Shared cerebral metabolic pathology in non-transgenic animal models of Alzheimer's and Parkinson's disease

Parkinson’s disease (PD) and Alzheimer’s disease (AD) are the most common chronic neurodegenerative disorders. PD is characterized by classic motor dysfunction symptoms due largely to the loss of dopaminergic neurons in the substantia nigra pars compacta (SNpc) and intracytoplasmic aggregates of the presynaptic protein α-synuclein (α-Syn) in formations called Lewy bodies (Davie 2008). On the other hand, AD is characterized by cognitive decline due to neuronal loss in the hippocampus (HPC) and temporal cortex, associated with pathological accumulation of aggregates of amyloid beta protein (Aβ) which forms extracellular plaques, and hyperphosphorylated tau protein which forms intracellular neurofibrillary tangles (Ballard et al. 2011). However, in the late stage of PD, the majority of patients also manifest cognitive deficits and dementia (Caballol et al. 2007; Gratwicke et al. 2018), while motor symptoms which also include parkinsonism are present in the majority of patients with end stage AD (O'Keeffe et al. 1996; Ballard et al. 2011; Levin 2019). Despite different phenotypes, AD and PD share some common pathological features, e.g. oxidative stress and neuroinflammation (Han et al. 2018). Metabolic changes presented as insulin resistant brain state (IRBS) have recently gained more and more attraction as the pathophysiological core of AD (Hoyer 2004; Chen et al. 2014; Defelice and Ferreira 2002; de la Monte et al. 2014), particularly after the results of epidemiological studies demonstrating that type 2 diabetes mellitus (DM) is a risk factor for AD (Santiago and Potashkin 2013). However, some studies provided evidence that IRBS is also occurring in PD patients (Athauda and Foltynie 2016; Yang et al. 2018) and that DM is a risk factor for PD as well (Biosa et al. 2018). Consequently, clinical trials have been designed to test the therapeutic potential of antidiabetic drugs, in particular intranasal insulin (Claxton et al. 2015) and incretins (glucagon-like peptide-1 [GLP-1] analogues) in AD and PD patients (Athauda and Foltynie 2016; Femminella and Edison 2014). AD and PD seem like two streets going in parallel to the same direction called neurodegeneration with some overlap in their paths, and the nature of this overlap and the point at which it starts in the course of their pathophysiology seem to be the crucial questions.

The pathology in both AD and PD starts years before the first clinical symptoms and is difficult to trace in living humans, which is why animal PD and AD models are valuable tools. Although one should keep in mind that none of the animal models is able to fully recapitulate PD and AD pathology and symptoms, they are currently the only tool available to study the underlying pathophysiology mechanisms of these two diseases. Based on the involvement of gene manipulation in designing of the models, they can be classified in transgenic (genetic) and non-transgenic models generated by injection of different chemical, mainly neurotoxic compounds. Each group has advantages and disadvantages, but genetic models generally represent the early onset forms of familial AD and PD, while the prevailing majority of AD and PD patients suffer from the sporadic and idiopathic forms of these diseases, respectively, for which non-transgenic animal models are more appropriate. This review aims to point out the similarities and differences in IRBS and related cerebral metabolic pathology between the non-transgenic PD and AD animal models. The representative non-transgenic AD and PD models will be compared from a perspective of similarities in the intracellular mechanism and target(s) of the toxicity of the compounds used to construct the models as well as the principles of their selectivity toward targeted neurons and non-neuronal cells. Finally, the alterations in insulin brain system and related interaction between insulinergic and dopaminergic transmission, particularly regarding the time frame of AD/PD pathology appearance in correlation with cognitive and motor symptoms will be pointed out.

Chemically induced non-transgenic animal models of AD and PD play an important role in defining critical disease-related mechanisms. Peripheral or central application of different substances, sometimes even in different parts of the brain (e.g. by intracerebroventricular [icv] or intrastriatal injection), with different mechanisms of action, results in different clinical symptoms which mimic those predominant in AD or PD patients. However, despite all these differences, these models also point to the common underlying molecular mechanisms of neuronal dysfunction (e.g. neuroinflammation, mitochondrial dysfunction and oxidative stress) when neurons and surrounding microglia and astrocytes are in stress following damage induced first by injection trauma and consequently by toxin-induced chemical reactions. It is still not clear whether neuroinflammation is the driven force of neurodegenerative disorders or is it a consequence of the metabolic dysfunction occurring earlier in the progression of diseases (Yin 2016). Based on its chemical structure, streptozotocin (STZ) might play multiple roles as a nonspecific cytotoxic compound which causes mitochondrial dysfunction, oxidative stress and neuroinflammation, as well as a selective metabolic toxin which causes damage to insulin-producing/secreting and insulin receptor (IR)-expressing cells (Correia et al. 2013; Eleazu et al. 2013). On the other side, 6-hydroxydopamine (6-OHDA) also demonstrates nonspecific toxicity by causing oxidative and mitochondrial damage and neuroinflammation, as well as selective toxicity to the dopaminergic neurons (Walsh et al. 2011). Despite of that, both e.g. STZ/AD and 6-OHDA/PD models seem to share IRBS and other insulin-related pathological features to a certain extent.

At the molecular level, PD (6-OHDA/MPTP)- and AD (STZ/Aβ_1–42)-causing neurotoxins share mitochondria as the same intracellular target of their toxicity (Walsch et al. 2011), while their targets greatly differ at the cellular membrane; PD-causing neurotoxins are both linked to DAT (Storch et al. 2004), whereas AD-causing ones are either linked to GLUT2 and/or IR (STZ) (Gai et al. 2004; Knezovic et al. 2017), or have no known preferences for membrane targets (Aβ_1–42). These differences are expected to lead to different selectivity towards particular neurons or non-neuronal cells which express DAT, GLUT2 or IR. Although at first sight this might imply that 6-OHDA/MPTP on one side, and STZ and Aβ_1–42 each on other sides, will target different neurons, literature does not entirely support this assumption. On the contrary, it suggests that, not only some neurons seem to co-express PD- and AD-neurotoxin targets (Fieglewicz et al. 2003), but such targets co-expression is seen additionally in non-neuronal cells like astrocytes (Przedborski 2000; Arluison et al. 2004), and astrocytes may then consequently damage neighbouring neurons (Fig. 1). In this regard, it could be speculated that a difference lays in the order in which particular cell types are being damaged with three possible scenarios; (1) neurons are attacked first which consequently causes damage to the neighbouring astrocytes, (2) astrocytes are the primary targets which then injure the neighbouring neurons, (3) neurons and astrocytes can be independently damaged in parallel. Due to selective targeting of DAT at the neuronal membrane, the first scenario most likely applies to the effects induced by 6-OHDA and MPTP. The second and the third scenarios seem to apply to STZ- and Aβ_1–42-induced toxicity. These two compounds enter the neurons but also the astrocytes which predominately express targets (STZ/GLUT2, IR) at the membrane (Dwyer et al. 2002; Arluison et al. 2004), or the toxin (Aβ_1–42) has preference for the astrocyte’s pore-stimulated Ca²⁺ influx with consequent astrocyte mitochondrial damage (Verghese et al. 2013). Therefore, keeping in mind the astrocytes’ region-dependent density and vulnerability to injuries (Xu and Zhang 2011), as well as differently expressing membrane signalling structures (Bordey and Sontheimer 2000), this again provides plausible evidence that, besides the dose, the site of central STZ/Aβ_1–42 application plays a key role in determining the final toxic effect.

Since early changes have not been fully characterized in AD and PD models discussed here, and because there are other pathological processes going on in AD and PD that have not been taken into account here (e.g. apoptosis and autophagy disturbances, etc.), it is difficult to draw a conclusion on the order of dysfunction appearance between the insulin and dopamine signalling and the causal relationship between the two of them. Published data indicates that both insulin and dopamine signalling have been impaired in AD models during the first hour following neurotoxin administration; IR expression is decreased in the cortex and hypothalamus, and accompanied by increased GLUT2 expression in the hippocampus and hypothalamus of the STZ-icv model (Knezovic et al. 2017), while at the same time-point dopamine levels are found decreased in the substantia nigra in the Aβ model (Mukhin et al. 2019). Such a comparison is not possible in PD models since insulin signalling in the brain has not been explored earlier than 7 days following 6-OHDA/MPTP treatment (Wilcox et al. 1989). Additionally, from that time-point onward, impaired insulin and dopaminergic signalling in the brain (and in the striatum and hippocampus in particular) have been detected both in AD and PD models (Salkovic-Petrisic et al. 2006; Lester-Coll et al. 2006; Grünblatt et al. 2007; Morris et al. 2008; Deng et al. 2009; Agrawal et al. 2010; Shonesy et al. 2012; Lee et al. 2014; Barilar et al. 2015; Knezovic et al. 2015, 2017; Hu et al. 2018; Rabie et al. 2018; Gupta et al. 2018; Wang et al. 2018a, b; Nassar et al. 2018). On one side, there is decreased IR expression in astrocytes and neurons already 1 h after STZ application, and STZ-induced mitochondrial damage manifested during 24 h following STZ-icv treatment (Amiri et al. 2017; Knezovic et al. 2017). On the other side, insulin mRNA in adult rats cannot be detected within 2 weeks post-STZ-icv treatment (Barilar et al. 2015). Therefore, it seems likely that IR signalling is the primary pathological event following STZ-icv treatment, while a decrease in insulin synthesis comes as a secondary pathology, contributing to aggravation and progression of neurodegeneration. This hypothesis is in line with the findings of disturbed PI3K/AKT signalling pathway in neurodegeneration (as reviewed elsewhere, Rai et al. 2019). The impaired signal is further transduced to GSK3 enzyme involved in dysregulation of AD-linked Aβ homeostasis and tau hyperphosphorylation (Martinez and Perez 2013), but also in α-Syn-mediated neurodegeneration in PD (Yang et al. 2018). 6-OHDA- and MPP+ -induced neurodegeneration is associated with increased GSK3β activity also in in vitro PD models (Wu et al. 2007). All this strongly suggests that impairment in the IR signalling cascade with a consequent IRBS condition is, actually, not unique only to AD, but instead, can be considered a common underlying mechanism in neurodegenerative disorders as evidenced in non-transgenic PD and AD models (Fig. 1).

This hypothesis on the common IRBS role as a contributor to neurodegeneration in AD and PD condition strongly agrees with a recently proposed role of α-Syn in neurodegenerative disorders (Riederer et al. 2019). For a long time, α-Syn has been thought of as a specific pathological hallmark of PD. However, a growing body of evidence has emerged showing its co-localisation and synergistic effects with other proteins prone to pathological misfolding (e.g. Aβ and hyperphosphorylated tau protein) in neurodegenerative disorders other than PD, which finally led to a conclusion that α-Syn is more likely a non-specific bystander and a contributor to neurodegeneration in general, than a specific etiopathogenic factor of PD (Riederer et al. 2019).

It might be speculated that in the case of IR being the primary target of toxicity (e.g. like in the STZ-icv model) and due to the widely spread IR expression in the brain on neurons, astrocytes, endothelial and inflammatory cells, the IRBS condition and its consequences would dominate, while dopaminergic signalling dysfunction could be below the threshold for clinical manifestations, as supported by, in general, lack of significant (or only a mild) locomotor impairment in STZ-icv treated rats (Table 1, Fig. 1). Considering the important role of insulin in the brain in synaptic plasticity, learning and memory functions, such a predomination of IRBS might be reflected in cognitive impairment consistently seen in the STZ-icv model from the 2-week time point onward (Knezovic et al. 2015), and even at earlier time-points in the Aβ model (Table 1) (Schmid et al. 2017). Recent evidence suggests that the neurotoxic properties of Aβ and related cognitive impairment are mediated by oxidative stress, neuroinflammation but also by the disturbed PI3K/AKT/GSK3 signalling pathway (Morroni et al. 2016; Schmid et al. 2017; Amin et al. 2015). Additionally, decreased immunoreactivity for the vesicular acetylcholine transporter (a marker of cholinergic neurons) was observed already 1 week after Aβ fragments treatment in 6–7 weeks-old rats (Zussy et al. 2011) as well as cholinergic neuronal loss found 3 weeks following STZ-icv administration to rats (Majkutewicz et al. 2016), which could contribute to cognitive impairment in both models. Although dopamine release is acutely disturbed after Aβ fragments administration (Trabace et al. 2007), the impact of this seems not to be capable of triggering significant locomotor deficits, at least not in observational periods explored so far (Guerra de Souza et al. 2018).

Cerebral glucose metabolism, explored and found decreased already within the first post-treatment week in the STZ-icv model, and at > 1-week time-points in Aβ/AD and in both PD models (Table 1) (Chen et al. 2018a, b; Knezović et al. 2017; Babic Perhoc et al. 2019; Jang et al. 2012; Silva et al. 2013) certainly contributes to cognitive impairments seen in these models, as does the glucose hypometabolism observed in the brain of AD patients (Chen and Zong 2013). Furthermore, GLUT changes seen in these non-transgenic models are quite resembling to those found in AD/PD patients, e.g. decreased levels of GLUT3 were found in the brain of AD patients post mortem (An et al. 2018; Liu et al. 2009) accompanied by an increase in GLUT2 expression which most likely occurs in astrocytes (Liu et al. 2008).

Following administration of DAT-targeting neurotoxins, IR seems to be a collateral damage due to the evidence of IR co-expression with tyrosine hydroxylase, a marker for dopaminergic neurons (Figlewicz et al. 2003). Thus, dopaminergic loss-related signalling dysfunction would be predominant over IRBS, and evidently manifested by PD-like motoric symptoms (Table 1). However, the impact of disturbed IR signalling pathway seems to be strong enough to trigger certain cognitive impairments, as observed in 6-OHDA/MPTP models starting from 2 weeks after treatment, and onwards (Castro et al. 2013; Grospe et al. 2018).

Trans-disciplinary preclinical/clinical research indicates that insulin signalling influences dopamine availability and turnover (Kleinridders et al. 2015; Stouffer et al. 2015) and a recent post-mortem study in the brains of mentally ill patients provided evidence that the expression of dopamine signalling genes was mediated by insulin signalling genes (Mansur et al. 2019). Additionally, human data points to changes in dopaminergic system found in AD patients post-mortem; significantly lower levels of DA, its precursor and its metabolite were found in the striatum and additionally in the cingulate gyrus, amygdala and raphe nuclei (Storga et al. 1996). These reductions in DA level correlated well with the duration of AD (Pinessi et al. 1987), and decrease in DAT expression (Allard et al. 1990). Furthermore, the number of striatal dopaminergic D1/D2 receptors was found decreased in AD patients (Pizzolato et al. 1996). In line with that, it is not unexpected that AD patients develop also symptoms of motor dysfunction (O'Keeffe et al. 1996; Albers et al. 2015). It seems likely that the reasons for motor dysfunction being generally not observed in STZ/Aβ animal models of sporadic AD, could be related to a too short post-treatment observational period, and to the lack of gene-environmental interactions that have been suggested to play a role in AD and PD onset, development and progression (Landrigan et al. 2005).

Although a growing body of evidence suggests that the prevailing sporadic AD and idiopathic PD are fundamentally metabolic diseases with characteristic neurodegenerative processes possibly being caused by IRBS and metabolic dysfunction (Hoyer 2004; Defelice and Ferreira 2002; de la Monte et al. 2014; Chen et al. 2017; Bloom et al. 2018; Athauda et al. 2016; Schelp et al. 2017; Yang et al. 2018); the cause(s) and characterisation of these metabolic alterations are still unknown. One way to understand the metabolic etiopathogenesis of sporadic/idiopathic AD and PD is to explore the onset, development and time-course of insulin signalling dysfunction and glucose metabolism in the brain of animal models that mimic these human diseases. These models are generated by administration of particular neurotoxins, some of which could be considered as environmental toxins (e.g. MPTP as a recreational drug (Lau and Meredith 2003), and STZ as chemically related to nitrosamine-based food preservative (Tong et al. 2009)), and thus, more likely to mimic real-life exposure which might be associated with the increased risk for developing sporadic/idiopathic forms of AD and PD. In line with that, it is of utmost importance to detect toxicity targets of these neurotoxins at the molecular, cellular and tissue/regional level and assess where and to which extent do they overlap.

Sharing of the molecular mechanisms (mitochondrial damage) and co-expression of the molecular targets (IR, GLUT2, DAT) on membranes of particular neurons as well as astrocytes might be the underlying mechanism for insulin-related metabolic alterations present in both AD and PD non-transgenic models. Dysfunction in IR and its downstream PI3K/AKT/GSK3 signalling pathway seems to be a key pathology shared by the models (Table 1), occurring either as a primary toxic event, and, thus, expressed to a larger extent with consequently severe cognitive deficits in non-transgenic AD models, or as a collateral damage, expressed to a lesser extent and leading to less dominant cognitive impairment in non-transgenic PD models (Fig. 1). Based on that, IRBS could be considered a common metabolic pathology in the representative non-transgenic animal models of both AD and PD which, instead of being a cause, seems to be a common contributor to neurodegenerative processes. Further research is needed to provide in depth characterization of the brain insulin and dopamine signalling interaction in particular models at the molecular, cellular and regional level, for more successful animal-to-human data translation.