University of Amsterdam
Institute for Interdisciplinary Sciences
Msc in Brain and Cognitive Sciences
Behavioral Neuroscience Track
Stress-Induced Synapse Dysfunction in
Alzheimer’s Disease
The relationship between psychological stress and AMPA-containing
glutamate receptor malfunction in Alzheimer’s Disease
Author:
Jochem Stormmesand
Student ID:
11286199
Supervisor:
Dr. Helmut Kessels
Co-assessor:
Dr. Harm Krugers
March 12, 2018
Adaptive behavior and memory formation are dependent on tuning of the strength of synaptic connections in the brain. Cellular models to explain this tuning are long term potentiation (LTP) and depression (LTD), which are respectively characterized by insertion and removal of glutamatergic AMPA receptors (AMPARs), and controlled by NMDA receptor (NMDAR) activation. Stress and stress hormones like cortisol mainly bind to glucocorticoid receptors. Glucocorticoids affect synaptic signaling and plasticity via Aβ that interferes with the signaling of these receptors and is thus implicated in AD. Here I review how amyloid β (Aβ), induced by stress, affects various AMPAR and NMDAR subtypes in Alzheimer’s Disease (AD). Short term corticosterone elevation increases the percentage of glutamate NMDA (GluN) receptor 2B as well as AMPA (GluA) subtype 2 in the synapse, which is beneficial for learning. However, prolonged corticosterone elevation can become detrimental and prevent LTP and learning. In AD, Aβ is speculated to remove GluA2/3 from synapses, or indirectly remove them via hyperactivation. Consequentially, GluN2B is removed and synapses are prone to be damaged. Finally, cognitive reserve models a neuroprotective reserve and the capacity for a functional buffer brought about by mental activity. Recent studies indicate that cognitive reserve likely functions by removal of the GluA3 AMPAR subtype upon cognitive activity. Therefore, the main binding site for Aβ is removed, preventing its interference in the synapse and the clinical manifestation of AD, even though the brain pathology progresses nonetheless. Furthermore, GluA1 insertion after learning is thought to protect the synapse against Aβ.
Contents
Abstract i 1 Introduction 1 1.1 Alzheimer’s Disease . . . 1 2 Synapses and AD 4 2.1 LTP vs LTD . . . 5 2.1.1 AMPA receptor . . . 5 2.1.2 NMDA receptor . . . 6 2.2 Amyloid β . . . 62.3 Relation between Aβ and AMPARs/NMDARs . . . 7
2.4 GluA1 vs GluA3 . . . 8
3 Stress and stress hormones 10 3.1 Corticotropin-releasing factor . . . 10
3.2 Stress and Aging . . . 12
4 Stress, Amyloid β and Plasticity 14 5 Cognitive reserve 16 5.1 Theoretical model . . . 16
5.2 Cellular Mechanisms . . . 18
6 Conclusion and Significance 20
Bibliography 20
Introduction
Memory formation is a key attribute for processing the world around us and allows us to remember daily information. Dysfunctional memory formation, as seen in Alzheimer’s Disease (AD), greatly impacts people’s lives. A prime suspect to cause AD is Amyloid β (Aβ). When Aβ is present in excess, it aggregates into oligomeric clusters, and these Aβ oligomers are known to have toxic effects on neurons and the connections through which they communicate, i.e. synapses. Not all people are equally susceptible to the detrimental effects of Aβ. A large amount of the current literature emphasizes the influence of psychosocial and physical stress on disease development (Dong and Csernansky, 2009; Lesuis et al., 2017; Mikasova et al., 2017). Chronic and early life stress (ELS) are both thought to enhance an individual’s vulnerability for the development of AD and other neurodegenerative disorders (Krugers et al., 2017; Ross et al., 2017). Although it is generally accepted that the progression of Alzheimer’s is a matter of synaptic malfunction, the exact mechanisms are not fully understood.
In this thesis I will first outline the urgency for understanding the pathology underlying AD. This will be followed by a description of the concepts related to memory formation and synaptic plasticity. Next, I will progress into the AD-associated pathological functioning of synapses and glutamate receptors and how these are influenced by Aβ. In the final chapters, I will explore the antagonistic effects of psychological stress and the theoretical cognitive reserve on AD-associated synaptic dysfunction.
1.1
Alzheimer’s Disease
AD is a neurodegenerative disorder characterized by Aβ formation, plaques, neurofibrillary tangles (NFT) and neuropil threads that are composed of hyperphosphorylated tau (tau-p) (Wirz et al., 2014). Furthermore, the disease can be distinguished by progressive loss of neurons and synapses, mostly in brain regions needed for cognitive processes, like temporal, parietal and frontal cortex, hippocampus and amygdala (Wirz et al., 2014). AD is the most prevalent form of dementia among the elderly, accounting for 70% of the cases (World Health Organization, 2012). Moreover, AD can be separated in familial, or early onset AD (about 5 to 10% of the cases (Wirz et al., 2014)) and the more common sporadic, or late onset AD. For the familial AD, mutations in three genes have been identified: Aβ precursor protein (APP), presenilin 1, and presenilin 2 (Price and Sisodia, 1998). These mutations suggest that familial AD corresponds to an increased production of Aβ. The origin of sporadic AD is under debate, but it is thought that Aβ accumulation in sporadic AD is a result of decreased capability for Aβ clearance from the brain (Deane et al., 2008; Kulstad et al., 2005; Verghese et al., 2013). In 2010, AD
CHAPTER 1. INTRODUCTION 2 affected an estimated 36 million patients and these numbers are rapidly increasing (Batsch and Mittelman, 2012). Many risk factors that are associated with an increased risk of sporadic AD development have been discerned over the years. The primary risk factor for AD is age. In addition, stress and having the ApoE ε4 allele are identified as major risk factors, as well as smoking, having a low degree of education, cardiovascular disorders and depression (Ross et al., 2017; Shankar and Walsh, 2009; Wirz et al., 2014). After the age of 65, the chances of AD development double every 5 years (Shankar and Walsh, 2009).
The ApoE ε4 allele has long been the only genetic risk factor directly linked to sporadic AD. A single allele for the ApoE ε4 isoform already subjects the individual to a two-fold increase in AD risk. Having two ApoE ε4 alleles corresponds to a 5 fold increased risk (Reitz and Mayeux, 2014). Various studies showed that AD risk increases with ApoE ε4 because it has a major impact on Aβ clearance in the brain (Deane et al., 2008; Verghese et al., 2013; Delano-wood et al., 2008). ApoE isoforms disturb Aβ clearance from the brain in different manners, with ApoE ε4 having the most disturbing effects on the clearance rate. Binding of Aβ to ApoE ε4 alters the clearance mechanism from the fast low-density lipoprotein receptor-related protein 1, to the very low-density lipoprotein receptor that internalizes the ApoE-Aβ complex slower at the blood brain barrier, as was shown in mice (Deane et al., 2008). The proposed mechanism by which ApoE isoforms directly but differentially influence Aβ metabolism have been challenged however. It was indicated that these effects are more likely due to the isoforms’ competition for Aβ clearance pathways (Verghese et al., 2013). Alternatively, a study evaluating ApoE ε4 and depression found that women with this allele were four times more likely to be depressed than controls without the allele (Delano-wood et al., 2008). This implies that ApoE ε4 might indirectly play a role in AD development, by for example causing depression and thereby increase the likelihood of AD development.
Tau-p, another biomarker for AD, is a product that progressively accumulates with aging, starting early in life. When Aβ, which is a physiological regulator of neuronal activity, starts to accumulate above a certain threshold, the phosphorylation and consecutive accumulation of tau increases rapidly and starts to injure and degenerate neurons, leading to synaptic deficits (Ross et al., 2017). Aggregates of Tau-p can be found as deposits inside neurons, forming NFTs that cause cytotoxicity (Hardy and Selkoe, 2002). NFTs are often used in post-mortem brain research to determine an individual’s stage and severity of AD, as its progression and density can clearly be tracked through the various brain regions. Accumulation of NFTs is thought to begin in the transentorhinal cortex, from there transferring to the entorhinal, hippocampal, temporal, parietal, and occipital cortices (Braak et al., 2006; Ross et al., 2017).
Alzheimer’s progression is most commonly divided into 7 stages (0-VI), called the Braak stages, which are based on the level of NFT pathology (Braak and Braak, 1991). Patients in stages I-II do not show any noticeable cognitive deficits but macroscopically NFTs can be found in the transentorhinal cortex. In stage III and IV patients show mild cognitive impairment (MCI) and NFT formation is detected in the limbic system including the hippocampus. In the final stages (V-VI), patients are fully demented and macroscopic examination of the brain always shows extensive NFTs throughout the neocortex (Braak and Braak, 1991; Stargardt et al., 2015).
Cognitive deficits and dementia may well arise long before any clinical diagnosis of AD can be reached (Stargardt et al., 2015). As the disease onset is a progressive process, an exact starting point of AD is nearly impossible to determine. The earliest signs of AD are often exhibited as minor mishaps in episodic memory. A gradual decline of declarative followed by non-declarative memory and then other cognitive functions slowly leads to full-blown dementia with cognitive and behavioral deficits (Shankar and Walsh, 2009). Risk factors can be used for
diagnostic purposes of the disease, with for example levels of tau-p and Aβ in cerebrospinal fluid (CSF) reflecting the levels of specific regions in the human brain (Reitz and Mayeux, 2014). However, since disease onset can happen long before clinical symptoms arise, it is im-portant to obtain an even better understanding of risk factors and biomarkers. Many biological processes malfunction in AD. The interaction between Aβ and synaptic dysfunction, one of the pathological hallmarks of AD, will be extensively discussed in this paper.
From the time of the first mention of Alzheimer’s disease in 1906 by Alois Alzheimer, over a century of research has not resulted in any disease-altering medication. Until now, the only treatment available for AD patients targets the devastating cognitive impairments in an effort to slow their progression (Ross et al., 2017), underlining the urgent need for a disease modifying treatment. It has been shown to be difficult to correlate cognitive impairments of AD directly to its neuropathological characteristics like amyloid plaques, neurofibrillary tangles and even neuronal loss (Coleman et al., 2004). Interestingly, synaptic loss was found to be a much better indicator for the severity of dementia and cognitive decline that accompanies AD, accounting for more or less 50% of the variance(Coleman et al., 2004; Stargardt et al., 2015). Additionally, for other types of dementia, including normal aging, neuronal loss showed to be the best pathological marker to measure cognitive decline as well (Coleman et al., 2004; Uylings and De Brabander, 2002). For these reasons, the following chapters explore synaptic functioning and malfunctioning related to AD.
Chapter 2
Synapses and AD
Neurons have the capacity for synaptic plasticity: the ability of a neuron or neuronal network to adapt to changes in its environment. This concept is most often described by means of Hebbian theory assuming that“persistence or repetition of a reverberatory activity (or “trace”) tends to induce lasting cellular changes that add to its stability” (Hebb, 1949). Among lay people, this plasticity rule is often referred to as ‘cells that fire together, wire together’, an important notion in learning and memory. This rule, however, should not be interpreted too strictly, since Hebb’s rule implies a causal relationship between one cell’s firing and the next instead of simultaneous firing implied by ’firing together’. Nevertheless, the strengthening of synapses in learning, memory and their weakening during the decay of these processes are often described in terms of Long Term Potentiation (LTP) and Long Term Depression (LTD) models, which are in accordance with the Hebbian theory (Shankar and Walsh, 2009).
Alzheimer’s disease is known as a neurodegenerative disorder. In AD, LTP is inhibited while LTD is facilitated, leading to synaptic degeneration. Santiago Ram´on y Cajal was one of the first to describe this link of synaptic degeneration and AD by stating that “dementia could result when synapses between neurons are weakened as a result of a more or less pathological condition, that is, when processes atrophy and no longer form contacts, when cortical mnemonic or association areas suffer partial disorganization” (Ramon y Cajal, 1928). The synaptic mal-function that is caused by disconnecting neurons explains the subsequent signs of dementia, which is the core characteristic of AD. The failure of synaptic functioning is widely associated to early AD pathogenesis and this often results in neuronal death (Coleman et al., 2004). The pro-cess of synaptic degeneration occurs via various mechanisms. It has so been shown previously that a neuron may lose its synaptic functioning whilst remaining alive (Coleman et al., 2004). In contrast, structurally healthy synapses can still be flawed for a failure in neurotransmitter synthesis, transport, release and reuptake, vesicle or receptor problems, or even downstream mi-croglia activation, reduced nicotinic receptors or metabolic activity (Coleman et al., 2004; Yao and Coleman, 1998). Neuronal death is a key cause of reduced synapse numbers in AD. Nev-ertheless, it cannot be the sole explanation for this issue, since synapse to neuron ratio reduces about 48% in AD symptom associated cortices (Bertoni-Freddari et al., 1996). AD associated synapse degeneration is thought to originate in transentorhinal/entorhinal cortex, since patho-logical symptoms initiate here. On the other hand, neocortex dysfunction is involved in the earliest clinical symptoms rather than transentorhinal malfunction. This clinical-pathological mismatch is maybe caused by a degeneration of neocortical synapses (Coleman et al., 2004).
In this chapter I will dive into the processes behind synaptic function and dysfunction related to learning and memory. Next, I will explore Aβ in relation to the synaptic dysfunction of various glutamate receptor subtypes in AD.
2.1
LTP vs LTD
Activity-dependent synaptic potentiation can be divided into three categories based on its duration: The shortest persisting tetanic activation is called post-tetanic potentiation (PTP), which lasts from milliseconds up to approximately two minutes, decaying rapidly (Bliss and Collingridge, 1993). Long-lasting tetanic activation is called LTP. The range in the middle of PTP and LTP is called short-term potentiation (STP). Generally, synaptic potentiation is classified as LTP if it lasts over an hour (Bliss and Collingridge, 1993). LTP is widely considered to be the biological substrate for memory formation. LTP can be expressed in a neuron by facilitating presynaptic neurotransmitter release or increasing the amount or functionality of postsynaptic neurotransmitter receptors (Bliss and Collingridge, 1993). LTP is input specific, meaning that a synapse will be potentiated if, and only if, it is active at a time when the region of dendrite on which it terminates is sufficiently depolarized. This notion was proven by applying low frequency, low intensity stimulation which should not provoke any form of LTP, and finding that it could still lead to strong LTP when repeatedly combined with depolarizing stimuli from an intracellular electrode (Bliss and Collingridge, 1993). Several lines of evidence suggest that LTP-like processes underlie memory formation. First, drugs or mutations that block LTP also inhibit memory formation in the hippocampus (Lynch, 2004). Second, changes observed in the hippocampal glutamate receptors of a rat during learning are the same as those observed in during LTP induced by high frequency stimulation (Whitlock et al., 2006). The best evidence that LTP underlies memory formation comes from Nabavi and colleagues, who found that fear conditioning could be inactivated and reactivated by LTD and LTP consecutively, pairing a foot-shock with optogenetic stimulation of auditory input into the amygdala (Nabavi et al., 2013). In AD research there is much attention for LTP, since cognitive deficits increase and memory formation is impaired, generally attributed to defective LTP. The hippocampus is considered to have a high impact role in LTP and deterioration of this specific brain area seems to be paired to cognitive decline and dementia development.
2.1.1
AMPA receptor
Fast information transfer across glutamatergic synapses depends on AMPA-type glutamate receptors (AMPARs). The number of AMPARs is one of the main regulators of the efficiency of signal transduction. Addition or removal of AMPARs explains LTP and LTD of synapse strength respectively (Lohmann and Kessels, 2014), of which the effects can last for hours, days, months or even longer (Lynch, 2004; Malenka and Bear, 2004). LTP and LTD are needed for learning, adaptation of behavior and updating the neuronal circuits involved in these processes (Kessels and Malinow, 2009). AMPARs are tetrameric molecules that consist of four AMPAR subunits. Four different AMPAR genes encode for subunits GluA1, GluA2, GluA3 and GluA4. Excitatory neurons in the mature rodent hippocampus predominantly express two types of AMPARs: GluA1/2 and GluA2/3 heteromers (Wenthold et al., 1996). GluA1-containing receptors are inserted into synapses upon the induction of LTP. Such increase in synapse levels of GluA1 AMPARs has been shown in vitro (Hayashi, 2000) and in vivo (Rumpel, 2005). Furthermore, GluA1-deficient neurons are greatly impaired in LTP (Zamanillo et al., 1999), whilst neurons that are GluA3 deficient are not (Meng et al., 2003), indicating that the subtype and ratio of these receptors are highly important for defining the plasticity of the neurons (Lohmann and Kessels, 2014). Trafficking of GluA1 happens in response to changes in synaptic activity and is stimulated by activation of NMDA-type glutamate receptors (NMDARs). GluA1
CHAPTER 2. SYNAPSES AND AD 6 trafficking can be initiated upon behavioral experience (e.g. fear learning) that activates a specific brain area (e.g. amygdala) and moves GluA1 into these synapses (Rumpel, 2005). Blocking GluA1 trafficking or knocking out GluA1 to stop GluA1-dependent LTP impairs learning, plasticity and memory formation (Rumpel, 2005). Importantly, GluA1 appears not to be required for all memory processes, leaving certain plasticity mechanisms intact when GluA1 is blocked (Zamanillo et al., 1999). In contrast to the knowledge of GluA1, little is known about the GluA3 subunit. GluA2/3 heteromers traffic to synapses independent of synaptic activity (Shi et al., 2001). They are present in large amounts at the synapse, although invisible due to inactivity at basal conditions. Upon activation via increases of intracellular cyclic AMP, the GluA3 channels activate and cause synaptic potentiation in CA1 in the hippocampus (Gutierrez-Castellanos et al., 2017).
2.1.2
NMDA receptor
Whilst AMPARs are required for fast synaptic transmission, NMDARs are coincidence detec-tors: they only open when glutamate release and postsynaptic depolarization coincide. The opening of NMDAR channels leads to calcium influx, activating Ca2+-dependent signaling path-ways. These pathways can lead to LTP, which is expressed by the addition of AMPARs at synapses. This type of LTP is also called NMDAR-dependent LTP. Similarly, LTD is also dependent on NMDAR activation. Initially, it was thought that LTD requires small amounts of Ca2+ to flow through the NMDAR. However, Nabavi and colleagues found that ligand
bind-ing to the NMDAR (in the absence of channel openbind-ing) is sufficient for the induction of LTD (provided that there is a basal level of Ca2+), indicating that NMDARs, besides an ionotropic function, also mediate metabotropic signaling (Nabavi et al., 2013). NMDARs are heterote-tramers formed by a variety of combinations of GluN1, GluN2A to D and GluN3 subunits all combinations accounting for distinct features of the receptor. Furthermore, the ratio of especially the GluN2A and GluN2B subunits varies per brain region, structure and even per synapse. This ratio has been implicated to change right after synaptic plasticity has been ini-tiated in young neurons in the hippocampus from predominantly GluN2B to GluN2A (Matta et al., 2011). Proportional changes are probably caused by a change in receptor trafficking through endo- and exocytosis or surface diffusion (Mikasova et al., 2017).
2.2
Amyloid β
Amyloid β and the amyloid cascade hypothesis are the leading model for AD progression. This theory entails that Aβ, cleaved from APP, causes AD by forming cytotoxic amyloid plaques and successively NFTs and tau-p (Yamamoto et al., 2015; Hardy and Selkoe, 2002). Under physiological circumstances Aβ is speculated to be involved as a regulator of the activity be-tween neurons. The production of Aβ peptides in the synapse is dependent on synapse activity, endocytosis, and localization of APP(Kang et al., 2007; Ross et al., 2017). APPs that are imme-diately found on the surface have a higher chance of being converted into the soluble, nontoxic APPα by α-secretase. With neuronal activity, there is increased APP processing by β- and γ-secretase and (Kamenetz et al., 2003). The explicit mechanisms behind Aβ’s physiological role are unknown, although it is thought that Aβ may modulate vesicle release and neurotransmis-sion via a negative feedback loop. Under normal circumstances, clusters of neurons fire together, leading to LTP in the postsynaptic neuron. Increased synaptic activity increases Aβ levels. In
turn, soluble Aβ inhibits synaptic transmission, creating a negative feedback loop to prevent overexcitation (Kamenetz et al., 2003). AD patients and mouse models, with overexpression of APP and subsequently enhanced Aβ levels, show abnormal patterns of neuronal activity. This indicates that the negative feedback normally induced by Aβ is somehow disturbed in AD. The consequences of a failure in the negative feedback loop are further shown in plaque deposition over the Braak stages: During stage I and II synaptic activity slowly increases to counteract the faltering synaptic efficiency caused by Aβ or perhaps other causes. Aβ excretion is managed by synaptic activity, so increased amounts of Aβ are released, leading to a further decrease of synaptic efficiency. To be able to uphold physiological cognitive functioning, synaptic activity levels need to be raised even more, resulting possibly even in an amplifying loop (Bossers et al., 2010).
So, when an overload of Aβ establishes, this homeostatic negative feedback loop may re-main constitutively active, which likely happens in AD. Aβ will interact with AMPARs and NMDARs to inhibit synaptic activity that could potentially affect many synapses when halting activity in a leading neuron in a top-down network (Ross et al., 2017). As a result of the aggregation of non-toxic, monomeric and soluble Aβ into still-soluble oligomers Aβ becomes toxic for its surrounding (Kihara et al., 1999; Yamamoto et al., 2015). The mechanisms for Aβ cytotoxicity remain unknown, but one hypothesis involves oxidative stress (Behl et al., 1994). Nitric oxide (NO) production is dependent on transcription factor NFκB, or happens via Ca2+ influx through NMDAR channels (Akama et al., 1998). Alternatively, Aβ has been
suggested to inhibit glutamate uptake. Glutamate initiates Ca2+ influx, activating NO
syn-thase to form NO. Eventually, NO acts neurotoxically on neurons by lipid peroxidation via free radicals (Kume et al., 1997; Kihara et al., 1999). Aβ, being one of the earliest pathological changes in AD, is thought to be the accelerating factor of age-related memory deterioration in AD, which is considered a sped-up version of regular aging (Shankar and Walsh, 2009). From here, the pathological effects that Aβ exerts on synapses will be discussed as it is thought to occur in AD.
2.3
Relation between Aβ and AMPARs/NMDARs
LTP dysfunction in AD is due to the presence of Aβ oligomers in hippocampal synapses, according to rodent models by Rowan and colleagues (Rowan et al., 2003). This Aβ-induced LTP impairment may underlie the cognitive decline that is found in early AD. When present in physiological concentrations presynaptically, one method by which Aβ can affect the neuronal excitability is by enhancing calcium permeability of the membrane, causing an increase in the probability of vesicle release (Ross et al., 2017). Postsynaptically, Aβ triggers an NMDAR-dependent signaling cascade that enhances AMPAR internalization, thereby mimicking LTD. Similar to LTD induction, this Aβ-induced synaptic depression is thought to be dependent on conformational changes of the NMDAR and not on calcium influx (Kessels et al., 2013).
At physiological concentrations Aβ will thus facilitate vesicle release and neurotransmis-sion. However, when the levels deviate from normal Aβ will inhibit synapse activity (Abramov et al., 2009). Aβ activates LTD inducing functions of GluN2B, as well as a selective loss of GluN2B responses, bringing about a shift from GluN2B towards GluN2A subunits (Kessels et al., 2013). A hypothesis is that Aβ or other amyloid related molecules bind to GluN2B causing a conformational change sending a signal downstream. Alternatively, a process that re-quires GluN2B activation and its conformational change is required for the Aβ driven signaling to continue (Kessels et al., 2013). Aβ was found to affect NMDAR that have GluN2B and not
CHAPTER 2. SYNAPSES AND AD 8 GluN2A. Possibly, GluN2A renders Aβ resistance whilst GluN2B increases Aβ susceptibility (Reinders et al., 2016a). Additional effects of Aβ on the synapse are reducing spine numbers, and inducing LTD via endocytosis of AMPA and NMDA receptors. For Aβ-induced LTD first NMDAR activation is needed, independent from ion flux through its channel (Kessels et al., 2013), which then triggers a signaling cascade with tau phosphorylation and finally AMPAR endocytosis (Kamenetz et al., 2003). Synaptic AMPAR endocytosis can cause a loss of NMDA responses and dendritic spine loss and therefore denotes an important role of AMPAR removal in the process of synaptic failure (Hsieh et al., 2006). However, it is also possible that LTP in AD is affected through other processes than Aβ, such as enzymes like protein kinase Mζ that accumulate in neurofibrillary tangles (Crary et al., 2006).
Next to a reduced number of AMPARs, Aβ also causes less ion flow through NMDARs (Reinders et al., 2016a). Presumably, Aβ removes AMPAR and NMDAR via similar pathways. This is thought because blocking the removal of AMPAR via endocytosis also reduces Aβ induced NMDAR depression (Reinders et al., 2016a). Furthermore, both NMDAR and AMPAR removal from synapses are independent on the calcium flux through the NMDAR, although Kessels and colleagues found NMDAR depression to be dependent (Reinders et al., 2016a; Kessels et al., 2013).
2.4
GluA1 vs GluA3
Mouse research looking into GluA3 has implied a major influence of this AMPA subunit in AD, since animals devoid of GluA3 seem not to develop synaptic deficits that are driven by Aβ (Reinders et al., 2016b). Animals without GluA3 express increased isolation-induced aggression and increased social behavior (Adamczyk et al., 2012). Recently, the focus has shifted to study GluA3’s involvement in AD pathology. GluA3, and not GluA1, was shown to be necessary for Aβ-driven synaptic depression by looking at the amount and ratio of AMPAR expression in hippocampal neurons that express the 100 amino acid C-terminal fragment of APP (APPCT 100),
an Aβ precursor (Reinders et al., 2016a). GluA3’s necessity in Aβ-driven synaptic dysfunction is further shown by the finding that APPCT 100decreases currents that flow through NMDAR in
both wildtype CA1 and GluA1-deficient CA1 neurons. However, this was not found in GluA3-deficient neurons (Reinders et al., 2016a). Furthermore, the reduction in GluN2A-containing NMDARs is only affected by GluA1 and not by GluA3 after subjection to APPCT 100 (Reinders
et al., 2016b). Neuronal spine loss, a consequence of the amyloid deposition, cannot be induced in the absence of GluA3 in the neuron (Reinders et al., 2016b). Moreover, AD is characterized by reduced LTP, which, again, is shown to be dependent on GluA3. This was found because LTP was induced similarly with or without GluA3 being present (Meng et al., 2003), however, if GluA3 was not expressed in the neuron, LTP could in vitro not be blocked anymore by Aβ-oligomers (Reinders et al., 2016b). APP/PS1 transgenic mice, expressing human APP and mutant presenilin 1, are used as a model for early onset AD. Animals in this model produce high levels of Aβ42, the suspected pathogenic form of the Aβ protein ending in residue
42. Moreover, they show reduced spine formation and cognitive deficits and they often die prematurely. Lacking GluA3, these animals show no fear memory deficits, no spine loss and no increased mortality as would normally be expected of these mice. In total, the research by Reinders and colleagues found that GluA3 is highly influential for Aβ-dependent synaptic dysfunction (Reinders et al., 2016b).
One of the current suggestions of the amyloid cascade hypothesis is that Aβ oligomers bind to GluA3. For this reason, GluA3 could be indicative for the expression of Aβ-induced synaptic
deficiencies in AD. It has been proposed that GluA3 is involved in the homeostatic scaling of synaptic strength, which is a process that dynamically adjusts synaptic strength in order to avoid network instability (Makino and Malinow, 2011; Turrigiano and Nelson, 2004). If a neuron does not receive synaptic input it will increase levels of GluA2/3 whilst overactivity results in reduced GluA2/3 levels. Synaptic and memory insufficiencies due to disturbance of the synaptic plasticity in AD are potentially caused by problems with the so called metaplasticity. This is a type of homeostatic balance in synaptic plasticity that coordinates the direction and magnitude of upcoming synaptic plasticity, dependent on previous activity of the neuron or synapse involved (Jang and Chung, 2016). In relation to this balance, studies have indicated that GluA2/3 subunits make up a constitutive pool. From this pool receptors can be taken and stored into depending on synaptic activity of the neuron. By continuously replacing the receptors, a stable baseline level of synaptic transmission can be maintained (Adamczyk et al., 2012; Takahashi et al., 2010). GluA1/2s make up a reserve receptor pool that is needed with increased activity in hippocampal neurons (Granger et al., 2012). In AD, Aβ is now thought to remove GluA2/3 from synapses even if the synapse has been active. Another possibility is that Aβ could cause hyperactivity which is countered by removal of GluA2/3 from the synapse. Following AMPAR removal with Aβ deposition, GluN2B is removed and synapses are likely to be damaged.
Looking at the receptor presence in relation to cognitive functioning, a study into genetic influences in MCI found genes encoding GluA3 and GluN2B to be negatively correlated to cognitive functioning (Berchtold et al., 2014). For this reason, having low levels of these two glutamate receptors is thought to prevent the symptoms of MCI. Alternatively, learning be-havior, which can be stimulated by mental exercise to yield LTP, has been shown to increase GluA1 subunit levels in the synapse and causes homeostatic removal of GluA2/3 (Hayashi, 2000; Makino and Malinow, 2011; Reinders et al., 2016b) and should thus reduce symptoms of MCI. As indicated, GluA1 forms to a major extend GluA1/2 heterodimers. Overexpression of GluA1 results in GluA1 homomers(Hayashi, 2000), which makes the subunits inwardly recti-fying. Homomeric GluA1 trafficking, which is under basal conditions prevented from synapse entry and only initiated upon learning could have protective effects in the synapse. It is sug-gested that GluA1 insertion after learning (Rumpel, 2005) would help to protect the synapse against Aβ’s binding opportunities to GluA3 (Reinders et al., 2016a). This is otherwise implied because blocking the GluA1 trafficking impairs many forms of learning (Rumpel, 2005; Humeau et al., 2007)
Chapter 3
Stress and stress hormones
Stress and glucocorticoids can facilitate memory formation. It was shown that acute stress promotes learning through an increase in AMPAR-trafficking (Conboy and Sandi, 2010). On the other hand, stress is one of the major correlates to neuropsychiatric disorders. More and more data indicate a connection between AD onset and advancement and the influence of physical or psychological stress on the HPA-axis (Dong and Csernansky, 2009; Lucassen et al., 2014). As one of the key causes of major depressive disorder, chronic stress initiates pathological mechanisms that predisposes patients with a stress related disorder like depression to develop neurodegenerative disorders such as AD (Dong and Csernansky, 2009; Ross et al., 2017). In a healthy environment, synapses change in accordance with environmental input stimuli (Ross et al., 2017). Many of these plasticity mechanisms are altered as a consequence of exposure to chronic stress, possibly increasing an individual’s vulnerability to develop neurodegenerative disorders like Alzheimer’s Disease. Among these changes are a reduction in negative feedback inhibition from the hypothalamic pituitary adrenal (HPA) axis, less dendritic arborization and spine density in the PFC and hippocampus, and the release of proinflammatory cytokines, which may suppress neurogenesis and promote neuronal cell death (McEwen and Morrison, 2013; Ross et al., 2017). These factors have all been found to play a role in stress-related psychiatric disease and AD as reviewed by Ross and colleagues. Furthermore, they have been observed in clinical and post-mortem studies of individuals with depression and AD (Ross et al., 2017).
3.1
Corticotropin-releasing factor
Corticotropin-releasing factor (CRF) and its receptors (CRFR) regulate glucocorticoid secretion (cortisol in humans, corticosterone in rodents) and are thereby one of the key regulators of HPA axis activity. Due to their widespread availability throughout cortex and hippocampus and even production and distribution at locations outside the hypothalamus, corticosteroids regulate all sorts of neural activities in processes like memory and anxiety that are sensitive to stress (Dong and Csernansky, 2009). Furthermore, corticosteroids are known to regulate excitatory transmission and plasticity in the hippocampus (Krugers et al., 2010). Glucocorticoids are key regulators of the dendritic spine density and morphology during stressful conditions and appear abundantly in PFC and hippocampus, making these areas to be easier altered under stress (McEwen and Morrison, 2013), and implicating the involvement of glucocorticoids in AD (Kulstad et al., 2005; Pedersen et al., 2001). However, it is important to keep in mind that both cortisol levels and HPA axis activity are thought to increase with regular aging.
Figure 3.1: Linking stress, Aβ and Alzheimers. Schematic diagram explaining the hypothesized influence of CRF in the regulation of Aβ. Stress stimulates hormonal and neuroregulator-like functions that can influence Aβ formation and deposition via various path-ways.
Hypercortisolemia and increased HPA axis activation are often associated to AD, but since AD is increasingly prevalent with age, the effects of high cortisol levels on AD need to be considered very carefully (Dong and Csernansky, 2009).
CRFRs are high affinity, G protein-coupled receptors that act on the HPA-axis as a response to stressful events. CRFRs enhance adenylate cyclase activity, and consequentially elevate intercellular levels of cAMP. Since the largest amount of adenylate cyclase is found in the cerebral cortex, which is highly associated to AD, it is hypothesized that CRFRs are of influence in AD pathogenesis (Dong and Csernansky, 2009). Similarly, the glutamate system, which is thought to be involved in Aβ formation and neurotoxicity, is found to influence CRFR expression (Dong and Csernansky, 2009). For this reason it is thought that CRF and other stress systems are possibly involved in the harmful effects that behavioral stress can have on synaptic activity, Aβ and AD.
A study by Kang and colleagues, showed that synaptic activity and vesicle release can be linked to Aβ release from neurons (Kang et al., 2007). Glutamate, which has an effect on over 80% of synapses, is thought to influence CRF and CRFR expression and function and to be involved in Aβ formation and its toxicity (Dong and Csernansky, 2009; Kume et al., 1997). For its regulatory role on neuronal activity with regards to stress, it is likely that increased amounts of glutamate, as a consequence of behavioral stressors, negatively affect synaptic activity over sustained periods of time, thereby affecting the physiological regulation of Aβ and maybe the risk to develop AD. Obviously, CRF and CRFR are not necessarily the only pathways by which stress affects Aβ and AD. The exact mechanisms remain unclear though with some contradicting findings: enhanced cortisol levels are linked to diminished Aβ degradation in old macaques caused by down-regulation of insulin degrading enzyme (Kulstad et al., 2005). In rats, increased corticosterone levels enhance the vulnerability of cholinergic neurons to Aβ (Abrah´am et al., 2001). Also in humans the findings are not always consistent when linking CRF to AD: some studies report reduced CRF in various brain regions in AD (Souza, 1988), whilst others indicate increased amounts of postsynaptic CRFRs (Behan et al., 1995; Dong and Csernansky, 2009). These findings are not necessarily contradictive though, and could be explained by insertion of CRFRs upon CRF deficits to restore baseline levels. Furthermore, the severity of dementia does not necessarily correspond to the level of CRF in many cortical regions (Davis et al., 1999). Such results imply that CRF might be temporarily enhanced in early AD.
CHAPTER 3. STRESS AND STRESS HORMONES 12 There are, various links associating stress with changes in HPA axis activity, neuronal activ-ity, and AD (Dong and Csernansky, 2009). Figure 3.1 ties changes of CRF, as a consequence of induced stress, to increased Aβ levels and subsequent neuron death. The figure does not touch upon the method by which Aβ is thought to cause the neuronal death, but is a good indicator that the relationship between stress and synaptic AMPAR malfunction exists, as I will explore in chapter 4.
3.2
Stress and Aging
Animal models are widely used to study the impact of environmental factors like stress on the course of AD pathogenesis via genetic manipulation. The Tg2576 mouse is one of the best-known animal models for AD, having a mutant gene hAPP Lys670-Asn, Met671-Leu with a hamster prion protein promotor. The model is reflecting AD since it shows overexpression of Aβ precursor protein with concurrent Aβ plaque deposition in cortex and hippocampus around 9-10 months. Furthermore, the animals exhibit behavioral deficiencies related to AD, reduced LTP in the hippocampus and synapse loss (Dong and Csernansky, 2009). Research has indicated a strong relationship between stress and an accelerated aging process that is typical for AD (Kang et al., 2007; Swaab, 1991). These two important effects in AD are well described in Figure 3.2. This figure gives a specific example of reduced resilience to stress due to diminishing plasticity of the brain with age: how the representation of the rat medial prefrontal cortex (mPFC) neurons adjusts from young to old age. At young age, stress inflicts apical dendrite shrinkage which can be overcome during a recovery phase, whilst at older age the neurons cannot recover from the shrinkage anymore (McEwen and Morrison, 2013; Ross et al., 2017) (Figure 3.2A). Thin spines are at young age affected by stress, causing a reduction in spine numbers. These are thought to be essential for healthy functioning of the PFC circuitry (McEwen and Morrison, 2013; Ross et al., 2017). The spines that remain are to a lesser extent able to change the circuit dynamics upon experience (Ross et al., 2017). However, since spine loss is an attribute of regular aging, at older ages the stress does not inflict any additional spine loss to regular aging (Figure 3.2B). It is suggested that the process by which resilience to stress is lost over time with regular aging is accelerated in AD (McEwen and Morrison, 2013). The dendritic and spinal failures with aging in rats correspond to the findings that human middle aged individuals who are exposed to chronic stress are more prone to develop AD than non-stressed controls (Johansson et al., 2010), and that exposure of non-demented, older individuals to chronic stress deteriorates their memory function if they have an apoE ε4 allele (Peavy et al., 2007). All of these findings underline the harmful effects of chronic stress and indicate that it leads to the development of AD associated symptoms.
.
Figure 3.2: Interaction between stress and aging in Layer 3 pyramidal neurons in the prelimbic mPFC. A) represents the effects of stress on the dendritic arborization and B) on spines. From top to bottom, young, middle-aged and old male rats are represented. A) For young rats, stress leads to dendrite shrinkage which can regrow after chronic stress stimulation has stopped. For middle-aged rats, this regrowth is limited and for the old males it is entirely absent. B) Due to chronic stress exposure, thin spines are lost in young rats. Middle-aged and old rats do not suffer from additional spine loss, likely because their losses are already caused by regular aging. Adapted from McEwen and Morrison, 2013.
Chapter 4
Stress, Amyloid β and Plasticity
It is well-known that stress can disrupt the activity of various brain regions (Ross et al., 2017). Asides from corticosteroid influence, stress in the form of locus coeruleus (LC) overactivation, increasing very widespread noradrenaline (NA) secretion, can leave the brain vulnerable to Aβ and plaque deposition (Ross et al., 2017). Sustained overactivity of neurons is thought to drive Aβ accumulation (Yamamoto et al., 2015), since synaptic activity is supposed to regulate the production of Aβ (Kamenetz et al., 2003). With this understanding, it can be argued that stress might affect Aβ levels in the synapse, thereby causing synaptic dysfunction. It remains questionable via what mechanism exactly stress affects synapses, as this is not necessarily Aβ-dependent as explained by Mikasova and colleagues: During stress, increasing concentrations of glucocorticoids activate high-affinity mineralocorticoid receptors (MR), and with high enough concentrations, the glucocorticoid receptor (GR). Both receptors act as transcription factors to regulate gene expression to exert long-term responses to changes in stress, and MR signaling has been implicated to act non-genomically too (Mikasova et al., 2017).
Under stressful conditions, corticosteroids initially induce glutamate release from the presy-naptic neuron, facilitating LTP of the glutamatergic synapses in the hippocampus (Wiegert et al., 2006). After a while, think of several hours, AMPAR mediated transmission in the postsynaptic membrane is increased via intracellular GR that in turn impairs the LTP process (Mikasova et al., 2017; Kim and Diamond, 2002). The time-dependent variations of the effect of glucocorticoids requires a differential mechanism of AMPAR membrane delivery and surface diffusion. This system mostly needs NMDAR activation, since NMDARs do not respond to single binding of glutamate, but concurrently require postsynaptic depolarization. For this rea-son, these receptors are highly important in LTP, as the Hebbian mechanisms demand that the pre-and postsynaptic activation should coincide (Hebb, 1949; Mikasova et al., 2017). Chronic stress reduces the amount of the main NMDAR subunits in the cortex and hippocampus in the long run (Mikasova et al., 2017). Furthermore, protein expression of GluN2B decreases after stress. Since GluN2B mRNA levels are generally not disturbed and only GluN2A in-creases are reported in chronically stressed animals, it is likely that these changes in NMDAR levels (GluN1, GluN2A and B mostly) are caused by transcriptional or transductional changes (Mikasova et al., 2017; Hoffmann et al., 2000). NMDARs diffuse along the plasma membrane which steadily provides receptors to the synapse. Corticosterone was shown to reduce GluN1 surface dynamics in the hippocampus (Mikasova et al., 2017). GluN1 remained much longer in the post synaptic area and the diffusion rate out of the synapse reduced. This was mostly due to an increase in the amount of immobile NMDAR compared to the mobile receptors (Mikasova et al., 2017). NMDAR activity is additionally thought to be necessary for synaptic depression, dendritic spine loss and a reduction in synaptic plasticity, as a consequence of oligomeric Aβ or
overexpression of APP (Kessels et al., 2013; Shankar and Walsh, 2009; Haass and Selkoe, 2007). Changes in NMDAR signaling to induce synaptic adaptation upon chronic stress exposure are caused by corticosterone that rapidly adjusts GluN2B-NMDAR surface dynamics and synapse content in glutamatergic neurons. These changes are in turn caused by a MR-like signaling pathway that stabilizes postsynaptic GluN2B-NMDAR. In turn, this process promotes LTP of AMPAR in synapses of the hippocampus (Krugers et al., 2010; Mikasova et al., 2017).
Corticosteroids have been shown to regulate synaptic plasticity via the ratio of GluN2A and GluN2B, likely at a transcriptional level (Mikasova et al., 2017). Changing the NMDAR ratio might be beneficial for optimizing the threshold for LTP. In AD, the GluN2A and GluN2B re-ceptors have differential effects over time. In short, it seems that acutely elevated levels of stress and thereby corticosteroids are beneficial for AD, whilst chronic elevation eventually becomes detrimental. More precisely, cortiscosteroids increase the percentage GluN2B receptors, possi-bly via a change in receptor trafficking (Mikasova et al., 2017). Corticosterone has immediate effects on surface dynamics of the AMPAR as well, by increasing GluA2 levels. By altering the glutamate receptor ratios, elevated corticosterone during learning can possibly aid learning and memory processes (Conboy and Sandi, 2010). However, if the corticosterone levels remain elevated over a longer period of time, their effects might actually revert and prevent LTP and learning after stress exposure (Jo¨els et al., 2006).
Chapter 5
Cognitive reserve
5.1
Theoretical model
Cognitive reserve is a theoretical model created to explain the discrepancy between brain pathol-ogy and clinical manifestation of AD (Stern, 2002; Hoenig et al., 2017). One of the first mentions was by Katzman and colleagues in 1988, when they examined a cohort of AD patients and found ten elderly patients that were considered healthy on cognitive level but showed extensive AD pathology with neocortical Aβ plaques in their brains post-mortem (Katzman et al., 1988). It was further established that these patients had plaque counts at 80% of demented AD patients, and choline acetyltransferase and somatostatin levels, normally reduced in AD, in between AD patients and controls. In AD the finding of significantly reduced brain weight has been at-tributed to the loss of large neurons specifically (Terry et al., 1981). Katzman and colleagues showed higher brain weights and a larger number of neurons in their 10 patients (Katzman et al., 1988). The conclusion that was drawn was that the patients involved might well have had AD but escaped loss of large neurons or possibly had larger brains initially, with more neurons and therefore a greater reserve to account for the neuron loss in AD (Katzman et al., 1988). With regards to aging in AD, around the same time Dick Swaab reviewed two opposing concepts of “wear and tear” versus “use or lose” of the brain’s neurons (Swaab, 1991). He proposed that AD etiology is often incorrectly ascribed to the ’wear and tear’ of the body, the decay over time associated to aging. It was suggested that cellular activity goes together with increased ’wear and tear’, due to free radical formation. The brain however, seemed to be an exception to this rule and require physiological activation of nerve cells to maintain healthy functionality and interfere with the aging process. This principle was named ’use it or lose it’, which was reflected by the relieving effects of cognitive exercise in AD patients. 30 years later, much more research has looked into how the phenomenon of cognitive reserve is accounted for in the brain.
As I explained in chapter 1, the processes of Aβ and Tau are closely related. Aβ triggers abnormal Tau phosphorylation leading to synaptic deficits (Wirz et al., 2014). The aggregated tau forms NFTs that lead to cytotoxicity and for this reason the literature often speaks of a tau burden in AD. Interestingly, certain individuals have neurons that are more resistant to Aβ. This means that they have neurons that can maintain functionality despite Aβ interference and its subsequent tau burden (Hoenig et al., 2017), which could explain the cognitive reserve. It is quite generally accepted that high cognitive reserve delays the clinical expression of dementia in AD, despite a growing tau burden that indicates normal development of the neuropathology. Many neuroimaging studies have assessed the cognitive reserve-induced structural and
tional variations in the brain (Bartr´es-Faz et al., 2017; Hoenig et al., 2017; Soldan et al., 2017), indicating for example a modified molecular architecture of the cortical regions that comes with high cellular demands (Bartr´es-Faz et al., 2017). Stressful life events increase the risk of devel-oping cognitive decline and AD (Dong and Csernansky, 2009). Contrarily, early lifetime factors can have a high impact on an individual’s cognitive health at a later stage in life. Examples of such factors that positively affect the developing brain are degree of education and challenges in one’s occupation. AD patients with high educational attainment have a profound lower tau burden compared to low educated AD patients. In a study examining patients with similar cognitive functioning, highly educated patients showed tau pathology in regions associated to Braak stage V and VI, whilst low educated patients showed tau pathology in regions associ-ated to Braak stage III and IV. The burden in the low educassoci-ated group was restrained to the temporal and posterior cingulate cortex, while the highly educated group showed distribution of tau pathology towards the frontal and parietal regions, normally only affected in advanced (stage V and VI) AD (Hoenig et al., 2017). For these findings it is suggested that cognitive re-serve acts by relieving the isocortical regions (Hoenig et al., 2017). Importantly, high education has been linked to better cognitive functioning irrespective of the amount of AD pathology. Education therefore does not seem to slow the pathology progress but is thought to affect the clinical manifestation of AD by slowing the cognitive decline that that results from for example the tau burden or amyloid deposition (Hoenig et al., 2017; Rowan et al., 2003). Hypotheses on how cognitive reserve works include increased functional connectivity by using extra brain areas, and increased dendrite density and enhanced vascular coupling (Hoenig et al., 2017). An interesting notion is that, in addition to education and occupation which are hardly malleable at older age, cognitive leisure activities can expand the cognitive reserve even at older age and thereby delay the onset of dementia (see Figure 5) (Hoenig et al., 2017; Soldan et al., 2017; Mistridis et al., 2017).
Figure 5.1: Relating how cognitive reserve delays clinical AD expression to AD-associated pathology (e.g. tau-p). Even though changes in educational attainment and occupational challenges are hard to induce after age 50, cognitive leisure can still increase the cognitive reserve and thereby delaying the onset of AD-associated cognitive decline. Adapted from Mistridis et al., 2017.
CHAPTER 5. COGNITIVE RESERVE 18 The theoretical concept of cognitive reserve has been translated to human research in recent years. It is extensively found that cognitive and physical activity and even social interaction have the potential to prevent AD development (Barnes and Yaffe, 2011; Erickson et al., 2012). Moreover, epidemiological studies have shown that physical activity reduces Aβ load and re-duces AD and dementia risk (Podewils et al., 2005; Rovio et al., 2005). The Verhaagen and Swaab labs investigated the transcriptional alterations associated with AD to find molecular evidence as to how an active, learning brain can be less susceptible to AD-related memory loss. To elucidate this, they extensively analyzed the gene expression profiles in human brain samples from individuals in different Braak stages of AD (Bossers et al., 2010).
5.2
Cellular Mechanisms
To unravel the mechanisms underlying cognitive reserve as found in the human studies above, I will next discuss animal studies that have been conducted to elucidate the cellular processes involved.
First, and similar to humans, it was shown that increased daily physical exercise reduces Aβ plaque load by improving the clearing of Aβ in the brains of AD-model mice (Adlard, 2005). Placing rodents in an enriched environment (EE) enhances their cognitive reserve. For animal studies of neurodegenerative diseases, raising mice in an EE has been shown to reduce Aβ levels and deposits as compared to animals caged in standard housing (Lazarov et al., 2005; Levi et al., 2003). An EE allows an animal to have more cognitive and physical activity and as a result multiple molecular effects are thought to aid to the degree of cognitive reserve observed. Among these molecular effects are increased expression of brain-derived neurotrophic factors, growth factors, synaptic proteins and these molecular changes are thought to support cognitive reserve via induction of synaptic plasticity, neurogenesis, dendritic branching and axonal transport (Levi et al., 2003; Nithianantharajah and Hannan, 2011). Although research has shown that cognitive deficits as a result of Aβ could be prevented by EE, Aβ plaque load decreased (Adlard, 2005; Lazarov et al., 2005) did not change (Valero et al., 2011) or even increased in such studies (Jankowsky et al., 2003). It is known that Aβ levels are found elevated in the brain pre-symptomatic of AD. Interestingly, it is thought that vascular changes, inflammation, oxidative stress, mitchondrial dysfunction and disruption of calcium homeostasis appear even before the plaques and tangles can be found in the brain (Nithianantharajah and Hannan, 2011). Enriched environments again help to lower all these levels as well as reducing the standard AD pathology. The effectiveness of enriched environment on cognitive stimulation is greatly dependent on the ApoE gene. Transgenic mice for human ApoE ε3 mice exposed to an enriched environment show improved learning and memory while transgenic ApoE ε4 mice do not improve (Levi et al., 2003). In all likelihood, two facets of EE act oppositely on Aβ levels. On the one hand, increased activity leads to increased Aβ clearance. On the other hand, increased neuronal activity could lead to an increase in Aβ peptide formation and secretion (Kamenetz et al., 2003). In this thesis, I am interested in the ability of neurons to remain intact in an active brain despite high levels of Aβ. Potentially, active neurons change their gene expression program to compensate for the effects of excess Aβ.
Recently, much research has been dedicated into understanding the mechanism that allows for a shift in the cognitive reserve with cognitive exercise as is shown in Figure 5. One of the targets for its potential role in the mechanism driving cognitive reserve is synaptic transcription factor Early growth response 1 (Egr1). Egr1 is thought to be involved in the formation of cognitive reserve for a number of reasons. First, Egr1 is upregulated after learning (Renbaum
et al., 2003). Second, Egr1 is a transcription factor involved in the regulation of synaptic proteins and regulates cleavage of APP to Aβ (Bossers et al., 2010). Third, Egr1 expression is found to be higher in cognitively healthy individuals that show excess Aβ (Bossers et al., 2010). For these reasons Egr1 was hypothesized to be upregulated in order to compensate for the elevated Aβ levels. This idea was proposed by Bossers and colleagues in 2010 because of the observation that after Braak stage 1, which is idle of both Aβ and AD symptoms, in Braak stage 2 Aβ levels rise but the emergence of clinical symptoms is postponed until braak stage 3 (Bossers et al., 2010). This phenomenon coincides with a rise in Egr1 levels up to Braak stage 2, after which the levels slowly decline again. This implies that Egr1 makes the synapses less sensitive to Aβ.
A very recent study by Kessels and colleagues showed that indeed Egr1 expressing neurons are less sensitive to Aβ-mediated synaptic depression (Kessels). Moreover, they found that Egr1 lowers the levels of GluA3 in the synapse. That Egr1 levels go up with learning suggests that mental exercise can lower the level of GluA3 and consequentially promote LTP and prevent LTD from occurring. Egr1 could thus be said to form a protective buffer against Aβ formation after learning, which could in the future hopefully be used preventatively in AD treatment. Alternatively, it could be theorized that it is LTP itself that yields the synaptic insensitivity to Aβ by increasing the amount of GluA1 and thereby reducing the relative amount of GluA3 in the synapse. Further studies into these mechanisms are required to substantiate these ideas.
The degree of cognitive reserve found in individuals is strongly dependent on how well an individual can learn and their consequential level of educational attainment (Barnes and Yaffe, 2011). Learning, in turn, is in part genetically determined but also majorly influenced by the early development of the brain: Rodent studies have shown that stress during early development can cause cognitive problems later in life (Krugers et al., 2010). Similarly, it is for example known that individuals born in the Dutch Hunger Winter of 1944-1945 are at higher risk to develop AD (de Rooij et al., 2010). For these reasons, stress is thought to influence cognitive reserve which was looked at by Lesuis and colleagues. They found that daily, 15 minute separations of pups from the dams to enhance maternal care at the early postnatal period, was beneficial for cognitive performance of the mice at a later stage in life (Lesuis et al., 2017). In addition to cognitive aids, these effects seemed to delay AD pathology too, as the animals had significantly reduced amounts of Aβ plaques in the hippocampus. Enhancement of the maternal care that animals receive increases their cognitive reserve, allowing for the development of a coping system to deal with stressful events more easily (Lesuis et al., 2017).
Importantly, cognitive reserve prolongs healthy cognitive function but does not diminish the effects that Aβ will later on have on the brain of the individuals. It will merely slow down AD development. It could therefore be that a higher cognitive reserve only masks the symptoms temporarily whilst the damaging of the brain continues. However, perhaps cognitive reserve does reduce Aβ’s effects on synapses, thereby still effectively targeting a potential cause of AD. Finally, given that GluA3 is essential for Aβ-induced synaptic depression (Reinders et al., 2016a) and that GluA3 reduction in cognitive reserve only delays but cannot stop the onset of AD (Stern, 2002) by disabling Aβ interference in the synapse (Lesuis et al., 2017), controlling the GluA1/GluA3 ratio in synapses could potentially be the direction for future AD treatment possibilities. Alternatively, it could be that during AD development, Aβ has, asides of GluA3, additional ways of interfering in the synapse and blocking its function or that GluA3 slowly returns to the synapse retrieved from the receptor pool. This last mechanism seems plausible, as Takahashi and colleagues indicated that GluA2/3 is constitutively replacing AMPARs in the synapse to maintain steady synaptic transmission (Takahashi et al., 2010).
Chapter 6
Conclusion and Significance
In this thesis I have tried to describe how AD is thought to be orchestrated by the dysfunction of a number of glutamate receptor subtypes which leads to diminished LTP, learning and memory. Aβ interference in the synapse is currently the leading hypothesis of AD pathology. The pathways involved in Aβ-induced AD pathology are presently studied and are thought to be dependent on the coinciding effects of decreased GluA1 and increased GluA3 as well as a ratio shift of GluN2B towards GluN2A. The findings above indicate that it seems that the recent literature has found a relation between AD and a decrease of GluA1 coinciding with an increase in GluA3. However, this relationship of GluA3 in AD has never directly been shown in humans. Substantiating the proposed mechanism in humans, however, a gene expression study found differing levels of GluN2B and GluA3 in MCI patients compared to normal aged controls and AD patients, by correlating GluN2B and GluA3 expression levels with Mini-Mental-State-Examinations (Berchtold et al., 2014). Furthermore, GluA3 has been implicated to be essential for delaying AD pathology by preventing Aβ formation in the synapse and forming a cognitive reserve. The model for cognitive reserve explains the delay of AD pathology and can be said to oppose the effects of stress on synapses: cognitive reserve decreases the harmful effects of Aβ in the synapse, while stress increases the vulnerability of synapses to Aβ. When assessing the influence of stress in AD, which is quite well established (Dong and Csernansky, 2009; Johansson et al., 2010), mostly the influence of stress on GluN2A/B is described. The stress research has hardly touched upon the AMPAR malfunctioning in AD, mostly considering the effect of short-term stressors on AMPAR in learning (Conboy and Sandi, 2010; Krugers et al., 2010). However, since chronic and early life stress are thought to have different effects than short term stress on cognitive reserve (Lesuis et al., 2017), and on Aβ deposition in the synapse (Reinders et al., 2016a), it is interesting to pursue further research into the effects of stress on AMPAR subtype GluA1 and GluA3. Glucocorticoids have already been shown to facilitate memory formation more via GluA2 than GluA1 subunits in mice, as reviewed by Krugers and colleagues (2010), which suggests that corticosteroids regulate synaptic activity via GluA2/3 rather than via GluA1/2 AMPAR. How this is affected by Aβ, in AD and in humans is however unknown and especially because the different receptor subunit combinations have different roles it remains to be determined what the exact role of glucocorticoids is in learning as well as in AD.
Abrah´am, I., Harkany, T., Horvath, K. M., Veenema, A. H., Penke, B., Nyakas, C., and Luiten, P. G. M. Chronic Corticosterone Administration Dose-Dependently Modulates Aβ(1-42)-and NMDA-Induced Neurodegeneration in Rat Magnocellular Nucleus Basalis. Journal of Neuroendocrinology, 12(6):486–494, dec 2001. ISSN 09538194. doi: 10.1046/j.1365-2826. 2000.00475.x. URL http://doi.wiley.com/10.1046/j.1365-2826.2000.00475.x.
Abramov, E., Dolev, I., Fogel, H., Ciccotosto, G. D., Ruff, E., and Slutsky, I. Amyloid-β as a positive endogenous regulator of release probability at hippocampal synapses. Nature neuroscience, 12(12):1567, 2009.
Adamczyk, A., Mejias, R., Takamiya, K., Yocum, J., Krasnova, I. N., Calderon, J., Cadet, J. L., Huganir, R. L., Pletnikov, M. V., and Wang, T. GluA3-deficiency in mice is associated with increased social and aggressive behavior and elevated dopamine in striatum. Behavioural Brain Research, 229(1):265–272, 2012. ISSN 01664328. doi: 10.1016/j.bbr.2012.01.007. URL http://dx.doi.org/10.1016/j.bbr.2012.01.007.
Adlard, P. A. Voluntary Exercise Decreases Amyloid Load in a Transgenic Model of Alzheimer’s Disease. Journal of Neuroscience, 25(17):4217–4221, 2005. ISSN 0270-6474. doi: 10.1523/JNEUROSCI.0496-05.2005. URL http://www.jneurosci.org/cgi/doi/10. 1523/JNEUROSCI.0496-05.2005.
Akama, K. T., Albanese, C., Pestell, R. G., and Van Eldik, L. J. Amyloid beta-peptide stimulates nitric oxide production in astrocytes through an NFkappaB-dependent mechanism. Proceedings of the National Academy of Sciences of the United States of America, 95(10): 5795–5800, 1998. ISSN 0027-8424. doi: 10.1073/pnas.95.10.5795.
Barnes, D. E. and Yaffe, K. The projected effect of risk factor reduction on Alzheimer’s disease prevalence. The Lancet Neurology, 10(9):819–828, 2011. ISSN 14744422. doi: 10.1016/ S1474-4422(11)70072-2.
Bartr´es-Faz, D., Gonz´alez-Escamilla, G., Vaqu´e-Alc´azar, L., Valls-Pedret, C., Ros, E., and Grothe, M. J. Cortical Regions Underlying Cognitive Reserve Effects in Elderly In-dividuals Are Characterized By a Distinct Molecular Architecture. Alzheimer’s & De-mentia, 13(7):P583–P584, 2017. ISSN 15525260. doi: 10.1016/j.jalz.2017.07.208. URL http://linkinghub.elsevier.com/retrieve/pii/S1552526017331242.
Batsch, N. L. and Mittelman, M. S. World Alzheimer’s report. Overcoming the stigma of dementia. Executive summary. Alzheimer’s Disease International., (Mci), 2012.
Behan, D. P., Heinrichs, S. C., Troncoso, J. C., Liu, X.-J., Kawas, C. H., Ling, N., and De Souza, E. B. Displacement of corticotropin releasing factor from its binding protein as a possible
BIBLIOGRAPHY 22 treatment for Alzheimer’s disease. Nature, 378(6554):284–287, nov 1995. ISSN 0028-0836. doi: 10.1038/378284a0. URL http://www.nature.com/doifinder/10.1038/378284a0. Behl, C., Davis, J. B., Lesley, R., and Schubert, D. Hydrogen peroxide mediates amyloid ??
protein toxicity. Cell, 77(6):817–827, 1994. ISSN 00928674. doi: 10.1016/0092-8674(94) 90131-7.
Berchtold, N. C., Sabbagh, M. N., Beach, T. G., Kim, R. C., Cribbs, D. H., and Cotman, C. W. Brain gene expression patterns differentiate mild cognitive im-pairment from normal aged and Alzheimer’s disease. Neurobiology of aging, 35 (9):1961–72, sep 2014. ISSN 1558-1497. doi: 10.1016/j.neurobiolaging.2014.03. 031. URL http://www.ncbi.nlm.nih.gov/pubmed/24786631http://www.pubmedcentral. nih.gov/articlerender.fcgi?artid=PMC4067010.
Bertoni-Freddari, C., Fattoretti, P., Casoli, T., Caselli, U., and Meier-Ruge, W. Deterioration threshold of synaptic morphology in aging and senile dementia of alzheimer’s type. Analytical and quantitative cytology and histology, 18(3):209–213, 1996.
Bliss, T. V. P. and Collingridge, G. L. A synaptic model of memory: long-term potentiation in the hippocampus. Nature, 361:31–39, 1993.
Bossers, K., Wirz, K. T., Meerhoff, G. F., Essing, A. H., van Dongen, J. W., Houba, P., Kruse, C. G., Verhaagen, J., and Swaab, D. F. Concerted changes in transcripts in the prefrontal cortex precede neuropathology in Alzheimer’s disease. Brain, 133(12):3699–3723, dec 2010. ISSN 1460-2156. doi: 10.1093/brain/awq258. URL https://academic.oup.com/brain/ article-lookup/doi/10.1093/brain/awq258.
Braak, H. and Braak, E. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathologica, 82(4):239–259, 1991. ISSN 0001-6322. doi: 10.1007/BF00308809. URL http://link.springer.com/10.1007/BF00308809.
Braak, H., Alafuzoff, I., Arzberger, T., Kretzschmar, H., and Tredici, K. Staging of Alzheimer disease-associated neurofibrillary pathology using paraffin sections and immunocytochem-istry. Acta Neuropathologica, 112(4):389–404, 2006. ISSN 00016322. doi: 10.1007/ s00401-006-0127-z.
Coleman, P., Federoff, H., and Kurlan, R. A focus on the synapse for neuroprotection in Alzheimer disease and other dementias. Neurology, 63(7):1155–62, 2004. ISSN 1526-632X. doi: 10.1212/01.WNL.0000140626.48118.0A.
Conboy, L. and Sandi, C. Stress at learning facilitates memory formation by regulating ampa receptor trafficking through a glucocorticoid action. Neuropsychopharmacology, 35(3):674– 685, 2010. ISSN 0893133X. doi: 10.1038/npp.2009.172. URL http://dx.doi.org/10.1038/ npp.2009.172.
Crary, J. F., Shao, C. Y., Mirra, S. S., Hernandez, A. I., and Sacktor, T. C. Atypical Protein Ki-nase C in Neurodegenerative Disease I: PKMζ Aggregates With Limbic Neurofibrillary Tan-gles and AMPA Receptors in Alzheimer Disease. Journal of Neuropathology & Experimental Neurology, 65(4):319–326, 2006. ISSN 00223069. doi: 10.1097/01.jnen.0000218442.07664.04.
Davis, K. L., Mohs, R. C., Marin, D. B., Purohit, D. P., Perl, D. P., Lantz, M., Austin, G., and Haroutunian, V. Neuropeptide Abnormalities in Patients With Early Alzheimer Disease. Archives of General Psychiatry, 56(11):981, nov 1999. ISSN 0003-990X. doi: 10. 1001/archpsyc.56.11.981. URL http://archpsyc.jamanetwork.com/article.aspx?doi= 10.1001/archpsyc.56.11.981.
de Rooij, S. R., Wouters, H., Yonker, J. E., Painter, R. C., and Roseboom, T. J. Prenatal undernutrition and cognitive function in late adulthood. Proceedings of the National Academy of Sciences, 107(39):16881–16886, 2010. ISSN 0027-8424. doi: 10.1073/pnas.1009459107. URL http://www.pnas.org/cgi/doi/10.1073/pnas.1009459107.
Deane, R., Sagare, A., Hamm, K., Parisi, M., Lane, S., Finn, M. B., Holtzman, D. M., and Zlokovic, B. V. apoE isoform – specific disruption of amyloid β peptide clearance from mouse brain. The Journal of Clinical Investigation, 118(12):4002–4013, 2008. ISSN 0021-9738. doi: 10.1172/JCI36663DS1.
Delano-wood, L., Houston, W. S., Emond, J. A., Marchant, N. L., P, D., Thal, L. J., Jeste, D. V., and Bondi, M. W. NIH Public Access. International Journal, 23(6):632–636, 2008. doi: 10.1002/gps.1953.APOE.
Dong, H. and Csernansky, J. G. Effects of stress and stress hormones on amyloid-\beta protein and plaque deposition, 2009. ISSN 13872877.
Erickson, K. I., Weinstein, A. M., and Lopez, O. L. Physical Activity, Brain Plasticity, and Alzheimer’s Disease. Archives of Medical Research, 43(8):615–621, 2012. ISSN 01884409. doi: 10.1016/j.arcmed.2012.09.008. URL http://dx.doi.org/10.1016/j.arcmed.2012. 09.008.
Granger, A. J., Shi, Y., Lu, W., Cerpas, M., and Nicoll, R. A. LTP requires a reserve pool of glutamate receptors independent of subunit type. Nature, 493(7433):495–500, dec 2012. ISSN 0028-0836. doi: 10.1038/nature11775. URL http://www.nature.com/doifinder/10. 1038/nature11775.
Gutierrez-Castellanos, N., Da Silva-Matos, C. M., Zhou, K., Canto, C. B., Renner, M. C., Koene, L. M., Ozyildirim, O., Sprengel, R., Kessels, H. W., and De Zeeuw, C. I. Mo-tor Learning Requires Purkinje Cell Synaptic Potentiation through Activation of AMPA-Receptor Subunit GluA3. Neuron, 93(2):409–424, 2017. ISSN 10974199. doi: 10.1016/j. neuron.2016.11.046. URL http://dx.doi.org/10.1016/j.neuron.2016.11.046.
Haass, C. and Selkoe, D. J. Soluble protein oligomers in neurodegeneration: lessons from the Alzheimer’s amyloid β-peptide. Nature Reviews Molecular Cell Biology, 8(2):101–112, 2007. ISSN 1471-0072. doi: 10.1038/nrm2101. URL http://www.nature.com/doifinder/10. 1038/nrm2101.
Hardy, J. and Selkoe, D. J. The amyloid hypothesis of Alzheimer’s disease: progress and prob-lems on the road to therapeutics. Science (New York, N.Y.), 297(5580):353–6, jul 2002. ISSN 1095-9203. doi: 10.1126/science.1072994. URL http://www.ncbi.nlm.nih.gov/pubmed/ 12130773.
Hayashi, Y. Driving AMPA Receptors into Synapses by LTP and CaMKII: Requirement for GluR1 and PDZ Domain Interaction. Science, 287(5461):2262–2267, 2000. ISSN 00368075.
