a dipeptidyl peptidase-4 (DPP-4) inhibitor, were evaluated in a rotenone-induced rat model of PD [28]. Linagliptin impeded rotenone-induced motor disorders and histological changes and significantly suppressed the activity of pro-inflammatory cytokines, such as tumor necrosis factor-α and interleukin-6 [28]. Dopamine-receptor antagonists haloperidol or SCH23390 have been used to model PD [29].

Another common agent used for inducing both rodent and primate PD models is 1-methyl-4-phenyl-1,2,3,6-tetrahydropyridine (MPTP) [30]. Interestingly, it was discovered that MPTP led to chronic PD in humans after several drug users accidentally exposed themselves to the compound in the early eighties [31, 32]. In contrast, usage in rodents comes with certain caveats. For unknown reasons, rats appear to be insensitive to MPTP [33, 34]. In mice, only certain strains are sensitive to MPTP [35, 36]. Moreover, factors such as mouse supplier, gender, age, weight and dosing amounts/schedule can influence lesion reproducibility [30, 35-37].

Figure 1. L-DOPA treatment increases dopaminergic cells in a monkey model of Parkinson’s Disease. Coronal sections of adult male cynomolgus monkey were stained for tyrosine hydroxylase (TH – brown; Nissl - blue) expression. A) Control MPTP treated monkey, B) MPTP monkey treated with vehicle, C) MPTP monkey additionally treated with L-DOPA. Note the increase in TH+ cells in panel C. From [1]

Thus far, five different primates have been utilized for PD research: African green, rhesus, cynomolgus, squirrel and marmoset monkeys [38]. MPTP causes a much more human-like PD syndrome in primates, although they also lack Lewy body development [39]. There is some species specificity as only African green monkeys appear to exhibit the typical PD rest tremor [40, 41]. As with mice, results may be variable due to the fact that researchers obtain their animals from a wide range of suppliers/countries [38]. However, despite these difference, PD primate models appear to be highly predictive of clinical trial success [42], mirroring some of the morphological changes that occur in humans. For example, L-DOPA appears to increase the number of tyrosine hydroxylase (TH) staining neurons — neurons which are important for dopamine synthesis — similarly in both PD humans and MPTP-treated primates (Fig 1) [1]. It should be noted that this synchronicity is not always the case. Factors that were demonstrated to be protective against 6-OHDA treatment in rodents, shown to alleviate neuronal loss in MPTP-treated primates and had initial Phase I clinical trial success, failed to show any efficacy in Phase II double-blind studies [43-46]. These results merely underscore that all animal models, regardless of how predictive they may appear to be, are simply best approximations for what may occur in humans. Another complicating issue is that gut microbes can metabolize L-DOPA, thus render any effects heterogeneous across Parkinson’s patients [47].

Figure 2. Knockdown of PINK1 leads to ommatidia and DN degeneration in Drosophila. Representative images of external eye phenotypes of 3-day-old flies expressing (A), control (GMR-GAL4 driver alone) and (B), PINK1 knockdown (GMR-GAL4/UAS-dPINK1 RNAi). A’ and B’ are corresponding images of DN (as identified by anti-dTH staining of 10-day old flies. This data suggests that PINK1 plays an important role in DN survival. From [2]

All the aforementioned models have been utilized to isolate PD therapeutics that alleviate disease symptoms but as previously mentioned, drug discovery with mammalian models can occur very slowly. Toward that end, PD animal models more amenable to high-throughput drug screening have been designed. For example, researchers have created symptomatic models in flies (rotenone) and worms (MPTP) [48, 49]. MPTP decreases dopamine content in both larval and adult zebrafish with corresponding decreases in locomotor activity [50-53]. To date, results from high-throughput drug screens using these types of models have not been published.

All previously mentioned PD models can be considered “cellular” models of PD with face validity as they mimic the disease symptoms but are not an accurate depiction of the underlying mechanism. As such, many researchers are turning to genetic models of the disease, incorporating mutated genes that have been linked PD. Multiple genes have been linked to both the familial and sporadic forms of PD: SNCA, lrrk2, Parkin, PINK1, DJ1, ATP13A2, PLA2G6, FBXO7, UCHL1, GIGYF2, HTRA2, EIF4G, GBA, MAPT, BST1, PARK16, GAK, and HLA [54]. Thus far, less than 20% of all PD have been linked to a genetic cause. As seen with the pharmacological PD models, not all genetic PD models express the typical hallmarks, e.g., Lewy bodies [55, 56]. Mutations of a mitochondrial serine protease HTRA2 cause essential tremor commonly seen in Parkinson patients [57].

Only two of the genes listed above (SNCA and lrrk2) have been clearly identified as autosomal dominant and therefore have been studied the most [58-61]. LRRK2 may also be involved in idiopathic PD [62]. Three clear autosomal recessive PD genes have been identified (Parkin, PINK1 and DJ1) with corresponding vertebrate and invertebrate animal models generated (for example, see Fig 2) [54].

Figure 3. Lack of a functional lrrk2 causes loss of DN in zebrafish. Zebrafish expressing morphants targeting the WD40 domain of the lrrk2 gene show decreased TH and DAT (dopamine transporter) gene expression. From [3]

A great deal of research has focused on the normal role of SNCA (α-synuclein) but it is difficult to ascertain its specific role given the disparate results from the various animal models. In mice, the knock-out data does not suggest that SNCA plays a crucial role in the development or maintenance of DN [63]. In addition, neither the multiple mice nor rat SNCA models demonstrate a progressive loss of DN [64]. There are conflicting reports in Drosophila as to whether overexpression of wildtype and/or mutant SNCA causes DN loss [65-67]. In worms overexpressing exogenous SNCA mutations, a food-sensing locomotor defect was observed which the researchers correlated to a loss of DN [68]. However, it can be somewhat difficult to interpret data resulting from either fly or worm experiments as neither normally express SNCA. Currently, there are no published reports of zebrafish expressing mutant SCNA (zebrafish do not express an SCNA ortholog), however, overexpression of human wildtype SCNA does cause neurotoxicity in zebrafish [69, 70]. It does appear, however, that SNCA plays a role in regards to dopamine homeostasis as exogenous SNCA was able to rescue mutant zebrafish lacking other synuclein paralogs [69]. Although no large scale drug screening tests have been published using a mutant SNCA model, a small scale drug study was conducted using flies expressing exogenous wildtype SNCA [71]. In zebrafish, a similar model was used to determine the efficacy of a putative therapeutic that would inhibit the formation of Lewy bodies [70].

As with SNCA knockout mice, DN cells appear to develop normally in mice lacking lrrk2 [55, 72] ; these mice do demonstrate slight differences in certain behavioral tests [55]. As seen with the mice, knockout models of an lrrk2 ortholog in worms do not affect overall neuronal development/survival, however, its effects in flies appear to be a bit more complex [73-75]. In regards to PD relevant lrrk2 mutants, a common PD-linked mutant form (G2019S) is toxic in certain cell lines [76, 77]. In contrast, mice overexpressing the same mutation develop normally but have only a slight age-related decrease in overall physiological dopamine concentration, unrelated to any obvious neuronal degeneration [78]. In addition, this decrease does not translate into the altered motor function in mice less than one year old. Similar results are observed in flies and worms incorporating the same mutation [79, 80]. However, the mouse researchers found this to be consistent with the clinical data as the percentage of G2019S carriers that eventually develop PD initially is fairly low (only 28% by age 59) but increases by the age of 69 (51%). The extension of this argument would then be that dopamine levels would decrease or neuronal degeneration would increase if the mice were allowed to mature further. Overexpression of similar pathologically linked lrrk2 mutants have not been introduced into zebrafish as of yet, however, fish which have had lrrk2 function blocked did show significant DN death and locomotor defects (Fig 5 [3] ). Lrrk2 mutant model has been used to evaluate the effect of LRRK2 inhibitor DNL201 on LRRK2 functions [81]. The effects of erucic acid, a peroxisome proliferator-activated receptors (PPARs) ligand, were evaluated in a PD zebrafish model [82]. Erucic acid treatment repaired the changes in the expression of proteins linked to cytoskeletal organisation, transport and localisation and improved locomotor functions and oxidant-damage in central nervous system and intestines [82].

Viral transduction of rodent brain with α-synuclein using a recombinant adeno-associated virus (rAAV) has been used extensively to model PD [83]. rAAV expressing either wild-type or mutant α-synuclein is able to transduce the dopaminergic cells of the substantia nigra. This model has been used a number of studies (see Table 1 of [83] for a comprehensive list). This model is attractive as it enables the overexpression of α-synuclein in the cells that are predominantly affected in the human disease. Furthermore, α-synuclein can be expressed readily in various wild-type, transgenic and knockout animals thus enabling an exploration of the effects of genetic background on PD pathology. The viral transduction of adult animal brain enables researchers to eliminate developmental effects that may complicate the analysis of transgenic and knockout models of PD. Different studies using the rAAV model have produced different phenotypic outcomes but common features are a progressive loss of striatal neurons and significant neuroinflammation (see [83] for a detailed discussion of the different phenotypes observed in various studies). By contrast, the rAAV model generates aggregates that do not fully recapitulate the structural and morphological features of the Lewy bodies seen in human PD [83, 84]. An important technical consideration of this model is that it requires injection of rAAV into the appropriate region of the brain. Whilst technically challenging, it does permit the targeting of specific, disease-relevant, brain regions.

Some groups have also used lentiviral vectors, rather than AAV, to drive expression of α-synuclein. Injection of recombinant lentivirus into the substantia nigra of mice resulted in the loss of dopaminergic neurons and motor dysfunction [85]. For a recent review of the use of viral vectors in PD research see [86].

Another PD model involves the injection into the rodent brain of pre-formed α-syn fibrils [87-89]. For example, researchers from Ionis Pharmaceuticals and Biogen tested the effectiveness of antisense oligonucleotides against SNCA in the reduction of SNCA in rodent preformed fibril models [88]. The fibril injection model leads to the formation of α-syn aggregates that resemble the Lewy bodies and Lewy neurites observed in PD patients [83, 90]. H Park et al showed that pathologic α-synuclein induced neurodegeneration through parthanatos-associated apoptosis-inducing factor nuclease (PAAN) activity [89]. Genetic depletion of PAAN prevented the degeneration of dopaminergic neurons and behavioral disorders [89]. PAANIB-1, a PAAN nuclease inhibitor, suppressed neurodegeneration induced by α-synuclein in vivo [89]. As with the rAAV model, the α-syn fibril model requires material to be injected into the brains of the experimental animals (with the same advantages and disadvantages as discussed above for the rAAV model).

There has been much interest in the mechanisms of propagation of PD within the brain. It appears that the mechanism of α-synucleinopathies such as PD may have much in common with prion-based diseases [91-93]. Models have been developed to investigate this propagation within the brain. In essence, rodents overexpressing human α-synuclein (either transgenically or via AAV transduction) receive a graft of neurons which are subsequently analysed for the presence of human α-synuclein [93]. E Martin-Lopez et al observed pathological changes and functional disorders in the olfactory system in transgenic α-syn-Tg mice expressing the human A30P mutant α-synuclein [94]. The transgenic animals developed severe α-syn pathology in the projection neurons of the olfactory pathway. Furthermore. neurogenesis in α-syn-Tg mice was reduced, and proteomic analysis showed synaptic endocytosis and exocytosis defects in the olfactory bulb [94]. In another version of this model, the brains of rats overexpressing human Htt via AAV transduction received a graft of embryonic dopaminergic neurones; endocytic uptake of human α-synuclein by the grafted neurones was observed which subsequently played a role in seeding the aggregation of endogenous rat α-synuclein [95]. This recapitulates the occurrence of α-synuclein aggregates in neurones transplanted into the brains of PD patients [96, 97]. Kam TI et al reported that pathologic α-syn could activate poly(adenosine 5'-diphosphate-ribose) (PAR) polymerase-1 (PARP-1), and PAR, in turn, could accelerate the formation of pathologic α-syn [98].

A recent study hypothesized that the tyrosine nonreceptor kinase-2 (TNK2) gene and systemic RNA interference protein-3 (SID-3), which is the sole TNK2 ortholog in C. elegans, are involved in dopaminergic and epigenetic signaling regulating neuronal homeostasis [99]. CRISPR-edited nematodes, endogenously expressing SID-3 analogous to TNK2 PD-associated single-nucleotide polymorphisms, showed dopaminergic neurodegeneration and SID-3 dysfunction [99].

Researchers often use mutliple models to address the underlying questions. For example, Fanning S et al applied yeast genetics, rat cortical neurons, worms, iPSC-derived neurons, neurons from PD patients and a mouse model to study the involvement of stearoyl CoA desaturase in synuclein neurotoxicity [100].

This article is derived from an earlier article "Animal Models for Huntington's and Parkinson's Diseases", and first appeared as a separate article in November 2019.