Early Failure of the Default‐Mode Network and the Pathogenesis of Alzheimer's Disease

The Resting Brain

In contrast to baseline brain activity, only a small percentage of neurons is required to respond to external stimuli. Such task‐related activation usually accounts for <5% of the total blood‐oxygen‐level‐dependent (BOLD) signal, while a very substantial amount of functional connectivity exists even in the absence of a task 1. Independent component analysis 2, 3 applied to the brain's resting‐state data (termed resting‐state functional connectivity MRI or fcMRI) identified at least eight major cortical networks: the primary vision network, the higher‐order visual processing network, the hearing network, the touch and movement network, the salience processing network, two networks for executive control and the default‐mode network (DMN) 4, 5. It seems that most of the baseline activity is mediated by neurons that are constantly active and participate in the DMN (also known as the resting‐state or task‐negative network) 6, 7.

Default‐Mode Network

Although the exact relationships between functional and structural connectivity are not always straightforward 8, the DMN is anatomically defined by structural interconnections among several epicenters that include parts of the medial prefrontal cortex (mPFC) and the posterior cingulate cortex and adjacent ventral precuneus and parietal cortex 9, 10, 11. The DMN undergoes developmental changes and is characterized by coherent neuronal oscillations at a rate lower than 0.1 Hz that become more consistent in children 9–12 years old and in older subjects 12. It appears that DMN functional connectivity reflects the level of consciousness, generates spontaneous thoughts, and integrates multimodal sensory and affective information. It is preferentially activated when individuals engage in internal tasks such as daydreaming, envisioning the future, and eventually retrieving episodic memories, while it shows negative correlation with brain systems that focus on external (environmental) visual signals 13. Interestingly, and in spite of their distinct evolutionary paths and their small size, even rodents possess a DMN that is broadly similar to the DMN of nonhuman primates and humans 14.

Various functional neuroimaging methods can be used in the study of DMN functional connectivity. FcMRI provides insight into the brain activity by the measurement of cerebral blood flow of certain brain regions while the subject is not performing any task 15. Positron‐emission tomography (PET) measures alterations in brain metabolism by the usage of specific radioisotopes like ¹⁸F‐fluorodeoxyglucose (FDG). Single photon‐emission computed tomography (SPECT) uses direct photon‐emitting isotopes rather than radioisotopes for the detection of blood flow that reflects the activity of brain regions 16. Additional functional neuroimaging methods, such as electroencephalography (EEG) and magnetoencephalography (MEG), can also be used in the study of functional connectivity 17.

How to Explain Differences between Familial and Late‐onset, Sporadic AD?

Dementia is a complex, multidimensional syndrome, which makes diagnosis and testing of novel treatments for its primary causes, most notably Alzheimer's disease (AD), difficult. Recent advances in functional and structural MRI, including diffusion tensor imaging (DTI) tractography, have enabled the precise delineation of multiple large‐scale distributed brain networks as well as identification of patterns of neuronal dysfunction in progressive neurodegenerative diseases. Moreover, PET radiotracers [¹¹C]‐labeled Pittsburgh Compound‐B (PiB), [¹⁸F]‐labeled florbetapir, [¹⁸F]‐labeled flutemetamol, as well as FDDNP (2‐(1‐{6‐[(2‐[¹⁸F]fluoroethyl)(methyl)amino]‐2‐naphthyl}ethylidene)malononitrile) have enabled detection of neuropathologic AD changes in vivo, either amyloid deposits/senile plaques (SP) alone or both amyloid and abnormally phosphorylated tau inclusions (neurofibrillary tangles, NFT, and neuritic plaques, NP), respectively 18, 19, 20, 21, 22. Two more amyloid‐imaging PET agents are in phase III clinical trials ([¹⁸F]‐labeled florbetaben and [¹⁸F]‐NAV4694) and two more that exclusively bind to tau deposits ([¹⁸F]‐labeled T808 and [¹¹C]‐labeled phenyl/pyridinyl‐butadienyl‐benzothiazoles/benzothiazolium [PBB3]). In a recent study that used the isotropic fractionator technique for counting neurons, Andrade‐Moraes et al. 23 however showed that the neuronal loss in hippocampus and neocortex, but not the presence of NFT and SP, is directly associated with dementia. This may explain the presence of NFT and SP in asymptomatic subjects (in whom reduced neuronal numbers have not been detected).

Late‐onset AD is the most common form of AD affecting over 90% of patients 24. It is also known as sporadic AD because there appears to be no single genetic factor or family link involved. The likelihood of developing late‐onset AD doubles every 5 years once a person has reached 65. So, early AD (also known as familial AD) and late‐onset AD differ because of the link to a family history and the existence of genetic mutations in specific genes in the familial forms. However, the symptoms and the progression of the disease appear to be the same even though the causes may be fundamentally different 24.

Mutations in genes for APP, PSEN1 (presenilin 1), and PSEN2 (presenilin 2) are associated with early‐onset familial AD. Since APP, PSEN1, and PSEN2 are directly involved in Aβ processing (APP as Aβ precursor, PSEN1 and PSEN2 as parts of γ‐secretase complex), impairment of their function leads to increased production of Aβ_1–42 24, 25. Thus, the central initiating factor in familial AD seems to be an increase in the amount and/or toxicity of Aβ, while the major proximate cause of neurodegeneration is the misprocessing of tau protein, most probably via interactions with Aβ (for an extensive review, see 26). Bhatia and Hall 26 even proposed a unifying hypothesis that explains the link between Aβ and tau in AD, and the relationship between familial and sporadic AD. Overproduction of Aβ near dendrites in familial AD causes the disruption of synaptic function and associated signal transduction, and also the production of toxic tau fragments from somatodendritic, nonmicrotubule associated tau. Interaction between Aβ and tau further leads to cell cycle re‐entry that activates dedifferentiation. This disrupts axons, dendrites, and synapses leading to tau delocalization from axons. On the other hand, events like axonal injury (caused by head trauma) can compromise the function of axons. Abnormally phosphorylated tau further accumulates in the soma and dendrites causing the disruption of somatodendritic signaling pathways and amplification of Aβ‐tau toxic interactions. The resulting event is tau aggregation. According to this hypothesis, changes in Aβ metabolism cause tau misprocessing that triggers additional neuronal changes, resulting in neurodegeneration 26.

In sporadic cases of AD, a more extensive interplay between predisposing factors, such as traumatic axonal injury, insufficient production of apolipoprotein E (dependent on APOE genotype) and decreased clearance of Aβ, and other known and yet unknown factors, probably occurs 26, 27. Using a novel stable isotope‐labeling tandem mass spectrometry (SILT MS) technique, Mawuenyega et al. 28 observed a decrease in clearance of Aβ from the brain of patients with sporadic AD. They hypothesized that impaired clearance of Aβ in addition to decreased Aβ catabolism, APOE4 genotype, impaired cerebrospinal fluid (CSF) transport, and transport across the blood‐brain barrier may be the cause of sporadic AD. Additionally, impaired brain glucose metabolism has also been shown to facilitate abnormal tau hyperphosphorylation in cultured cells and in mouse models of sporadic AD by at least two different mechanisms, abnormal glycosylation of tau 29, and impaired insulin receptor signaling cascade 30.

Why is the DMN Particularly Relevant for Studying Dementia and AD?

Both cognitive performance and the functional connectivity of the DMN decrease in aging [31, for review see 32]. When compared to younger adults, elderly subjects show a gradual decrease in task‐induced deactivation in the posterior cingulate cortex, as well as parietal and frontal cortical regions, suggesting an age‐related reduction in the ability to suspend DMN activity when the experimental condition requires focused attention (mostly through activation of the dorsal attention network [DAN]) 9, 33, 34, 35. In one of the studies, BOLD signal was corrected for grey matter atrophy but the observed decreases in DMN functional connectivity remained significant, indicating that aging‐related changes in connectivity are not solely associated with reductions in grey matter volume 36. Decreased DMN connectivity has also been documented in patients with mild or preclinical AD, particularly in posterior cingulate cortex and precuneus (it was again present when corrected for grey matter atrophy) 37, and was associated with severity and progression of AD 38, 39, 40.

Other reasons to study DMN in dementia and AD are DMN neuronal vulnerability to early metabolic reduction in very early AD 41, early deposition of Aβ in DMN hub regions, and abnormalities in fcMRI of DMN. In addition to deposition of Aβ measured by both ¹¹C‐labeled PiB‐PET 42 and more amyloid‐sensitive 43 ¹⁸F‐labeled florbetapir‐PET 44, abnormalities in fcMRI of DMN may represent early functional impairment associated with risk for both mild cognitive impairment (MCI) and AD. It is considered as a promising candidate for a biomarker of preclinical dementia, and potentially a valuable tool in dementia prediction, prognosis, and prevention 45, 46, 47, and in identifying AD risk populations 48. The recent study of Reiman et al. 49 demonstrated that fMRI changes are detected in young adults with predisposition of early‐onset familial AD (PSEN1 E280A mutation carriers). Greater activation of the hippocampus and parahippocampal regions in addition to less posterior cingulate and precuneus deactivation can be detected very early, even before the Aβ deposition. Chhatwal et al. 50 also showed that changes in DMN occur very early in PSEN1, PSEN2, and APP mutations carriers. Decreased precuneus, posterior cingulate, and parietal cortices activation occurs in these patients. Metabolic reduction in AD was reviewed by Sheline and Raichle in the context of the DMN 51. The regions with the highest aerobic glycolysis almost totally overlap with DMN areas. As the greatest amount of energy produced in the brain is through the oxidative phosphorylation (coupled to aerobic glycolysis), these recent findings on decreased aerobic glycolysis in DMN regions are in line with the previous studies on impaired aerobic glycolysis 52 and mitochondrial function in AD 27.

Vulnerable Cortical Neurons in AD

Both age‐associated cognitive impairment and AD are characterized by vulnerability of the same cortical circuits 53. However, the neurobiological substrates for functional impairment in aging differ in important ways from those in AD 54, 55. While age‐associated cognitive impairment is mediated mostly by synaptic alterations in otherwise more or less intact circuits, AD is characterized by more extensive structural pathological changes and neuron death 53. Although no satisfactory explanation for the neuronal specificity of the degenerative process in AD is available, it has been established that an identifiable subpopulation of large, cortically projecting, glutamatergic pyramidal neurons in high‐order association regions of the neocortex is preferentially affected in the course of the disease 53, 54. These neurons share a particular morphological and neurochemical phenotype characterized by their size, distribution and connectivity, and their high expression of dephosphorylated epitopes of neurofilament proteins. They are affected by NFT formation at very early stages of AD, degenerate at a faster rate than other pyramidal neurons and shrink significantly in the process 56, 57, 58, 59.

Concurrent and Early Involvement of Subcortical Nuclei in the Pathogenesis of AD

Three views on anatomical selectivity and early involvement of the large projecting neurons of subcortical nuclei 60 in pathogenesis of sporadic, late‐onset AD have recently been proposed. Based mostly on a large series of postmortem brains, Braak and Del Tredici 61, 62 hypothesized that the earliest pathological changes in AD are the appearance of hyperphosphorylated tau (as revealed by antibody AT8) within the nerve cells of brainstem nuclei, particularly those of the noradrenergic LC and the serotonergic DRN, and those in the cholinergic NB. These changes occur in the context of a total lack of brain Aβ deposition, even in the form of diffuse deposits. According to this hypothesis, pretangle projection neurons drive the formation of Aβ, release of Aβ from nonjunctional varicosities of their axons, and accumulation of Aβ in their projection areas in the form of extracellular plaques. Mann and Hardy 63 criticized this tentative set of events from two major points: (1) while antibody AT8 recognizes tau hyperphosphorylated at Ser202 and Thr205 residues and is thus an excellent antibody for detecting “generic” tau, it does not differentiate between different isoforms of tau (either with 3‐ or 4‐repeat binding domains), suggesting that those changes observed by Braak and Del Tredici in younger people might not represent “Alzheimer's tau” changes, but may reflect other tauopaties (certain forms of early‐onset dementia including frontotemporal dementia, progressive supranuclear palsy, corticobasal degeneration, etc.), consequence of head injury, viral conditions, etc. and (2) they restated arguments speaking in favor of the amyloid cascade hypothesis so that “the formation of Aβ is the primary driving force behind the pathogenesis of AD with tau pathology following as a consequential, or at least secondary, event”. The third view on possible mechanisms of AD pathogenesis formulated by Attems and Jellinger 64 is closer to Mann and Hardy's view in that subtle LC tau pathology in younger individuals (as revealed by Braak and collaborators using AT8 immunocytochemistry) does not necessarily represent AD‐associated tau pathology per se, but may be a reversible condition. In addition, Attems and Jellinger stressed that only the simultaneous presence of both Aβ and moderate to severe amounts of tau pathology in a specific topographical pattern constitutes a valid index for the neuropathological correlate of AD.

These views did not consider, however, recent in vivo brain imaging findings, particularly in relation to an early decrease in the functional connectivity of DMN in patients with preclinical AD 65, 66. Here, we discuss how these new findings may reconcile the previously proposed mechanisms. We also extend speculations on a more coherent pathogenetic series of events in both familial and late‐onset, sporadic AD.

Correlation of Aβ Deposition by PiB PET and DMN Functional Connectivity

It has been shown that DMN cortical hubs are the same cortical regions that harbor high amounts of Aβ deposits 42 and that clinically normal subjects with high amyloid burden in the DMN display significantly reduced functional correlations within it, even when controlled for age and structural atrophy 65. Due to the disruption of the DMN functionality and its interactions across multiple regions, including the hippocampal formation (particularly between the posterior cingulate cortex and hippocampal formation), the hippocampal formation becomes functionally disconnected from DMN in the preclinical phase of AD 65. These findings were confirmed by Sheline and coworkers 67, who found that the congruence of DMN functional connectivity and PiB uptake overlap was highest in the posterior cingulate cortex, precuneus, and mPFC, again suggesting that early manifestations of Aβ toxicity can be detected using fcMRI 67.

Further studies also noted that PiB uptake is always generally more diffuse than the confines of the DMN 68. In a recent cross‐sectional cohort study, Wang and coworkers found that among 207 older adults, the decreased CSF values of Aβ_1–42 (<500 pg/mL) and increased CSF values of p‐tau₁₈₁ (more than 80 pg/mL) were associated with significant reduction in DMN functional integrity 66. The most prominent decreases in functional connectivity were observed between the posterior cingulate and medial temporal regions, and were not attributable to age or structural atrophy 66. These findings suggest that both Aβ and tau pathology affect DMN integrity before the clinical onset of AD.

Because processing of the amyloid precursor protein (APP) is activity‐dependent 42, where regional increases in neuronal activity are associated with regional increases in the concentration of Aβ in the interstitial fluid 69, it can be speculated that due to their constant activity, DMN neurons produce and release more Aβ than occur elsewhere in the neocortex, thus leading to an increase in production, oligomerization, and aggregation of Aβ as well as tau hyperphosphorylation, which presumably is caused by the released Aβ oligomers 70, 71, 72.

It is well established that the cells generating Aβ are chiefly projection neurons, either cortical or of those of brainstem nuclei 60, 61. We have recently proposed that the current clinical criteria for diagnosis of AD are focused mostly on cognitive deficits produced by dysfunction of hippocampal and high‐order neocortical areas. At the same time, noncognitive, behavioral, and psychological symptoms of dementia such as disturbances in mood, emotion, and appetite, together with confusion, agitation, and depression, and deficits in wake–sleep cycle have been less considered. The early occurrence of these symptoms suggests brainstem involvement, and more specifically of the serotonergic nuclei. Several recent reports, including ours, drew attention to the possibility of selective and early involvement of raphe nuclei, particularly the DRN, in the pathogenesis of AD 60. In fact, recent data suggested that plaque formation in the brain of APPswe/PS1ΔE9 mouse model of AD causes the deterioration of sleep–wake cycle and loss of diurnal fluctuation of Aβ measured in the interstitial fluid 73. As it is a well‐established fact that lesions of the raphe nuclei cause insomnia and are directly associated with the sleep regulation 74, the pathology of raphe nuclei (at pretangle stage) could be associated with Aβ increase in two ways: first, through Aβ release from raphe projection axons 61, 62 and second, through possible sleep deprivation that leads to inadequate clearance of Aβ in sporadic AD 73, 75, thus creating a vicious circle (less sleep causing more Aβ to be formed and more Aβ accumulated leading to less sleep). Recent findings confirmed and broaden these initial observations. Namely, it has been demonstrated that even cognitively normal individuals with biomarker evidence of preclinical AD have worse quality of sleep and sleep efficiency than control individuals 76.

As DMN connectivity persists during light sleep because self‐reflective thoughts do not abruptly cease but rather decrease gradually as a person falls asleep, what particularly matters is the duration of the deepest stages of slow‐wave sleep (SWS) during which the activity of DMN is virtually absent and cerebral metabolic rate declines by 43.8% in comparison with wakefulness and REM sleep 77, 78.

Since the introduction of a staging system for extracellular amyloid deposits and intraneuronal neurofibrillary changes by Braak and Braak in 1991 79, it has been confirmed that (1) in the first stage of AD (stage A), amyloid deposits are first encountered in the isocortex and particularly in the basal portions of the temporal, frontal, and occipital lobe (while the hippocampal formation remains devoid of amyloid), (2) AD cases with severe neurofibrillary changes consistently show high densities of amyloid deposits, while those rich in amyloid do not always turn out to be affected by neurofibrillary changes (and only a marginal percentage of cases with moderate neurofibrillary changes do not reveal the presence of amyloid), (3) accumulation of this pathological material begins before the appearance of clinical symptoms (preclinical phase), (4) since deposition of amyloid is among the first changes seen in the cerebral cortex, this has led to the assumption that they precede the development of neurofibrillary changes, and (5) in preclinical AD cases, dystrophic neurites are characterized by the accumulation of neurofilaments, including dephosphorylated forms, and these abnormal processes have the morphology of the proximal and distal segments of physically damaged axons 80. Thus, the deposition of Aβ fibrils in the extracellular space inflicts physical or mechanical stress to nearby axons, triggering the neuronal reaction to injury in the form of neurofibrillary degeneration 81.

Within this framework, we hypothesize that subcortical nuclei first release Aβ in the temporal, frontal, and occipital isocortex (as shown by Braak and Braak 79, stage A of amyloid deposition), and as a secondary event, retrograde and anterograde neurofibrillary changes develop or accelerate in DMN neurons due to the physical compromitation of their axons in the regions where passing through fibrillary amyloid deposits (Figure 1). This also explains why the behavioral and neuropsychiatric symptomatology described is not strongly correlated with amyloid accumulation or DMN alterations in early AD and could also help to explain differential vulnerability to neurodegeneration within cortical regions of the DMN early in the disease. An additional reason for later frontal cortex involvement during the development of AD may be its apparent decoupling from other parts of DMN during sleep, particularly deep, SWS sleep 77.

Figure 1.

Simplified scheme of Aβ release from large projecting neurons of subcortical nuclei—noradrenergic locus coeruleus (LC), serotonergic dorsal raphe nucleus (DRN) and cholinergic nucleus basalis (NB) complex that could induce neurofibrillary changes within vulnerable glutamatergic pyramidal neurons in DMN hub regions (medial prefrontal cortex, posterior cingulate cortex, ventral precuneus, and parietal cortex).

Conclusion

Several recent studies have revealed that DMN cortical hubs exhibit high amounts of Aβ deposits in AD 42, 66, 67 and that clinically normal subjects with high amyloid burden in these regions have significantly reduced functional correlations within the DMN. It is therefore possible that due to DMN neurons' enhanced activity‐dependent processing of APP, these neurons may produce and release more Aβ than occur elsewhere in the neocortex. In turn, this can lead to an increase in production, oligomerization, and aggregation of Aβ in a hierarchical sequence 82, as well as tau hyperphosphorylation, which presumably is caused by the released Aβ oligomers. As the cells generating Aβ are probably not only DMN cortical, but also some brainstem projection neurons in LC and DRN, as well as the magnocellular NB, this may add to the complexity and variability of AD‐related pathological changes. Namely, DMN cortical regions are richly innervated by long projection fibers from the LC, DRN, and NB. Unlike the cerebellum and striatum (in which neurofibrillary changes do not develop in spite of the presence of amyloid deposits), these cortical projection areas contain glutamatergic projection neurons. This may make these regions more vulnerable, and deposition of Aβ fibrils in the extracellular space by physical or mechanical stress to nearby passing axons may cause AD‐related tau pathology, that is, neurofibrillary degeneration of DMN projection neurons. Within this framework, it can be concluded that familial AD is caused by an increase in production, oligomerization, and aggregation of Aβ and abnormal processing of tau protein, most probably via interactions with Aβ (not excluding other contributing factors), whereas in sporadic AD, a more complex combination of predisposing factors, including decreased Aβ clearance, as well as traumatic axonal injury, APOE genotype, disruption of the sleep–wake cycle, and yet unknown factors, likely induces pathological changes of DMN projection neurons.

Conflict of Interest

The authors declare no conflict of interest.