Introduction
Alzheimer’s disease (AD) is a major neurodegenerative disorder often related to the deposition of amyloid β-peptide (Aβ) plaques in brain tissue followed by formation of neurofibrillary tangles (Murphy and Levine
2010) and is associated to symptoms such as memory loss, alterations in mood and behavior and have been associated with, dementia, disorientation and aphasia (Jahn
2013). Parkinson’s disease (PD) is the second most prevalent neurodegenerative disorder that causes death of dopaminergic neurons in substantia nigra pars compacta of the midbrain, which leads to the decline in the synthesis of dopamine (Mhyre et al.
2012). It is also characterised by a large number of motor and non-motor features and by the increase in incidence above the age of 65 (Mahlknecht et al.
2015). The clinical manifestations of PD include resting tremor, muscular rigidity, bradykinesia, depression and postural instability (DeMaagd and Philip
2015). According to the Alzheimer’s Association, National Institutes of Health in the United States of America spends $480 million on Alzheimer’s research compared to $3 billion on HIV/AIDS, $4 billion on heart disease and $6 billion on cancer. PD affects 1–2 per 1000 of the population with its prevalence increasing with age and affecting about 1% of the population above 60 years (Tysnes and Storstein
2017). Therefore, discovery and characterization of accurate biomarkers play a crucial role in disease prediction (Padmanabhan et al.
2017). Both, genetic and environmental factors, have been identified related to the risks of developing AD (Killin et al.
2016). If the condition can be detected earlier, effective treatment can help manage the condition although, often, fewer symptoms manifest in the early stages resulting in late detection of the disease. The biomarkers till date change with the factors and illness, and it may not be expressed all the time. One of the main medications given for PD patients is L-Dopa (Hardebo and Owman
1980). Although, several experimental models to analyse the disease pathology have been developed, there are aspects regarding the pathological cascade of both sporadic and familial conditions (Golde et al.
2013; Ferreira et al.
2015) that need to be addressed. A reason behind this is the limitations in conducting in vitro and in vivo experiments with neurons (Polikov et al.
2007) and the complexity in neuronal circuits. For example, in the case of AD, it is difficult to extract β amyloid aggregates from neural specimens, and previous reports indicate aggregates in saline cause toxicity in vitro (Esparza et al.
2016; Giorgetti et al.
2018). Although there is evidence that amyloid β oligomers contribute chronic neurological manifestations, they are difficult to be detected by conventional staining techniques (Ferreira et al.
2015).
There are some drugs available to manage the symptoms, despite decades of research, no treatment has been reported to completely halt the disease conditions yet (Padmanabhan et al.
2017). Since the disease mechanisms are poorly understood, especially in case of major proteins such as alpha synuclein (αS), amyloid β and tau that form fibrils, plaques and tangles, it is difficult to turn them off and patients are diagnosed late or remain undiagnosed (Bendor et al.
2013; Razzokov et al.
2019). Studies have also suggested the accumulation of Aβ and αS in the brain during the natural process of aging (Li et al.
2004). As these aggregated proteins lead to signal cross-talk within the brain, it may be associated to further signalling cascades which initiates onset of the disease. This includes several abnormal cell damage events such as mitochondrial dysfunction, oxidative stress, hyperphosphorylation of tau, increased neuro-inflammatory responses, decreased neuroplasticity and neurogenesis, neurodegeneration etc. (de JR de Paula et al.
2009). Recent studies have shown that during neuroinflammation, the release of TNFα by activated astrocytes and microglia was linked to AD and PD (Reddy and Seth
n.d. ; Olmos and Lladó
2014).
Inflammation in the brain may be promoted by invasion of pathogens, spinal cord injury, aggregation of misfolded proteins, tangles, plaques. The cellular insults activate microglia and initiate the production of proinflammatory cytokines to protect neurons from tissue damage initially, which results in neurodegeneration (Amor et al.
2010). Tumour necrosis factor alpha (TNF훼), a major proinflammatory cytokine, has been known to have an important role in neuroinflammation and glutamate mediated excitotoxicity related to AD and PD (A Frankola et al.
2011; Olmos and Lladó
2014). It has been reported that TNFα along with surface receptors, are present in healthy brain at low levels, and high levels in diseased states (Santello and Volterra
2012). Under normal physiological conditions, the production of excess TNFα is controlled by inhibiting activation of microglia (Wang et al.
2015). During inflammation, astrocytes and microglia become activated through initiation of proinflammatory triggers and cytokines release due to T cells infiltration (Liberman et al.
2019). Aβ and αS have been reported as key proteins that trigger neuroinflammation in AD and PD respectively (Tweedie et al.
2012). Amyloid plaques have been known to cause neurodegeneration due to the toxic effect of Aβ (Pasinetti and Hiller-Sturmhöfel
2008). Hyperphosphorylation of tau forms neurofibrillary tangles that again leads to intracellular lesions in the brain (de JR de Paula et al.
2009). In vitro and in vivo studies have shown that mutations in genes code of amyloid precursor protein, presenilin 1 and 2 trigger plaque formation (Żekanowski et al.
2003). In earlier studies, the involvement and presence of TNFα around Aβ plaques have been reported in the post-mortem brain tissue of both transgenic AD mice and AD conditions (Chang et al.
2017). A study had reported the role of neuroinflammation and TNFα signalling in the early stages of AD and its role in neurodegeneration through accumulation of plaques, neurofibrillary tangles and elevations of misfolded or mutated proteins/genes involved (Montgomery and Bowers
2012). Experiments on both animal and human PD brain tissues suggested that abnormal levels of TNFα released by high microglial activation can lead to dopaminergic cell death (Yiannopoulou and Papageorgiou
2013). Elevation in accumulation of αS aggregates activates microglia in turn increasing the production and release of excess TNFα (Zhang et al.
2018). Both Aβ plaques and αS aggregation can lead to elevated levels of TNFα that eventually results in the progression of AD or PD pathology (Decourt et al.
2016). Cellular and molecular changes implicate increased TNFα levels by microglial activation in both AD and PD, which suggests a commonness in TNFα signalling pathway in the progression of both these diseases (A Frankola et al.
2011). Several studies have indicated changes in oxidative stress due to increased level of reactive oxygen species (ROS) which was induced by TNFα signalling in AD and PD (Fischer and Maier
2015). In both AD and PD conditions, accumulation of abnormal αS and Aβ lead to oxidative stress that trigger the apoptotic pathway (Singh et al.
2019).
Brain’s insulin sensitivity has been studied (Blázquez et al.
2014) and insulin has been known to regulate cellular mechanisms inside the brain (Plum et al.
2005). Insulin has also been documented to regulate various brain functions such as neuroprotection, synaptic plasticity, memory, and reward recognition (Ferrario et al.
2018). The cellular links between insulin resistance and neurodegeneration in PD related pathological mechanisms have been previously discussed (Athauda and Foltynie
2016). Impaired insulin signalling pathway has also been identified as a critical pathological factor contributing to the development of AD (Hölscher
2014). Impaired insulin signalling has been known to be associated to the formation of plaques, tangles, increased oxidative stress, and important factors facilitating neurodegeneration in AD (Rad et al.
2018). Experimental models have shown the critical role of insulin signalling pathway in degradation of Aβ and αS and blocking them led to formation of toxic fibrils and plaques (Sharma and Singh
2016). Insulin signalling and neuroinflammation may be co-related, with an imbalance possibly inducing an elevation in inflammatory cytokines including TNFα as observed during AD and PD (Yang et al.
2018). There is a need to re-analyze the roles and relationships of TNFα signalling, activation of glial cells, oxidative stress, insulin resistance and accumulation of misfolded/mutated protein aggregates that are linked to each other, share common genes and possibly signalling pathways, that may lead to neurodegeneration in both AD and PD pathology.
Modelling disease related signalling pathways in silico helps in understanding the experimentally-relevant relationships between individual proteins, interactions and related perturbations which are crucial for analysing disease mechanisms and for mapping appropriate therapeutic targets (Vidal et al.
2011; Hao et al.
2018). Modelling complex biological pathway networks including their cellular and molecular components, and interactions (Ji et al.
2017) can help connect critical factors statistically relevant as common signaling mechanisms or phentotypic functions to both disorders. Developing computational models can aid reproducing disease pathways and predicting dynamical behaviours essential for approprite protocol design and experimental testing and to map clinical symptoms to molecular processes going through cellular and circuit functions (Conradi et al.
2007; Bartocci and Lió
2016). Using biochemical systems theory (BST), sub-cellular reactions and biochemical pathways were modeled using ordinary differential equations (ODE) for reconstructing signalling dynamics in this study (Savageau et al.
1987). All biochemical reactions involved in disease-related signalling pathways were expressed mathematically using ODE and rate equations were computed using computational tools (Bartocci and Lió
2016).
The objective of this modeling exercise was to map major genes or proteins involved in disease mechanism, the reactions affected by the mutation of these genes and the difference in reactions when compared with healthy controls, action ofpotential drugs. In literature, BST models on oxidative stress and inflammation in insulin resistance were already available for PD condition (Braatz and Coleman
2015). These models explore some of the important pathways involved in PD and the treatment options. Most of the initial conditions for the model parameters were assigned as relative values rather than real data. With the need to model crosstalk and critical networks relevant to neurodegeneration identified by more recent studies, we have incorporated the cross-talk between insulin resistance, oxidative stress and neuroinflammation related to TNFα signalling in normal, AD and PD conditions (Fallahi-Sichani et al.
2011; Sasidharakurup et al.
2020; Su and Wu
2020). The parameteric values relating to biological states and initial conditions for this model were manually extracted from literature on disease models. In a previous study, we had modelled the role of TNFα mediated glutamate excitotoxicity and neuroinflammation (Sasidharakurup et al.
2020) and the variations in TNFα levels during both healthy and diseased conditions were analyzed. Some autocrine loops co-involved in the activation of TNFα were also modeled to study how TNF훼 stimulates its own release. To extend the relationships between AD and PD, this present study focuses on developing a model of TNFα related pathways regulated by neuroinflammation, oxidative stress and insulin resistance during neurodegeneration. In addition, few of the crucial feedback loops involved in TNFα signalling and their emergent properties maintaining the disease condition needed to be incorporated. Aimed towards building a tool for designing experimental interventions and connecting to clinically relevant biomarkers, common cellular components found in both AD and PD conditions, that trigger TNFα signalling such as Aβ, αS, tau phosphorylation, calcium, glutamate etc. were modelled in this paper.
Discussion
Although the role of TNFα pathways were independently associated with neuroinflammation, oxidative stress and insulin signalling, the involvement of autocrine loops and interdependency of biochemical reactions and their correlations in disease onset and progression, as modeled in this study, are critical to understand AD, PD and neurodegeneration.
In this study, the contribution of cellular reactions involved in the pathways related to the neurodegeneration processes that leads to Alzheimer’s and Parkinson’s have been modelled to understand the emergent properties. The modelling shows that mutations in some of the proteins such as αS, Aβ and tau share common pathophysiology in both AD and PD. These proteins along with TNFα, ROS and other kinases can induce oxidative stress in neurons that trigger apoptotic pathways. Simulations suggested insulin was a key factor that could trigger and modulated common signalling pathways observed in AD and PD such as neuroinflammation and oxidative stress. It is also associated with the variations in cellular concentrations of αS, Aβ and tau and led to accumulation of toxic cellular oligomers.
Several factors led to microglial activation. and included TNFα, suggesting the role of TNFα-induced neuroinflammation in activation of glial cells that lead to neurodegeneration. The simulations also highlight feedback loops, oxidative stress and insulin pathway in the brain regulated by TNFα. Feedback interaction of TNFα with its receptor TNFR1 induced its own release matching experimental studies (Olmos and Lladó
2014). Increased concentrations in TNFα and its receptor due to the feedback mechanism could be attributed to diseased states. Increases in the level of TNFα in the model led to production of excess glutamate that consequently led to an increase in TNFα concentration level has been observed.
In neurological conditions, simulations suggested a prolonged activation of microglia. Although TNFα and other cytokines come to homeostasis inside cells in time, glial cells can stay active for a longer period. In diseased condition, this prolonged activation of microglia may lead to a release of proinflammatory cytokines including TNFα and interleukin 6 beta (IL-6β) that could disrupt the downstream cellular signalling processes leading to cell death as suggested in experiments (A Frankola et al.
2011). Like relevant clincal markers. Simulations showedelevated levels of mutated AD and PD related protein aggregation in diseased conditions compared to normal conditions. A high level of αS, Aβ and tau protein; key factors in the formation of fibrils, plaques and tangles were related and could induce oxidative stress in the cell. Increase in concentration levels of TNFα, IL-1β and calcium levels have also been observed in diseased conditions. This could be the attributed cause of activating microglia and could lead to production of TNFα and IL-1β, as indicated in simulations.
The insulin signalling pathway in the brain is regulated by cross-talk between several other signalling pathways including TNFα signalling leading to neuroinflammation and oxidative stress. Along with major mutated proteins in AD/PD such as αS, Aβ and tau, dysregulation in these signalling pathways can cause an insulin resistance in the brain. The results also have shown an increased level of ROS in diseased state. The simulations suggest low insulin could lead to high inflammation rate, oxidative stress and cell death when compared to control. It may be predicted that TNFα, ROS and insulin act as reliable biomarkers for both PD and AD.
The control condition may be indicative of the relative concentration changes in TNFα, insulin, glutamate, TNFR1, calcium, ROS, αS, P38, microglial activation and rate of cell death. This may be used as a prediction template for AD/PD relating conditions of neuroinflammation, oxidative stress and insulin resistance. When compared to diseased state, rate of microglial activation, αS, Aβ, P38, ROS formation and cell death rate, tau hyperphosphorylation, oxidative stress and cell death were considerably low.
Given all the three conditions (inflammation, oxidative stress and insulin resistance), diseased state can be identified with high concentration elevations in TNFα, αS, Aβ and tau compared to control. Given the correlations among the feedback loops, PD can be distinguished from normal conditions through high relative concentration differences in TNFα, glutamate, calcium and rate of cell death during neuroinflammation compared to control. In diseased state models associated with oxidative stress, there could be high activation of microglia, increased concentration levels of calcium, ROS, cytochrome c, and proinflammatory caspases was observed compared to control. Insulin involvement in disease state can be identified with high inflammation rate, oxidative stress and cell death compared to control.
The predictions relate experimentally observed concentrations to parameters seen during clinical measurements. The study correlated Aβ toxicity to potential clinical features such as delusions, hallucinations, seizures attributed with tau toxicity. This matches with recent studies; both familial and sporadic PD patients report symptoms related genes indicating abnormal accumulation of αS, Aβ, tau toxicity and other neuroinflammatory cytokines as mentioned in this model and hence the model can be used to predict changes in biomarkers for both prodromal and preclinical diagnosis of the disease (Popescu
2016; He et al.
2018). This model may serve as a design framework for altering experimental interventions. Although cerebro-spinal fluid was the main source of data for several parameters related to initial conditions, given the predictions from the data it may relate to changes in substantia nigra, blood, serum, blood plasma, brain cortex and hippocampus for labelling and further analysis.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.