Network dysfunction in Alzheimer's disease: refining the disconnection hypothesis

Abstract

Much effort in recent years has focused on understanding the effects of Alzheimer's disease (AD) on neural function. This effort has resulted in an enormous number of papers describing different facets of the functional derangement seen in AD. A particularly important tool for these investigations has been resting-state functional connectivity. Attempts to comprehensively synthesize resting-state functional connectivity results have focused on the potential utility of functional connectivity as a biomarker for disease risk, disease staging, or prognosis. While these are all appropriate uses of this technique, the purpose of this review is to examine how functional connectivity disruptions inform our understanding of AD pathophysiology. Here, we examine the rationale and methodological considerations behind functional connectivity studies and then provide a critical review of the existing literature. In conclusion, we propose a hypothesis regarding the development and spread of functional connectivity deficits seen in AD.

Keywords: Alzheimer's disease; brain networks; resting-state functional connectivity.

FIG. 1.

Schematic of Alzheimer's disease (AD) staging. Progression of AD is modeled in terms of amyloid abnormalities [measured via cerebrospinal fluid or PiB positron emission tomography (PET)], neurodegeneration (tau pathology or hippocampal atrophy), and cognitive dysfunction (measured by neuropsychological testing or clinical exam). The ordering of these biomarkers is based on prominent models in the field (Jack et al., 2010, 2013). Two classification schemes [Clinical Dementia Rating (CDR) and NIA] are incorporated here. The CDR indexes the presence or absence of dementia and scales its severity. The NIA criteria stage the presence of pathology in the absence of overt clinical symptoms. The absence of any of these three biomarkers is termed “healthy aging.” NIA1–3 is termed “preclinical AD” and CDR>0 is termed “clinical AD.” PiB, Pittsburgh compound B.

FIG. 2.

Summary maps of the extent and magnitude of amyloid and tau deposition and functional connectivity disruption parametric on disease severity. In both amyloid and tau columns, color indicates intensity of deposition as previously described (Braak and Braak, ; Thal et al., 2002). Amyloid patterns in preclinical AD are based on longitudinal PiB PET studies (Okello et al., ; Villemagne et al., 2011). Since tau imaging data do not currently exist (i.e., no longitudinal follow up in pathological studies), it is impossible to determine the topography in patients who would eventually develop clinical AD, so that each data point is omitted. Clinicopathological correlation for amyloid (Braak and Braak, 1991) and tau (Nelson et al., 2009) is based on previous studies. Functional connectivity can either be increased (pink/purple) or decreased (green-grey). No formalization for intensity of functional connectivity deficits exists, so colors indicate number of consecutive disease stages in which greater abnormalities are detected. For example, if FC in a particular region decreases in two consecutive AD stages, the value would be −2. This attempts to synthesize the entire literature but is particularly based on Zhou et al. (2010), Brier et al. (, Wang et al. (2013b). These effects are compared with age-matched controls. While aging alone is associated with some change in functional connectivity (Andrews-Hanna et al., 2007), observed changes due to AD are above and beyond changes typically associated with aging. Note: The hippocampus (HC) is shown in a box labeled “HC,” because it is not visible on these surface projections. FC, functional connectivity.

FIG. 3.

Network models of brain function in AD. (A) A sample of a potential brain network containing four distinct functional regions (A–D). Region A receives inputs from other parts of the brain not depicted (large arrow) and projects to regions B and C (small arrows), which also receive inputs from other brain regions not depicted (large arrows). Brain regions (B and C) then project to region D (small arrows), which also receives inputs from other parts of the brain (large arrows). Each of these brain regions is healthy (denoted by green color) and the connections are intact (solid black lines for arrows). (B) Disconnection model. In this model, a pathological insult (lightning bolt) affects a single brain region (region A) that is completely destroyed (red color). The connections from region A to regions B and C are rendered non-functional (dotted line). In this model, the insult to region A is a fully resolved event and region A no longer contributes to the network. Regions B and C still receive inputs from the rest of the brain and communicate with region D. The observable deficits are strictly related to region A and do not progress. (C) In the proposed model, a similar insult occurs in region A; however, the region is not rendered completely silent (partial green color) and continues to communicate with regions (B and C) in a disrupted way, leading to induction of dysfunction in regions (B and C). In addition, it is known that histopathology can spread along physical connections (blue star). These two processes may overlap and may be additive (e.g., region B). The newly affected regions then provide disrupted output to region D.